TW200541280A - Transmitter predistortion circuit and method therefor - Google Patents

Abstract

A digital communications transmitter (100) includes a digital linear-and-nonlinear predistortion section (200, 1800, 2800) to compensate for linear and nonlinear distortion introduced by transmitter-analog components (120). A direct-digital-downconversion section (300) generates a complex digital return-data stream (254) from the analog components (120) without introducing quadrature imbalance. A relatively low resolution exhibited by the return-data stream (254) is effectively increased through arithmetic processing. Distortion introduced by an analog-to digital converter (304) may be compensated using a variety of adaptive techniques. Linear distortion is compensated using adaptive techniques with an equalizer (246) positioned in the forward-data stream (112). Nonlinear distortion is then compensated using adaptive techniques with a plurality of equalizers (226) that filter a plurality of orthogonal, higher-ordered-basis functions (214) generated from the forward-data stream (112). The filtered-basis functions are combined together and subtracted from the forward-data stream (112).

Description

200541280 IX. Description of the invention: [Technical field to which the invention belongs] [0001] This patent is a predistortion circuit and compensation method and a digital radio frequency communication transmitter. On May 6, an application was filed in the United States, Xu County 1 () Qian 735, and the part of the scale "out-of-wire control surface communication transmitter and its design method" was extended and continued by Yu Gangyi, the inventor of this patent. The application was filed in the United States on May 27, serial number 10/766. Both are incorporated herein by reference. [0002] This patent is related to the trade of Yan Man You 10 / 971,628, filed in the United States on October 22, 2004), "Methods for Compensating Linear Distortion of Women's Frequency Communication Transmitters in Trade and Material Roads, and Gamma Power This good watch style is related to the current, poor, pre-distortion circuit and digital RF communication transmitter compensation method for non-linear distortion "(serial number 10 / 766,779, filed in the United States on January 27, 2004). All were invented by the inventor of this patent. [0003] The present invention is generally related to digital radio frequency communications. More specifically, the present invention relates to controlling and reducing errors in digital communication signals generated by analog parts of the transmitter. [Prior Art] [0004] Digital communication transmitters can take advantage of low prices and use digital processing for communication signals in large numbers. Even relatively wideband communication signals can be expressed digitally and processed accurately with digital methods at reasonable prices. The digital representation of a signal is obtained by providing a stream sample that is suitable for its bandwidth and meets the required resolution. However, the communication signal of digital representation has traditionally been converted into an analog form, upscaling, filtering Zoom in for analog parts for transmission. 200541280 [] 7 pieces of analogy are not like digital parts, they can only have limited accuracy, and even if the accuracy of the analogy is poor, it is still more expensive, and to achieve greater accuracy, it costs more than $ The trend of communication transmitters is to extend the digital processing to the antenna as much as possible instead of analog processing, and then broadcast the RF signal from the antenna. [0006] Two trends are recorded, which require amplifying the Lai Wei formula and using analog parts that are more convenient and less accurate. The use of modulation forms that require linear amplification has its needs. Because of the secrets, they are allowed to convey more information at a fixed time, fixed bandwidth, and fixed transmission. The use of cheaper parts is always the ideal goal, but the demands of large markets and / or highly competitive markets are also important goals. [_71 Linear power amplifier is the most responsible and most power-consuming analog component in the transmitter, but when the linear converter cannot fully copy and amplify its input signal, it will cause signal distortion. In addition, the distortion is more serious when the silk-priced and low-power amplifier is made. [0008] A type of power amplifier distortion that has received considerable attention is its non-linearity. Non-linearity is a particularly obvious characteristic of a linear power amplifier, and it refers to the degree to which the output of the power amplifier cannot be linearly related to the input of the amplifier. The non-linearity of RF transmitters is particularly troublesome because of the problem of spectrum regeneration. Perhaps the RF input of an amplifier may be completely within a specific electromagnetic frequency spectrum, but the non-linearity of any amplifier will cause intermodulation, and the RF output of the amplifier will span a large electromagnetic frequency spectrum. [0009] The transmitter should preferably be able to use the maximum bandwidth allowed by the statute in order to transmit information efficiently. Therefore, spectrum regeneration often leads to transmitter violations. To avoid violating the regulations, 6 200541280 linear power amplifier, it is best to precisely increase the amount of miscellaneous rails with miscellaneous amounts. Digital communication transmitters are facing another trend: standards and regulations continue to tighten the transmitter operation spectrum, so the need to reduce the non-linear spectrum regeneration of power amplifiers is more urgent than ever. [0010] One method to solve the non-linear spectrum regeneration of the power amplifier is to use a higher power amplifier and operate the higher power amplifier with a larger backoff. Backoff is the degree to which an amplifier produces a weaker signal than it can produce. Generally speaking, when the power amplifier operates below its maximum capacity, it will become more linear; and a larger backoff will maintain the amplifier operation in a more linear range. This method not only requires the use of more expensive and higher power amplifiers, but also typically requires the amplifier to be operated in a less efficient range, which causes the transmitter to consume more power. This problem becomes more significant when the communication signal exhibits a higher peak-to-average power ratio, such as when several digital communication signals are mixed before being amplified. Furthermore, mixing several digital communication signals before amplification is common in applications such as mobile phone base stations. [0011] Another method to solve the non-linear problem of power amplifiers is through digital predistortion. Digital predistortion has been used for digital communication signals to allow the use of less expensive amplifiers and to increase the effectiveness of more expensive amplifiers. Digital pre-distortion means that when the communication signal is still in digital form, digital signal processing is performed before analog conversion. The digital signal processing test distorts the digital communication signal in an accurate manner so that the result of the communication signal can be as accurate as possible after adding the error of the linear amplifier and other analog processing. The non-linearity of the amplifier is corrected by digital pre-distortion, so that lower power and cheaper amplifiers can be used. The amplifier can also be more efficient and reduce the low frequency spectrum regeneration. Moreover, since digital predistortion is performed through digital signal processing, it can perform any premiss function with great accuracy at a reasonable price. 7 200541280 [0012] Although the previous predistortion technology has been somewhat successful, the results are limited, and the recent legal requirements are more stringent, resulting in insufficient traditional predistortion technology. [0013] The predistortion technology needs to know the way in which the analog part distorts the communication signal in order to make an appropriate anti-predistortion transfer function and accurately compensate the distortion produced by the analog part. The more accurately the traditional pre-distortion technology uses the feedback from the output of the power amplifier to obtain real-time information, the more accurately it reflects the actual analog parts and operating conditions. [0014] Traditionally, in order to monitor this feedback signal, a large amount of processing needs to be performed to obtain a #distortion conversion function. Then, the inverse distortion conversion function is calculated and converted into a program, and a digital predistorter is placed. In many traditional applications, the transmitter needs to transmit a preset series of training data to reduce complexity and improve the accuracy of the large amount of processing required to derive the distortion transfer function. The traditional predistortion technology with less accuracy or narrow frequency may at most only construct a _digital predistorter as a simple communication signal filter H to perform the inverse conversion function. But in many more accurate and often more expensive traditional systems, the digital predistorter itself includes one or more wire detectors. This table is used to define the predistortion program, so that the digital predistorter can Add to communication signal •. _] U-stage application of technology to compensate for memory effects is more complicated and more expensive. In general, the memory effect refers to the behavior of a power amplifier in one environment that is different from another environment. For example, the gain of a power amplifier and its conversion characteristics may vary as a function of video rate, power amplifier bias condition, temperature, and part aging. In order to solve the memory effect, pre-missing in and out of order includes multiple recording silk and Xu? Dealing with soft inverse transformation memory functions combined with inverse conversion functions and changing the predistortion instructions accordingly becomes more complicated. 200541280 [0016] A wide range of traditional pre-distortion techniques suffer from a variety of problems. The use of training sequences is particularly undesirable because it requires the use of frequency spectrum for control, not for load purposes, and this usually increases complexity. Generally speaking, in the path of the feedback signal and in the design of the predistorter, the complexity of the processing is increased to improve the accuracy, but a large increase in complexity often only improves a little accuracy. The increase in the processing complexity of the feedback signal is not desirable because it will result in increased transmitter costs and power consumption. Following the traditional digital pre-distortion technology, the cost of digital pre-distortion quickly reaches or even exceeds the cost of using a higher power amplifier operating at a higher backoff to achieve essentially the same result. Therefore, digital pre-missing has traditionally been more practical in high-end applications, and even then, it can only be of limited effectiveness. [0017] More precisely, the use of conventional techniques for processing feedback signals suffers from particularly annoying problems. Traditionally, an inverse operation is used to form an inverse conversion function for the digital predistorter program. Although the inverse operation itself may be somewhat complicated and more serious, it is sensitive to small errors in the feedback signal. Even small errors through an inverse operation can cause very inaccurate results for the inverse conversion function. Φ [0018] Using the traditional predistortion technology, the feedback signal should be captured very accurately to accurately calculate the inverse conversion function. Using traditional predistortion technology, this requires a high-precision analog / digital conversion circuit (A / D) 'to capture the feedback signal, and then connect to a high-resolution, low-error digital circuit to process the feedback signal. What's more troublesome is that the feedback signal usually exhibits an expanded bandwidth due to the non-linearity of the power amplifier due to the spectrum regeneration. To accurately capture the extended bandwidth of the feedback signal using traditional techniques, the A / D should also be composed of high-speed circuits. However, such high-speed and high-resolution A / Ds are often expensive and high-power parts. They will negate the power savings of any power amplifier with digital pre-distortion, except for the highest-order applications. [0019] In order to avoid the use of high speed and high resolution, some traditional predistortion techniques continue to use gods that only process parts outside the width of the feedback signal. However, the power outside of the feedback signal bandwidth only indirectly explains the distortion of the analog parts, which increases the error and reduces the accuracy again in the inverse conversion function. [0020] Even when the traditional design uses a high-speed, high-resolution dove capture feedback signal, it is still unable to control other error sources, so that the inverse conversion function may cause considerable φ inaccuracy after the inverse operation. ^ The phase jitter of the clock increases the error, as it would happen before any analog processing before A / D conversion. Moreover, the traditional habit requires digital communication signals with complex signals of in-phase and quadrature parts, and the feedback signals of this part before A / D conversion are traditionally processed separately. The quadrature imbalance introduced by any feedback 彳 § number will cause further errors, and after the inverse operation, it will cause significant errors in an inverse conversion function. [0021] The linear distortion of communication signals introduced by analog parts is believed to be another source of error that plagues traditional digital predistortion techniques. Linear distortion refers to the error of the signal, which is reproduced or introduced by the power amplifier and falls within the bandwidth. Examples of linear distortion include the imbalance of quadrature gain, phase, and group delay. In addition, when the frequency of the communication signal becomes wider, the frequency-dependent gain and phase change impose greater linear distortion effects. More examples of linear distortion include certain types of signal image and intermodulation. Linear distortion is often considered a more benevolent error than non-linear distortion because it does not cause spectral regeneration. In general, the receiver has compensated linear distortion after transmitting the channel, and at the front end of the receiver, further linear distortion has been added to the analog parts, but in at least one example, the communication system has been combined into a receiver to determine some linear distortion Correct the partner, then send the 200541280 shooter for some corrective actions. [0022] The linear distortion of a linear transmission communication signal needs to be reduced because it will reduce the amount of linear distortion that the receiver must compensate for and improve performance. Moreover, as the bandwidth of the communication signal becomes larger, it is more necessary to reduce linear distortion. However, the improvement method used by a receiver to set the transmitter should not work because it cannot separate channel-induced distortion from the transmitter-induced distortion. Because multiple paths often impose dynamic effects on the transmitted RF communication signals, this method is often unsuccessful. In addition, it wastes spectrum of emission control data, not load data, and it requires many receivers to achieve compatibility. [0023] Traditional transmitters are not only unable to solve the linear distortion problem, they are also believed to cause further errors when characterizing non-linear transfer functions. Most algorithms that convert raw data to a transfer function are based on a model of the amplifier that is reasonably accurate under control conditions, but using this model to derive the transfer function using a linear distortion number, especially over a wider bandwidth, may violate Control status. Therefore, the conversion function derived here is believed to be relatively inaccurate, and any inverse conversion function calculated for the digital predistorter may be very inaccurate because of this. [0024] In some digital communication applications (such as mobile phone base stations), a broadband communication signal is composed of several independent narrowband signals' to form a broadband communication signal by frequency multiplexing. This situation poses a challenge to the predistortion circuit. In this application of multiple narrowband signals, some narrowband signals will show weaker than other signals. In general, communication system specifications insist that all transmission channels must meet a minimum vector size (EVM) or signal-to-noise ratio. Claim. Therefore, there is a need to use predistortion and other transmission processing that meet the requirements of this weak and strong channel at the same time. [Summary of the Invention] 11 200541280 [0025] At least one embodiment of the present invention can provide an improved transmit predistortion circuit and method. [0026] In another embodiment of the present invention, a quantization error compensator may be provided to compensate for the quantization error caused by the analog digital conversion circuit (A / D) in the feedback signal generated by the analog transmitter component. [0027] Another embodiment of the present invention may provide a procedure for compensating the distortion of the feedback signal path before using the feedback signal path to cancel the distortion generated by the analog transmitter parts. [0028] The present invention also provides at least one other embodiment, which can provide a method of responding to the relative strength of the frequency multiplex communication channel to offset the distortion generated by the analog transmitter parts. [0029] These and other benefits are realized in the form of a method of compensating for distortions generated by analog transmitter parts. This method requires obtaining a forward data stream that is configured to transmit multiple frequency division communication channels. The previous data stream is processed through the analog transmitter parts. After the forward data stream is affected by the analog transmitter parts, it responds with the return data stream. Each communication channel responds to the forward data stream with its communication signal strength, and each communication channel responds to the forward data stream with its error signal. The distortion generated by the analog transmitter parts is offset by the response of the aforementioned communication channel to the strength of the communication signal and the strength of the error signal. [0030] These and other benefits are realized in the form of another predistortion circuit that compensates for distortions produced by analog transmitter parts. The predistortion circuit includes an adaptive equalizer configured to receive a forward data stream and generate a processed forward data stream. A digital / analog converter (d / a) is coupled to the adaptive equalizer and is configured to convert the processed forward data stream into a forward analog signal for transmission through an analog transmitter part. An analog / digital converter (A / D) receives an analog signal returned from the 12 200541280 analog transmitter part. The returned analog signal responds to the forward analog signal, and the A / D is set to generate a returned original digital data stream. An a / D compensation section is used to receive the forward data stream. This A / D compensation section is set to compensate the error generated by the A / D into the original digital data stream and returns a return data stream. A control circuit is coupled to the adaptive equalizer, and the control circuit is configured to change the processed forward data stream in response to the returned data stream. [Embodiment] [0031] FIG. 1 shows a block diagram of a digital communication radio frequency (Rp) transmitter 100, which is configured according to the present invention. The transmitter 10 () is a transmitter that can be used in a mobile phone or a mobile base station, but the transmitter 100 can also be used in other applications. [0032] The transmitter 100 provides a plurality of digital information streams 102 to correspond to the plurality of digital modulators 104. In the application of the mobile base station, the information stream 102 may transmit the information to be transmitted to multiple users, and the different information streams 102 may be related to others or may not be related at all. [0033] The modulator 104 may perform any kind of digital modulation, but the advantages of the present invention can be realized from the form of the modulation. The amplitude and phase angle of this modulation are used to digitally convey information This modulation typically requires the use of linear high-power amplifiers (HPA, S). Examples of such modulation include any quadrature amplitude modulation (QAM), code division multiple acquisition (CDMA), orthogonal frequency division modulation (OFDM), multiple input multiple output (MIM0) systems, and so on. In the preferred embodiment, the modulation information output from the modulator 104 uses a complex information flow to digitally convey the message. Those familiar with this technology will understand that a complex information stream consists of two parallel streams. Figure 丨 uses the traditional name to plot one of the streams as an in-phase ① stream, and the other as a quadrature (Q) stream, reflecting the orthogonal relationship between the two streams when they are processed and merged downward. Although not indicated here, the modulation 13 200541280 may include a pulse-forming filter configured to be intersymbol-interference (ISI) known to those skilled in the art 'to minimize it, and after modulation Other forms of signal processing. [0034] In one of the preferred embodiments, the modulator 104 is coupled to the merging part 106, and multiple complex modulated information streams are combined into a single digital communication signal, which is referred to herein as the complex Forward Information Flow 108. For ease of explanation here, the complex forward information flow 108 and all subsequent variations of the downstream processing between the merged portion 106 and an antenna are used for transmission processing. Hereinafter, the forward information flow is collectively referred to as the following discussion. The difference is the return flow of information traveling in the opposite direction. Even if the information stream 102 is a narrow-band information stream, the combined complex forward information stream 108 can also be regarded as a wide-band information stream. One of the consequences of merging different modulation information streams is that the peak-to-average ratio of the complex forward information stream 108 will increase, so that the downstream must cope with higher linear amplification. [0035] One of the outputs of the merged portion 106 is coupled to the peak reduction portion 110, and the peak reduction portion 110 reduces the peak-to-average ratio of the forward information flow 108 so that the complex peak of the result reduces the forward information flow 112 It will make the downstream only need to deal with smaller linear amplification. In the preferred embodiment, the peak reduction section 110 uses a peak reduction or peak reduction technique that only produces distortion in the frequency band of the forward information stream 108. Therefore, where the complex peak reduces the forward information flow 112 or other application peak reduction results, there will be no significant spectrum regeneration. [0036] In addition, the peak reduction part 110 needs to use the peak reduction under control so as to respond to the peak reduction feedback signal 114. In particular, the feedback signal U4 may provide a residual non-linear EVM value that can be converted by the peak reduction portion 110 into a critical value, which indicates the minimum amplitude that the forward information flow 108 needs to display before any peak reduction is used. Generally speaking, when the size of the forward information stream 108 exceeds this threshold a lot, a large amount of peak reduction will be applied to 200541280 on the forward information stream 108. The increase in peak reduction may be achieved by reducing the size of the threshold. The peak reduction is applied to the forward information flow 108. This will add greater internal bandwidth distortion and enter the peak to reduce the forward information flow. The effect of 112. Techniques suitable for peak reduction are described in U.S. Patent Nos. 6,104,761 and 6,366,619, both of which are incorporated herein by reference, but techniques not described herein may be used. [0037] In a preferred embodiment, the feedback signal 114 represents the amount of residual non-linear distortion transmitted from the transmitter 100 to the radio frequency communication signal 116. The development of the feedback signal 114 is discussed below. In one of the preferred embodiments, the peak reduction portion 11 is operated so that when the amount of non-linear distortion exceeds the peak reduction amount added to the forward information flow 108 is in accordance with a preset value Will increase. The design requirement of the transmitter 100 is that under normal steady-state operation, the amount of non-linear distortion of the RF communication signal 116 will not be too large, and the entire error vector size (EVM) is slightly smaller than the maximum allowed by the system specifications. However, abnormal operating conditions may cause excessive non-linear distortion, and as a result, spectrum regeneration exceeding the regulations and EVM specifications may result. [0038] Therefore, the feedback signal 114 has the ability to control the amount of distortion of the radio frequency communication signal 116, so that most of the distortion is not within the bandwidth and a few is not outside the bandwidth. The feedback signal 114 allows the peak reduction portion 110 to increase the peak reduction, and then causes the HPA 136 to operate with a larger backoff. Operating the PPA 136 with a larger setback has the effect of reducing nonlinear distortion and radiation outside the bandwidth. However, increasing the peak reduction will also increase the distortion in the bandwidth. Therefore, the total distortion can be maintained at a predetermined value, but its characteristics will shift from outside the bandwidth to within the bandwidth. [0039] If there is a peak reduction portion 110, it can be used as the source of the previous information flow 12 of the linear and non-linear predistortion circuit 200. The predistortion circuit 20 uses a number of characteristics to deliberately generate linear and non-linear distortion into the forward information stream 11: 2 through the number 15 200541280 bit signal method. These multiple features are discussed below. The feedback signal 114 is generated by the predistortion circuit 200, and after the predistortion circuit 2000 processes it, the 'forward information stream 112 becomes a complex orthogonal balanced equalized forward information stream 118. For the broadband signals represented by forward streams 108 and 112, forward stream 118 now represents an ultra-wideband signal. The forward manager 118 not only conveys the wide ship signal reduction, but also the anti-inversion distortion generated by the predistortion circuit, and this predistortion circuit 200ffl compensates for the nonlinear distortion generated by the analog part 12 in the future. [00401 In the embodiment, a selectable peak reduction control signal 114 can be provided to the predistortion circuit from the peak reduction portion m. The peak reduction system signal 114 represents the amount of noise in the forward information stream 112 added by operating the peak reduction portion 110 through estimation, measurement, or calculation. In the presently preferred embodiment, the value reduction control signal 114 is used to convey short-term average noise, and this short-term average noise is added to the complex information stream output from the modulator 104 as described above for each independent modulation, # The amount of waves determined by the low-pass wave energy at the same time. The use of the peak decrement signal 114 'is discussed below with FIG. 29. [0041] The analog part 12 includes a digital / analog converter (D / A) 122 for each point of the complex forward information flow 118. The frame 122 converts the forward information stream 118 from digital to analog signals, and the subsequent bidirectional communication > fs number processing is analog processing, and is affected by the characteristics of the analog processing and is erroneous. For example, two different D / A 122s may not have the same gain, and they may produce slightly different delays. The difference between this gain and the delay may cause linear distortion of the communication signal. Furthermore, as long as complex signals are not separated, pulses and parts are separated, Zwei may add different frequency responses, making the linear distortion worse due to the increase in frequency phase gain and delay imbalance 200541280. And 'the frequency-dependent gain and delay imbalance worsens because the communication signal bandwidth becomes wider. [0042] The two complex signals of the analog signal pass through the two low-pass filters (LpF, s) of D / A 122. 124 «LPF's 124 can be a slightly different linear distortion with slightly different gain and delay shifts in addition to slightly different frequency dependent characteristics. From LPF, S 124, the complex signal of these two analog signals leads to a direct orthogonal up-conversion section 126. The up-conversion part 126 mixes these two complex signals with a seismic signal, and obtains it from a 12g seismic signal in a manner familiar to the person skilled in the art. Another linear distortion in the manner of gain and delay imbalance can also be generated, and the leakage of this earthquake can produce unwanted DC displacement. In addition, the up-conversion section 126 combines two different complex signals and passes the combined signal, which is now called a radio frequency analog signal 13 and reaches a band-pass filter (BPF) 132. For cost reasons, the 126 part is preferably up-converted, at least up to a frequency below about 2.5 GHz. For higher frequencies, multi-stage upscaling can be used. [0043] BPF 132 is in the radio frequency analog signal, now called analog signal 130, which is configured to block unwanted sidebands, but also in communication signals, which are now called radio frequency analog signals 134, 41 and generate a phase angle. Delay. The radio frequency analog signal 134 drives a power amplifier 136, which is also traditionally referred to as a high power amplifier (HPA). The HPA 136 is coupled to the antenna 138 and generates an amplified radio frequency analog signal, referred to above as the radio frequency communication signal 116. [0044] HPA 136 may be the source of several linear and nonlinear distortions in the communication signal. Figure 1 is HPA 136 using the Wiener-Hammerstein RF amplifier model. It can be used to explain some of these distortions, at least ideally. Control status of the signal. according to

Wiener-Hammerstein's HPA model, HPA 136 is like an input band-pass filter (BPF) 17 200541280 140, followed by a memoryless non-linearity, and labeled amp 142 in Figure 1, and then connected to the output band-pass filter ( BPF) 144. Amp 142 produces an output signal, which may be an input higher-order complex polynomial function. Each of the BPF's 140 and 144 may produce a linear distortion, but may be a small non-linear distortion. On the other hand, amp 142 is a source of non-linear distortion. [0045] In a preferred embodiment, the linear and non-linear predistortion circuit 200 receives at least three or four analog input signals. One of them is the local oscillation signal, and the rising frequency part 126 is used as the rising frequency. The other is an optional feedback signal, which is output from at least one or two sections of the complex signal of D / A, s 122. This output is shown in the figure. 1 is labeled as the fundamental frequency (BB) signal 123; the other analog input signals are the feedback signals derived from the radio frequency analog signal 134, which is used as the input signal of the HPA 136, and the radio frequency communication signal 116 through a direct coupling element 115, which is used as [0046] By monitoring this feedback signal, the linear and non-linear predistortion circuit 200 learns how to apply predistortion to reduce linearity and then non-linear distortion. Although there are several different sources of distortion, the actual properties of the analog parts that cause the distortion change slowly. This allows the circuit 200 to perform a prediction and convergence procedure to determine the appropriate predistortion characteristics, and this procedure allows a slow convergence speed. . The use of prediction and convergence procedures reduces the complexity of processing and reduces the sensitivity of the feedback signal to errors. Furthermore, the use of a slow convergence speed allows the circuit 200 to reduce the actual error of the feedback signal 'and obtain accurate predistortion characteristics. Since the error of the feedback signal can be tolerated, the feedback signal can be processed using a low-resolution circuit, so the purpose of reducing circuit parts and saving power is achieved. [0047] FIG. 2 shows a block of a first embodiment of a linear and non-linear predistortion circuit 200 200541280 FIG. 'First embodiment' is referred to as a linear and non-linear predistortion circuit 1800, which is connected to FIG. 18 and It is discussed below; and the third embodiment, referred to as a linear and non-linear predistortion circuit 2800, is concatenated with FIG. 28 and discussed below. A complex forward information stream u] configured to convey digital information is added to input port 202 of circuit 200. Compared with the return information flow discussed below, the forward information flow U2 shows a lower resolution, as indicated by the letter "H," as shown in Figure 2. Those familiar with this technology will understand the resolution decision at least among them The part is represented by the number of bits sampled by each forward information stream 112. For more south-resolution information streams, it is common to use # samples with more bits than low-resolution information streams for transmission. Similarly The forward information stream 112 shows relatively low levels of error from quantization noise, phase angle error, etc. As discussed above, any previous information stream 112 is basic, and the signal to the analog part 120 is also considered to be A form of forward information flow. Since the forward information flow 112 flows through the predistortion circuit 200, it can maintain the characteristics of high resolution and low error level. [0048] In a preferred embodiment, the forward direction The information stream 112 is connected to the rate doubler 204. In this preferred embodiment, the forward information stream 112 only transmits the digital communication signals at the base frequency and needs to flow at an information rate that supports Nyquist conditions. In the preferred embodiment, Continued processing of the forward information stream 112 will generate higher frequency components to compensate for non-linear distortion. Therefore, the rate doubler 204 will increase the information rate for the highest frequency to be generated at least equal to and preferably greater than the Nyquist condition. So far, the forward information stream 12 can be regarded as an ultra-wideband information stream, and the rate doubler 204 can be performed by a person familiar with this method using interpolation; or if the nonlinear compensation is to be omitted, the rate doubler 204 It can be omitted. [0049] The forward information stream is output from the doubler 204 and passed to a high-pass wave filter (HpF) 205, which is configured to filter only the DC component. The high-pass filter 205 preferably has to join the return information stream. 200541280 The same filtering characteristics of another high-pass filter, as discussed below. The high-pass filter 205 can also be located before the rate doubler 204, as shown in the following figure, and associated with FIGS. 18 and 25, or at other equal positions. [0050] A rate-increasing complex forward information flow 206 flows from the high-pass filter 205 to the delay portion 208, the basic function generation portion 160, and a thermal change evaluation The estimated part is 1700. The basic function generating part 1600 is used to connect with the non-linear compensation. If the non-linear compensation is omitted, the basic function generating part can also be omitted. The basic function generating part 1600 generates multiple complexes. Basic Function Information Flow® 214, and each of the Basic Function Information Flows 214 reacts to χ (η) · | χ (η) | κ, where χ (η) represents the forward information flow received by 160%, and K is an integer greater than or equal to one. Therefore, the 1600 part generates multiple higher harmonics from the forward information stream 206. The basic function information stream 214 provides the highest order basic function information stream 214 (ie, Has the largest κ value), the information stream 214 is connected to the input of a multiplexer (MUX) 222. The basic function generating part 16⑻ will be discussed in detail with FIG. 15 and FIG. 16 below. [0051] Similarly, the thermal change evaluation section 1700 is used to link with the non-linear compensation. If the “non-linear compensation” is omitted, the thermal change evaluation section can also be omitted. In general, the heat exchange evaluation section 1700 generates a -thermal difference signal (Δ_Η_ 216, which is used to explain the relative power in the forward information flow 206 'to characterize the instantaneous change in the accumulation of ΗΡΑ 136 relative to the long-term average heat. The thermal difference signal 216 is then used to affect the basic function information flow 214 to compensate for the thermal §memory effect of a typical price ρα 136. The thermal change evaluation section 1700 will be discussed in more detail below with FIG. 15 and FIG. 17 [ 0052] In a preferred embodiment, all basic function information streams 214 display a time delay equal to tf in the basic function generation section 1600. Therefore, a complex forward information stream output from the delay section 208 20 200541280 218 has the same clock in each basic function information stream 214, including the highest-order basic function information stream 214 '. The complex forward information stream 218 output from the delay section 208 is connected to a merging circuit 22 and a night job. The second input of the converter 222. The merging circuit 220 is due to the complex subtraction circuit of FIG. 2, each section of the complex signal path has a subtraction element, and the complex forward information flow 218 is connected to the positive input terminal of the subtraction element. ] All basic function information flow 214 To a non-linear predistorter 224, and if the non-linear compensation is omitted from the transmitter 100, the complex basic function information flow can also be omitted. The non-linear predistorter 224 includes multiple equalizers (EQ) 226, The first equalizer 226 provides the basic function information flow 214. Fig. 2 shows the relationship between the equalizer 226 and the one or two basic functions, one or three basic functions, etc., to (κ + 1) basic functions. Each equalizer 226 is a multiple equalizer, as shown in FIG. 12 for equalizer 1200, and the output of each equalizer 226 and equalizer 1200 are combined to form a complex filtering basic function information stream 230. Information Stream 230, used as a non-linear predistortion compensation stream, is connected to the subtraction input of the merging circuit 220. [0054] For the purpose of this description, any equalizer such as one of the equalizers 226 is programmable. Filtering The device determines how it changes the signal it processes by specifying its coefficients. In the preferred embodiment, it is intended to use a wavelet of multiple complexity. Each equalizer 226 may have a contact or any greater than Its number. The adaptive equalizer is configured to determine its own The wave coefficient, and the coefficient of constant inclination is continuously changed. It is an equalizer that accepts the filter wire without changing the coefficient until the next burning. However, as discussed below, it is hot in some places. The difference signal 216 may cause the non-adaptive equalizer filter coefficient to change. [0055] In a preferred embodiment, its equalizer 226 is a non-adaptive equalizer. However, when coupled to the -adaptive engine 1300 At the time, the combination of the equalizer 226 and the non-adaptive equalizer will form a new equalizer 21 200541280. Each equalizer 226, other equalizers included in linear and non-linear predistortion circuits, and adaptive The engine 1300 belongs to the equalizer part. In a preferred embodiment, the adaptation engine 1300 selectively couples or decouples with multiple equalizers located in the equalizer section 234 at any time to determine filter coefficients by executing an estimation and convergence procedure. FIG. 2 depicts the selective matching or decoupling of the non-linear predistorter 224 and the 236 characteristic within the adaptive engine compensation. The thermal difference signal 216 is one of the signals connected to the adaptive engine 1300, and the thermal difference signal 216 is also the input of the non-linear predistorter 224-. The equalizer 226, the adaptation engine 13⑽, and the estimation and convergence program executed are discussed in detail below, as shown in Figure U_13. [0056] The output of the merging circuit 220 provides a complex non-linear pre-distortion forward information flow. In the embodiment of the present invention, the forward information stream 238 drives a differential mode time positioning part. If the linear compensation is omitted from the transmitter 100, the time positioning part 800 may also be omitted. Time positioning part 800 of the forward information flow 238! In the Q complex band, a different amount of time delay is added to compensate for the -reverse difference time delay '. This difference time delay may be generated through the analog part 12o. The time positioning part 800 is discussed in detail below, as shown in Figures 5 and 8. The output of the time localization part 800 is used to locate a forward information flow 242 when a complex difference is generated to drive a linear predistorter 244. On the other hand, the time positioning part 8⑻ may be located after the linear predistorter 244 if necessary, instead of being front as shown in FIG. Also, if linear compensation is omitted from the transmitter 100, the predistorter 244 may also be omitted. [0058] The linear predistorter 244 performs several adjustments on the forward information stream 242. For example, the linear predistorter 244 performs an orthogonal balance adjustment portion. Therefore, the linear predistorter 244 adjusts the gain and phase angle of the I and Q complex frequency bands of the complex forward information stream 242, and independently adjusts the I and Q frequency bands of the complex forward information stream 2 magic 22 200541280 so as to be orthogonal. Balance can be done. In addition, the linear predistorter 244 compensates for fresh quadrature gain and chirp imbalance. Therefore, even the wideband and ultra-wideband communication signals discussed above are orthogonally balanced through the linear predistorter 244. [0059] In a preferred embodiment, the linear predistorter 244 uses a complex equalizer to implement this, and this equalizer can be configured as an equalizer 120, but it is likely that there will be more connections. point. If there are enough contacts, the differential mode time positioning part 800 can be omitted entirely. The equalizer 2 is labeled EQF, where the subscript "F" indicates that the equalizer 246 will pass forward information. As discussed above for the equalizer move #, the equalizer 246 is used as part of the equalizer section 234. Also, the equalizer 246 needs to be a non-adaptive equalizer, so that it will become an adaptive equalizer when coupled to the adaptive engine 1300 through the 236 characteristic. By appropriately selecting the forward data wave (that is, the wave coefficient of the forward equalizer eQf) to the equalizer 246, the linear predistorter 244 can compensate the predistortion generated by the gamma ratio component 120. The forward filter coefficient is determined by the-training process, which will be discussed below, as shown in Figure 17 and 19. After training, in addition to correcting the phase angle and gain imbalance and distortion of the I and Q segments that are convenient with frequency, the forward filtering can also be used as an orthogonal balance coefficient or parameter. Spring GG6G] The linear predistorter 244 generates a complex orthogonal balanced equalized forward information stream 18 and passes it to the analog part i20. The forward information flow 118 must maintain the characteristics of high resolution and low error level as upstream information. It has been distorted in the preferred embodiment to compensate for the non-linear and linear distortion of the communication signal that has not been added by the analog part 120. Furthermore, it must support the above discussions, including the fundamental frequency signal plus the super-bandwidth of higher harmonics, but other facts may still benefit from compensating for only linear distortion or only non-linear distortion. [0061] Referring to FIG. I, the feedback from the analog part 12 is obtained through the feedback signals m and 134 23 200541280. The feedback signal 117 is obtained from the radio frequency analog signal output from the HPA 136, and the feedback signal 134 is obtained from the radio frequency analog signal input from the HPA 136. Returning to FIG. 2, the feedback signals 117 and 134 are sent to a feedback part 248 of a linear and non-linear predistortion circuit 200 of a multiplexer 250. The feedback section 248 also includes a digital down-conversion section 300, which is used to receive the output from the multiplexer 25. The down-frequency section 300 also receives the same local oscillator signal from the local oscillator 128, which is used by the up-frequency section 126. The frequency reduction section 300 first reduces the frequency of the feedback signal 134 for use in training the linear and non-linear predistortion circuit 200 to compensate for various types of linearity loss caused by the input signal of the HPA 136. Then, the frequency reduction section 300 frequency-reduces the feedback signal 117 for use in training the linear and non-linear predistortion circuit 200 to compensate for various types of linear distortion generated by the signal output by the HPA 136. The frequency reduction section 300 will be discussed further below, as shown in FIG. 3. [0062] The frequency reduction section 300 generates a multiple return information stream 254. As shown by the letter "L" in Fig. 2, the 'return information stream 254 shows a low resolution and a high degree of error compared to the multiple forms of the forward information stream. For the convenience of the discussion here, all information flows from the analog part 120 and based on the return information flow 254 are regarded as a form of return information flow. ® [0063] The recurrent information stream 254 drives an adjustable attenuation circuit 256. The adjustable attenuation circuit 256 must be used as a trimmer or cursor to control or decide how to attenuate the return information stream 254 to compensate for the attenuation provided by the HPA 136 to the forward propagation communication signal and the attenuation provided by the coupler 115. The adjustable attenuator 256 can be implemented by a multiplexer. [0064] The adjustable attenuator 256 generates an attenuated complex reverse information stream 258 to a equalizer 260, which can be configured as equalizer 1200, but most likely will have more contacts. Figure 2 adds the label EQR to the equalizer 260, and the subscript "R" indicates that the equalizer 260 filters the return information stream. As discussed in 24 200541280 and other equalizers 226 and 246, the equalizer 260 is part of the equalizer part 234, and the equalizer 260 must be a non-adaptive equalizer. When it passes through the 236 characteristic When connected to the adaptation engine 1300, it becomes the -adaptive equalizer. Appropriately adjust the return filter coefficients that enter the equalizer 26 (that is, the filter coefficients of the return equalizer EQR). The linear distortion mainly generated by the HPA 136 itself will be compensated. And the form of the linear distortion generated by this HPA 136 itself Does not contaminate subsequent training to compensate for non-linear distortion. The return filter coefficient is determined through the training process, which will be discussed below, as shown in Figure 1M4. # [_5] Huahua 11 260 produces-equalizes the return information stream 262 to maintain relatively low resolution and high error as discussed above. Using low-resolution processing of return streams can result in power and part savings. [0066] An output of one of the multiplexers 222 is used to drive a common-mode time positioning section 700. The time positioning part 700 is selected according to whether it is the multiplexer 222, and the same amount of delay is added to the I and Q complex frequency bands of the forward information stream 218 or the highest-order basic function information stream 214 '. In addition, the number of delays when the time positioning part 700 is inserted can be adjusted. The time positioning part 700 generates a time delay to forward the information stream 266, and the time positioning part 700 is programmable, so that the time positioning of the stream 266 can be consistent with the return information stream 262. The time positioning part 700 will be discussed in more detail below, as shown in Figure 5-7. [0067] The time delay complex forward information stream 266 is connected to a phase conversion part 1000 and the first data input of a multiplexer (MUX) 270. The phase-conversion part 1000 rotates the time delay complex forward information stream 2 by a variable amount, and generates a positioning complex forward information stream 272. The phase conversion part 1000 is programmable, so that the information stream 272 can be in phase with the return information stream 262 to compensate for the time when the analog part 12 of 25 200541280 filters 132, 140, and / or 144 are brought. Delay. Phase inversion will be discussed in more detail below, as shown in Figures 5 and 9-10. [0068] The positioning complex forward information stream 272 is connected to the second data input of the adaptation engine 300 and the multiplexer 27, and the positioning complex forward information stream 272 and the equalization return information stream 262 are connected to a complex The combining circuit 274 forms two subtraction elements at the edge of FIG. 2. The merging circuit 274 subtracts the return information stream 262 from the forward information stream 272 to form an error signal or error stream 276. The equalized return information stream 262 and error signal stream 276 are connected to a multiplexer (mux) 278, and the thermal difference signal 216 is the same. In addition, the error stream 276 is connected to the third information input terminal of the multiplexer 270 and the adaptation engine 1300, and a difference coefficient (△ • coeff) signal 279 generated by the adaptation engine 1300 is connected to the fourth of the multiplexer 270 Information input. [0069] The outputs of the multiplexers 270 and 278 are each connected to a correlation engine 280, in particular, the outputs of the multiplexers 270 and 278 are supplied to different information input terminals of a complex multiplier 282, and An output terminal is coupled to an input terminal of an accumulator 284. Through multiplexers 27 and 278, multiple different information streams can be related to each other in the correlation engine 280. The multiplier 282 performs a basic correlation operation, and the correlation results are integrated in an accumulator 284. One of the information flows performed by the correlation engine 28o is based on the return information flow and displays the low resolution and high error levels as described above. [0070] In a preferred embodiment, the accumulator 284 must allow a large number of accumulations (e.g., samples between 216 and 224) in order to be able to process several times the number of samples before making a decision based on the correlation results. The low resolution and high error level of the "like" return information stream can be ignored, so that the result after integration can have a smaller effective error level. -In general, as long as the noise is more or less unrelated, the noise variation of the sampled signal decreases as the square root of the average number of samples increases. Therefore, for example, the effective error level of the return information stream is reduced. , Which is equivalent to the accumulation of fine error levels. Samples with a value of approximately 106 increase by 10 bits of resolution (ie, about 60 dB). Network diagram 2 has a controller 286 and several inputs and outputs. The simplified block diagram of 2 does not show that 'the input and output surfaces are connected to multiple sub-parts of the linear and non-linear predistortion circuit 2' to provide control data here and read data from there. For example, the controller controls the multiplexers 278 and 270 'to specify which information flow or signal must be associated at the start of the correlation engine 28o, and the output from the accumulator 284 of the correlation engine 28o is connected to the control器 shift. A person familiar with this technology may use any conventional microprocessor or provide it to the controller 286 ′. Therefore, the controller 286 may execute electronic software commands stored in a memory section (not shown). In an embodiment, the controller 286 may provide control functions to the linear and non-linear predistortion circuit 200 and other parts of the transmitter 100. The work performed by the controller 286 and the linear and non-linear pre-distortion circuit 200 due to the control effect of the controller 286 will be further discussed in FIGS. 4-6, 9, η, and 14-15. [0072] FIG. 3 shows a block diagram of a digital down-conversion section 300, which is suitable for the linear and non-linear predistortion circuit 200 of the transmitter 100. [0073] Part 300 receives a radio frequency analog input from the multiplexer 250 and connects this input to a programmable analog attenuator 302. The control input of the attenuator 302 determines the amount of attenuation provided by the attenuator 302 and is provided by the controller (C) 286. The attenuator 302 must be used as a rough adjustment in conjunction with the adjustable digital attenuator 256 to attenuate the signal level of the return information stream 254, to compensate for the gain and facet of the communication signal that was previously transmitted before adding ΑΡΑ m. Provided attenuation. [0074] One output of the attenuator 302 is coupled to an input of an analog / digital converter (α / 〇) 304. In addition, the same local oscillator used by the up-conversion part 126 is used to input a 3⑻ part And received at the synthesizer 306. The synthesizer 306 must be configured to multiply the cost of the oscillator frequency by four, and divide the result of the product by an -odd number, such as 2N ± 1, where N is a positive integer that satisfies the Nyquist condition of the above ultra-wideband signal and is usually greater than or Equal ten. As a result, _304 performs a direct down-conversion through sub-harmonic sampling. [0075] In an embodiment, the 300 part may include an average power calculator (not shown), which will respond to the return information stream 254. The average power must be maintained at a certain level, so the analog attenuator 302 can be adjusted to the optimal A / D 304 load in response to the average power, and the adjustable digital attenuator 256 can then be adjusted to be close to equal to the analog attenuator 302 plus The mutual gains of all. This keeps the overall gain at a constant. [0076] The direct sub-harmonic sampling down-frequency conversion process performed by A / D 304 requires a / d 300 to be capable of fast conversion. In addition, the sub-spectral wave sampling process tends to add thermal noise from several fundamental frequency harmonics to the final fundamental frequency signal, thus increasing noise compared to other types of down-frequency conversion. Although these factors are a problem in many applications, the 300 part is not too much of a burden because, as discussed above, only low resolution is required here. In addition, the low resolution requirement of A / D 304 does not have a special burden on the phase angle-noise in the clock generated by the synthesizer 306, or the aperture error unique to A / D 304. This low resolution requirement can be allowed due to the operation of multiple estimation and convergence procedures discussed below. The final averaging effect can reduce noise, phase angle errors, and / or aperture errors. 28 200541280 [0077] In particular, the A / D 304 only needs to provide a resolution of at most four bits less than the forward resolution of the forward information stream 112 through the linear and non-linear predistortion circuit 200. In one embodiment, A / D 304 may be implemented with a resolution that provides only one or two bits. As discussed above, multiple techniques, such as prediction and convergence procedures and integration, can be used to convert increased processing time to reduce the effective error level of the return information stream. Therefore, since the multiplication of samples before making a decision based on the feedback signal, and without a sample of one or even a small or medium-sized population, has a significant impact on the decision based on the feedback signal, the low resolution is effectively increased. The high quantization error φ and high thermal noise error do not cause significant problems for the linear and non-linear predistortion circuit 200. [0078] In a preferred embodiment, the linear and non-linear predistortion circuits 2000 are generally manufactured using a semiconductor substrate 1 *. However, the high speed requirements of the a / D 304 and the synthesizer 306 can be provided using a silicon germanium process comparable to the CMOS process. [0079] The processing of the feedback signal upstream of the A / D 304 has been achieved through the use of analog technology 'and is therefore affected by the inaccuracy of the analog process. However, A / D 304 provides a digital information stream, and subsequent processing is not affected by the inaccuracy of the analog. This digital information stream uses the complex feedback signal as the characteristic # as the combined signal of the I and Q bands. The subsequent processing is to place the sub-harmonics appropriately at the fundamental frequency and separate the I and Q bands of the complex signal. Although this is an independent follow-up to the complex signal and Q-band, it is processed digitally, so there is no linear distortion due to quadrature imbalance, and / or gain and phase angle characteristics of various frequencies. [0080] In particular, the digital information stream from the A / D 304 is connected to a demultiplexer (DEMUX) 308, which is used to divide the information stream into even and odd sampled information streams. One of these even and odd sampled information streams is delayed only in the delay element 310, and the other is converted in the conversion part 312. 29 200541280 The output of the 7G part 310 and the part 312 is filtered by a high-pass filter (11? 1 ^) 314 to remove the direct current into a wound 'and then used as the return information stream 254 as a whole. Of course, when they propagate through the 300 part, the information flow rate is slower, and the clock signals have been properly divided (not shown) to support the reduced information rate. The high-pass filter 314 is matched with the high-pass filter 205. [0081] FIG. 3 depicts a form of a complex digital harmonic sampling down-converter, which is suitable for use as a digital down-frequency section 300. But those that are discarded in the design can be used for direct digital subsampling down-conversion '. Although frequency reduction is necessary, because this does not produce different analog errors # Entering the 1 and Q bands, which can cause linear distortion, in higher frequency applications (such as higher than 2. 5 GHz) frequency reduction can be divided into two phases, and the first phase is analog frequency reduction. In this case, the distortion caused by the analog frequency reduction in the first stage will be less obvious, because it is used for a relatively narrow bandwidth, which is only a part of the carrier frequency. [0082] FIG. 4 shows a first embodiment of a transmitter distortion management process 400. Process 400 and its sub-processes and the sub-processes it includes are performed by the controller 286 executing software familiar to those in the same industry. The second embodiment of process 400 is discussed below and is referred to as process # 1900, as shown in FIG. A third embodiment of process 400 is discussed below and is referred to as process 3200, as shown in FIG. [0083] The process 400 may be initiated when the transmitter 100 is turned on, or any time after the transmitter 100 is in operation. Generally speaking, the analog component 120 adds distortion to the RF communication signal 116 from various sources; in other words, the RF communication signal 116 can be regarded as showing multiple distortions instead of a single distortion. Not only are there differences between linear and non-linear distortions, there are many different causes for linear distortions. Process 400 trains linear and non-linear predistortion circuits 200 to compensate for the worst loss one by one 30 200541280 True, training is performed by using an estimation and convergence program so that complex processing can be avoided, and the feedback signal is Sensitivity can be reduced, but calculation of forward conversion functions and inverse functions can be avoided. [0084] Process 400 first performs subprocess 500 to compensate for linear distortion added upstream of HpA 136. [0085] FIG. 5 shows a flowchart of a sub-process 500. The process 500 first performs an initial job 502. In particular, both the forward equalizer 246 and the reverse equalizer 260 set filter coefficients so that they pass through without changing the forward and reverse information streams, respectively. The adaptation engine 1300 is decoupled from all equalizers. The adjustable attenuators 256 and 302 are adjusted to a gain equal to one (that is, no gain and no attenuation), and the selection control value is provided to the multiplexer 250A, so that the RF analog feedback signal 134 (RIM) is connected to the down-frequency section. 300. The basic function is to control the predistorter 224 to produce a certain value regardless of the input. The multiplexer 222 is controlled to connect the forward information stream 218 to the time positioning part 700. Correlation Engine (CE) 280 is added to multiplexers 278 and 270 to select an appropriate value, and is configured to correlate the "ideal" delay forward stream 266 with return stream 262. The time-delay forward information stream 266 can be considered ideal because it has not been distorted by the predistortion circuit 200 or the analog part 120. Time positioning Part 700 and 800 perform time positioning, set it to the middle value, and remove the thermal difference signal 216 processing function. At this point, the linear and non-linear predistortion circuit 200 is ready to begin training linear compensation. [0086] After job 502, job 504 starts a secondary process 600 to execute an estimation and convergence procedure. In particular, the sub-process 600 executes this routine in task 504 to provide a programmable delay component provided by the common-mode time positioning section 700. Therefore, the sub-process 600 will now delay the time 31 200541280 to delay the forward information flow 266 and the return information flow 262 for time positioning. After work 504, work 506 again starts sub-processing 600 or a comparable process to re-execute the time positioning estimation and converge the program, but this time for the programmable delay provided by the differential time positioning section 800. element. At work 506, the sub-processing 600 performs time positioning on the I and Q frequency bands of the forward information flow 238. [0087] FIG. 6 shows a flowchart of a process 600 that can be used for each of the tasks 502 and 504, and the time positioning part 700 and 800. In the common mode operation 504, the time positioning part 700 control can adjust the time delay added to the delay complex forward information flow 266, but in the differential mode work 506, the time positioning part 800 control The relative delay added to the time-positioned return information stream 262 and one of the Q bands can be adjusted. [0088] The sub-process 600 performs a task 602 to correlate the correlation engine (CE) 28o to the "ideal" time-delay complex forward information stream 266, and returns the information stream 262 at a suitable choice for multiplexers 270 and 278. [0089] Next, work 604 sets association convergence conditions. The convergence condition determines how many samples the correlation engine 280 needs to converge to a correlation result for correlation and integration. As discussed above, processing more samples can lead to a greater increase in the effective ruggedness of the return information stream, or a reduction in error levels. Therefore, an increase in program processing time translates the return information stream into a reduced effective error level. Through work 604, the convergence rate is controlled to achieve the preset effective return error level, which is smaller than the error level of the return information stream. One example is that it is possible to process a sample of 106 to achieve an improvement in the signal-to-noise ratio of 60 volumes. Of course, the sub-processing 600 does not need to set different convergence conditions in different situations, but the correlation engine 280 can use the same 32 200541280 conditions for all settings to do hardware settings. In this case, the task 604 is performed by the correlation engine 280 instead of the controller 286. [0090] After work 604, the sub-process 600 performs a query work 606, which determines when the correlation engine 280 has converged to a correlation result. During work 604, the correlation engine 280 processes multiple multiple samples. By using a preset delay element, a correlation is made between the return information stream and the delay direction ^ stream, and the initial value of the set delay element is an intermediate value.丨 ⑽91] When the correlation result occurs, the initial work 6G8 is followed by—the initial step of the large step positive displacement • Beginning estimation ’Bribery—coming soon_binary search program. "Big, step refers to the calculation time of the upcoming binary search program, relative to the time delay added by the previous association." Positive, displacement refers to the time delay of the upcoming calculation, which will be greater than the previous An arbitrary value. After work 608, a work 610 adjusts the programmable time positioning hardware (part 700 or part 800) to reflect the current step size and displacement direction. [0092] FIG. 7 shows a block diagram of an embodiment of the common-mode time positioning part 700. This embodiment uses a relatively simple hardware application to achieve accurate results, and thus needs to be used. But _ is that although the time positioning part 700 A has a linear and non-linear predistortion circuit 200 and provides suitable results, people in the same industry can design their own embodiments. The time positioning section 7 includes a minimum delay element 202, which can receive a complex information stream from the multiplexer 222. The longest delay element 702, the only non-programmable element, is used to add the delay of a whole clock, which is equivalent to the combined circuit 200, time positioning part 800, linear predistorter 244, analog part 120, feedback part The minimum delay imposed by combining 248, attenuator 256, and equalizer 26. The one-clock multiple junction delay line 704 is driven by the minimum delay element 702, and each segment of the complex signal has an equal 33 200541280 delay on the delay line. Although FIG. 7 depicts eight contacts 706, persons in the same industry can provide any number of contacts 706. The contact 706 is coupled to the data input point of the multiplexer 708, and an output of the multiplexer is connected to an input point of the complex interpolation 710. The interpolator 710 may use a Farrow or other architecture and add equal delay to both ends of the complex signal. One of the outputs of the interpolator 710 provides the delay complex forward information flow 266 ', and the controller (0286 provides control input points to the multiplexer 708 and the interpolator 710. A clock number 712 is also provided to the minimum delay component 702, delay line 704, and interpolator 710. The clock 712 is preferably synchronized with the data rate of the forward and return information streams. [0093] When work 610 is used for common mode time positioning part 700 (that is, During the working period of 504), the time positioning part 700 can be adjusted by providing appropriate control inputs for the multiplexer 708 and the interpolator 71.-The entire part 714 includes the delay line 704 and the multiplexer. 708, and is used to provide an integer multiple of the delay of the clock 712, such as the control data specified by control $ 286. Part of the section 716 includes the interpolator 710 and is used to provide the clock 712-cycle delay. Time positioning Part 700 can use the control multiplexer 708 to achieve an arbitrary multiple of time delay, and control the interpolator 71 can achieve the fraction of the delay. I _circle 8 shows the time of the money. Block diagram and this embodiment can achieve accurate results because it uses a relatively simple hard-zhao design. The result is needed. But although the time positioning part can provide the suitable results of the linear and non-linear predistortion circuit 200, people in the industry can design their own feasible lines. Differential time positioning part can The common-mode time positioning part 700 is similar to 'but with a different effect as shown in Figure ，. One of the complex forward information streams 238 is connected to a clock delay and the other is a foot' shown in Figure 8 The Q pin 'is connected to the fixed delay element m. The delay element 8 (H is configured to realize the delay of 34 200541280 B delay element 802! / 2. Although it depends on 8_ is the delay with human health points Line 802 'People in the same industry can design any number of contacts. Contacts connect to the data input points of Duoyi 808, and the multiplexer has one output connected to the input of the interpolator 810. Point. The interpolator 810 can be implemented using FarrOW or other architectures. One output k of the interpolator 810 is used for complex direction > section I of the stream 242, and one of the delay elements 804 output provides information Segment Q of stream 242. The controller (Q 286 provides control inputs to the multiplexer 808 and interpolator 81 And the clock "812" is also provided to the delay line 802, the delay element 804, and the interpolator 81. The clock% 812 is preferably synchronized with the data rate of the forward information flow. [0095] When working 610 is used for the differential mode time positioning part 8 hours (that is, when working at 506), the time positioning part 800 can use the way of providing suitable control inputs to the multiplexer 808 and the interpolator 810. An integration part 814 includes a delay line 802 and a multiplexer 8⑽, and provides a delay of an integer multiple of the clock 812, such as control data specified by the controller 286. Part of part 816 includes interpolator 810 and provides a fraction of clock 812. If you want to preset the time delay of any fraction of the positioning part 800, you can control the multiplexer 808 to achieve it, and the fractional part @part time delay can be achieved using the control interpolator 810. [0096] Returning to FIG. 6, when task 610 adjusts the time positioning hardware and reflects the new delay according to the old delay and the current step size and polarity, inquiry work 612 is performed. During work 612, the correlation engine 280 performs its correlation and integration operations until the correlation condition is reached. When job 612 determines that the association condition is met, a query job 614 determines whether the current association result is greater than the maximum association value recorded since the start of process 600. If the current correlation result is not greater than the previous correlation, then work 616 makes the estimated size of the steps the same as the former, but changes 35 200541280 the positive and negative values of the displacement ’and then program control continues with work 618. If the current association result is not greater than the previous association, a job 620 estimates the size of a step from the previous step size, usually 0.5 to 1. 0 times the size of the previous step, and the polarity displacement is also estimated, then the program control continues to work 618. [0097] Work 618 decides whether the prediction and convergence formula has converged to a common mode delay value or a differential mode delay value, so as to maximize the correlation between the negative signal flow and the return information flow. Convergence can be monitored by the current step size, and if the current step is smaller than the resolution of the interpolator 71 or 81, then φ is judged to have reached convergence. When task 618 determines that delay convergence has not occurred, program control returns to task 610. At work 610, the estimation of the previous time delay is changed according to the current step size and displacement pole, and then the association process is repeated. [0098] When work 618 determines that delay convergence has occurred, the sub-process 600 has been completed. At this time, the time delay complex forward information stream 266 has been positioned in time with the complex information stream 262. In addition, the linear compensation process 500 may continue to perform another positioning process, which is a prerequisite for the actual embodiment to perform actual linear compensation. Lu [0099] Referring back to FIG. 5, after starting the secondary process 600 twice, once for the common mode time and once for the differential mode time, respectively, work 504 and 506, a process of 9㈧ will be performed to execute A prediction and convergence formula is obtained, and the phase angle of the positioning complex forward information flow 272 is transformed into the same phase as the time delay complex forward information flow 266 through the phase angle rotator 1000. [0100] FIG. 9 shows a flowchart of a sub-process 900. The sub-process 900 includes a job 902 for controlling the multiplexer 270 'so that the correlation engine 280 can be coupled to "ideally, locate the complex forward information." A correlation operation is performed between the stream 272 and the return information stream 262. Then, work 204 removes 36 200541280 and selects the CORDIC unit. Work 904 connects to a specific piece of hardware to perform the phase angle in the preferred embodiment. Rotator 1000. [0101] Fig. 10 shows a block diagram of an embodiment of the phase angle rotator part 1000. This embodiment is useful because it can use a relatively simple hardware design to achieve accurate results. But although the phase angle rotator 1000 can provide suitable results to the linear and non-linear predistortion circuit 200, people in the same industry can design their own flexible embodiments. The phase angle rotator includes a quadrant selection unit 1002, followed by The CORDIC unit 1004 is connected in series. Figure 10 only depicts two of the CORDIC Lu unit 1004, labeled 1004 and 1004, but the remaining units should have the same architecture as the unit 1004 '. Any number of CORDIC units 1004 can be included, and the preferred embodiment has 6 to 16 units 1004. If you enclose 10 units of 1004, it can provide 0: 112 degrees of accuracy. [0102] Delay forward information The stream 266 is received at the quadrant selection unit 1002, and the feet of the complex information stream are received at its own selection inversion circuit 1006, while FIG. 10 depicts a multiplier. The selection inversion circuit 1006 Independent control by the controller 286 is reversed, or the information flow is passed through unchanged. The units 1002 and 1004 are plotted in the figure as the terminals of the latch 1008, and the control circuit 1006 is displayed The combination of various inversions and passes can save four possible quadrants, and the unit 1002 can shift the incoming information stream 266 by 0, 90, 180, or 270. [0103] In each CORDIC unit 1004 In each unit, the I and Q pins that flow into the complex information stream are connected to the shifter 1010. Figure 10 depicts the shifter 1010 as a multiplication circuit, because the shifter 1010 is executed by the inverse of the power of two Mathematical multiplication. In the first Cordic unit of 2004, the shifter 1010 can be omitted because they will flow into the The material is shifted to the right by zero and the operation multiplied by 37 200541280 one is performed. At the first CORDIC unit 1004, it is displayed with the next unit. The shifter w shifts the machine-in data to the right by the previous unit. Add one bit. Therefore, the figure shows the displacement unit 1G1G on the element surface, which is multiplied by Q 5, and the three c0ROCIC units_ are actually fine G · 25, so that is the same.- People in the industry may think that it is not necessary to implement the shifter 1010 with actual parts, and only replace it with a wire. [0104] In each CORDIC unit 1004, the output of the shifter 1010 is connected to a selectively enabled circuit 1G12 'as shown in several AND gates in the embodiment of the figure, in which the feet of the complex information flow There is a brake. The other inputs of each AND gate are controlled by the controller 286. Therefore, the controller 286 either passes the output of the shifter 1010 completely or adds a zero value. [0105] In the I pin of each CORDIC unit 1004, a subtractor 1014 subtracts the output of the selection enabling circuit 1012 from the Q pin flowing into the information stream. An adder 1016 in the q pin of each 〇RDIC unit 1004 selects the Q pin of the incoming information stream and the I pin of the incoming information stream to enable the output of the circuit 1012. From the subtracter 1014 and the adder 1016, the I and q pins flow out of the CORDIC unit 1004 through the latch 1008. [0106] Each CORDIC unit 1004 rotates its incoming complex information stream in a gradually small angle manner, as shown in the following example: Table I-10 CORDIC unit phase angle rotator multiplier 1. 0 0. 5 0. 25 0. 125 0. 063 0. 031 0. 016 0. 008 0. 004 0. 002 Angle (degrees) 45. 0 26. 6 14. 0 7. 125 3. 576 1. 790 0. 895 0. 448 0. 224 0. The rotation of each unit is only slightly more than the% of the rotation of the previous unit. Therefore, for the resolution of 1000 included in the phase angle rotator part, the number of CORDIC units determines the resolution by selectively combining 38 200541280 CORDIC units 1004, anything between 0. _90. The range can be reached. [0107] Although not shown, a scaling stage can be used to compensate for the scaling of the amplitude of the signal processed by the CORDIC unit 1004. In one embodiment, each CORDIC unit 1004 can be set to a positive or negative value to maintain its scaling at a certain value for different rotations. [0108] Referring to FIGS. 9 and 10, the operation 104 disables the selection enable circuit 1012 in each unit 1004, so that a signal from any pin of the unit 1004 will not be cross-linked to another pin. Therefore, the CORDIC unit 1004 does not rotate due to work 904. After work 904, a work 906 sets the convergence conditions. As discussed in the above work 604, setting the convergence conditions can control the rate of association convergence to use a low-resolution return information flow to reach a preset effective error level. By task 906, the increase in program processing time translates into a return flow of information that reduces the effective error level. [0109] After work 906, a work 908 is selected to adjust the control input at the selected inverter 1006 to control the other quadrant. The phase rotation part 1000, which affects the rotation, represents an estimate of the required phase rotation to bring the positioning and forward information stream 272 to the same phase as the return and return information stream 262. [0110] After work 908, the correlation engine 280 integrates between the forward and backward information flow 272 and the return information flow 262 based on the current phase transfer estimate. A query work 91 determines whether the convergence conditions of the above work 906 are met. When the convergence conditions are met, a query work 912 determines whether all four quadrants have been selected. If the number of quadrants tried has been less than four, the correlation results are stored and program control returns to work 908 until all four quadrants have been tested. [0111] Although work 908, 910, and 912 describe one embodiment of the quadrant evaluation, another alternative embodiment is that one of the forward and reverse information flows and the reverse information flow can be compared with other information flows. 39 200541280 Each foot is associated, and the result of the previous association sub-processing, such as sub-processing 600, can be used. The quadrants can then be selected based on the relative size and polarity of the correlation results. [0112] When all four quadrants have been tested or otherwise evaluated, a job 914 selects from the four quadrants the quadrant that will or should have the greatest correlation, and sets the selected inverter 1006. Then, a job 916 selects the next most obvious CORDIC unit 1004 by selecting the unit 1004. In the first calculation of work 916, a displacement of 45 was selected. CCoRIC Zhuo Yuan 1004. At this point, another estimation of the phase inversion to bring the positioning forward forward information flow 272 into the return return information flow Lu has completed, and the correlation engine 280 performs its correlation and integration work. [0113] After work 918, a query work 916 determines whether the convergence conditions set by the work 906 described above are met. When the convergence conditions are met, a query 920 decides whether the maximum associated record from the start of the sub-process 900 to the present has increased by the most recent estimate. If not found, a job 922 deselects the current CORDIC unit 1004. After work 922 and work 920 find that the maximum correlation value increases, a work 924 determines whether the nearest and smallest CORDIC unit 1004 is selected. As long as there is still a small untested CORDIC unit 1004, program control will return to # job916. [0114] After job 924 determines that the last CORDIC unit 1004 has been evaluated, sub-process 900 is complete. At this time, the processing 900 has measured all the CORDIC units 1004, and has selected all the units 1004 that generate the phase transition evaluation to the maximum correlation, such as the correlation engine 280 decider. This process brings the positioning complex forward information stream 272 to the same phase as the complex return information stream 262, reaching the accuracy set by the convergence conditions used by the correlation engine 280, and the number of CORDIC units 1004 included in the phase conversion part 1000. 200541280 [0115] Referring to the previous figure 5, after the sub-processing 900 is completed, a job 508 optimizes the gain adjustment provided by the adjustable attenuator. Therefore, task 508 executes a suitable optimization program to provide the addition and / or reduction of programmable attenuation in the coarse and fine adjustment attenuators 302 and 256, respectively. The optimization program must be able to make attenuation adjustments to minimize the cumulative difference between the forward stream 272 and the return stream 262. The optimization procedure may use techniques similar to those discussed above, as shown in Figure 6-10, or other techniques that may be used. [0116] After working 508, the linear and non-linear predistortion circuit 2000 is now sufficiently trained and ready to face the compensation problem of linear distortion generated by the analog part 120 more directly. So far, in the combination circuit 274, "ideally, the time and phase angle of the forward information flow and the reverse information flow coincide with each other. Therefore, the error information flow 276 now represents the distortion produced by the analog part 12. But as explained above The error information 276 is from the return information stream, at least part of which was formed, and shows a high degree of error and low resolution. Now start a secondary process ι〇〇 for the forward equalizer 246 to perform pre- Evaluate and converge the program. [0117] 11 shows a flowchart of the sub-process 1100. The sub-process 1 100 has been adapted to a special embodiment of the non-adaptive equalizer 1200 and a special embodiment. The engines are said to operate together. Figure I2 shows a block diagram of a representative non-adaptive equalizer suitable for several parts of the linear and non-linear predistortion circuit 2000 and a sub-process blue. The forward equalizer 246 can be configured similar to a non-adaptive equalizer, but it is likely to have more contacts. Of course, Figure 13 shows-the proper match with the _Lin Qingteng adaptive engine test shown in @ 12, and linear and non- Linear predistortion circuit 200. But the same People in the industry can identify other non-adaptive equalizers. The embodiment of the adaptive engine test and sub-processing can be designed to achieve many of the goals of the 200541280 existing invention. [0118] Referring to FIG. 12, the non-adaptive equalizer 1200. There are only three contacts drawn for the convenience of a complex equalizer, but people in the same industry can discern that the number of contacts can be easily increased or decreased depending on the particular application. I and Q of the complex input information flow The pins are connected to the nodes 1202 and 1204, and the equalizer 1200, or the equivalent, can be used in multiple locations of the linear and non-linear predistortion circuits 200, such as the equalizers 226 and 246. , And / or 260, etc. Therefore, the true identity of the complex input information flow varies depending on the place of use. • [❶119] 1 · Node 1202 is coupled to and drives the clock contact delay lines 1206 and 1208. And the Q_point 1204 is connected to and drives the clock contact delay lines 121〇 and 1212. The delay line 1206 drives the in-phase direct path 1216 of the equalizer 1200; the delay line 12ι drives the equalizer 12〇 〇 Orthogonal direct path 1216, delay line 1208 drives equalizer 1200 The phase-to-orthogonal cross path 1218, and the delay line 1212 drives the orthogonal to the in-phase cross path 1220 of the equalizer 1200. [0120] Each contact 1222 of each delay line 1206, 1208, 1210, and 1212 drives its own The first input of the multiplier 1224, and the output of the multiplier 1224 drives the adder 1226. An output of the phase 0 angle path 1214 is provided by the sum of all multipliers 1224 on its path to the positive of a subtractor 1228 The input is quadrature to one of the in-phase paths 1220 and is provided by the sum of all multipliers Π24 on its path to the negative input of a subtractor 1228. An output of the quadrature path 1216 is provided by the sum of all multipliers 1224 on its path to the first input of an adder 1230, and an output of the in-phase to quadrature path 1218 is provided by its The sum of all multipliers 1224 is provided to the second input of an adder 1230. One output of the subtractor 1228 provides the j-pin of the complex output information stream, and the output of the adder 1230 provides the 42-200541280 Q-pin of the complex output lean stream. [0121] The contacts 1222 of the in-phase and quadrature direct paths 1214 and 1216 have the same filtering coefficients and are provided by the multiplier 1232 through an optional heat adaptor unit 1234, and each of the contacts 1222 has an output. Figure 12 shows two heat adaptor units 1234, and only one of them is marked with details. If the heat adaptor unit 1234 is omitted, the filter coefficient of each contact is directly provided by the multiplier 1232. This filter coefficient output is coupled to the second input point of two corresponding multipliers of the direct paths 1214 and 1216. Similarly, the cross paths Pig and 1220 have the same filter coefficients as each contact 1222, provided by the multiplier 1236 through an optional heat adaptor unit 1234, and each contact 1224 has an output. This filter coefficient output is coupled to the second input point of two corresponding multipliers of the cross paths 1218 and 1220. [0122] Multiplexers 1232 and 1236 receive filter coefficients from the adaptation engine 1300 at feature 236 or from the controller 286. When the heat adaptor unit 1234 is included, the thermal coefficient is also received from the adaptation engine 1300 or the controller 286. The controller 286 also controls the selection inputs of the multiplexers 1232 and 1234. The slave controller 286 or the adaptation engine is used to connect the filter coefficient and the thermal coefficient. The adaptation engine 1300 is used to couple and decouple the equalizer 1200. . When filter coefficients and selectable thermal coefficients are provided from the controller 286, the equalizer 1200 operates in a non-adaptive mode. In the non-adaptive mode, a set of direct filter coefficients and direct thermal coefficients are set by the controller 286 to the direct paths 1214 and 1216, and a set of cross filter coefficients and cross thermal coefficients are set by the controller 2 to the cross paths 1218 and 1220. Unless the controller 286 changes the program, any set of filter coefficients will not change, but the filter coefficients can be selectively adjusted in the heat adaptor unit 1234 in response to the thermal difference signal 216. In the preferred embodiment, the optional adaptor unit 1234 and the non-adaptor 43 200541280 adaptive equalizer 226 are included, but in other applications may include this and other equalizers, or from all equalizers. Omitted. [0123] Each heat adaptor unit 234 includes a multiplier 1238 of each contact and an adder 1240 of each contact. The thermal difference signal 216 is coupled to the first input of each multiplier 1238. For each contact, the multiplexer 1232 or 1238 provides the second input of the thermal coefficient "?" To the contact of the multiplier 1238. The relative output of the multiplier 1238 is coupled to the first corresponding input of the adder 1240. Moreover, for each contact, the multiplexer 1232 or 1238 provides a filter coefficient "w" to the second input of the contact of the adder 1240. The output of the adder 1240 provides the filter coefficient of the heat adaptor unit 1234. Output. Therefore, the filter coefficient has a displacement, maybe positive or negative, and responds to the thermal difference signal 216 with the weight of the thermal coefficient. [0124] When the adaptation engine 1300 provides a filter coefficient and an optional thermal coefficient, the equalizer 1200 operates in adaptive mode. In adaptive mode, the adaptive engine 1300 provides at least one direct and alternating filter coefficient and thermal coefficient, and as long as the equalizer 1200 maintains its adaptive mode 'these filter coefficients The groups with the thermal coefficient will change continuously. _ [0125] With reference to 囷 13, in one embodiment the adaptation engine 1300 is configured to accommodate a partial equalizer to reduce linear and non-linear predistortion circuits 2 The number of parts of 〇〇. Especially when the adaptive engine 1300 is connected to the -non-adaptive equalizer 12 00, it will be connected to the direct path i2i4 and 1216 or the cross-route control 1218 and 1220, but will not In order to be consistent with the three-contact re-equalizer 1200 of FIG. 12, FIG. 13 shows a three-contact circuit. However, people in the same industry can discern that the number of contacts can be easily adjusted according to their particular application. Increase or decrease. [0126] The "ideal" positioning complex forward information flow 272 and the Q pin are respectively supplied to the clock connection 44 200541280 delay time lines 1302 and 1304, and three delay lines are drawn for convenience. Contacts. When the adaptive engine 1300 operates in its mode coupled to the direct path 1214 and PM of the non-adaptive equalizer 1200, the I and Q pins of the error information stream 276 are connected to the delay elements 1306 and 1308. The delay elements 1306 and 1308 are each configured to delay the time from the error information stream 276 to the midpoint of the contact delay lines 1302 and 1304. When the adaptation engine 13⑻ is operating in the mode coupled to the cross path i2i8 and its mode, the error information flow 276 and the Q pin are supplied to the clock components 1306 and 1308, respectively. The first input of the corresponding in-phase multiplier i3i2 is connected from the contact 131 of the in-phase delay line 1302 and the corresponding input of the quadrature delay line 130 The first input of the corresponding positive father phase multiplier 1316. The output from the in-phase multiplier 1312 is coupled to the first input of the corresponding adder 1318, and the output from the quadrature multiplier 1318 is coupled to the corresponding adder 1318 through the selected inverse component 1320. The second input. [0127] The selected inverse element 1320 is shown in FIG. 13 as a multiplier, and one input of the multiplier is controlled by the controller 286. The controller 286 inverts the output of the weighted quadrature signal when the adaptive engine 1300 is operating in its cross paths 1218 and 1220 coupled to the non-adaptive equalizer 1200, but when the adaptive engine 1300 is coupled to When the modes of the direct paths 1214 and 1216 operate, the 'orthogonal multiplier 1316 does not output an inverted weighted orthogonal signal. Those in the same industry will recognize the inverse element 1320 that does not require the use of a multiplier to perform the selection. Similarly, people in the same industry will recognize that quantizing the error signal 276, the ideal positioning signal 272, or both into a single bit or a-/ 0 / + value, can reduce the complexity of the adaptation engine 1300. In this alternative design, the above multiplier can be replaced with a simpler circuit. [0128] The various outputs of the adder 1318 will be noisy because they are based on the return information stream 45 200541280. These outputs are coupled to the first input of the corresponding multiplier 1322, and the second input of all multipliers 1322 is coupled to the controller 286. The controller 286 provides a convergence factor "μ," to determine how much the filter coefficient can be changed compared to the clock period to the clock period. In a preferred embodiment, μ uses a small value to prevent addition The multiplier 1318 allows considerable variables to have a great impact, and output noise from any single element or even a group of elements of a considerable size. [0129] Each output of the multiplier 1322 is coupled to the corresponding adder 1324. One input, and the output of each adder 1324 is coupled to the corresponding single-cycle delay element 1328 through the first data # input of the corresponding multiplexer (mux) 1326. The controller 286 provides the multiplexer 1326. The second data input and selection control input. The delay element 1328 can initialize the preset filter coefficients from the controller 286, but under normal adaptive operating conditions, each adder 1324 will leave the former in the corresponding delay element 1328. The coefficient value is added to the change value of the filter coefficient. In addition, for each contact, the output of the adder 1324 provides the filter coefficient w "output by the adaptation engine 13 in the characteristic output. When operating in its adaptive mode, the controller 286 provides a filter coefficient "w" to the equalizer 1200, and can be read by the controller 2. _ [0130] The subsequent processing of the wave coefficients is directed to the thermal-related memory effect, especially the data wave coefficient “w” output from each adder 1324 is connected to the corresponding IIR • crosser circuit. The waver circuits each include a subtractor circuit 1330, a multiplier 1322, -adder 1324, and a single-cycle delay 1336. The output from each adder 1324 is tapped to the positive input of the corresponding subtractor circuit. The input of each subtractor circuit 133 is used for the output of the crossover, and is lightly connected to the first input of the sensitivity multiplier 1322. The second in-round adaptation of each sensitivity multiplier 1322 receives the miscellaneous numbers provided by the controller 286 ^ The output of each multiplier 提供 is provided to the corresponding first input of the 2005 461280 generator 1334, and each of the adders 1334 The output is clocked through the corresponding delay element 1336. Each input of the delay element 1336 is connected to the corresponding adder CU4 and to the negative input terminal of the corresponding subtractor circuit 133. [0131] The adder 1334 provides an average coefficient output of each filter circuit, which is expressed as m; the wave coefficient w is considered as a long-term average value or a transition signal. The subtracter circuit measures the difference between the current instantaneous value of the filter coefficient "w" and the long-term average value from the previous clock cycle. The coefficient sensitivity factor γ determines the sensitivity of the long-term average value to the effect of the instantaneous filter coefficient. A small Y value is required to make the average value reflect a longer period, and it is less sensitive to the filter coefficient of any -clock period. The output of the subtractor circuit 1330 provides a coefficient difference information stream 1338. For the indirect points in the adaptation engine 1300, the coefficient difference information stream 1338 forms a coefficient difference signal, and is selectively connected to the correlation engine 280. [0132] When the HPA 136 passes through the average filter coefficients related to heat, it can be changed with the change of temperature to determine the change of the wave. Therefore, the subsequent principle of appropriateness is implemented in the poor resource domain—the book of Wei Qianying. On the _ side, each coefficient difference information stream is connected to the positive input terminal of the corresponding subtractor circuit. Each subtracter circuit mo is connected to the first input of the corresponding multiplier 1342, and the output of each multiplier 1342 is connected to the first input of the corresponding convergent multiplier 1344. The output of each retractor 1344 is connected to the first input of the corresponding adder 1346, and the output of each adder 接 is returned to the same adder 透过 through the corresponding single-service delay element ⑽. The two inputs thus form an integrator from the adder 1346 and the delay element ⑽. Furthermore, the output of each adder 1336 is connected to the first input of the corresponding multiplier 135 °, and the output of each multiplier 47 200541280 is connected to the negative input of the corresponding adder 134 °. The thermal difference signal 216 drives the second input of all multipliers 1350 and all multipliers 1342, and the controller 286 provides a convergence value λ to the second input of the convergence multiplier 1344. [0133] The output of the adder 1346 provides a thermal coefficient α, which is output from the adaptation engine 1300 at characteristic 236. When operating in its adaptive mode, the thermal coefficient α is provided to the equalizer 12⑻, and can also be read by the controller 286. After a period of time, the thermal coefficient 〇1 of the thermal difference signal 216 converges to a more accurate estimation of the sensitivity of the change in the filter coefficient "w". As discussed below and shown in FIG. 17, the thermal difference signal 216 characterizes changes in the HPA 136. Therefore, the thermistor coefficient α, the thermal signal, and the filter coefficient are used together in the adaptor unit 1234 to remove the correlation between the possible thermal change of the price ρα 136 and the change of the equalizer filter coefficient. In other words, determining the thermal coefficient α, when it forms the thermal difference signal 216, it will cause the thermal signal to become greatly correlated with the corresponding coefficient difference signal. [0134] FIG. 13 depicts all single-cycle delay elements 1348 as having a clear input driven by the output of the controller 286. This input allows the controller 286 to initialize the delay element 1348 to a zero state and eliminate heat treatment. [0135] In one embodiment of the adaptation engine 1300, the integration and output operation (not shown) can be performed on the thermal difference signal 216 and the coefficient difference signal 1338 to slow down its data rate. This may be allowed because thermal changes occur at a slower rate than the rate at which the data is processed through the transmitter 100 one by one. Reducing the data rate here can save the power of the downstream coefficient difference signal 1338. [0136] Referring to FIG. 11 above, the sub-processing 1100 operates with the equalizer 1200 and the adaptation engine 130048200541280 to perform a low-resolution and high-error characteristic of a tolerable error information stream 276. Level estimation and convergence formula. When the process 1100 is operated for the initial stages of linear distortion compensation and non-linear distortion compensation, the heat treatment is cancelled by operating the initial work of 502. The heat treatment can be cancelled by setting the single-cycle delay element 1348 to display a zero value and setting the convergence value χ to zero. This will cause the adaptor unit 1234 to fail, but cancelling the heat treatment is an unresolved point associated with the equalizer 1200 and used as a forward or return equalizer 246 and 260, and this can be omitted in the preferred embodiment Heat adaptor unit 1234. [0137] The process 1100 performs a job U02 to lock the adaptation engine 1300. The adaptation engine 1300 can be locked by providing a convergence factor μ = 0 to the adaptation engine 1300. The adaptive wave coefficient "w" of the adaptive engine 1300 'transmission characteristic 236 cannot be changed. After work ι102, a work 1104 initializes the mode of the adaptation engine (AE) 1300 to determine the filter coefficients of the direct paths 1214 and 1216 of the equalizer 1200. At this time, the choice of the direct paths 1214 and 1216 and the parent fork paths 1218 and 1220 is arbitrary. The adaptation engine 1300 can initialize the adaptation of the direct path filter coefficients via an inversion circuit 1320 that controls the selection so that they do not reverse the weighted orthogonal signals they process. After working 1104, the sub-process 1100 starts a program 1106. Here, the estimation and convergence program of the 1/2 complex equalizer 1200 is used to determine a set of chirp coefficients.丨 0138] Of course, there is no need to adapt the engine 1300 to adapt to only a part of the path of the equalizer. If the adaptation engine 1300 is configured to adapt to all the paths of the equalizer 1200 at the same time, it is due to the figure The adaptation engine circuit of 13 is substantially doubled, but some reverse circuits 13 and 1104 can be omitted. In this case, the adaptation engine 1300 performs addition in circuit 1318 and the other half of adaptation engine 1300 performs subtraction in circuit 1318. 49 200541280 Tao 9] In particular, after working 1104, a work · initialization of the adaptation engine (outline wave coefficient. The work circle can initialize a single-cycle delay element 以 to display the currently used equalization road control settings. Filter coefficient. However, once initialized and in other situations, a single-cycle light component can be set to a desired value, duty value, or not set at all. [0140] After work 1108, the work 1100 will adapt to the engine (ae) The 1300 handle is connected to the% part of this non-adaptive equalizer 1200. The handle is to control the multiplexer 1232 or 1236, as the case may be, to select the wave coefficient from the adaptive engine instead of the controller. • [〇141] Next, a job 1112 sets the convergence conditions. In one part, the estimation and closing formula is given, and the adaptive engine (AE) blue is unlocked. Part of the convergence conditions is set and the adaptive engine is unlocked 2 It may be achieved by providing a positive value of the adaptation engine to the convergence value μ. This value is preferably a fraction that is very small, and the convergence condition determines how many samples the adaptation engine 1300 should process before it can be considered to have converged to one. The answer of the wave coefficient group. As discussed above, the more effective the sampling process, the greater the effective resolution, or the error level will be reduced in the return information flow. The increase in the number of program offices has transformed into reducing the error of the return information flow. Level. Through work 1112, the speed of convergence can be controlled to achieve a preset effective return error level lower than the error level of the return information stream. In one embodiment, the convergence variable μ is initially set to a higher value However, it decays with time. This method allows fast convergence to a close result, and then slows down the convergence rate to achieve a smaller final tracking error. [0142] After work 1112, the adaptation engine 1300 will perform a LMS, prediction and convergence formula, in which the estimated value of the filter coefficients is continuously changed to minimize the error signal. LMS, prediction and convergence formula is repeatedly changed in the filter coefficients, to minimize the error signal. And 50 200541280 and make the error signal flow from the forward information flow unrelated, this action also increases the forward and return information flow. More specifically, the filter coefficient After adjustment until the result from the HPA input signal 134 or HPA output signal 117, according to the current state of the multiplexer 250, it becomes a substantially uncorrelated signal (for example, as close to white noise as possible). [0143] At this time, a Work 1114 confirms whether the wave coefficient determined by the adaptation engine 1300 is converging. Work 1114 is performed together with work 1112 to set the convergence conditions. Using smaller μ values, the longer time used by work 1114 further increases the effective analysis. And further reduce the effective return information flow error level. Job 1114 can only decide whether there is enough time to reach convergence, or job 1114 can monitor the filter coefficients generated by the adaptation engine 1300, and when the filter and wave coefficients are measured without consistency When changing morphology, make sure that convergence has occurred. [0144] After work 1114 determines that convergence has occurred, a query work 1115 decides whether the secondary process 1100 has been started to include the decision of heat treatment and filter coefficients. There are no heat treatment, forward and return equalizers 246 and 260, and an equalizer 226 that initially determines the coefficient calculation. In these cases' program control will be handed over to job 1116. The case of heat treatment will be discussed below with FIGS. 15 and 17. [0145] Working 1116 locks the adaptation engine 1300 by setting the convergence coefficient of the wave coefficient and the thermal convergence factor = 0. Then, a job 18 reads the characteristics 236 of the adaptation engine (where) 1300, and reads a set of filters Coefficient and thermal coefficient. After work 1116 and 1118, a job 112 has set the set of wave coefficients to the subject non-adaptive equalizer 1200, and a job 1122 has decoupled the adaptive engine (AE) 1300 from the non-adaptive equalizer 120. . When the sub-process 1100 is used to determine the thermal coefficient, the work 1120 also sets the subject non-adaptive equalizer 1200 to a set of thermal coefficients. 51 200541280 [0146] Decoupling can be achieved using controller data input selected in subject multiplexer 1232 or 1234 instead of using adaptation engine data input. At this time, the adaptive engine 1300 has determined a set of filter coefficients and possibly a set of thermal coefficients, and the filter coefficient group and the thermal coefficient group have been set in the non-adaptive equalizer 1200, so the adaptive engine 1300 is now One can determine another set of filter coefficients and complete the procedure 1106. The filter coefficient group and the thermal coefficient group that have just been determined are preferably static, but the filter coefficient group may continue to be adjusted in the non-adaptive equalizer due to the thermal difference signal 216 and the thermal coefficient.丨 0147] After work 1122 and program 1106, a work 1124 starts the adaptation engine 1300 mode to determine the cross path coefficient of the subject non-adaptive equalizer 1200. The adaptation engine 1300 can use the inversion circuit 1320 selected by the control to start the crossover wave coefficients so that they can invert the weighted orthogonal signals they process. Next, a job 1126 repeats the procedure 1106 with other filter coefficient sets. After work 1126 completes the other filter coefficient sets, the sub-processing 1100 is complete. [0148] Referring to FIG. 5 above, when the sub-process 1100 is completed, the sub-process 500 is also completed at the same time. At this time, the linear predistorter 244 has previously set the forward filter coefficient to the forward equalizer 246 to compensate the linear distortion of the RF analog signal 134 at the input of the HPA 136 caused by the analog part 120. In particular, the forward equalizer 246 can now compensate for frequency-dependent gain and phase imbalances, and it can also compensate for differences in gain and delay caused between the pins of the secondary communication signal. Therefore, the closer the RF analog signal at the input of the HPA 136 is, the better, and the linear distortion caused by the analog part 120 upstream of the M 136 is responsible for the predistortion generated by the hybrid predistorter 244. [0149] With reference to the high-pass filter 205 in the previous figure and the high-pass filter 314 in FIG. 3, each filter has a very small effect. In particular, in order to block the value flow component, a small amount of energy close to DC is also blocked by the high-pass viewer 314 in the return path. But this is the same for DC in the spectrum. An overview circuit is used to drive the forward equalizer 246 to match the return path. Therefore, the adaptation engine 1300 decides to control the linear distortion of the forward information flow toward the wave coefficient, except for approaching the direct capture here. The Nantong filter 205 removes this near-DC energy only from the forward information stream so that all the amount passed through the forward equalizer 246_ is used to control linear distortion.

[0150] After 500 cycles of processing, and one processing is performed 1. To extend through the compensation of 136 linear distortion. Since a considerable undistorted signal is present at the input of the i36, the HPA 136 now puts a signal that is more in line with the control status of the model designed by the jjpA model. Furthermore, at this time, there is no non-linear compensation and forward information flow, and the considerable amount of undistorted signal presented to the HPA 136 includes only frequency components within the bandwidth. [0151] FIG. 14 shows a flowchart of a sub-process 140, which includes a job 1402, a multiplexer 250 for RF analog signals generated from the input of the HPA 136 to the output of the HPA 136. , Control the feedback signal provided to the down-frequency part 300. Since the feedback signal now propagates through the HPA 136, compared to the feedback signal from the input of the HPA 136, it will experience more delay and phase inversion. Next, a job 1404 starts a sub-process 600 to execute an estimation and convergence procedure for the common-mode time positioning part 700. Finally, this "ideally, the time delay complex forward information flow 266 is brought back to time positioning and the return information flow 262. Here, no further time positioning is needed, because the two feet of the complex communication signal are processed before the HPA 136 processing. Both have been merged. Because the same analog parts (ie, ΑΡΑ 136) handle the two legs of the merged signal, there will be no opportunity to form a further difference in orthogonal imbalance. 53 200541280 [0152] After work 1404 'A work 1406 was started Process 900 to execute an estimation and convergence program to reposition the phase of the positioning complex forward information stream 272 and the phase of the returning complex information stream 262. Finally, this "ideal" positioning complex forward information stream 272 is taken Go back and return to stream 262 to locate. [0153] After the phase repositioning of job 1406, a job 1408 provides optimized gain adjustments by adjustable attenuators 302 and 256 in a manner similar to that performed by job 508. After work 1408, a work 1414 starts a secondary process 110 to execute a forward equalizer 246 estimation and convergence program to increase the correlation between the output RF analog signal 117 of M and the forward information flow. . Finally, the forward filter coefficients set before the forward equalizer 246 are updated to compensate for the linear distortion generated by? 136. In particular, this linear distortion can be generated by the input band-pass filter (BPF) 140 and the output band-pass filter (BpF) 140 of the ΗPA mode. However, here, linear compensation covers wideband signals that do not include non-linear parts. [0154] After job 1414, a job 1416 controls the multiplexer 222 to connect the highest-order basic function information stream 214, instead of the forward information stream 218, to the adaptation engine 13 '. As discussed above, _ the highest-order basic function gift 214 presents the same clock as the forward information flow 218, and in the preferred embodiment, the 'basic function generation part 16' does not perform rotation. The most basic: lack of basic function information Processing of quadrature phases of stream 214. Therefore, no further time or phase positioning is needed to call the most basic function information stream to and return ^ bits. For the purpose of linear compensation, a 'significant difference' between the basic function information, stream 214, and forward information stream 218, will show the ultra-wideband discussed previously. [0155] With reference to the ηρα model depicted in FIG. 1, & 142) may produce 54 200541280 non-linear distortion, causing components outside the bandwidth to be processed by the output bandpass data wave (bpf), and may encounter Linear distortion. In order to compensate the linear distortion of the output band-pass filter (BpF) 44, a job 1418 starts the sub-process 1100 again to execute an estimation and convergence program. However, in this sub-processing 1100, the return equalizer 260 is activated to adjust the return information flow, so that the association between the HPA analog output signal 117 reflected by the return information flow and the ultra-wideband, highest-order basic function information flow 214 'can be maximized. In the forward-propagating signal, the higher-order terms do not have a significant degree until the output of the memoryless nonlinear part of the Wiener-Hammerstein HPA model (that is, • amp 142). However, when these higher-order terms pass the output BPF 144 of the Wiener-Hammerstein HPA model, they may be subject to linear distortion. Therefore, the compensation for linear distortion covers the wider bandwidth that the output BPF 144 must handle. [0156] This operation further captures the linear distortion that appears at the output of ραα136, but does not adjust the HPA output signal, but adjusts it in the return signal path to allow subsequent training of non-linear compensation to rely on linear distortion compensated signals. After working in 1418, the return wave coefficient is determined through the estimation and convergence program and set to the return equalizer 26. Moreover, the return information flow is like an exact copy of the output of the memoryless non-linear part (ie, _142) that can be achieved by modifying the Wiener_Hammerstein I model. [0157] Next, a job 1420 controls the multiplexer 222 to connect the forward information stream 222 to the adaptation engine 1300 instead of the highest-order basic function information stream 214. Then, a job 1422 starts the sub-process 1100 again to execute an estimation and convergence program. This time, since the return equalizer 26 has set linear distortion for the output of the bandpass wave filter Η4, a secondary process of 1100 is initiated to the forward equalizer 246 to remove the possibility of being between the return and forward information flow. Any associations. This operation is 55 200541280 and is specifically aimed at compensating for the linear distortion that the input band-pass filter (BPF) 140 may produce. [0158] After working 1422, the secondary process 1400 is completed, and the linear and non-linear predistortion circuit 200 has been trained to compensate for linear distortion. Since the almost ideal signal is provided to the HPA 136, the HPA 136's amplified signal now closely matches the control design of the amplifier model design. Furthermore, the source of linear distortion after amp 142 has been compensated, so it is now possible to conduct nonlinear distortion training without being eroded by too much linear distortion. [0159] Referring to the previous FIG. 4, after the sub-process 14 ′, the transmission distortion management process 400 now performs a job 402 to start a process 1500. The sub-process 1500 is used to compensate for the nonlinear distortion generated by the HPA 136. More precisely, the sub-process 1500 at work 402 compensates for non-linear distortions without compensating for thermally induced memory effects. [0160] FIG. 15 shows a flowchart of a secondary process 1500. In general, the subprocess 150 is configured to be compatible with the Wiener-Hammerstein HPA model. In particular, it is assumed that the non-linear distortion is a form of higher harmonics of the amplified signal. In this model, because of the linear compensation discussed above, the signal amplified in the narrow mouth 142 is now driven by the basic function generating part 160. "Ideally, the signal fits tightly." Also, the basic function generating part 160 is generated. Higher harmonics of this signal. The non-linear predistorter 224 filters these higher harmonics and then combines them with the ideal signal in a subtractive manner. [0161] The sub-processing 1500 includes a generator from the basic function For the basic skill month b generated by 16⑻, choose the next basic function job 1502. In the first calculation of job 1502, any one of the basic functions, from the second basic function to the κth basic function, can be selected. Otherwise For job 1502, it is best to select the basic 56 200541280 function that has not been selected for the longest time before the previous calculation of job 1502. The next job is to determine the filter coefficient of equalizer 226 and set the equalizer η. Filter coefficients to train equalizers reserved for selected basic functions. [0162] In the preferred embodiment, the basic functions are actually orthogonal to each other. The way of intersecting, the chirp wave added to the basic function will have the smallest impact on other basic functions. Furthermore, when the filtering of the basic function and the wave change, these changes are less likely to affect other basic functions. [0163] FIG. 16 shows a basic function suitable for linear and non-linear predistortion circuits 200. A block diagram of an embodiment capable of generating part 1600. This embodiment is necessary because it can use a relatively simple hardware design It achieves the substantially orthogonal basic function. Moreover, it will respond to a high-resolution, low-error input information flow, and the result can provide a high-resolution, low-error input information flow. However, although the basic function The generating part 1600 provides suitable results for the purpose of the linear and non-linear predistortion circuit 200. Those in the same industry can design alternative embodiments. [0164] An amplitude circuit 1602 and a multiplier 1604 can receive complex pre-distortion. To information stream # 206, the amplitude circuit 1602 generates a scalar information stream representing the size of the complex forward information stream 206, and is connected to multipliers 1604, and multipliers 1606 and 160. 8. Figure 16 shows that the basic function generating part 1600 is divided into cells 1610, and each cell generates a basic function. Multipliers 1604, 1606, and 1608 are combined with different cells 1610. In general, each basic function will affect the X ( η) · | X (η) | κ responds, where χ (η) represents the forward information stream 206 received by part 1600, and K is an integer greater than or equal to one. Multipliers 1604, 1606, and 1608 The output is X (η) · | X (η) | κ information flow. 57 200541280 [0165] But in order to achieve substantial orthogonality, each basic function is equal to a near-weighted X (n) 丨 X (η) | ϊ ^ The sum of the stream, and all lower-weighted X (n). | X (n) | Kf streams. Therefore, the output of the 'multiplier 1604 is directly used as the second basic function and provides one of the complex basic function information streams 214. The multiplier 1606 multiplies the multiplier 1612 by a coefficient w22, and the output of the multiplier 1604 is multiplied by a coefficient by the Wzi multiplier 1614. The adder 1616 adds the outputs of the multipliers 1612 and 1614, and the adder 1616 is used as the third basic function and provides another complex basic function information stream 214. Similarly, the output of multiplier 1604 is multiplied by multiplier 1618 φ by a coefficient W3 !; the output of multiplier 1606 is multiplied by multiplier 1620 by a coefficient w32; and the output of multiplier 1608 is multiplied by multiplier 1622 by a coefficient W33. The outputs of the multipliers 1618, 1620, and 1622 are added to an adder 1624, and the output of the adder 1624 is used for the fourth basic function and provides another complex basic function information stream 214. In the preferred embodiment, these coefficients are determined at the design stage using the Gram-Schmidt orthogonal technique, or any orthogonal technique by anyone in the same industry. Therefore, these coefficients remain unchanged during the operation of the transmitter 100. However, under certain conditions, these coefficients cannot be guaranteed when transmitter 1 () 0 is operating. [0166] People in the same industry know that the basic function generation part ι6⑻ can be expanded and added to the cell 1610 to provide any number of basic functions. Furthermore, people in the same industry know that pipeline stages can be added as needed to suit the clock characteristics of the parts in question, and confirm that the basic functions are made from virtually equal clocks. The larger the number of basic functions, the more appropriately the non-linear distortion can be compensated. But adding a lot of basic functions will inevitably deal with very wideband and ultra-wideband signals. The preferred embodiment is expected to use 2-5 basic functions, but the invention does not have such a requirement. 58 200541280 [0167] Referring to Figure 15 above, after selecting the basic function in job 1502, job 1504 is either with or without heat treatment. If job 402 starts the sub-treatment 150, job 1504 does not use heat treatment. Next, a job 1506 calls the process u⑻ to execute an estimation and convergence procedure to determine the appropriate filter coefficients for the non-adaptive equalizer 226 associated with the selected basic function. During initial and linear compensation, the selected non-adaptive equalizer may have failed by setting all filter coefficients to zero. During work 1506, the coefficients of this non-adaptive equalizer 226 have been determined to minimize the correlation between the forward data and the error information stream and maximize the correlation between the forward and return information streams. Into. When orthogonal basic functions are used, the increase in the correlation between forward and return information flow has no impact on other basic functions. [0168] After work 1506,-ask work · determine whether all basic functions have been touched by the secondary processing. As long as there are other basic functions waiting for processing, program control will return to work 1502 to determine the filter coefficients of the remaining basic functions. . #Work confirms that all basic functions have been processed, and the next processing is completed at 1500. [0169] Referring to FIG. 4 above, a job is started after job 502 starts a sub-process of 150,000. # Degrees starts a sub-process of 150. This process_ is used to train non-linear compensation with heat treatment. Therefore, the human-processed thermal difference signal of the work of 1500 plus 4 will be suppressed. Referring to FIG. 13, the heat treatment may be selected to use a single delay element 1348 adapted to the engine. [0170] FIG. Π shows a block diagram of an embodiment suitable for linear and non-linear predistortion circuits 2 ', which represents a thermal change estimation section 17'. This embodiment has its needs because it is configured with a thermal difference signal 216 to respond to instantaneous changes in the long-term average power displayed by the forward information stream, and it uses a-relatively simple hardware implementation. But although the thermal change estimation section 17 provides suitable results for the purposes of linear 59 200541280 and non-linear predistortion circuit 200, those in the same industry can design alternative embodiments that work. [0171] The thermal processing amplitude determination circuit 1702 in the thermal change estimation section 1700 receives the complex forward information stream 206. The magnitude of the complex signal is determined in the circuit 1702, so that an amplitude scalar signal is driven-programmable_positioning part ·. In the embodiment, the amplitude determining circuit provides a signal flow of amplitude value in response to the magnitude of the complex forward information flow 206. In another embodiment, the circuit 1702 provides the signal flow of the amplitude value to the promotion Greater than a power. [00172] The programmable time positioning section 1704 accepts programmable input from the controller (c) 286, and the programmable time positioning section 1704 can be configured similar to the description of Figures 7 and 8 above. In other words, part 1704 allows the controller to change the delay of the information flow in the amplitude value of part 286. ^ 1704 provides-the amplitude of the information flow delay time to an embodiment of the iir wave filter. [0173] The output of the IIR filter at an adder 1706 provides an average amplitude output, but this average amplitude output is not the output of the thermal change estimation part. This average amplitude output can provide a representation of the long-term average amplitude city's present time, and this signal is connected to a single period delay element. Its output provides a previous representation of the long-term average amplitude signal. The difference of this long-term average amplitude signal is that the human meter is connected to the first input of the adder ， and connected to the negative input of a subtraction circuit mo. From the _location part Lan Zhizhen _ Lai Liu is provided to the positive input terminal of the subtraction circuit 171G, and the output of the subtraction circuit 提供 provides a thermal difference signal addition, and this is the output of the thermal change estimation part _. The thermal difference signal 216 receives the input from the receiver-receiver 1712 and the controller (C) 286 provides a receiver value η to the second input of the convergence multiplier 1710. One of the convergence multipliers 1712 outputs to the second input of the adder Satoshi. 200541280 matrix] θ Here, the long-term average amplitude signal reflects the average amplitude in time with the forward information flow 206, or a power greater than-, and at each clock cycle, the fraction of the instantaneous amplitude value. The value of Yang Yang in this recording agency is ^^, and the smaller money η makes the long-term average amplitude signal less able to respond to the instantaneous magnitude value. Furthermore, the thermal difference signal 216 characterizes the deviation of the instantaneous magnitude value from the long-term average amplitude signal. _] Refer to Fig. 4 and Fig. 15 above, in the work of transmitting distortion management. The filter coefficient of Huayi 226 is continuously adjusted. Moreover, during operation 404, the thermal sensitivity of the equalizer is adjusted by the thermal difference signal 216. Work 1506 started through work 404. Each calculation-time W starts to equalize the estimation and converge the program to determine the filter coefficient and the thermal coefficient. [0Π6] Reference circle 1 When the processing process converges to a result of a set of wave coefficients, the inquiry work 1115 starts. Work 1115 determines whether heat treatment is to be included. During work 404, when the heat treatment is to be included, the program control advances to a work 28, 30 and U32. Tasks 1128, 1130, and 1132 are optional tasks and are used to execute the -time program control of this path for initialization purposes, and then only occasionally execute this program. In one embodiment, tasks 1128, 1130, and 1132 are set in the next processing 1500, and are performed only in the first calculation of the program loop. [0177] Work 1128 couples the engine (CE) 280 and the difference coefficient signal 279 to the multiplexers 270 and 278 'to use an appropriately selected method to couple to the associated thermal difference signal 216. Then, job 1130 performs a time positioning optimization operation. In particular, the thermal difference signal 216 is delayed due to more accurate estimation until convergence is reached. At this time, the result that the thermal difference signal 216 is associated with the difference 200541280 coefficient signal 279 can be obtained. An optimization program similar to the one discussed in Figure 6 above. Other optimization programs can be used in job 113. At this time, the thermal difference signal 279 has reached the time position in the middle of the adaptation engine. The thermal change of HPA w, as shown in the scene of the forward information flow, changes the filter coefficient of the intermediate point to the maximum possible degree. [0178] After Job 113, Job 1132 performs another optimization operation. At work 113, the convergence values η and γ are optimized. The convergence values η and γ determine the sensitivity of the long-term average to instantaneous changes in power and intermediate filter coefficient signals. The "convergence value" and γ are preferably small positive values in order to be sensitive enough to instantaneous changes in the call. However, the optimization of the convergence values η and γ is performed by gradually correcting these values until the maximum correlation is observed in the correlation engine 280. The result is obtained. [0179] Next, part of a job 1134 is the hot part of the estimation and convergence program and the unlocking adaptation engine (ΑΕ) 13⑻ sets the convergence conditions to perform the thermal coefficient processing and filter coefficient processing. Convergence The setting of the conditions and the unlocking of the adaptation engine 1300 can be done simultaneously by providing a positive convergence variable λ of the adaptation engine 13000. Of course, this value is preferably a very small fraction of one. The convergence condition determines the adaptation How many samples does the engine 1300 need to process to determine the result that converges to a set of thermal systems. As discussed above, a larger number of samples can lead to an increase in the effective resolution of the return information stream or a reduction in the error level. Program The increase in processing time can therefore be converted into a reduction in the effective error level of the return information stream. By working 1134, the speed of convergence can be controlled to achieve a predetermined The effective return error level is lower than the error level of the return information stream. In one embodiment, the convergence variable λ is initially set to a slightly higher value, but it decreases with time. [0180] After working for 1134, the adaptation engine is 130%. Now execute two minimum mean square, estimation and convergence programs. One of the programs continuously changes the estimated value of the filter coefficients to minimize the 2% error in the information flow. 62 200541280 The difference signal minimizes. The other program continuously changes the thermal coefficient. In order to minimize the error signal provided by the difference between the thermal difference signal 216 and the difference-associated signal 1338. Both the LMS and the prediction and convergence procedures repeatedly update the filter coefficient and the thermal coefficient to minimize their error signals. [0181丨 At this time, a job 1136 is asked if the thermal coefficient determined by the adaptation engine 1300 has converged. Job 1136 and job 1134 jointly set the convergence conditions. Job 1136 may only decide whether there is enough time to reach convergence or job 1136 may monitor Adapting the thermal coefficient generated by the engine 1300, and when the wave coefficient was not found to have a consistent change in shape, it was decided that convergence had occurred. [0182] When work 1136 determines that convergence has occurred, the thermal coefficient α has been determined, and when it is multiplied by the thermal difference signal 216, this thermal signal becomes the largest correlation with the corresponding difference coefficient signal 1338. This At that time, the program control works 1116 to lock the adaptive engine 1300, obtain the filter coefficient and the thermal coefficient from the adaptive engine 1300, and set these coefficients back to the subject non-adaptive equalizer 226. Then the adaptor unit 1234 Then in response to the thermal difference signal 216, the filter coefficient is adjusted to be weighted with the corresponding thermal coefficient to compensate for the heat accumulated or lost at Ηρα 136. [0183] Referring to Figure 4 above, the linear and non-linear predistortion circuit 200 has been The linear and non-linear distortions generated by the analog part 120 are compensated, but the predistortion circuit 2000 cannot remove all the distortions from the HPA analog amplifier signal in, and there will be some residual amount. This residual distortion is contributed to the error vector magnitude (EVM). There are two forms of residual distortion that contribute to EVM, one is linear and the other is non-linear. The overall EVM generated using the transmitter 100 should be kept as low as possible, so that the reception of communication signals is better. However, the industry standard is configured to achieve acceptable reception, but allows 63 200541280 EMU. These two forms of residual distortion that contribute to EVM may to some extent be EVM. , Non-linear distortion

true, And does not substantially contribute to spectrum regeneration.  therefore, In some applications, It may be necessary to detect • EVM caused by μ non-linear distortion has increased ’distortion in this form, Swap into more benevolent band distortion.  [0185] Therefore, After working 404, A job 406 obtains a residual nonlinear EVM value.  This residual nonlinear EVM value, Is after linear and nonlinear compensation caused by nonlinear distortion,  Estimated or residual distortion remaining in the HPA RF analog amplifier signal 117. For example, In operation 406, the residual non-linear EVM value can be obtained by controlling the multiplexers 270 and 278. So that the error information stream 276 is associated with itself in the correlation engine 280, Then make at least two associations.  One of these two associations, Is to measure the error signal from the analog signal, This is the input of the HPA 136. The other is to measure the signal from the analog, That is, the input of QPA 136, The resulting error signal.  of course, Clock, Phase positioning and gain adjustment, Before making the associations, Both can be performed as described here. Suitable convergence conditions are used for the operation of two associations, So that the effective error level of the error information stream 276 can be greatly reduced as described above.  [0186] Job 406 may then evaluate two associations, The residual non-linear EVM value is obtained. This difference is mainly due to the memoryless nonlinearity 142 of HPA 136 and represents the non-linear distortion 64 200541280. Although __ multiple sources will record the results of each _, These filament sources are mostly common to all associated operations. therefore, The difference between these two associations produces a residual nonlinear EVM value, This value and the source of the noise are essentially independent of each other.  [0187] After working 406, A job 408 assessment, When the residual non-linear EVM value is too large compared to a preset value. Excessive values may cause ageing but not expired HPA 136, Aged power supply, Operates at extreme temperatures or a variety of other conditions. If the residual nonlinear EVM value is too large, Then, the operation 408 provides the peak reduction feedback signal 114 to the peak reduction portion 110. The feedback signal 114 is obtained based on the residual non-linear EVM value of Yuzuo 406 above. As mentioned above,  The peak reduction portion 110 changes the amount of peak reduction it adds to the forward stream. especially, When the value of the residual nonlinear EVM is too large, The amount of peak reduction will increase, So that ρρα ι36 can operate under a larger rollback, This results in reduced non-linear distortion. Increasing the amount of peak reduction also increases linear distortion, But it should also reduce some non-linear distortion. The transmitter 100 will thereafter operate at a lower non-linear distortion, Reception is moderately weakened, But spectrum regeneration can be virtually avoided. In addition, Task 408 may trigger an alert or send control messages automatically, To indicate the condition of the residual nonlinear EVM.  [0188] After working 408, Program control will return to any sub-processes and tasks within process 400. So that the processing and work can be repeated at any time according to the appropriate schedule.  [0189] The embodiments of the predistortion circuit 200 and the transmitter distortion management process 400 discussed above, A / D 304 in DDC300 produces only negligible small amount of distortion, When displaying the ultra-wideband feedback signal added to the A / D 304 processing according to the frequency effect, Can provide favorable results. Only when the amplitude of the quantization error caused by phase noise or aperture jump in A / D 304 is not related to the unrelated error 65 200541280 will present obvious problems, Because the estimation and convergence procedures discussed above for processing feedback signals can tolerate such errors, Noise and jitter.  [0190] However, even at a low resolution, The high error a / d 304 may also be a complex component,  And the overall cost of the predistortion circuit 200, It may be further reduced by allowing the use of the less complex A / D 304 which will produce some distortion into the feedback signal. If this distortion is not compensated,  Would be incorrectly interpreted by the transmitter distortion management process 400 as the analog transmitter part 120 producer.  therefore, In addition to removing the above sources of distortion, Equalizer 226, 246 and 260 will set the contact value that can also produce unwanted distortion in the forward information flow. The unwanted distortion is the inverse of the distortion produced by A / D.  [0191] FIG. 18 shows a block diagram of a second embodiment of the linear and non-linear predistortion part 2000, which is hereinafter referred to as a predistortion circuit 180 of the transmitter 100. The pre-distortion circuit 1800 is configured to compensate for the linear and non-linear distortions discussed above with Some a / d induced distortion. By using the predistortion circuit 1800, The transmitter 100 can even use an inexpensive A / D to produce significant distortion, The feedback signal into its processing.  [0192] The predistortion circuit 1800 is configured to be mostly like the predistortion circuit 200, And most of the discussion of the predistortion circuit 200 described above also applies to the predistortion circuit 180, In the block diagram, The same reference numerals between 2 and 18 indicate similar parts. but, For the sake of convenience 'some parts of the predistortion circuit 200, Such as the gain adjustment circuits 30 and 256, Thermal change estimation circuit 1700 and a circuit that generates a clock signal for A / D 304, They are omitted in Figure π.  People in the same industry understand that these parts still need to be included in the predistortion circuit 18⑻, And actually use it, As discussed above with Figure 2-17.  66 200541280 [0193] Pre-distortion circuit ι_ also includes rate multiplier 204 This is used to generate the speed-up complex forward information flow 206. Forward information flow 206 drives the generation of basic functions, Delay element 208,  -The real part of the delay time of the worms, The common mode time positioning part 700 shown in the phase 2 is shown.  [0194] The basic function generating section 1600 provides a majority of basic function information streams to the non-linear predistorter 224, A settable delay portion corresponding to a majority of 700. The non-linear predistorter 224 includes an analog transmitter part compensator 1803, And contains a majority of equalizers and merge circuits 228, As discussed in Figure 2 above, However, the merging circuit 228 is omitted from FIG. 18 for convenience. The analog transmitter component compensator 1803 is used to deal with the analog transmitter component 120.  Fig. 18 Equalizer 226 is injected into EQ | ^ PA, The subscript "k" represents the basic function of the equalizer 226 and the subscript "HPA" indicates that the equalizer 226 is suitable for compensating the non-linear distortion generated by the HPA 136. The non-linear predistorter 224 provides a complex filtering basic function information stream 230 to the negative end of a combined circuit 22o. The delay element 208 provides a complex forward information stream 218 to the positive end of a combining circuit 220. The merging circuit 220 provides a complex non-linear pre-distortion forward information flow 238 to the forward equalizer 246 and 0 to a settable delay portion, As discussed above, but, 囷 18 draws the equalizer 246 and the delay part 800 in a different order. The forward equalizer 246 is also included in the analog transmitter component compensator 1803. The time delay part 800 is actually equivalent to the differential mode time positioning part 800 discussed in Figure 2-17. [0195] The time delay part 800 provides complex orthogonal balance equalization forward information flow 118 to digital / analog conversion (D / A's) 122, The remainder of the analog drive part 120 is driven. As discussed above,  D / A's 122 preferably presents a resolution much higher than a / d 304. In Figure 18, The box labeled 67 200541280 “XPF” includes a low-pass filter 124, The up-conversion section 126 and the band-pass filter 132 of FIG. The output of one of the band-pass filters 132 provides a radio-frequency analog signal 34 to the multiplexer 250 and the HPA 136, The RF analog signal 117 comes from M! % Output, Connected to the multiplexer 25. Unlike the predistortion circuit 200 discussed in FIG. 2 above, D / A, One of s 122 also directly generates the fundamental frequency signal 123, Connected to the multiplexer 250. The fundamental frequency signal 123 is a one-bit filtered signal. Because it does not pass the filter provided by the analog transmitter part 120. therefore, It does not suffer the distortion imposed by the filter.  [0196] In one embodiment, D / A 122 is actually equivalent to the other parameters in terms of resolution, In another embodiment, The D / A122 that generates the fundamental frequency signal 123 has a higher resolution and / or quality than other D / A122s. In another embodiment, The third D / A (not shown) is dedicated to drive the base number 123 ′ but also; ϊ Need to transfer other analog launch age pieces 12 (). The driving base frequency signal I23 WD / A should have high resolution and high quality. because, There is a more detailed discussion below,  D / A is used to create compensation for A / D 304, This side will be limited by any distortion caused by D / A. fortunately, 咼 D / A with high resolution and high quality has low price products everywhere.  [0197] The settable delay part 700 provides the delay before moving the information flow to the phase conversion part. And the phase conversion part 1000 provides positioning complex forward information flow 272, . Forward Information Stream 272,  Upward frequency magnetic 272 ' is up-to-digital ± up conversion (DUC) miscellaneous. Duc part 1806 digitally forwards the information stream 272, Up to ⑽, Where Fs is the sampling frequency.  One of the outputs of part 1806 drives a real number conversion part 1808.  [0198] Each settable delay portion 700, , Configured similar to the delay part 700, And each of them is connected to its own phase-change part, . Turn to some recitations, Are configured to resemble the phase conversion part, And turn to part 1_, Each provides—the basic function information of positioning flows to _non-linear pre-68 200541280 distortion suppression 224 non-linear pre-distorter 224, Preferably configured similar to the non-linear predistorter 224,  However, it is included in the A / D compensation section of 1805. especially, Non-linear predistorter 224, Including a majority linear equalizer 226, , And one of the equalizers 226, Dedicated to the basic functions of the independent wave, the equalizer 226 ′ is labeled as eqI Where the superscript "k" stands for equalizer 226,  Related basic functions, And the subscript "trip, , Represents equalizer 226, The outputs of the equalizer 226 used to compensate for the distortion produced by the dove are all merged together. As discussed above (not shown), Then non-linear | ± pre-distortion H 224 ’generation—complex basic function wave information flow 18G9, And provide up to a digital ® upscaling (DUC) ^ injury 1810. This in turn drives the real-to-real conversion part 1812.  [〇199] Real number conversion, The brain and 1812 each converted their complex forward information flow into a real information flow, Use technology familiar to people in the same industry, Four pairs of samples from each group of complex forward to information flow. Face and job choose it! , q and 0 samples. The real number conversion part 1802 is coupled to a settable delay part 700, , And this can be configured to be substantially similar to the configurable time delay portion 700. Delay part 700, The surface is connected to a fixed delay part ⑻4 ′, and this execution is substantially equal to the phase conversion part 1000 and the surface, Added delay. The delay section calls for the provision of -delay forward information flow to _fixed delay element ^ delay element 1818 plus a delay substantially equal to the digital upscaling part 1806.  [〇2〇〇] real number conversion part of the surface and delay elements 1818 1820, And this is included in the trip compensation part. Swap some of the first output of the lion, Connected to-linear distortion compensator concept. The linear distortion compensator body is provided by a linear equalizer, And _ 18 is labeled EQk, Where superscript "丨, , The representation is a linear operator, And the subscript ‘fen, Indicates that the equalizer tripod is used to compensate for the distortion produced by the trip. In a preferred embodiment of the present invention over 69 200541280, The equalizer 1824 is preferably configured similar to the equalizer 226, 246, 260 and 226, ,  But the equalizer 1824 only needs to process a real information stream instead of a multiple information stream. And the number of contacts may be different. but, Such as equalizer 226, 246, 260 and 226, , The equalizer 1824 is preferably configured as an adaptive equalizer, Either directly or through operation of the adaptation engine 1300. Therefore, As discussed in more detail below with reference to Figures 19 and 24, The equalizer 1824 is adjusted to compensate for the linear distortion produced by the A / D 304.  [0201] The second output of the switching section 1820 is coupled to a quantization error compensator 2200, _ This compensator is also included in the A / D compensation section 1805. Generally speaking, The quantization error compensator 2200 ignores the compensation for the magnitude of the quantization error. However, the quantization error compensator 2200 symmetricizes the compensation error. Quantization error compensator 2200, This is discussed in more detail below with Figures 21-22. The second embodiment of a quantization error compensator 2200 that symmetricizes and compensates for quantization errors. This is discussed in Figure 31 below.  [0202 丨 Real number conversion section 1812 The outputs of the linear distortion compensator 1822 and the quantization error compensator 1826 are added together in a combining circuit 1828. A version of the data stream is provided to one of the inputs of the merge circuit 1830 before being output from the merge circuit 1828, The merging circuit 1830 provides a compensation point,  In order to merge the processed forward information stream with the return information stream output from A / D 304.  [0203] One of the merging circuits 1830 in turn provides an A / D compensation return information stream 1832 to the direct digital down-conversion section 1834. In this second embodiment, DDC 1834 contains only DDC 300 parts 308 discussed above from the first embodiment, 310 and 312. Generally speaking, A / D 304 effectively down-converts the sampled feedback signal to an intermediate frequency (IF) signal jy4, Where & Is the sampling frequency. DDC 1834 generates a return flow 254, This is essentially a reply letter 200541280 located at the fundamental frequency. As discussed in the foregoing first embodiment, The returning information stream M may present a lower error and lower forward information stream than the forward information stream. The return stream 254 drives the return equalizer,  In turn, these equalizers generate an equalized return information stream 262, As discussed in Figure 2-17 above. The return equalizer 260 is also included in the analog transmitter component compensator 1803.  [0204] In this second predistortion circuit 1800 embodiment, As discussed in Figure 217 above,  Controller 286, Adaptation engine 13〇〇, Adapt engine 13⑻, The correlation engine 28 must be coupled to multiple parts of the predistortion circuit brain, To control the direction and timing of information flow, And handle different versions of rebate _ news flow.  [0205] FIG. 19 shows a flowchart of a second embodiment of the transmission distortion management process 400 performed by the transmitter 100, This embodiment is referred to as Process 1900. Process 1900 is different from process 400 described above,  In that it contains extra sub-processing, Used to compensate for some forms of distortion produced by A / D 304. Process 1900 is discussed in more detail below.  [0206] FIG. 20 shows a typical analog / digital converter model 2000, Such as >  Ming A / D 304. Model 2000 shows multiple sources of distortion that A / D 304 can produce. An input analog signal 2002 is provided to the amplifier 2004, The amplifier 2004 of FIG. 20 is labeled "NL AMP", Let the amplifier 204 be a possible source of non-linear distortion. One output of amplifier 2004, Drive a Low Pass Filter (LPF) 2006. LPF 2006 is a possible source of low distortion,  Because the "knee" of this filter is usually very high above the frequencies we are interested in. One output coupling of LPF 2006 is one of switch 2008, A sample and hold circuit 2010 is then driven. The sample and hold circuit 2010 is similar to a low-pass filter. Can produce a lot of linear distortion. The sample and hold circuit 2010 drives an adder 2012 through a switch 2014. On the adder 2012, Probably 71 200541280 Adding DC displacement. Although DC displacement is usually an unwanted result, But it does not necessarily cause distortion-related problems. The adder 2012 drives a quantizer, The quantizer 2016 digitizes the analog voltage captured by the sample and hold circuit 2010. And provide digital output to add. Quantizer 2016 can be a source of several different errors.  [0207] FIG. 21 shows the characteristics of quantization and quantization error for an example of one-bit resolution wd. The two-bit resolution characteristic is not a requirement of the present invention. Figure 21 shows the two-dimensional spatial representation of the possible input analog voltages of all A / D on a straight line 2102, This is an example of how A / D could digitize the input analog voltage on line 2104, The result of the quantization error is shown in trace 2106. The -line three digits to the left of @ 21 represent the traditional two's complement representation of quantized output, And @ 21 之右 — another binary representation of the quantized output represented by the binary number, An embodiment suitable for use in a preferred embodiment. This 1/2 bit shift representation is compared to two's complement notation, Use one to increase the bit resolution, But does not include the zero state or any other even state, And have the same non-zero positive and negative states.  [0208] The quantizer 2016 of the A / D model 200 is characterized by exchanging a critical value of 2108. One _ two-bit A / D preferably has three exchange thresholds 2108 at zero and in the full 1/2 value range (FS / 2), The larger resolution a / d has more switching critical value 2108. When an input voltage is slightly lower than the switching threshold 21G8, A / D will output a code, When #input people voltage is slightly higher than the switching threshold 21G8, A / D will output another code. On the exchange threshold 21 () 8, The quantization error suddenly jumps from -local minimum to -local maximum. If all exchange threshold toilets are placed accurately, Then the absolute values of all local minimums and all local maximums are equal. The magnitude of the quantization error may cause an A / D distortion in a certain RF communication application. As shown in Figure 27_32 72 200541280. But in other applications, The magnitude of the quantization error is zero due to the estimation and convergence formula of the return information stream π%.  [0209] Whether or not the magnitude of the quantization error will cause a problem, If the exchange threshold is not set, Will cause asymmetry. Figure 21 shows one such asymmetry, The actual + FS / 2 exchange threshold 2108 has been shifted from the ideal position to the negative direction. But the —FS / 2 exchange threshold is in place. The asymmetry of this quantization error, Another kind of distortion is produced in the signal processed by pass 304. If uncompensated, Then this distortion will cause equalizer 226, 246, The contact coefficient with 26〇 _ is not accurate. The interesting asymmetry is about DC displacement, And this can but need not be equal to zero. The use of 4-bit displacement representation further enhances symmetry, Because each digit of the displacement representation has a corresponding positive value, Represents a corresponding analog input. in other words,  This encoding method is symmetrical at zero.  [0210] The predistortion circuit 1800 compensates the a / d quantization error, But more specifically, a quantization error compensator 2200 is used to compensate for asymmetry. Fig. 22 shows a first embodiment of a quantization error compensator 2200. Generally speaking, The quantization error compensator 2200 allows the formation of an effective exchange threshold 2108 at an ideal position. At least as accurate as d / A 122. Another embodiment of a quantization error compensator 2200 is discussed in FIG. 31 below.  [0211] Referring to FIG. 22, In the combining circuit 2202, a positive displacement is added to the analog feedback input signal driving the A / D 304. This positive displacement is not a requirement, But it can be used to simplify hardware. The positive displacement is preferably slightly larger than the maximum value of an actual exchange critical value 2108 which will shift in the negative direction of the ideal exchange critical value. therefore, A positive displacement has the effect of moving all actual exchange thresholds 2108,  Compared with the analog input signal, it presents a negative error effect. This negative exchange critical error results in some A / D digital outputs with 73 200541280 analog inputs, Showing too positive values, But this too positive output can be corrected by adding only negative displacement. Adapting A / D 304, And add the LSB of the two's complement representation from A / D 304, To provide the above-mentioned 1/2 bit displacement representation, And set this bit to "1" forever. As discussed above, The output of A / D 304 is connected to a merge circuit 1830.  [0212] In this embodiment, The controller (c) 286 is configured to monitor the compensation return information stream 1832 output from the combining circuit 1830. The controller 286 is also configured to write data to the logger 2204, 2206 and 2208. Recorder 2204, 2206 and 2208 output, Respectively coupled to the comparators 2210, Positive input of 2212 and 2214. Comparator 2216, 2218 and 2220 negative inputs,  All are driven by the output of the switch 1820. Comparator 221〇, The negative inputs of 2212 and 2214, Adapted to receive-FS / 2, 0, And + FS / 2, Where "FS" refers to the entire range. Comparator 2216, The 2218 and 2220 losers, It is also driven by the output of the switch 1820. From the comparators 2210 and 2216, the "greater than, , Output, When the positive input is greater than the negative input, And coupled to an input of an AND 2222, An active signal is generated; The “greater than output” from the comparators 2212 and 2218 is coupled to the input terminal of an AND gate 2224; And the output from the comparators 2214 and 2220 is greater than Coupled to the input of an AND gate 2226. From AND gate 2222 2224,  And the output of 2226, Is coupled to the input of OR gate 2228, And an output terminal of the OR gate 2228 is connected to the zero data input of the multiplexer (MUX) 223. An "〇, , Value is provided to zero input of multiplexer 2230, And one ‘]’ value provides at most one input of data. Output of one of the multiplexers 2230, Through a delay element 2234, a displacement value information stream 2232 to the parallel port circuit I828 is provided. Among them, this information stream is merged with the real number conversion part ⑻2 and equalizer.  Adding the information stream with the delay το 2234 will cause the quantization error compensator to display the same 200541280 delay ′, as shown by Ehwa Zhai 1824. As discussed above, The output of the merging circuit is 8, Tap to a negative input terminal of the merge circuit 1830.  [0213] With reference to FIG. 19 above, Process 1900 initially performed a secondary treatment of 23㈨, _ Sub-processing works with quantization error compensator 2200, To symmetric a / d quantization error. In the following discussion of the second embodiment of the quantization error compensator and FIG. 31, The secondary processing of 2300 compensates the magnitude and symmetric quantization of the quantization error. FIG. 23 shows a flowchart of the secondary process 2300.  [0214] The secondary process 2300 is configured to execute a power-on form, Or when the transmitter 100 does not transmit data #. Sub-process 2300 first performs a job 2302, To initialize a predistortion circuit 180.  Work 2302 can, For example, Set the basic function generator 16⑻, So that only zero values are output.  The multiplexer 250 should be set such that the baseband (BB) inverse signal 123 is connected to A / D 304. The equalizer 1824 is preferably set to output only zero values, And the switch 1820 is preferably controlled to pass through the delay part 700.  The fundamental frequency path is connected to a quantization error compensator 2200. and, Recorder 2204, 2206, And 2208 is best set to the largest negative value. In this state, Analog transmitter part 12 except for D / A 122 driving the fundamental frequency feedback signal 123, No distortion was generated into the signal monitored by the a / D 304. same, At the compensation point of the merge circuit 1830, The output of A / D 304 has no effect. Make recorder 2204, 2206 and 2208 exhibit the same maximum negative value, which prevents the quantization error compensator 2200 from affecting the output of the A / D 304.  [0215] After working 2302, A job 2304 identifies the actual exchange threshold used by an A / D 304. The first exchange threshold, For example, May be -FS / 2 threshold, And the positive displacement added to the merging circuit 2202, Make the actual exchange threshold less than the identified ideal threshold. Next, A job 2306 causes the D / A 122 to output an analog value. Because the resolution of D / A 122 75 200541280 is higher than that of A / D 304, Such ratio values are output with high accuracy, And it is directly fed back to the A / D 304 through the multiplexer 250.  [0216] Next, after waiting for a considerable time, A query 2308 determines whether the A / D output value has been converted from the previous value. Suppose job 2308 is unaware of the conversion action. Then job 2310 increases the output value to the information stream before the South resolution by one LSB. And the program flow returns to 2306 to output this new and slightly larger value. The program flow stays at work 2306, The circles of 23〇8 and 2310, Until outputting a value that causes the A / D output to be converted into a new output code. Due to the positive displacement, the exchange critical value presents a negative error value. The output of A / D 304 will now display a positive error value.  [0217] An actual exchange threshold has been determined, A job 2312 then records the actual switching threshold, And a query 2314 determines whether the previously perceived actual exchange threshold is the previous threshold. As long as there are other exchange thresholds waiting to be detected, The program flow returns to task 2304. To detect another actual swap threshold. When job 2314 determines that the last actual exchange threshold has been detected, Then a job 2316 sets the recorder 2204, 2206 and 2208 are the actual exchange thresholds. In a variant embodiment, The actual exchange threshold set by work 2316, It may be% LSBs or 1 LSB less than the detected actual exchange threshold and recorded in job 2312. at this time, The sub-process of 2300 was completed. The actual exchange threshold has been measured in the accuracy range provided by D / A 122.  [0218] In the next operation, The forward information stream driving D / A 122 is also provided to comparator 2210,  2212, 2214, 2216, 2218 and 2220 (Figure 22). Whether the value of the forward information flow is between the ideal and the actual exchange threshold, Can be detected by comparator 2210, 2212, 2214,  76 200541280 2216, 2218 and 2220 with AND gate 2222, 2224 and 2226, By combining the circuits 828 and 1830 to provide a displacement value of 1, To compensate the output of a / d 304. At last, Symmetry of quantization error can be achieved. For the effective exchange threshold 2108 used by A / D 304, When more positive than any DC displacement, A / D 304 also uses a critical ratio of 2108, Any DC displacement is more negative, And the average of the thresholds of the exchange of correction and more negative, That is equal to the DC displacement. More precisely,  For the resolution of D / A 122, The actual exchange threshold 2108 is converted to a more effective ideal parent exchange threshold 2108 '. This sets a near-zero DC displacement, And make all effective exchanges ϋ critical value 2108, Be symmetrical around zero.  [0219] Although the embodiment of the quantization error compensator 2200 of the circles 22-23 described above is based on the fact that the transmitter 2 does not transmit data and the process 2300 operates, This is not a requirement of the invention. In a variant embodiment, The transmitter 100 may identify the actual exchange threshold 2108 while transmitting the data. In this modified quantization error compensator 2200, The forward information flow can be monitored in the quantization error compensator 2200 for a period of time, And record the value of the maximum forward information flow of each A / D output state. The actual exchange threshold 2108 can then be determined to be a value slightly less than or equal to the maximum record. In another variant embodiment, The value of the largest and smallest forward stream of each A / D output state can be recorded when transmitting a large amount of data. Then the actual exchange threshold can be determined by averaging the maximum value of one state and the minimum value of the next state.  [0220] Slightly back to process 1900, After completing the secondary process 2300, A sub-process of 2400 starts to execute, To compensate the linear distortion produced by A / D 304. Referring to Figure 20, A / D 304 may produce linear distortion mainly through the sample and hold circuit 2010 and secondary through LPF 2006.  [0221] FIG. 24 shows an example of a flowchart of a secondary process 2400. The secondary processing 2400 is performed at any time when the data is transmitted by the transmitter 100 And it is better to set the quantization error compensation detection state 2200 to compensate the A / D quantization error in the second processing 2300. After compensating for the quantization error distortion produced by the source of linear distortion downstream of 304, Quantization error distortion is unlikely to hurt a resolution that is properly compensated for linear distortion. therefore, During the 24h period of the sub-process, The quantization error compensator 2200 is preferably enabled and operated.  [0222] Subprocess 2400 performs an initial job 2402, The initial predistortion circuit 1800 is used to perform the second process 2400. Work 2402 can control multiplexer 250, So that the fundamental frequency (BB) feedback _ 佗 number 123 can be connected to A / D 304. The switch 1820 can be controlled to pass the delay portion 700, The forward information stream can be connected to a linear distortion compensator 1822. The equalizer 1824 of the linear distortion compensator 1822 is initialized to a required state, To pass without filtering. and, One can adapt to multiplexer 2500 (Figure 25), So that when equalizer 1824 is used as an adaptive equalizer,  The appropriate ideal positioning and error signals connected to the adaptation engine 1300, Connect directly to the equalizer ι824.  [0223] FIG. 25 shows a signal used in combination with the predistortion circuit 1800, A block diagram of a multiplexer 2500 to drive multiple adaptive equalizers' including the contacts of the equalizer 1824. These contacts can be driven by the equalizer engine 1300. The workaround is, Multiple equalizers, Including the equalizer 1824, Can be configured as an adaptive equalizer. Figure 25 omits the sign of the complex signal for convenience. But people in the same industry know, The complex signal can be wired through the multiplex section 2500 if necessary. Generally speaking, Error signal 276 at the equalizer contact of the adaptive engine 1300,  Is in the subtractor circuit 274, Generated by subtracting the return flow from one version of the forward flow. The return information stream is connected to the subtraction circuit 274 through the multiplexer 2502. And multiple versions of the forward information flow are connected to the subtraction circuit 274 through a multiplexer 2504. The ideal positioning signal 272 also drives a number of 78 200541280 adaptive equalizer contacts' including the equalizer 1824. The ideal positioning signal 272 is obtained from the forward data signal through appropriate wiring 'of a multiplexer 2506. The multiplexing section 2500 is configured to direct the appropriate forward and return information flow, In order to generate suitable ideal positioning and error signals 272 and 276. In this embodiment, The high-pass filter (HpF) 314 has been combined with the embodiment of FIG. 2, And supplied to the downstream subtraction circuit 274. therefore, The error signal 276 is mostly generated directly by HPF 314. and, A delay element 2508 is added after the multiplexer 2506. The addition of the delay element 2508 is about equal to the delay of the HPF 314. So that the error signal 276 and the ideal positioning signal 272 can maintain the positioning with time.  [0224] With reference to FIGS. 24 and 25, Work 2402 can initialize the multiplexer part 2500, To select "〇 ， , Multiplexer input. These choices are input through the multiplexer 2502 input subtractor 274 and through a delay element 2510 and the multiplexer input M compensation return information flow.  therefore, The error signal 276 is basically provided by the combining circuit 1830. The ideal positioning signal 272 is provided by the delay-to-information stream 1816 through the multiplexer 2506 and a delay element 2512. The delay element 2510 is inserted by the DDC 1834 and the return equalizer 260 into a fixed time equal to the delay of the entire signal >  Delay. The delay element 2512 is composed of DDC 1834, Return equalizer 260, The digital upconverter and equalizer 1824 inserts a fixed delay equal to the overall signal delay. The delay of the delay elements 2510 and 2512 causes the error positioning signals 276 and 272, To deal with what happens later, Maintain time positioning, In this process, ‘different parts will be transferred into the signal path’. [0225] After working 2402 ’, it ’s best to perform 2600 for 2400. As discussed above, Or a similar process to perform an estimation and convergence procedure causes the forward and return information flow to be time-positioned at the compensation point provided by the merging circuit 1830. Time positioning can change the delay element. 79 200541280 Xiao Cheng joined Cheng Gengyan Qian Li, _Supervisor 8) Predicted output. The RMS estimator 2514 has a wheel whose input is connected to a subtractor 274, This output reflects the timing of the compensation point. The RMS estimator 2514 preferably performs the same function as the correlation engine 28.  And configured to accumulate an estimated abdominal 8 value for a large number of samples, As discussed in Correlation Engine 2 above.  At that time Yanshu 700 ’was set up; ^ At Gu Zawei 2514 _ 丨 —at the minimum face value,  Time positioning is reached. In a modified embodiment, The correlation engine 28 can be used to find the maximum correlation between the data and the returned data signal before the compensation point.  [0226] After execution of the secondary process 600 from the secondary process 2400, The second process 2400 is the second process 1100 ’. To run an estimation and convergence program, And this program solves the contact coefficient of the equalizer 1824. After the secondary process 1100 is completed, Its coefficient is determined by And set to equalizer 1824, And the coefficient just determined makes the return information flow output from a / d 304 and the forward information flow get the maximum correlation result. at this time, The equalizer 1824 has been adjusted to compensate for the linear distortion produced by A / D 304, And the secondary process 2400 has been completed.  [0227] Referring to Figure 19 above, After the secondary process 2400 is completed, Process 1900 then proceeds to process 500, As discussed above, To compensate for linear distortion generated upstream of HPA 136. During sub-process 500 and subsequent sub-processes, The quantization error compensator 2200 and the linear distortion compensator 1822 remain in the set and operating state. So that when these subsequent compensations occur, Use A / D distortion compensation for sub-processing.  [0228] During work 502 of the sub-process 500, The multiplexer 250 is switched to connect the RF feedback signal 134 to the A / D 304. The radio frequency feedback signal 134 is an up-conversion form of the fundamental frequency feedback signal 123. In addition, there is no distortion included in the fundamental frequency feedback signal 123. therefore, Work 502 is best transferred to 200541280 converter 1820, The forward information stream is passed through the delay element and the digital up-converter to the quantization error compensator 2200 and the linear distortion compensator. Although the up-conversion part 126 is not required, it is better not to up-convert to Fs / 4. As is the case with digital upconverters,  304 performs a sub-sample down-conversion and centers its output at Fs / 4. therefore, The up-conversion part 126 and ⑽ 304 work together as if performing up-conversion to Fs / 4. The quantization error and linear distortion compensation previously determined for the fundamental frequency can now be used at Fs / 4.  [0229] In addition, The initial work 502 should control the multiplexing part 2500, In order to select the multiplexer input terminal marked as τ. These choices, The return information stream 262 is connected to the subtractor 274 through the multiplexer 2502 and the forward multiplexer is connected to the forward information flow 272 through the delay component. To form an error signal 276. The ideal positioning signal 272 consists of a forward information stream 272, It is provided through the delay element 2516. The delay element 2516 adds a value equal to DDC 1834, Return equalizer 26〇,  Digital upconverter delete, Fixed delay with the signal delay added by the equalizer,  To maintain time alignment with other processes, And in this process, Different parts are routed into the signal path.  [0230] After working 502, The second processing is 500, and then the common mode and differential mode time positioning is adjusted by setting a delay of 700 and ca. As discussed in Figure M above, And adjust the phase conversion part to position the phase as discussed in Figure 9-10 above. then, The subprocess 500 performs an estimation and convergence procedure to solve the contact coefficients for the forward equalizer 246. at this time, Цρα 136 The linear distortion generated by the upstream stream to the information stream has been compensated.  [0231] Referring again to FIG. 19, After the secondary process 500 is completed, Process 1900 followed by a process of 2600, To compensate for the non-linear distortion generated by the A / D 304. Reference circle 20, _3〇4 81 200541280 Non-linear distortion is generated mainly through the operation of NL amp 2004.  [0232] FIG. 26 shows a flowchart of one of the sub-processes 2600. The processing 2600 is preferably in the predistortion circuit 1800 to be set to compensate for quantization error distortion, A / D linear distortion, And HPA 136 upstream linear distortion. at this time, The RF feedback signal 134 has been adjusted to remove linear distortion, And there is no substantial amount of nonlinear distortion on the path of the radio frequency feedback signal 134. therefore, Any non-linear distortion mainly comes from A / D304.  [0233] Process 2600 includes an initialization job 2602, It is used to set the predistortion circuit 1800 to determine the corrective action of the A / D nonlinear distortion that needs to be compensated. Work 2602 can convert the multiplexer 250 ’to connect the RF feedback signal 134 to the A / D 304, And the control switch 1820 passes the forward information flow through the delay element 700, The digital upconverter 806 is connected to a quantization error compensator 2200 and a linear distortion compensator 1822. and, The multiplex section 2500 can be controlled to select the multiplexer input terminal indicated by "2" in FIG. These choices connect the return information stream 262 to the subtractor 274 through the multiplexer 2502, While forward information flow 272, Through delay element 2516 and multiplexer 2504, To form an error signal 276. The ideal positioning signal 272 is provided by one of the basic function information streams 1804, If the stream marked D2, There is a delay through the delay element 2518. The delay element 2518 adds one equal to DDC 1834, Return equalizer 260, Digital upconverter 1806 and first equalizer 1826, Added to the fixed delay of the overall signal delay, To maintain the positioning of _ with other processing, Different parts of other processes are converted into the human signal path. Initializing Yuzuo 26G2, you can also choose the basic function generator 1600 to generate basic functions. But the equalizer 226 has no choice but to generate a zero stream of information. Palpitations equalization of basic functions H 226, , For example, the EQ ^ of the d2 basic Wei f news stream, It is best to set the initial value to any other equalizer 226 for processing, It is best to initialize to output a zero stream.  82 200541280 [0234] After initialization work, The second process is to perform κ at a glance to set the delay part 700 ”and the phase conversion part. . Work can, But not needed, Use the estimation and convergence program to determine the appropriate delay and phase settings. If you use this program, They can then be configured substantially as discussed in Figure VII above. However, the delay part 700 ”and the phase transition part have a fixed relationship to the delay part 700 and the phase transition part, respectively. This fixed relationship is determined by the relative delay of the components of the forward flow path. therefore, Partial delays, It can be set using only the displacement determined in advance by the delay portion 700 and the phase conversion portion 1000. The purpose of this program is to make the forward information flow on this path reach the merging circuits 1828 and 1830 in time positioning and propagate through the delay sections 700 and 700 'before.  [0235] The next treatment is 2600, and the next treatment is 1100. To perform the estimation and convergence procedure of the equalizer 226 '. After finishing the secondary treatment, eql equalizer move, Set the coefficient, While making the second-order basic function filtering, So that it is most relevant to the presentation of the return flow from W304. The second-order distorted parts of this return stream are then minimized.  [0236] FIG. 26 shows an example, The predistortion circuit 180 uses three basic functions.  _ So, In order to make this case the secondary processing of 2600 repeats the secondary processing of 11,000 more: I think eq2a① equalizer 226 ’and EQK + equalizer 226, Find the coefficient. In the next calculation of processing u⑻, it is best to control the multiplexing section 2500 and select the multiplexer shown in Figure 25 with "3, , Input with "4". Both options connect the return information stream 262 to the subtractor 274 through the multiplexer 2502, And forward stream 272, Through the delay element 2516 and the multiplexer 2504, An error signal 276 is formed.  In the choice of ‘3, The ideal positioning signal 272 provided by the basic function information stream 1804 is labeled & , Delayed by the delay element 2520. While at "4, , In the selection, The ideal positioning signal 272 provided by the basic function information 83 200541280 stream 1804 is labeled Dk + i, Delayed by the delay element 2522.  The delay elements 2520 and 2522 each add the same delay as the delay element 2518. People in the industry know that there is no need to prescribe the use of several sets of basic functions. After 1100 times of necessary processing, Processing 2_ is complete, And the nonlinear predistorter looks like, The non-linear distortion generated by the dove 304 has been set to compensate.  [0237] With reference to FIG. 19 above, After performing the secondary processing at 2600, All the substantial forms of distortion produced by a / d 304 have been compensated. therefore, As discussed above, The remaining processing 19⑻ part chase >  The corresponding part of the trace processing 400. The secondary processing 1400 is used to compensate linear distortion through M 136,  therefore, As shown in Figure 14, The initialization work 1402 in the secondary process 1400 controls the multiplexer 25 to connect the RF feedback signal 117 output from the HPA 136 to the input of a / d 300. Positioning of time and phase, Because HPA 136 monitors forward information flow 272, And the feedback information stream 1832, And readjusted.  [0238] Then, This process was performed 100 times for three times. The first calculation of 1 100 was started at work 1414, Used to control the multiplex section 2500 to select "5, , So much work.  ϋ Enter, this has the same effect as selecting "1". The forward coefficient of the forward equalizer 246 is determined during the first calculation. The second calculation of processing u⑻ occurred in job 1418 but the previous job 1416 can control the multiplexing part 2500 to select the multiplexer input marked "6" in 圊 25. This selection sends the return information stream 262 through the delay element 2522 and the multiplexer 2502 to the subtractor 274 and the highest-level basic function (that is, D-call) to form the error key m. The ideal positioning signal 272 is also provided by the highest-order basic function (that is, Dk + 0 is provided through the delay element 2522. The return coefficient of the equalizer 260 is determined during the second calculation. The third processing · calculation is in work 84 200541280 Occurs as M22, but the previous job 1420 can control the multiplexer input of the multiplexer in the measurement selection map μ in the month 1 or 5. During the third calculation, the forward equalizer 2 you forward coefficient It is readjusted. [0239] After 14 hours of sub-processing, process 1900 performs work 402, in fact as discussed in Figure 4 and Figure 15 above. Work 402 performs sub-processing 15 years to compensate for the price of 136 The linear distortion does not include the memory effect of thermal induction. The sub-process 1500 calculates and connects different basic functions to the adaptation bow, and executes the sub-process 1100 to perform the -estimation and convergence program.疋 Equalizer coefficients. After the three basic functions discussed above, for these calculations, the multiplexing part 2500 can be controlled to select the multiplexers marked "7", "8", and "9" in Figure 25, respectively. Input. Each selection is through multiplexer 2 5G2 connects the return information stream 262 to the subtractor and the forward information stream 272 through the delay element 2516 and the multiplexer 2504 to form the error signal 276. In the choice of 7, the basic function of D2 is marked The information stream 1804 provides the ideal positioning signal 272 'delayed by the delay element 2518 and determines the coefficient of the EQ? PA equalizer 226. In the choice of 8, the marked basic function information stream 1804 provides the ideal positioning The signal 272 is called for delay through the delay element 2522, and determines the coefficient of the equalizer 226 of EQjpA. In the selection of "9 ,," the basic function information stream 1804 marked with DK + 1 provides the ideal positioning signal 272. The delay element 2518 delays, and determines the coefficient of EQK + {pA equalizer 226. However, people in the same industry know that there is no stipulation that these basic functions must be used. [0240] As discussed above for the processing of 400, after work 402, A job 404 repeats processing 1500 times, but this time the memory benefit of thermal induction must also be compensated. Then, after job 404, jobs 406 and 408 get a residual EVM value, and use this value to adjust the peak reduction After work 85 200541280 and 408, any subsequent processing and work in processing 1900 towels can be repeated as needed to allow the predistortion circuit 1800 to provide compensation in time and temperature. [0241] Figure 27 shows Multiple chicks are used to convey the example of four multiplexed channels. Refer to Figures i and 27. In this example, four modulators generate four independent information flows 2700. Each independent information flow 2700 is located at the fundamental frequency when generated by its individual modulator 104. In other words, each characteristic is extended to -3 using a bandwidth with a center point of 0 out. Between 8 MHz and + 3 · 8 MHz. [Lu 0224] In this example, the combiner 106 uses frequency division to combine the four independent information streams to generate a complex forward information stream 108, as shown in trace 2702 in the figure. Two of channel 2704, labeled "A" and "B" in 囷 27, are centered on negative frequencies (ie, _7 · 5 MHz and -2.5 MHz) 'and two of channel 2704 , Marked as "C" and "D" in Figure 27, are centered on a positive frequency (ie +2. 5 MHz and +7.5 MHz). Use negative and positive frequencies to represent frequency division of labor. Channel 2704 allows the use of lower clocks. ^ If all channels 2704 are frequency-characterized with the same polarity, The forward information stream 108 must be processed using this clock.  ® [0243] Trace 2702 is the spectral characteristic of the wideband forward information stream 108. Because it is processed downstream of the combiner 106 in the transmitter 100. For illustration purposes, One of the channels 2704, This example assumes channel B, Weaker signal than other channels, In this case, A wideband digital communication signal configured to include multiple discrete frequency division channels, And these discrete channels show varying signal strengths with each other, Application used to represent a mobile base station with other digital communications. But those in the same industry know that the invention is not limited to this particular application, And this particular application is not limited to any particular number of channels 2704, Nor is it limited to any particular restrictions on relative channel strength.  86 200541280 [] __ of track 2 losing fortune, Its towel forward information flow is configured to include multiple 7-bit division of labor channels, And the signal strength of the discrete channel depends on the presentation of the signal strength. If the complex information flow m is perfectly up-converted in the up-conversion part, the in-phase part of this up-conversion will initially frequency-synthesize the sum and difference of each channel. , The orthogonal part of the up-scaling will also initially shift the frequency and divide each channel into the sum and difference of the frequencies. The upsampling depends on the orthogonal part of the meeting and is based on ", , How the in-phase and quadrature parts merge, The sum of the frequencies will completely cancel each other out or the difference in frequencies will all cancel each other out / Xiao. in other words, There is no image signal at the full upscaling of _ negative signals.  [0245] However, it is unlikely that the operation of the upsampling section 126 will have a complete upsampling. Although the forward equalizer 246 is one of the 800 targets with the difference time localization portion, Is to try to balance the in-phase and quadrature parts of the complex forward information flow 118, But inevitably there will be some residual imbalances. This residual imbalance, At 126 liters of tenderness, It will make the miscellaneous term 丨 now the radio frequency analog signal 130 ′ because the orthogonal terms of opposite complex parts will not completely cancel each other out. and, Because channel A is on the negative frequency used by channel D, And channel B is at the negative frequency used by channel c, The video signal 2708 will fall within the frequency band. in other words, The image signal 2708 from channel a falls on channel D in the radio frequency analog signal 130; The image signal 2708 from channel b falls on channel C in the radio frequency analog signal 130; The image signal 2708 from channel c falls on channel B in the radio frequency analog signal 130; The image signal 2708 from channel D falls on channel A in the RF analog signal 130. The video signal 2708 is unnecessary. Because they represent errors in occupied channels, Noise or interference.  [0246] Trace 2706 illustrates, The video signal 2708 is weaker than the signal they are video.  87 200541280 Therefore, When the channels are at mutual video frequencies, Approximately equal to the intensity of the communication channel 2710, For channels A and D ', the image problem can be easily managed by the embodiment discussed in Figures 1-26 above. The receiver is tuned so that the receiving channels A and D can successfully modulate their signals. The error signal strength 2712 caused by the video signal and the communication signal strength 2710 of channels A and D are relatively weak.  [0247] But when the channels fall on each other's video frequency, And at very different intensities,  There will be anxiety about images, For example, channels B and C. especially, Trace 2706 is an example,  Among them, the weaker channel B ’s communication signal strength 2710 is stronger than the stronger signal ’s 2712 error signal strength. Maybe weaker. Once tuned, the receiver receiving channel c can easily modulate its signal, Because the error signal strength of channel C is 2712, Is caused by the image of channel B, Compared with the 2710 signal strength of channel C, it is very weak. on the other hand, A receiver receiving channel B upon tuning may not be able to successfully demodulate its signal. The error signal strength at channel b is 2712, Caused by the image of channel C, Compared with the 2710 signal strength of channel B, it is very strong.  [0248] FIG. 28 shows a block diagram of a third embodiment of the linear and non-linear predistortion section 200, This is hereinafter referred to as the predistortion circuit 2800 of the transmitter 100. The predistortion circuit 2800 is configured to meet the requirements of the error vector size (EVM) and / or noise ratio (S / N) of the weak and strong channels transmitted by the transmitter 100, Perform pre-distortion and other transmitter processing.  [0249] The predistortion circuit 2800 is configured mostly like the predistortion circuits 200 and 1800, And the above discussion about the predistortion circuits 200 and 1800, Most of them apply to the predistortion circuit 2800. The same reference numbers can be referred to Figure 1. 2, 18 and 28 similar parts. But for the sake of convenience, Some parts of the predistortion circuits 200 and 18⑻, For example, the gain adjustment circuits 302 and 256,  88 200541280 Thermal change estimation circuit 1700, And generate a / D 304 clock signal, Rate multiplier 204, Controller 286, Adaptation engine 1300, Circuit of correlation engine 28, And some delay stages are omitted in Figure 28. People in the same industry know that these parts may still add predistortion circuits and use them extensively as discussed in Figure 2-26.  [0250] The rate increasing complex forward information flow 206 drives a non-linear processing section 2802. Common mode time positioning part 700, The positive input terminal 22 of a summing circuit. The non-linear processing part 2802 includes the part discussed above, It is useful for nonlinear distortion caused by HPA 136 and ^^ 々. These parts include the basic function generating part 16⑻, Equalizers 226 and 226, (See Figure 丨 Magic, etc. As discussed above, A complex signal output from the processing section 2802 is coupled to the negative input terminal 220 of the combining circuit. Merging circuit 220 provides forward information flow 238 to forward equalizer 2 you, And to the differential mode time positioning part 800. The forward equalizer 246 generates-processes the forward information stream 118.  Although not shown in Figure 28, An alternative embodiment could be a merge circuit 220 downstream of the forward equalizer 248, Instead of upstream as shown in Figure 28. In this modified embodiment, The forward equalizer 248 can then operate at a lower clock.  [0251] As the embodiment discussed above, The forward equalizer 246 is connected in series to the analog transmitter part 120. Time positioning, Or time delay, Some 800s provide forward information streams 118 to digital / analog converters (D / A ’s) 122, And this transforms the forward information stream into a forward analog signal, The remaining analog transmitter parts 120 are merged. As discussed above, D / A122 preferably presents a resolution much higher than a / d304. The box labeled "xpF" in @ 28 includes the low-pass vessel 124, Up-conversion part Π6 ^ From the band-pass wave · 132 of Figure 1. Since the forward information stream continues to be processed through the analog transmitter part 120, A radio frequency analog signal 134 89 200541280 is provided from one of the band-pass filters 132 to the multiplexer 250 and to the HPA 136. The RF analog signal 117 is derived from the HPA 136,  Connected to the multiplexer 250. As discussed in Figure 18 above, One of the D / A 122 also directly generates the fundamental frequency signal 123, And connected to the multiplexer 250. But it is also possible to use a dedicated D / A (not shown), To generate a fundamental frequency signal 123. The output of the multiplexer 250 is connected to one of the inputs of the A / D 304. The D / A of the driving fundamental frequency signal 123 should preferably have high resolution and high quality. because, As discussed more closely below,  D / A is the compensator used to establish A / D 304, And this compensation is limited by any distortion caused by D / A.  [0252] One difference between this third embodiment and the embodiments discussed above is that The time positioning section 700 directly provides the ideal forward information flow 272 for positioning, The phase converter 1000 is placed in the return information stream 262. But included in the a / d compensation section 1805 includes an inverter 2804, There are 272-driven inputs for forward information flow, And generate forward information stream 272, . The inverter 2804 inverts a phase rotation from the phase rotation provided by the inverter 1000.  [0253] In a variant embodiment, The common mode positioning part 700 can be divided into an integer part 714 (Figure 7) and two (unlabeled) fraction parts 716. One of the score parts 716 can operate the forward direction information flow, And feed the-input of the subtraction circuit 274, The other fractional part 716 then operates to return the information stream. The tie-in sends another input of the subtraction circuit 274. The two fractional parts 716 are preferably controlled to produce fractional delays that are equal but opposite to the midpoint of the clock. Then any linear distortions produced by these two / knife portions 716 can be equal to each other, And remove any effects that this distortion may have on the equalizer contact adjustment.  [0254] The μ-direction information stream 272 'drives the digital upscaling part i8Q6 (pulse ^) and _ fixed delay το delete', and this element plus _ is actually equivalent to the delay of the recorded upscaling part ⑽G Fixed 200541280 time delay. Duc section 1806 will forward information flow 272, Digitally upscales to an Fs / 4 intermediate frequency (IF), Where Fs is the sampling frequency. The output of the DUC part 1806 and the delay element 1818 is coupled to the switching part 1820, And this part is included in the A / D compensation part 1805. The exchange part 1820 has an output which can be coupled to the linear equalizer 1824 through the real number conversion part. Marked as EQ in Figure 28; vd ’where the superscript" 1 "represents a linear operand, While the subscript "a / d, , It means that an equalizer 1824 is provided to compensate for the distortion caused by the A / D. The output of the exchange section 1820 is also converted by the real number. The section 1808 drives the quantization error compensator 3100, This is also included in the a / d compensation part 1805. Generally speaking, The quantization error compensator 3100 compensates the magnitude and asymmetry of the quantization error.  The quantization error compensation 3100 is discussed in detail in FIG. 31 below.  [0255] Non-linear processing part 2802 The output of the equalizer 1824 and the quantization error compensator 3100, Add together in the merge circuit 1828. Output the previous version to the information stream from the merging circuit 1828, Provided to the negative input terminal of the merge circuit 1830. Merging circuit 1830 provides compensation points,  Let the processed stream return the original digital information stream to the information stream and output from A / D 304, Merge, An A / D compensation return information stream 1832 is formed. but, Fig. 28 shows that the resolution of the returned original digital data stream 304 'has been adjusted in the resolution adjuster 2806 before this merging if necessary. The resolution can be adjusted by performing the% bit shift notation discussed in Figures 21-22 above and / or by increasing the resolution to approximately the same resolution as the forward information stream. In some embodiments, No actual action needs to occur in the resolution adjuster 2806.  [0256] The A / D compensation return information stream 1832 delivers a direct digital down-conversion part (DDC) 1834. In this third embodiment, DDC 1834 includes only parts 308 of DDC 300 discussed from the first embodiment above, 310 and 312. Generally speaking, A / D 304 effectively down-converts the sampled 200541280 feedback signal into a real signal at Fs / 4 intermediate frequency (IF). Where Fs is the sampling frequency. DDC 1834 generates a return flow 254, And this is a complex signal that actually lies at the fundamental frequency. The return information stream 254 drives the return equalizer 260, The equalizer then drives the phase inverter 1000. An output of one of the phase inverters 1000 generates a return flow information stream 262, This information stream is sent to the subtraction circuit 274,  The first input of a phase estimator 2808, The first input of a difference delay measurer 2810.  [0257] The output of the subtraction circuit 274 produces an error information stream 276, This information stream is transmitted to a spectrum management part 2900 and a spectrum management switch 2814. The ideal forward information stream 272 is sent to the phase estimate 2808, Differential delay estimator 2810, Spectrum management part 2900 and switch 2814. An output from one of the phase estimators 2808 is coupled to a control input of an inverter 2804. and, An output from the differential delay estimator 2810 is coupled to a control input of the differential mode time positioning section 800.  [0258] As in the embodiments discussed above, Equalizer 246, 260, 1824, And other equalizers that may be included in the predistortion circuit 2800 are preferably programmable equalizers that can become adaptive equalizers when coupled to the adaptive engine 1300, Or an adaptive equalizer that includes a coefficient adaptive circuit.  [0259] The difference delay estimator 2810 is a hardware block, The same result as that of the time positioning sub-process 600 can be achieved. Generally speaking, The delay estimator 2810 closes a feedback loop, This circuit drives a different time delay imposed by the differential-mode time positioning part 800. therefore, The delays added by Section 8 are continuously adjusted at any time. Differential delays between the in-phase and quadrature parts of the forward information flow have their needs, Because image problems are particularly sensitive to differential delays.  [0260] Similarly, The phase estimator 2808 is a hardware block. The same result as the phase positioning 92 200541280 processing 900 can be achieved. Generally speaking, The phase estimator 2808 closes a feedback loop which drives the required phase rotation, The return information stream 262 and the forward information stream 272 in the subtraction circuit 274 are positioned. The phase rotations added by the phase inverter 1000 and the phase inverter 28004 are preferably equal to each other. But the direction is opposite. and, These phase rotations are adjusted dynamically and continuously. Dynamic adjustment of phase rotation has its needs, Because the difference between the forward and return information streams is positioned in phase, the calculation of the difference delay is more accurate.  [0261] The spectrum management part 2900 is also directed at the image issues discussed above. This problem occurs when multiple frequency-division communication channels are forwarded to the stream.  [0262] FIG. 29 shows a representative spectrum management part 2900 financial block suitable for linear and non-linear predistorter 2800, The miscellaneous part of the signal receiving circuit 2902 receives the forward information stream 272 and the signal signal measuring circuit 29q4 receives the error information stream 276. Information streams 272 and 276 are complex signal information streams. However, plural signs are omitted in FIG. 29.  [0263] In the signal strength measurement circuit 2902, Forward gift 272 is sent to an multiplexer 2906, And this divides the forward bribe 272 into multiple discrete communication signals, Its discrete communication signal 2908 corresponds to the communication relay 2704. The signal strength measurement circuit also includes-detection of small and large circuits for each of the frigates. The lying size circuit 2910 identifies the communication signal strength 2710 for the channel 2704. Similarly, In the signal strength measurement circuit,  The error data U76 is sent to the inter-multiplexer 2912, And this divides the forward information flow μ into a plurality of discrete communication signals 2914, The discrete communication signal 2914 and the communication channel 2704 have a one-to-one correspondence. The impact strength measurement circuit 2904 also includes a detection circuit for measuring various miscellaneous communication signals. System size · 2916 _ Tao articles identification track reliability following. In a better embodiment than 93 200541280, the respective detection size circuits 2910 and 2916 measure the power of their individual discrete or error communication signals 2908 or 2914.  [0264] The outputs of the detection magnitude circuits 2910 and 2916 are sent to an error vector magnitude (EVM) calculator 2918. The EVM calculator 2918 calculates an EVM statistical value EVM for each communication channel 2704. Generally speaking, These EVM calculations, Is the error power obtained from the error information stream 276, Divide by the channel power 272 obtained from the forward stream. However, the object of the present invention can also be achieved by using the calculation of other relative communication signal strength 2710 and relative error signal strength φ 2712 in channel 2704. The EVM calculator 2918 passes the EVM statistics to a gain controller 3000.  [0265] The EVM calculator 2918 does not need to assume that the forward stream 272 is absolutely "ideal" without transmitting the wrong stream of information. In the embodiment of the transmitter including the peak reduction portion 11 (Figure 丨), The forward information stream 272 may contain some distortion caused by the peak reduction portion no. Distortion added to the forward stream 112 from the peak reduction portion 110 should be the responsibility of the EVM calculator 2918. Therefore, The peak reduction control signal 114 'transmits a short-range average noise signal added to the forward stream 112. Preferably, the peak reduction control signal 114 'transmits short-range average noise to each independent modulated complex information stream output from the modulator 104 (Fig. 1). For example, the amount of process energy after low-pass filtering. In this embodiment, EVM calculator 2918 for each channel 2704, It is therefore possible to calculate the response of the EVM to the total RMS value of the peak reduction noise obtained from the control signal 114 'and the error noise obtained from the detection magnitude circuit 2916.  [0266] In general, Gain controller 3000 has a scaling factor of 292.0, 2922 and 2924,  Used to scale the relative impact of channel 2704, And instructs the forward equalizer 246 to adjust the coefficient to reduce the correlation between the direction and the error information flow. more specifically, The gain controller 30.0 performs an inclination to emphasize the influence of the discrete communication signal 2908, which has a weak coefficient in the forward equalizer 246,  The process of weakening the influence of the discrete communication signal 2908 with a strong coefficient of the forward equalizer 246 is adjusted. Stronger signal 2908 tends, But not necessary, Showing a lower EVM, The weaker signal 2908 tends to show higher EVM. therefore, In a preferred EVM embodiment,  This program uses the metric system or its equivalent. The signal 2908 with the higher EVM is emphasized relative to the person with the lower EVM and its coefficient is adjusted in the forward equalizer 246.  Lu [0267] Four scaling factors 2920 are provided to the first input of the multiplier 2926. The second input of the multiplier 2926 is adapted to receive four discrete communication signals 2908, The output of the multiplier 2926 provides an inverse multiplexer 2930, one of the scaled discrete communication signals 2908. The anti-multiplexer 2930 performs the inverse operation of the inter-multiplexer 2906 and forms a combined communication signal 272 ". The combined communication signal 272 "also transmits four frequency multi-function communication channels 2704 and generally corresponds to the forward information stream 272 at the information rate and resolution. However, the spectrum content of the combined communication signal 272 "has been changed to a more powerful EVM. And generally weaker 'on lower EVM channels, And in general • Stronger channels.  [0268] Four scaling factors 2922 are provided to the first input of the multiplier 2932. The second input of the multiplier 2932 is adapted to receive four discrete communication signals 2914, The output of the multiplier 2932 provides an inverse multiplexer 2936, one of the scaled discrete communication signals 2934. The anti-multiplexer 2936 performs the inverse operation of the inter-multiplexer 2912 and forms a combined communication signal 276 ". The combined communication signal 276 "also transmits four frequency multi-function communication channels 2704 and generally corresponds to the error information stream 276 at the information rate and resolution. However, the spectrum content of the combined communication signal 276 "has been changed to a strong EVM of 95 200541280, And generally weaker, On lower EVM channels, And generally stronger channels.  [0269] Four scaling factors 2924 are provided to the first input of the multiplier 2938. The second input of the multiplier 2938 is adapted to receive four discrete communication signals 2914, The output of the multiplier 2938 provides an inverse multiplexer 2942, one of the scaled discrete communication signals 2940. The anti-multiplexer 2942 performs the inverse operation of the inter-multiplexer 2912 and forms a combined communication signal 276 ".  [0270] Under normal operation, The forward equalizer 246 adapts its coefficients in response to the merged communication number 272 "and the merged direct path error signal 276" or the merged cross path error signal 276, , The relationship between The direct paths 1214 and 1216 or the cross paths 1218 and 1220 are adjusted according to whether the coefficients are adjusted. [0271] FIG. 30 shows an example of the operation of the gain controller 3000 for the spectrum management section 2900. The gain controller 3000 can be A separate controller element is implemented in hardware dedicated to providing similar functions.  [0272] The program executed by the gain controller 3000 includes a job 3002, Used to identify the channel 2704 with the largest EVM. in other words, Used to identify the worst channel. This is the channel where EγM needs most improvement. After working 3002, A query work 3004 decides whether the calculated coefficients are direct paths 1214 and 1216 to the equalizer 1200, This is also used as the forward equalizer 246,  Or give cross paths 1218 and 1220. If Job 3004 measured 1218 and 1220,  Then work 3008 selects the scaling factor now used for the cross path. If working in 2004, Russia measured 1214 and 1216, Then work 3006 selects the scaling factor now used for the direct path. Gain controller 3000 can be configured to switch back and forth 96 200541280 at any time to calculate the zoom factor between the direct and cross paths. Switching can happen on a fixed schedule or based on an error, It was detected that the EVM system has improved significantly due to the previous update of the zoom factor data. The workaround is,  The gain controller 3000 can be configured to focus on the direct path first, Equalizer coefficients that lock the direct path, Then turn to the intersection path. or, The gain controller 3000 can use other switching programs designed by people in the same industry.  [0273] After working 3006 or 3008, A query 3010 determines the scaling factor 2920 currently generated for the identified channel 2704, 2922, Or whether 2924 is at its preset maximum level. As long as these current scaling factors are not at their maximum values, Then a job 3012 increases the scaling factor by a predetermined amount, Program control then returns to work 002. When the scaling factor of other channels is not changed ’, increase the scaling factor of the worst EVM channel, This reinforces the impact of channel 2704, which has the worst EVM, And weaken the impact of the remaining channels 2704.  [0274] As 3010, it is determined that the zoom factor of the current worst channel is at or above its maximum allowable level, A program loop adjusts the zoom factor used by all channels. especially, A job 3014 identifies the first channel during the first calculation of this circuit, Or identify the next channel in subsequent calculations. And after working 3014, A job 3016 reduces the scaling factor of this identified channel by a predetermined amount. After working 3014, A query determines whether the scaling factor for this channel is now at or below the minimum level. If the smallest level is detected, Then work 3020 will minimize the scaling factor. After Wei Zuo face and when Wei Zuo Tong determines that the scaling factor of this channel is not at its minimum value, Then one asks whether the working mosquito program loop has been adjusted on the channel. If red—channel is not processed, Hall control returns to work 3014 When the previous channel has been processed, Program control then returns to work 3002.  97 200541280 [0275] As a result of scaling the gain of channel 2704 used to adjust the feedback from the forward equalizer 246, The coefficients are changed in a manner that drives the individual EVMA of the communication channel 2704 to approximately equal values. Channels 2704 with higher EVM can achieve a greater reduction in EVM than channels 2704 with lower EVM. The distortion generated by the analog transmitter part Π0 is more offset in response to channel 2704 with higher EVM than channel 2704 with lower EVM.  [0276] Notwithstanding the discussion presented above, This is an example of using a program executed in the gain controller 3000. People in the same industry can design alternative and equal programs to achieve essentially the same results. E.g, Gains for all channels may be initially set to low or minimum values. Then the gain of channels with higher EVM can be increased as needed, So that the EVM of all channels is maintained at a substantially equal level, And this rank should be as small as possible.  [0277] Referring to the previous figures 27-28, the track 2714 indicates that multiple frequencies are currently being transmitted to the information stream when the communication channel of the division of labor and one or more channels are particularly stronger than the other of the video channels, An increased worry that can occur. In order for the first loop to perform satisfactorily, It is best to output the trajectory of the result. A feedback signal can be changed in a predetermined manner. In the embodiments discussed above with reference to Figures 28-30,  EVM value calculated by communication channel 2704, The RF communication signal 117 is preferably amplified as a result of a change in the coefficient of the forward equalizer 246, There is a predictable relationship.  [0278] But the quantification that occurred at A / D 304, And especially in the A / D model 2000 quantizer 2016 (Figure 20) may undermine the calculation of EVM. Quantization is a non-linear operation that can produce intermodulation between channels 2704. Some intermodulation phenomena fall within the frequency band. and, Since the resolution of A / D3〇4 has been reduced, Quantization errors and intermodulation in the frequency band become worse. Trace 2714 shows a case where the weaker channel B uses a lower resolution A / D 304 and results in a greater signal strength than the communication signal 2710. In this case, The EVM measurement of channel B by the spectrum management part 2900 is discussed above. There is no obvious relationship between the change in response to the channel B response and the feedback signal that emphasizes the EVM value measured in channel b. therefore, The quantization error compensator is configured to compensate the quantization error magnitude and the quantization error asymmetry. By compensating for the quantization error magnitude and the asymmetry of the quantization error, Intermodulation in the frequency band can be greatly improved, And the EVM measurement performed by the spectrum management part 2900 for the weaker channel 2704 will follow the output change.  [0279] FIG. 31 shows a quantization error compensator 3100, A block diagram configured to compensate for the magnitude and asymmetry of a quantization error generated by a two-bit a / d 304. People in the same industry know that the invention does not require the use of a binary A / D 304, And the example of the quantization error compensator 3100 can be extended to A / D of any accuracy.  [0280] In general, The quantization error compensator 3100 includes a quantization simulator 3102 and a difference circuit 3104. The quantization simulator 3102 is configured to simulate the operation of the quantizer 2016 from the A / D304. especially, The quantization simulator 3102 includes a recorder 3106 executed by the A / D 304 for each switching threshold 2108 (FIG. 21). For a two-bit A / D 304, Three loggers 3106 can be included. Each recorder 3106 is configured to store a value provided by the controller 286. The recorder 3106 is preferably set to the actual exchange threshold value measured according to the A / D 304 of the work 2316 of the secondary process 2300. Or from another process that can achieve the same result.  [0281] Data output from the recorder 3106 ', Contains intermediate exchange threshold 2108, Connect to the negative input of a comparator 3110. In this binary example, Data output from other recorders 3106, Connect to the data input terminal of a multiplexer (MUX) 3112. The output of the comparator 3110 drives one of the multiplexers 3112 to select the input. One of the outputs of the multiplexer 3112 drives a negative input terminal of a comparator 3114 99 200541280. Depending on the status of the switch lion, either in baseband or radio frequency,  The forward information flow is connected to the positive input terminal of the comparator and to the positive input terminal of the difference circuit 3. The output of the comparator 3114 provides a highest bit, The other output of the comparator 3114 provides -the lowest bit 7L to -the resolution adjuster 3116. The resolution adjuster 3116 performs a resolution adjuster 2806 similar to the output at A / D 304. In some embodiments, The resolution adjuster 3116 does not need any specific action to take place. An output of the resolution adjuster 3116 produces a quantized analog information stream 3118, Represents the output from the quantization simulator 3102.  Lu [0282] The quantized analog information stream 3118 drives one of the negative inputs of the difference circuit 3104. A control recorder 3120 adapts to receive from a control input of one of the controllers 286, And one output drives one input of an AND function element 3122. A quantization error information stream 3122 is provided from the output of the difference circuit 3104, And an output of one of the elements 3122 provides a quantized error information stream 3124, Represents the output of the quantized error compensator 3100. The control recorder 3120 and the AND element 3122 provide functions with and without the quantization error compensator 3100. When not selected, The quantization error compensator 3100 has no effect on the operation of the transmitter 100.  _ [0283] When set to the actual exchange threshold 2108, The quantization simulator 3102 faithfully simulates the quantizer 2016 in the A / D model 2000. The quantization simulator 3102 converts the forward information traffic into a base frequency or radio frequency form. And as a result, the intermodulation generated by the tracking A / D 304 is generated. In addition, Quantization error due to asymmetry and / or quantization error, Reflected in Quantitative Analog Information Stream 3118. Quantify the extent to which the simulated information flow 3118 cannot match the forward information flow, Provide an estimate of the errors generated by the quantizer 2016, Whether it's from asymmetry, Quantization error magnitude or intermodulation.  [0284] Therefore, The quantization error information flow 3124 characterizes the extent to which the quantized analog information flow 3118 cannot cope with 100 200541280 forward information flow. As in the second embodiment discussed in FIG. 18 above, This A / D error, along with other A / D compensation factors, In the merge circuit 1830, the original digital information stream 304 is returned,  Minus, To compensate the A / D 304 error. Since the quantization error compensator 3100 has to operate, The intermodulation 2716 of all communication channels 2704 is reduced to a level lower than the communication signal strength 2710. Including weak channel B, This is shown as trace 2718 in FIG. 27.  [0285] FIG. 32 shows a flowchart of a transmitter distortion management process 400 performed by the transmitter 100. This third embodiment is hereinafter referred to as process 3200. Processing 3200 is different from processing 400 and 1900, As discussed above, It includes several tasks to compensate for the above-mentioned image signal and intermodulation problems.  [0286] With reference to FIGS. 28 and 32, Assume that processing 3200 starts from a power-on or reset event. Process 3200 performs a few basic initial tasks first, And may work in any order ^ one work 32〇2 invalidates the price ρα 136, So that the transmitter loo has no obvious transmission. Once working, set the appropriate threshold on the recorder, It is called to use the quantization error facet. -Turn 32〇6 to lock the forward equalizer (EQF) 246 and return equalizer (EQr) 26g And confirm that the equalizer applies 260 execution-unit conversion functions, So that they will not have any impact on the transmitter 100. Locked. Maybe ’for example, Achieved by setting the convergence factor "μ" (Figure π) to near zero, Whereas the unit conversion function may, For example, The edit coefficient is: _ • Way to reach. Similarly, -Work 3208 locks all to compensate for view 136, And equalizer (Ed 226, Equalization with the silk distortion produced by censorship, Erding into a zero conversion function, So that the daggers will not have any effect on the transmitter 1GG. Zero conversion functions are possible, Give Taking the setting coefficient as: · Qing " · Way to reach. — Work slaves — compensation for some lions, Take 101 200541280 through just to >  The fundamental form of the stream, And a job 3212 sets up a spectrum management switch,  To pass forward information stream 272 and error information stream 276, And the adaptation coefficient for the forward equalizer 246, in other words, The operation of job 3212 can prevent the spectrum management part 2900 from having any impact on the operation of the transmitter 100. A job 3214 does not use delay and phase estimators 2810 and 2808. So that the differential delay and phase adjustment are not used automatically. and, One job switch multiplexer 250, The fundamental frequency signal 123 is connected to the input terminal of a / d 304. Basic initialization is complete after work 3216.  [0287] After basic initialization, Process 3200 performs a job 3218 to use the secondary process. Adjust the common mode time positioning 70. Or use another function to achieve the same result. Task 3218 generally takes the subtraction circuit 274 to bring forward information stream 272 and return information stream 262 into each other's time position. Next, A job 3220 compensates for A / D quantization errors and asymmetry. Job 3220 preferably performs a secondary process 2300 or another function that produces the same result. Sub-process 2300 evaluates A / D 304 to measure the actual exchange threshold 2108, The actual exchange threshold value 2108 is set to the quantization simulator 3102. Work 3220 can pass a scan test signal through the forward path of the transmitter 100. After working for 3220, A job 3222 causes the transmitter 100 to process a frequency multiplexed communication signal, This makes one channel stronger than the other channel at its image frequency. in other words, The transmitter 100 can start generating a communication signal, Although this signal may not be transmitted from the transmitter 100, Because HPA 136 has not been used.  [0288] Next, A job 3224 starts a linear A / D equalizer (EQi〇) 1824, To compensate for linear A / D distortion. Job 3224 can perform secondary processing 2400 or use another function to achieve the same result. then, A job 3226 locks or greatly limits the frequency of the linear A / D equalizer 1824 102 200541280 wide, So that the coefficients of the equalizer 1824 will not continue to be significantly adapted.  [0289] After working 3226, A job 3228 switches the multiplexer 250 to the RF analog signal 134 to the input of the A / D 304. The return information stream 262 is further delayed from the forward information stream 272. therefore, A job 3230 repositions the clock between the forward information stream 272 and the return information stream 262. Job 3230 can perform sub-processing 600 or another function that may produce the same result, To restore the required time positioning. Bandpass filter (Figure U adds a significant phase inversion to the return analog signal now input to A / D 304. therefore, A working 3232 starts the phase estimator 2808, To close the feedback loop, And maintain the phase positioning between the forward information flow and the return information flow from a / D 304. and, A job 3234 starts the differential delay estimator 2810, To close the feedback loop, The clock position of the difference between the in-phase (I) and quadrature (Q) components of the complex forward information flow through the analog transmitter part 120 is maintained.  [0290] After working 3234, A work 3236 sets the A / D compensation switch 1820, In the form of an IF through the forward information stream 272. at this time, The quantization error compensator 3100 and the A / D linear equalizer 1824 begin to operate on communication signals in the form of IF. Just like the characteristics of forward information flow,  Very similar to the subsampling form of A / D 304 on RF signals. The quantization and linear distortion errors produced by a / D 304 are now compensated. here, The internal tone of 2716 is reduced, However, the video signal may still remain on the weak channel 2704 and weaken the EVM value.  [0291] A job 3238 then starts the forward equalizer (EQF) 246, Causing the forward equalizer 246 to perform prediction and convergence program processing 1100, Or any function that produces the same result,  And the coefficient of the forward equalizer 246 is adapted to a value that minimizes the correlation between the forward information stream 272 and the error information stream 276. This reduces distortion and further reduces errors in the return flow, And 103 200541280 further reduced the internal tone of 2716. Next, A job 324 () sets the spectrum management switch 2814, To combine communication signals 272, , Wash with combined error signal, , To forward equalizer 2 you,  For the use of the adaptation factor. The spectrum management part 2_ now starts to affect the operation of Lai 100. The heading equalizer 246 continues to adapt its coefficients, But now the frequency spectrum of the forward error stream is changed. This form of spectrum change can emphasize weaker channels. As shown in the track 2718 of FIG. 27,  The intensity of the error 仏 2712 may increase in stronger channels, But the error signal strength 2712 can be reduced in weaker channels. On Zhao, All channels 2704 are more in line with EγM specifications, And the balance can be reached when the EVM is substantially equal to all channels 2704. At a sufficient time to allow the coefficients to adjust to equilibrium, A job 3242 locks the forward equalizer 246, To prevent further adjustment of the coefficient.  [0292] Next, One job 3244 performs non-linear A / D compensation sub-processing 2600, Or another function that can achieve the same result, To compensate for the non-linear distortion generated by the A / D 304. Then a job 3246 switches the multiplexer 250 to the amplified RF analog signal 117, And the predistortion circuit 2800 can now begin to compensate for the distortion produced by the HPA 136. A job 3248 started ΗΡΑ # 136, So that the HPA 130 starts to generate a signal. Since the return analog signal now uses a different path ’, it will cause increased latency, It also interferes with the time positioning between the forward and return streams. After a proper warm-up, One job 3250 repeats job 3230, 3232 and 3234 to reposition the clock and phase. then, A job 3252 starts the return equalizer (EQR) 260,  To compensate for the linear distortion generated by the HPA 136 in the return equalizer (EQR) 260, While a job 3254 locks the equalizer 260, To prevent any significant readjustment of its coefficient.  [0293] At this time, A very low distortion signal should be presented to the non-linear point marked with _ 142 of 104 200541280 in memory i, Due to the work previously performed in the distortion management process 3200. therefore, The predistortion circuit 2800 is now ready to compensate for the non-linear distortion generated by the HPA 136, And process 3200 performs a job 3256. Job 3256 performs QPA compensation to perform job 402, And this job is called 1500 times, Or another function that can achieve the same result. then, The forward equalizer 246 restarts at work 3258, To follow up the relative changes in the signal strength of channel 2704. After work 3258, A job 3260 is performed to repeat at any time some or all of the work previously performed in Process 3200 'to track heat and / or aging. In addition, Work 3260 may include work to manage peak reductions, In response to the calculation of the EVM value as discussed above.  [0294] To summarize the above, The invention provides an improved transmitter and distortion circuit and method. A quantization error compensator is provided to compensate quantization errors generated by analog / digital circuits (A / D). This circuit monitors the feedback signal generated by an analog transmitter part. A process provides that before using the feedback signal path to cancel the distortion produced by the analog transmitter parts, It is used to compensate the distortion caused by a feedback signal. and, Distortion from transmitter parts, Cancel in a way that reflects the relative strength of the frequency multi-function communication channel.  [0295] Although the present invention will be described in detail with reference to the preferred embodiments, However, people in the same industry immediately discovered that many modifications can be made here without departing from the spirit of the present invention or the scope of the accompanying statement. E.g, Differential time positioning part 800 or phase rotation part 1000 can be omitted. Especially when the forward equalizer 246 has many contacts. or, Part 800 can be performed in different ways, For example, by using independent phase-locked loops to generate the clock signals for the I and Q pins. The adaptation engine 1300 can be configured as an adaptation engine, And all the paths in a multiplexer operate simultaneously, Instead of the two paths described above, As an in-and-out individual information stream to determine the overall adaptability of one of the filter coefficients 105 200541280 etc., or it can replace a non-adaptive equalizer, etc. Even power and chip area will increase due to wire. Here are a series of gorge parts and components, There are Xu Xi can be merged in-. In one embodiment, An equalizer that provides linear distortion compensation for a trip can be cascaded with the return information stream, It's not like this. A gain ramp equalizer can be added to the output to allow equalizer full power, Provide linear distortion surface for added a ^ d or other distortion parts. and, Although the discussion presented above is in terms of the fitness factor of the forward equalizer ’, Mention the use of channel management, The channel management part can also be used to adjust other adaptive equalizer coefficients. These and other modifications and uses are obvious to those in the same industry, All are included in the scope of the present invention.  [Schematic description] [0296] Refer to detailed description, Patent scope and drawings, The present invention can be more completely understood.  [0297] FIG. 1 shows a block IS3 of a digital communication transmitter configured according to the teachings of the present invention,  [0298] FIG. 2 shows a block diagram of the first embodiment of φ of the linear and non-linear predistortion parts plotted in the transmitter of FIG. 1;  [0299] FIG. 3 shows a digital down-conversion part suitable for the linear and non-linear predistortion part of the transmitter of FIG. 1;  [0300] FIG. 4 shows a flowchart of a first embodiment of the transmission-distortion management performed by the transmitter of FIG. 1;  [0301] FIG. 5 shows a flowchart of the secondary processing due to the process of FIG. 4, This process is used to compensate the linear distortion generated by an upstream high power amplifier (HPA);  106 200541280 [〇3〇2 丨 FIG. 6 shows a flowchart of the sub-processing of the sub-processing shown in FIG. 5 and FIG. 14.  And the execution of the process—an example of a program for time positioning estimation and aggregation;  [] FIG. 7 shows a square Jt ghost diagram of a common mode time positioning part suitable for the widely used linear and non-linear predistortion parts of one of the transmitters of FIG. 1;  [] FIG. 8 shows a block diagram of a differential mode time positioning part suitable for the linear and non-linear predistortion parts of the transmitter depicted in FIG. 1;  [0305] FIG. 9 shows a flowchart of a sub-process due to the sub-processes of FIGS. 5 and 14. And • This process is used to execute the -common mode phasing estimation Juqin program;  [0306] ® 10 shows a block 相 suitable for the phase conversion part of the linear and non-linear predistortion part of the transmitter of FIG.  [_] 囷 11 shows a drawing in Figure 5, 囷 14 map of the 15th process of the 15th process, And this process is used to execute the first-rate estimation aggregating program;  1_8】 ® 12 Display-Record the block diagram of the equalizer of the multiple parts of the linear and heterogeneous predistortion of the transmitter.  . [⑽09] ® 13 Display—A block diagram of the adaptation engine part of the linear and heterogeneous predistortion part of the transmitter suitable for inspection. [0310] Figure 14 shows the flowchart of the sub-processing due to 囷 4. , And this process is used to compensate the linear distortion generated by HPA; [0311] @ 15 显 tf- is a flowchart of the secondary process of the processing of FIG. 4, and this process is used to compensate the nonlinear distortion generated by HPA; [0311] 0312] FIG. 16 shows a block diagram of the basic generating part of the linear and non-linear predistortion part 107 200541280 suitable for the transmitter of FIG. 丨 [0313] FIG. 17 shows a transmission suitable for use in FIG. 1 The block diagram of the linear and non-linear pre-distortion part of the receiver represents the thermal prediction part; [0314] FIG. 18 shows a second embodiment of the linear and non-linear pre-distortion part of the transmitter shown in FIG. Block diagram; [0315] FIG. 19 shows a flowchart of a second embodiment of the transmission distortion management performed by the transmitter depicted in FIG. 1; [0316] FIG. 20 shows an analog / digital converter (A / D ) Model; [0317] Figure 21 shows the characteristics of two-bit A / D quantization and quantization error. Table example; [0318] FIG. 22 shows a block diagram of the first representative quantization error compensator used in the linear and non-linear predistortion part of the transmitter of FIG. 1; [0319] FIG. Z3 shows the result from the figure The flowchart of the sub-process of the process of 19, and this process is used to control the quantization error compensator drawn in FIG. 22 to compensate for the asymmetry of the quantization error; [032〇] FIG. 24 shows the process drawn in FIG. 19 This process is used to compensate the linear distortion generated by A / D. [0321] The second embodiment of the linear and non-linear predistortion shown in FIG. 25 and FIG. 18 is used. The squares of the multiplexed part are used to generate the interface to drive the adaptive equalizer. [0322] The flow chart of the processing of 26 display and 19 processing is provided to compensate the A / D generated. Non-linear distortion; [] Figure 27 shows a circle bound to several spectrum charts, which is used to determine the characteristics of a number of communication signals that transmit multiple frequency 108 200541280 multiplex channels. [0324] B 28 shows a block diagram of the third embodiment of the linear and non-linear predistortion part of the transmitter due to the figure; [0325] 129 shows the tf-generation fabric transfer and makes the Lin diagram Block diagram of the linear and non-linear pre-distortion parts that depend on; [0326] FIG. 30 shows a flowchart of the operation of a representative gain controller used in the spectrum management part shown in FIG. 29; iG327] ® 31 The block diagram of the second representative quantization error compensator of the Wei-like miscellaneous predistortion of the bribe @ 丨; [0328] FIG. 32 shows the transmission distortion management performed by the transmitter of FIG. 1 The flow chart of the third embodiment; [Key component symbol description] 102 Digital information flow 106 Merging part 110 Peak reduction part 114 Peak reduction feedback signal 115 Direct coupling element 117 Feedback signal 120 Analog part 123 Base frequency (BB) signal 126 Direct orthogonal upscaling section 130 RF analog signal 134 RF analog signal 138 Antenna 142 Amp 104 Digital modulator 108 Complex forward information flow 112 Peak reduction Forward information flow 114 Peak reduction control signal 116 RF communication signal 118 Complex forward information 122 digital / analog converter (D / A) 124 low-pass filter (LPF'S)

128 oscillating signals 132 BPF 136 High Power Amplifier (HPA) 140 Input Band Pass Filter (BPF) 144 Input Band Pass Filter (BPF) 109 200541280

200 linear and non-linear predistortion circuit 1800 predistortion circuit 2800 predistortion circuit 202 input port 204 rate doubler 205 high-pass filter 206 rate increase complex forward information flow 208 delay section 1600 basic function generation section 1700 heat exchange evaluation section 214 Highest order basic function information flow 214 'complex basic function information flow 216 thermal difference signal 218 complex forward information flow 220 merge circuit 222 multiplexer (MUX) 224 nonlinear predistorter 226 equalizer 230 information flow 238 complex non-linearity Pre-distortion forward information stream 234 equalizer section 1200 equalizer 1300 adapts to the positioning of the engine 242 complex difference forward information stream 246 multiple equalizer 250 multiplexer 256 attenuation circuit 260 multiple equalizer 266 delay complex forward Information flow 800 time positioning section 244 linear predistorter 248 feedback section 254 complex return information stream 258 attenuation complex reverse information stream 262 equalization complex return information stream 270 multiplexer 272 positioning complex forward information stream 276 error stream 279 difference Coefficient signal 282 Complex multiplication 274 Merging circuit 278 Multiplexer 280 Correlation engine 284 Accumulator 110 200541280 286 Controller 300 Frequency reduction section 700 Time positioning section 302 attenuator 304 analog / digital converter (A / D) 306 synthesizer 310 delay element 312 delay element part 314 high-pass filter (HPF'S) 400 transmitter distortion management process 404 work 406 work 408 work 1900 process 3200 process 500 Secondary process 502 Initial work 504 work 600 times processing 700 time positioning part 800 time positioning part 506 work 508 work 602 work 604 work 606 work 608 initial work 610 work 612 work 614 inquiry work 616 work 618 work 620 work 702 minimum delay element 704 Clock multiple contact delay line 706 contact 708 multiplexer 710 interpolation 712 clock signal 714 whole part 802 clock delay line 804 fixed delay element 111 200541280

806 contact 810 interpolator 816 part 904 work 908 work 912 query work 916 work 922 work 1000 phase angle rotator 1004CORDIC unit 1006 selection inversion circuit 1010 shifter 1014 subtractor 1102 work 12021-node 1208 clock contact delay Line 1212 clock contact delay line 1216 in-phase direct path 1224 multiplier 1228 subtracter 1232 multiplier 808 multiplexer 812 clock 900 processing 906 work 910 work 914 work 920 work 924 work 1002 quadrant selection unit 1004 ' Unit 1008 latch 1012 circuit 1100 times processing 1200 equalizer 1206 clock contact delay line 1210 clock contact delay line 1214 phase angler 1222 contact 1226 output drive adder 1230 adder 1234 heat adaptor unit 112 200541280

1236 multiplexer 1240 adder 1304 clock contact delay line 1308 delay element 1312 in-phase multiplier 1316 quadrature phase multiplier 1320 inverse element 1324 adder 1328 delay element 1332 sensitivity multiplier 1336 single-cycle delay Element 1340 subtracter circuit 1344 convergence multiplier 1348 single cycle delay element 1400 processing 1404 work 1408 work 1416 work 1420 work 1500 processing 1504 work 1238 multiplexer 1302 clock contact delay line 1306 delay element 1310 contact 1314 contact 1318 adder 1322 multiplier 1326 multiplexer (MUX) 1330 subtractor circuit 1334 adder 1338 coefficient difference information stream 1342 multiplier 1346 adder 1350 multiplier 1402 work 1406 work 1414 work 1418 work 1422 work 1502 work 1506 Job 113 200541280 1508 Inquiry Job 1602 Amplitude Circuit 1606 Multiplier 1612 Multiplier 1616 Multiplier 1620 Multiplier 1624 Adder φ Π02 Circuit 1706 Adder 1710 Subtraction Circuit 1802 Real Number Conversion Section 1804 Basic Function Information Stream 1806 Digital Up Conversion (DUC) 1809 Basic function information flow 1812 Real number conversion part 1816 Delay forward information flow 1820 Exchange 1824 equalizer 1828 merge circuit 1832A / D compensation return information stream 1900 processing 1600 basic function generation section 1604 multiplier 1610 cell 1614 multiplier 1618 multiplier 1622 multiplier 1700 thermal change estimation section Π04 programmable time positioning section 1708 single-cycle delay element 1712 convergence multiplier 1803 analog transmitter part compensator 1805 compensation part part 1808 real number conversion part 1810 digital up conversion (DUC) part 1814 fixed delay part 1818 fixed delay element 1822 linear distortion compensator 1826 quantization Error compensator 1830 Merging circuit 1834 Direct digital down-conversion part 2000 Analog / digital converter model 114 200541280 2002 Input analog signal 2004 amplifier 2006 Low-pass filter (LPF) 2008 Switcher 2010 Sampling and retention circuit 2012 Adder 2016 Quantizer 208 exchange threshold 2200 quantization error compensator 2202 merge circuit 2204 recorder 2206 recorder 2208 recorder 2210 comparator 2212 comparator 2214 comparator 2216 comparator 2218 comparator 2220 comparator 2222 AND gate 2224 AND gate 2226 AND gate 2228 OR Gate 2230 Multiplexer (MUX) 2232 Information flow 2234 delay element 2300 processing 2302 work 2304 work 2308 query work 2310 work 2312 work 2314 work 2316 work 2400 work processing 2402 initial work 2500 multiplex part 2502 multiplexer 2504 multiplexer 2506 multiplexer 2508 delay element 2510 delay element 115 200541280

2512 time delay element 2516 time delay element 2520 time delay element 2600 times processing 2604 work 2702 trace 2708 video signal 2710 error signal strength 2802 non-linear processing part 2808 phase estimator 2900 spectrum management part 2904 signal strength measurement circuit 2908 discrete communication signal 2914 Signal strength measurement circuit 2918 EVM calculator 3100 quantization error compensator 2922 scale factor 2926 multiplier 2938 multiplier 2942 anti-inter-multiplexer 3004 inquiry work 2514 root mean square (RMS) estimator 2518 delay element 2522 delay element 2602 Initial work 2700 Information flow 2704 Channel 2710 Communication channel strength 2800 Pre-distortion circuit 2804 Inverter 2810 Differential delay estimator 2902 Signal strength measurement circuit 2906 Multiplexer 2912 Inter-multiplexer 2916 Detection circuit 3000 Gain controller 2920 Scaling factor 2924 Scaling factor 2932 Multiplier 2940 Discrete communication signal 3002 work 3006 work 116 200541280

3008 work 3012 work 3016 work 3020 work 3100 quantization error compensator 3104 difference circuit 3110 comparator 3114 comparator 3118 quantized analog information flow 3122 AND function element 3200 processing 3204 work 3208 work 3212 work 3216 work 3220 work 3224 work 3228 work 3232 work 3236 Job 3240 job 3010 query job 3014 job 3018 job 3022 query job 3102 quantization simulator 3106 recorder 3112 multiplexer 3116 resolution adjuster 3120 control recorder 3124 quantization error information stream 3202 job 3206 job 3210 job 3214 job 3218 job 3222 job 3226 work 3230 work 3234 work 3238 work 3242 work 117 200541280 3244 work 3246 work 3248 work 3250 work 3252 work 3256 work 3258 work 3260 work 118

Claims (1)

200541280 10. Scope of patent application: 1. A predistortion compensation circuit (200, 1800, 2800) of a transmitter, which is used to compensate the distortion of analog transmitter parts (120) of digital communication transmitter (100). The predistortion circuit includes: a source (202) of a complex forward information stream (112) for digitally transmitting data; an equalizer section (234) coupled to the source of the complex forward information stream (2) 〇2) to provide an equalized forward information stream (118) and transmit the equalized forward information stream (118) to the analog transmitter part (120); _ a digital harmonic sampling down-converter (300), the analog transmitter part (120) receives the feedback signal (117, 123, 134), and provides a complex variable return information stream (254); and a controller (286), coupled to the frequency reduction The converter (300) is coupled to the carburetor portion (234) and configured to cause the carburetor portion (234) to compensate the distortion generated by the analog transmitter part (120). 2. The predistortion compensation circuit for the transmitter according to item 丨 of the patent application scope, which also includes: a seismic input port (128), and an analog transmitter part (120) receives a seismic signal 'the seismic signal display A part of the analog transmitter uses a local oscillator frequency for upconversion; a synthesizing circuit (306) is used to synthesize a clock signal that is equal to the local oscillator frequency multiplied by 2N ± 1 divided by 4, where N A positive integer suitable for Nyquist conditions for a bandwidth to compensate for distortion; and wherein the digital sub-harmonic sampling down-converter (300) includes an analog digital converter (304) for sampling at a rate determined by the clock signal Feedback signals (117, 123, 134). 200541280 3. The predistortion compensation circuit of the transmitter according to item 1 of the scope of the patent application, wherein the forward information stream of the complex change shows a forward resolution; and the return information stream of the complex change shows a return resolution minus The aforementioned forward resolution. 4. The predistortion compensation circuit of the transmitter according to item 1 of the scope of the patent application, which further comprises a programmable delay element (700), which is coupled to the forward source of the complex information (202) and the frequency reduction. Between the converters (300), the programmable delay element (700) is used to generate a time delay complex change forward information flow (266), and the time delay complex change forward information flow (266) is temporarily associated with the complex Change back to information flow (262) positioning. 5. The predistortion compensation circuit of the transmitter according to item 5 of the scope of patent application, wherein: the predistortion circuit further comprises a correlator (280) having an input coupled to the programmable delay element (700), and The downconverter (300) also has an output coupled to the controller (286); the controller (286) and the above-mentioned correlator (280) are configured to execute an estimation and convergence program (600) to convert The time-delayed forward information flow and the return-backed information flow are brought to time positioning. 6. The predistortion compensation circuit of the transmitter according to item 4 of the scope of the patent application, wherein: the programmable delay element is the _-th delay element, which adjusts the common-mode delay between the information streams returned to Wei; and The transmitter further comprises a second delay element _), which is connected to the above-mentioned complex forward information flow and the downconverter, and the second delay element is used to adjust the delay to the differential mode. 7. The predistortion compensation circuit of the transmitter according to item 4 of the scope of the patent application, wherein the complex forward information flow is transmitted through the above predistortion circuit in response to a clock signal; and the programmable delay element ⑽) includes -Integral part ⑽), and delay at least a part of the complex forward information flow to several integer periods of the clock and include-fractional part (716), and delay 200541280 the complex forward information flow to the The fraction of the clock. 8. The predistortion compensation circuit of the transmitter according to item 1 of the scope of patent application, wherein: the predistortion circuit compensates the linear distortion of the analog transmitter part (12); the equalizer part (234) is a A quadrature balance adjustment section (244); and the controller (286) is configured so that the equalizer section (234) compensates the quadrature gain and phase imbalance generated by the analog transmitter part (120). 9. · A method for digitally compensating the linear distortion produced by the analog transmitter part (12) of the digital communication transmitter (100). The method includes: quadrature balance (244, 1100) — The complex forward information flow responds to the orthogonal balance parameter to generate a balanced complex forward information flow (118); provides the balanced complex forward information flow (118) to the analog transmitter part (12); at the transmitter (100), down-convert (300) a feedback signal received from the analog transmitter part (120) to generate a multiple return information stream (254, 258, 262); and processing (1300) The information stream is returned to generate the orthogonal balance parameter. _10. According to the method of 9 items in the scope of patent application, where: the orthogonal balance action is performed by a first equalizer part (234); and the processing action surface-frequency varies, and the analogy depends on the part The resulting quadrature gain and phase are unbalanced. 11. The method according to item 1G of the patent application, wherein the chemistor part (234) includes a first equalization configured to filter the complex forward information flow and generate the balanced complex forward information flow. And a second equalizer (260) configured to filter the return return information stream. 200541280 12. The method according to item u of the patent application scope, wherein: the analog transmitter part includes a power amplifier (136) driven by a power amplifier input signal (134) and generates a power amplifier output signal (117); the feedback The signal is a first feedback signal derived from the input signal of the power amplifier; the method further includes, after down-converting the first feedback signal (500), a second feedback signal output from the power amplifier ( 14〇〇) to generate the return flow of information; and the processing action causes the first equalizer (246) to compensate for #linear distortion in the power amplifier input signal (134) to avoid a feedback signal, The first equalizer (246) is then compensated for linear distortion at the power amplifier output signal (117). 13. The method according to item 9 of the scope of patent application, wherein the frequency-reducing action (3⑽) is performed by a digital chirp sampling frequency-reducing device. 14. The method according to item 9 of the scope of patent application, wherein the processing action controls or more estimates and convergence procedures (600, 900, 1100) to generate the orthogonal balance parameter. 15. The method according to item 9 of the scope of patent application, wherein the analog transmitter part (120) includes a band-pass filter (132) driven by an upconverter (126), and the band-pass filter (132 ) Adding a band-pass filter delay; the method further includes rotating (100, 900) the phase of the complex forward and return information streams relative to the other phase to compensate for the band-pass filter Delay. 16. The method according to item 9 of the scope of patent application, wherein: the analog transmitter part (120) includes a power amplifier (136) driven by the analog transmitter input signal (134), and generates a power amplifier output signal (117 ); 200541280 The feedback signal is-the first feedback signal derived from the power amplifier input signal (4); and the method additionally includes, after the first feedback signal is down-converted (500), the feedback signal The second feedback signal derived from the output signal of the power amplifier is down-converted (1400) to generate the multiplexed return information stream. 17. The method according to item 9 of the scope of patent application, wherein: the orthogonal balance action is performed by an equalizer (246); The filter coefficient is the orthogonal balance parameter, and the filter coefficient causes the equalizer to compensate linear distortion generated by a part of the analog transmitter part upstream of a power amplifier; the processing action (1300) includes the compensation The linear distortion produced by the portion of the analog transmitter part upstream of the power amplifier changes (1414) the filter coefficients to compensate for the linear distortion generated by the power amplifier. 18. The method according to item 9 of the scope of patent application, wherein: the providing action includes providing a digital / analog converter (122) with a first resolution; and the down-converting action (300) includes providing a An analog / digital converter (304) of the first resolution and the second resolution. ^ A method for managing distortion in a digital communication transmitter (100), wherein at least a part of the distortion is generated by an analog transmitter part (120), the method includes: obtaining a forward information stream (112) configuration Digital information is successfully transmitted; training a linear predistorter (244) to compensate for the linearity loss of the analog transmitter part (120) 200541280 true; and training-a non-linear predistorter (22 sentences to compensate for the analog transmitter part Ii) Non-linear distortion generated. # 20. The method according to item 19 of the scope of patent application, wherein: the linear predistorter (244) includes an equalizer (246), and the non-linear predistorter includes a second equalizer (226) ; The linear predistorter training action includes operating the first equalizer in an adaptive mode ⑽) • to compensate the linear distortion; and the nonlinear predistorter training action includes operating the second equalization in an adaptive mode (226) to compensate for the non-linear distortion. 2i. The method according to item 19 of the scope of the patent application, wherein the non-linear predistorter training action occurs after the linear predistorter training action 400. 22. The method according to item 19 of the scope of the patent application, wherein the training action of the linear predistorter includes determining the wave coefficients of the equalizer (246), and the equalizer filters the forward information stream. • 23 · According to the method in the scope of application for patent No. 22, which additionally includes ·· Using a digital-human skin sampling downconverter, a counter L number obtained from the analog transmitter part (117, 123, 134) Frequency reduction (300) to generate a return information stream (254, 258, 262); and processing (1300) the return information stream to generate the chirp coefficient. 24. According to the method in the scope of the patent application, which includes: Pre-distortion H training and non-linear pre-realizer training actions for this line—return information stream obtained from this analog 200541280 shooter part (120) ( 254, 258, 262); the forward information stream (112) presents a forward resolution; and the return information stream (254, 258, 262) presents a return resolution that is less than the forward resolution 25. According to The method of claim 19 in the scope of patent application, wherein: the analog transmitter part (120) includes a power amplifier (136) driven by a power amplifier input signal (134), and generates a power amplifier output signal (117); and The training operation of the linear predistorter includes frequency-reducing (300, 500) of the power amplifier input signal (134) and then frequency-reducing (7) of the power amplifier output signal (3), 1400, 150,000.