TWI330467B - A new five-bit 128-tone sigma-delta modulation d/a and a/d scheme and device for uwb-ofdm transceiver - Google Patents

Description

1330467 IX. Description of the invention: [Technical field of the invention] The present invention relates to a five-bit 1 28-SDM digital-to-analog ratio analogy and analog-to-digital number method for an orthogonal frequency division multiplexing ultra-wideband transmission machine and The device 'especially an ultra-wideband communication system with high speed, low power and large bandwidth transmission characteristics is very suitable for digital home and vehicle collision avoidance radar. Since the FCC of the United States passed legislation on February 14, 2002, setting the frequency band of the ultra-wideband communication system to 3.1 GHz to 10.6 GHz and opening it to commercial use has immediately become the focus of wireless communication operators. The UWB technology is divided into two categories: Texas [nstruments of orthogonal frequency division multiplexing ultra wideband (UWB-OFDM) and Motorola's direct sequence spread spectrum ultra wideband (DS-UWB)] and the present invention is applicable to UWB-OFDM. UWB-OFDM has the advantages of high spectral efficiency, multipath attenuation and adaptability, and has been supported by many international manufacturers and domestic players. The UWB-OFDM transceiver differs from the Narrowband NB-OFDM transceiver in that the D/A and A/D converters of the UWB-OFDM transceiver must have both high sampling rate and low power consumption. Therefore, the D/A and A/D converter bit resolution cannot be too high. In other words, the UWB-OFDM communication system must use a lower bit resolution D/A and A/D converter to transmit/receive a peak-to-average ratio OFDM signal. Existing UWB-OFDM transceivers use 528MHz sampling rate - five-bit D/A and A/D converter to transmit/receive 128 QPSK modulated subcarriers because the subcarriers of the OFDM signal use 16-QAM. In the 64-QAM high-order modulation, the peak value of the OFDM signal is too high, which affects system performance. How to reduce the quantization noise of 9 1330467 bit resolution of UWB-OFDM transceiver D/A and A/D converter, and generate the SQNR needed to transmit/receive high-order modulated OFDM signals, satisfying the BER of UWB communication system The requirement is an important issue that must be solved for UWB-OFDM transceivers. In order to avoid interference with existing communication systems, the FFC has strict regulations on the power spectral density of the transmitted ultra-wideband signals. The maximum power spectral density of the ultra-wideband signal transmitted by the FFC is 41.3dBm/MHz. For the entire operating frequency band, the maximum average power of the ultra-wideband signal is about 0.5m W, so the UWB communication system is limited to short-distance applications. UWB-OFDM is the latest solution to the IEEE 802.15.3a (TG3a) standard [Note 1]. The TG3a standard features wireless ultra-wideband transmission with low complexity, low cost, low power consumption and high data rate. To meet consumer demand for size, price and power consumption, single-chip ultra-wideband transmitters must use very few external components. For the D/A and receiver A/D converters at the transmitting end, if the UWB-OFDM signal with high PAPR needs to obtain a higher SQNR, more bit quantizers are needed, thus improving D/A and A/. D converter hardware complexity and power consumption, so reducing the PAPR of OFDM signals has become a problem that UWB-OFDM transceivers must solve. Because it is more difficult to implement large dynamic range circuits with lower operating voltages, UWB-OFDM transceivers must handle the potential low cost and low power consumption architecture with more complex digital signals to alleviate the difficulty of implementing analog circuits. . The present invention proposes a quantizer method for designing D/A and A/D converters with a W-SDM modulator architecture, which greatly improves the PAPR effect, and quantizes the dynamic range and quantizes of the D/A and A/D converters. The parameters such as signal-to-noise ratio, circuit volume and circuit operating speed are trade-offs, and flexible design is required for various application requirements. [Note 1] A. Batra et al., "Multi-band OFDM physical 1330467 layer proposal f〇r IEEE 802.15.3a ^ » Sept. 2004, available at http://www.nlultibandofdm.org. • As shown in Figure 1 (4) 'Basic D/A converter contains (1). Bit quantizer (ii). ό Bit voltage stage decoder ((1)) sample/hold (sample/hold > S/Η) Circuit (iv). Smooth low-pass filter. In order to improve the accuracy of IFFT, you can use the fixed-point value higher than ^ bit • (flXetl number) or the multiplication and addition of floating-point values for IFFT processing, ijIFFT The output OFDM symbol signal is higher than the floating point, value or fixed point value represented by the Δ bit. Before the 0FDM symbol input & the d/a converter, it must be quantized to the bit by the bit quantizer. The value is in accordance with the number of bits of the D/A conversion. The & bit voltage stage decoder converts the value represented by the input 6-bit digit signal into a corresponding analog voltage signal, which is sampled and maintained by the S/Η circuit. A period of time, this is the sampling interval of the D/A converter, DM The maximum output bandwidth of the converter output analog signal is 〇.5/; called /2^sm: interval ', the quantized value of the time sampling quantizer is used as the output value, and the time is maintained, so the ideal S/H system pulse The response can be expressed as ' 1

S/H 0 can be obtained, otherwise the system response of the ideal S/H system is the Fourier transform frequency response 1^30467 ~ (7): Wne (A), (-M) = ^ lexp (a) into) (2) t The output signal of the head S/H is not - a strict band limit white. Therefore, the S/Η output signal must pass through a low-pass smoothing filter or a compensation chopper) to suppress the excess: frequency. Width 2 with ideal compensation filter for people, rabbits

(/) , | / | < 0.5 / ^ , | / | > 〇 - 5 乂 (3) As shown in Figure 1 (b) 'Basic A / D converter contains (1)

(4) Coded sample/hold pot circuit (4). Pure element quantization J = j ratio t唬 will pass the anti-aliasing low pass filter first will exceed +0 circuit every second. Time sampling anti-aliasing low pass filter

The circuit is binary and maintained. In the body, the 71 quantizer quantizes the output value of the S/H circuit to 〇, ±Δ, , ±^-2δ — 2/>

==Voltage signal ' g Yuan II encoder will be represented as the corresponding 6-bit fixed-point value signal according to the A-round voltage. The two of the linear Midreads seen in the general price are open. .. function as an example, as shown in the figure - the thousand... t-significant of the transfer number of a vacant device 〇 = 3) the round letter of the second movement sorrow W+lM/2' (,, + ^ 唬 amplitude according to the closest The quantization level 〇, ±δ/2,, :Π—(2,1)Λ/2. When the quantized input signal is within the range of the quantized-to-noise, the quantization noise of the A/D converter output signal For even distribution in [+0.5Δ' _05 is the round-in signal amplitude exceeds the quantitative crying, sentence quantification, if only the maximum output 2M ^ : dry circumference, the Bezier finite element quantizer output corresponds to 6 The bit fixed value 2: ... bit A / D value, so that the A / D converter rounds the maximum value of the letter and the minimum bandit .5 Δ, called overload (overload) ^ noise will be beyond the Finite bit D/A and A/D conversion ω

Input range, average sentence quantization noise is changed: In fact, the larger the re-weighting and the A/D converter quantization unit solution, in order to meet the system D/A, the converter input range is = two requirements D/A Quantification with the A/D converter 5^ 生^ 有限 有限 动 辄 转换器 转换器 转换器 转换器 转换器 转换器 动态 动态 , , , , , , , , , , 动态 动态 动态 动态 动态 动态 动态 动态 动态 动态 动态 动态 动态because

Small, so the appropriate use of smaller movements: the probability of the amplitude of the amplitude is better than the ^ ^ centroid 槐 置 来 削减 削减 削减 OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM OFDM 量化 量化Uniform Quantization Generated by the Large Dynamic Range Quantizer = Ming revealed that the #音-SDM circuit replaces the d/a and the a/d converter, and according to the known W-IFfT output 〇FD] The y^f variance (σ2) and the D/A and A/D converter bit resolution 6 determine the quantization unit Δ, so that the quantized variance of the quantized OFDM signal is minimized, so the W-SDM circuit quantizes 〇 The FDM signal 'can get the highest SQNR. A/D舆D/A converters with both UWB-OFDM transceivers: Basic ϋ WB-OFDM transmitter and receiver (transfer) architecture is shown in Figure 3. The binary data passes through the nebula encoder. Converted to a complex 1330467 subcarrier signal, the binary data can be selected according to the nebula coding method, QPSK, 16-QAM or 64-QAM. Each sub-multiple subcarrier signal is converted into an OFDM signal by serial-to-column conversion, and the AMFFT is converted into an OFDM signal, twice time spread (repeating the previous OFDM signal once), and converted to analogy by a high-speed office D/A. UWB-OFDM signal. The UWB-OFDM signal passes through the Additional White Gaussian Noise 'AWGN' channel. Under the condition of time synchronization, the ideal receiver is to receive the received UWB-OFDM signal through the high-speed unit A/D. Converted to a digital OFDM signal, after two times of despreading processing, the signal-to-noise-ratio (SNR) of the output OFDM signal can be increased by 3 dB, and the despread OFDM signal is input into AA-FFT demodulation. To solve the complex subcarrier signal, the binary data is finally solved via the nebula decoder. The D/A converter of the existing OFDM transmitter has low sampling rate and high bit resolution, and can efficiently perform IFFT conversion using fixed-point numerical operations consistent with the number of D/A converter bits. The IFFT output value can be directly input. d/A converter. The D/A converter of the UWB-OFDM transmitter has a high sampling rate (528 MHz) and a low bit resolution (5-bit). The quantization noise of the 128-IFFT output OFDM symbol is too large using a five-bit fixed-point numerical operation. It seriously affects receiver performance. Both U WB-OFDM transceivers use twice the time spread to increase the signal-to-noise ratio (SNR) of the received signal of UWB-OFDM. However, the quantization noise of UWB-OFDM signal cannot be reduced over time. The AMFFT output OFDM symbol signal is converted into analog OFDM signal waveform by the d/A of the device. Currently, the OFDM transmitter has two ways to design the AMFFT. The first method is that the AMFFT input signal is the Δ bit.

14 1330467 疋 point value 'Using the addition and multiplication of the 6-bit fixed-point value~ AMFFT circuit, the output 〇FDM signal is the fixed-point value of the ς ς Μ t t t t t t t t t t t t t t t t t signal. Using a fixed-point value of the same bit length as the D/A converter to operate the a/_IFf circuit, the higher the resolution of the D/A converter bit of the OFDM transmitter and the smaller the number of processed IFFT points (yv), the fixed point The numerical operation ^ scare the SQNR will be closer to the SQNR of the floating-point AMFFT circuit, and the output value does not need to be quantized to the fixed-point value, but the fixed bit, the force method and the fixed-point multiplication process The resulting quantization noise variance increases as the number of AMFFT points W increases. Therefore, the fixed-point numerical operation circuit is simpler and more efficient than the floating-point numerical operation//-IFFT circuit, so it is suitable for low-speed, high-bit resolution d/a and A/D converters and the number of subcarriers. Fewer 〇FDM transceivers. The first way is to use the addition and multiplication of floating-point values "_1{?1: the output of the OFDM symbol is a high PAPR floating point value, 〇 FD 2, the meta signal is first cut too much by the limiter After the amplitude, the input 里 bit chemizer quantizes to a bit fixed point value, and the input 6 bit D/A is converted into an analog OFDM signal. The probability that the AMFFT output OFDM symbol signal will generate a large amplitude value will be small. This is because the OFDM signal will be large enough when the # is sufficiently large, and the larger the variance of the ''{7〇]^ symbol signal is, the larger the pApR is. The larger the dynamic range representing the amplitude of the OFDM symbol signal. For fixed-bit D/A and A/D converters, the larger the quantization input range is quantized, the larger the information will be. In order to reduce the dynamic range of the 0FDM signal, it is possible to reduce the amplitude of the OFDM waveform by using the field to obtain more High SQNR. The disadvantage of this method is that a more complex floating-point numerical operation circuit is used, and the 〇FDM symbol signal of the AMFFT circuit that has a floating-point value must be subjected to appropriate reduction and quantization processing, and then input D/A or a/d. The converter can get a higher SQNR. The invention discloses a method for constructing a time-frequency equivalent interpolation/reduction of a sound-SDM circuit, [a method of outputting a floating-point value OFDM symbol signal of a chitin circuit, and outputting an OFDM signal variance (Y) according to a known AMFFT. The bit resolution of the D/A and A/D converters (W determines the quantization unit Δ, so that the quantized variance of the quantized OFDM signal is minimized, and the highest SqNr is obtained. There are five UWB-OFDM receivers. The bit A/D converter must be able to effectively reduce the power consumption of the A/D converter with a large resistor and a comparator. The 6-bit A/D converter must use a comparator to perform the voltage division of the y resistor. S / Η sampling voltage comparison, 妒 -] comparator output can not - 1 voltage level, after the latter encoder quantifies the circuit sampled voltage to 6-bit fixed-point value, as shown in Figure 4. The number of resistors and the number of comparators will increase with the number of bits in the A/D converter. The operating current of the comparator is fixed, so the larger the number of band-stop resistors, the greater the power required to divide the voltage. Maintain: Same working current. As shown in Figure 5, there is UWB_ The FDM transceiver uses 128-IFFT/FFT to modulate/demodulate the DMFDM signal, and transmits/receives a small QPSK UWB-OFDM signal with pApR through a 528mHz sampling rate-five-bit D/A and A/D converter. 'Make SQNR large enough to approach the ideal D/A and A/D converter (converted 1[~8_〇17]〇1^ signal without deuterated noise or SQNR infinity) UWB_〇FDM Bit error rate (Bit err〇r rate ' BER) of the transceiver. If the 5-bit D/A and a/d converter are used to transmit/receive the BER of the 丨6_qAM or 64_qAM UWB-OFDM signal with a large PAPR The use of 1330467 in the uwB-OFDM transmission machine - the BER of the ideal D/A and A/D converter will be larger. [Explanation] Based on solving the above-mentioned shortcomings of the prior art, the present invention is applied to the positive six-frequency division. The method and device for the five-bit 128-sound-SDM digital-to-wide ratio and analog-to-digital conversion of the multiplexed ultra-wideband transmission machine, the main purpose of the present invention is that the OFDM signal with high=average ratio is constructed by time-domain equivalent frequency interpolation.彳^广 减 32 times 128 sounds, SDM circuit is changed to 〇, +△, , , 1 ^一··· ±16Λρ are all layers of quantized signals, then the quantized signal can Converted from five-bit d/a to analog signal' and converted from five-bit A/D to digital signal, avoiding the reduction of waveform amplitude to generate chopping noise. The quantization unit of 128-SDM circuit is known according to The AMFFT outputs the OFDM signal variance (y) and the bit resolution of the D/A and A/D converters (l〇g232 = 5), so that the analog signal output by the D/A converter's can obtain the highest SQNR. The added construction ^ time domain equivalent frequency interpolation / reduction of 32 times the 128-sound-SDM circuit, only need to add an additional 3 / VL sub-multiple addition operation - no complex multiplication is required. Another object of the present invention is to replace the D/A and A/D converter five-bit quantizer with a time-domain equivalent frequency interpolation/reduction 32-times 28-tone-SDM circuit, which is referred to as five-bit 128. -SDM D/A and A/D converter, and determine the quantization unit Δ according to the known AMFFT output OFDM signal variance and the bit resolution of the D/A converter, so that the quantized variance of the quantized OFDM signal is minimized. Therefore, the highest SQNR can be obtained by transmitting/receiving UWB-OFDM signals using a five-bit 128-SDM D/A and A/D converter. The third object of the present invention is to use a five-bit 128-SDM D/A ϋ WB-OFDM transmitter with a time domain equivalent frequency interpolation of two times 1330467 -, and a five-bit 128-SDM D/A The converter output signal achieves the effect of generating a double-time spread-spectrum UWB-OFDM signal, and the U WB-OFDM receiver uses a method of reducing the time-domain equivalent frequency by a factor of two to simultaneously reduce the effect of quantization noise. SQNR can • Increase the 5dB by over twice the time spread spectrum method. The five-bit A/D converter of the UWB-OFDM receiver must use a large number of resistors and comparators to effectively reduce the A/D converter power consumption. The five-bit 128-SDM A/D converter disclosed in the present invention adopts Parallel processing analogy SDM circuit frame | Configuration instead of five-bit A / D converter quantizer, the best choice between the circuit operating speed and volume, you can choose the appropriate over-sampling ratio according to the circuit process, by improving the circuit The operating speed reduces the circuit volume, which in turn improves the power consumption of the five-bit A/D converter of the UWB-OFDM receiver. In order to further clarify the present invention, it will be helpful to review the review by the following illustrations, illustrations, and detailed description of the invention. [Embodiment] The detailed structure of the present invention and its connection relationship will be described in conjunction with the following drawings to facilitate an understanding of the audit committee.

The main architecture of the present invention is to convert two-bit data into a complex sub-carrier signal through a nebula encoder, and the complex sub-carrier signals of each # point are serially converted into parallel and then input into a V-fast inverse Fourier transform to be converted into an orthogonal sub-segment. The frequency multiplexed signal is further activated by the five-bit digital quantizer and the five-digit digital/analog converter of the above components, and the five-bit 128-SDM D/A 18 1330467 UWB-OFDM transmitter is constructed in the time domain and the like. The effective frequency is twice, and the five-element m-SDMDM output signal reaches the intersection of the UWB_〇FDM signal, and the UWB-OFDM connection is constructed to reduce the time-domain equivalent frequency by a factor of two. Sqnr can be used to solve the spread spectrum of the orthogonal frequency division multiplexing signal rounds back to the two-bit data. After the acquisition, the solution is solved by the raw cloud decoder (A) 5-bit 128-sound-SDM circuit = SDM circuit is shown in Figure 6 (4), the moon of the falling-sound circuit =, ··· The first block diagram is shown in Figure 6 (b) . Define the quantization of the quantizer (4) β ")» — ♦), then the ~-sound circuit of the wheel illusion ° coffee) H Λ 0 cut (4) + coffee). From Figure 6 (8), K dq, n — N,, so d祆n, nv :: n, , hair coffee — (4) (n) ~ e(n) - - A^), V/7 = 0, 1,.. #} yv/ _ 1 e Correction is defined as the quantization axis of the read circuit. 1 tone - let the output signal of the circuit be ^q(n) - -^(«) + e.(/7)5 V « = 0,1,..., from the z transformation of (4) Μ -1 (5) (6) .(2) E(2)~E{z)z^' =(lzs)E(z) = Hn(z)E(z) where, state, let ζ =/.2;γ<7 / Λ/, the noise of the i'f' noise filter is called ##-SDM circuit 7 = 0,1,···, A/- I Substitute //"(z) Dispersive frequency response. '/2< 2sin(^£)e^Z- , Vq = 0,1,..., Μ (7) where L = from / yv. i 1 , λ //Λ, r

Hn(2L) (8) From (7), we know that //(0)=仏(4) //"((/V - 1) ')= 〇, ie "D = 0,v“〇,i”,.,yv The corpse juice is g = 0, I , 〇r /xr ··· ' (W - 1) A is called the "sound-sDM electric noise chopper's zero frequency. If (9), especially (1),., ^ (from ~ 〗) after the output from the -IFFT from the point 〇 FDM symbol signal,

..., Χ (^~丨) is modulated by the Y-SDM circuit, and the output signal has NRZ with positive or negative (±〇.5Λ) (N〇n return is better than zer〇Ms, so &quot The sound_8 chest circuit is a one-bit quantizer, which can quantize the ΛΖ-IFFT output value into the NRz signal of the bit. From (5), we know that D('iq) = X^ + E: (9) yq = - ] (Q\ Substituting (6) and (7) into (9) can get % (from Bu X (slave), V divination, 1 to -1, then the -IFFT output value is quantized into one by vV-SDM circuit The NRZ signal of the bit is at the zero frequency 〇, L,%, (" 20 1 △ quantization noise is zero, when Ml is Bub ~ 1330467 ο V 々 =: 〇1, but the implementation is #音_ The input signal of the SDM circuit ^ illusion and the output of the nrz "number a (w) can not be the same. <(") = especially (... and the actual situation is not, the character, this is because 仏 (slave) = Jr (from) It is obtained by the (2) conversion of 2, and it is not considered to be a zero-point noise caused by the initial value of the loop of the sound-SDM circuit. In practice, the output signal of the sound_SDM circuit is at zero. The frequency will have residual quantization noise, called zero-point quantization noise. The back-end response of the [Sound-SDM circuit must be fed back to the loop front by the loop back end of the "sound SDM circuit, but when the w-syn_SD] VI circuit is turned on or reset (reset this time) ;=〇), there is no signal at the back end of the loop <(—Λ〇, a yV+1),..., pointed (―丨) and Μ ' <-yv+ 1),...丨) can be returned The front end, so the implementation of the W a -SDM circuit must be turned on or back to the loop back when the system is turned on /\P), dq (H, ..., (a \, and d (- N), dd + 〇,. .., d(-)) sets the initial value. Because the loop signal <(〇), 〇), ... and #〇), #1) of the "sound-SDM circuit after power-on or reset, will follow Input signal x(0), x(l), ... and change 'General input signal x(〇) ' X(l) ' is an unknown random variable, so the quantization of the SDM circuit when it is turned on or reset Noise e(— Ν)= <(—(9)"), e( ^ + 1) = dq{~ N + \) - d(~ N + \) ^ ... > e(~\)= The sharp noise (6), <丨, generated by the sharp (-丨)-d(-1) and the first input SDM circuit input signal χ(〇), 41),··., — 1), ..., e(A" — I) won't Etc. Similarly, quantizing the noise <0), <]),..., -1) and the second stroke "sound-SDM circuit input signal 乂(^), + 1) ' ·.·, x( 2tV-1) The quantized noise e(^), 〆"+丨),..., e(2yv-1) will not be equal. According to this, ez(0) = e(0) - e (~ 21 丄WU407 V) Jokes, e7(\) = e(\) - e(\ ~ N) _ a... Therefore, the implementation of the 7^-SDM circuit quantization noise (10)) training will not be zero. If [cough/W, then the fixed-zero frequency in the v-sdm circuit is 〇, 1, u, (~-1) L quantization noise & (0) = 仏 (〇), five private) = five ,〇),', 〗 〖)/〇= £,·(#- 1), called zero-point quantization noise, ... (10) _)' A(1)'...' £, (yv-!) "(〇), e, (1),, converted value, where e, (〇), e, (1),, 1) are the quantized noise e(M-N) - e(MN + of the last V point of the A/sound-SDM circuit, respectively 1) > , e(A/_ 〇^7V^--SDMt path initial quantization noise e(-, <__ Y+n difference, ie "...' ten i) (Π) (12 Ei{n) = e{MN + n)-e{nN\ ^n = 〇,...yN_x Substituting (10) into (9)

Dq (^) = X{kL) + E^k^Vk = 〇,],._ι From 02) It can be seen that the actual output __ circuit output signal will have residual zero quantization noise at zero frequency, zero quantization The variance of the signal is proportional to the PAPR of the input signal. Hey. Because the OFDM signal is modulated into an NRZ signal by the w_SDM circuit and has less quantization noise at the zero frequency, 22 0 1330467 does not lose the generality, so that the input OFDM signal of the input voice-SDM circuit is none (0). None (1), ..., for M-1) is the M-IFFT output value of Z(0), fork (1), ..., Z(Ml), and the AM complex complex subcarrier signal to be transmitted is in the Arpeggio-SDM circuit. 7V zero frequency, ie X(kL) = X(k), \/k = 0, \,..., N-\ (13) So the W-SDM circuit quantizes the noise power ratio SQNR to zero frequency The quantized variance of the complex subcarrier signal power to the zero frequency, ie

SQNR E[X\k)] Νσ; E[Ef(k)) Νσ: (14) where ¥ and < are the variance of the OFDM signal variance and the zero quantization noise, respectively. It is generally assumed that the quantization noise of the W-SDM circuit is a uniformly distributed random variable and avoids the quantizer nonlinearity to simplify the difficulty of analyzing the nonlinear system. To avoid the overload quantization noise caused by the nonlinearity of the quantizer completely (overload) The quantization noise cannot be uniformly distributed.) The input signal maximum value Xmax of the W-SDM circuit is less than or equal to the output maximum value of the W-SDM circuit quantizer 0.5A, that is, xmax < 0.5A, where A is W sound. - The quantization unit of the SDM circuit. When xmax S 0.5A, the quantized noise e(/7) of the 7V tone-SDM circuit is evenly distributed at [-0.5A, + 0.5Δ]. Let the initial quantization noise of the W-SDM circuit be zero. From (11), the zero-point quantization noise of the V-sound-SDM circuit e/〇) = e(M- W + w) will also be 23 1330467 evenly distributed. In Bu Yi. 5Δ, +〇·5Δ]. Then e/(the variance of 岣 is σ 2 1°·5Δ ο 1 A2 W' Ζ Ί, υ〇.5 Δ (15) Substituting (1 5) into (1 4) can be obtained (16) SQNR = 丨 2 Bi , υ〇.5Δ From (16), if you want to avoid overload completely, and have the most + quantized noise variance, then △ = 2Xmax, substituting (16), squeezing SQNR = 3 σ

<ax PAPR (17) where PAPR=<x/~2, since the OFDM signal amplitude is not constant σ_ϊ < Xmax ', so PAPR must be greater than ·—, then )8) SQNR < 3 = 4.8 dB, for xmax = 0.5Δ From (1 8), it can be seen that when quantizing the singularity Δ > 2xmax, the quantization signal SQNR < 4.8 dB is quantized. It does not meet the requirements of the UWB-OFDM communication system. Therefore, the #音-SDM circuit uses a frequency domain interpolation f-channel and a frequency domain reduction circuit, and the #音-SDM circuit can obtain an SQNR higher than the above: 4.8 dB. The signal of the input of the AA complex subcarrier signals χ(〇), 24 1330467 ι(ι) ′ ..., x(yv — 1) to be transmitted is interpolated i - 1 zero ’

X{k\\fq II Lk 0 , \/q Φ Lk yk = 〇x...,N-\ (19) Because the interpolation action is before M-IFFT, it is called frequency interpolation. The M-IFFT output signal derived from (19) is x{m) = χ(η + IN) = -x(n)y〇<n<N-\,〇<l< L-\ ( 20) From (20), the amplitude of the OFDM signal jf(m) in the frequency domain interpolation is reduced. • It is 1 / L times the original signal x(A). Therefore, the quantized unit S = of the lower-level _SDM circuit can be reduced to 丨/z times of the original quantized unit. Then the quantization noise variance of the W-SDM circuit is reduced to the original φ 7 times. Therefore, the frequency-domain interpolation L times the "sound-8〇]^4 circuit's quantized signal-to-noise ratio SQNR upper limit will increase to the original B 2 times, that is, SQNR < ^ = ^ (2〇1〇g]〇L + 4 S) dB (21)

L L From (21), the upper limit of the SqnR can be increased by 6 dB for each doubling of the frequency domain interpolation magnification. The frequency-interpolated 1 times ofdm signal is modulated by the yy-SDM circuit to become the μ point NRZ signal field 25 8 1330467 (22) dq{m) = x{m) + e:(m), Vw = 0,1, ..., A/ _ j is known by (10) and 09) (22); the v-sound _SDM circuit is modulated from the point NRZ signal after M_FFT demodulation, after reduction, I can get at "zeros Subcarrier signal with frequency value

Dq(kL) = X(kL) + E:(kL) (23) ~ X{k) + E^k), Vk = 0,1,...jVv_j Because the reduction action is after the FFT, so weigh Reduced for frequency. The larger the frequency domain interpolation/reduction rate is; the higher the ratio is, but relatively, the one-bit NRZ signal must be at the same time; Therefore, the frequency domain interpolation/reduction of one-bit vowel-S=variable circuit can be regarded as super-multiple-like-position constructed in time-domain equivalent frequency interpolation/reduction L = 32 times〗 28: :SDM Modulation circuit ό = 5-bit digital analog converter ^ D / A) As shown in Figure 7, five-bit coffee surface = containing the five-bit quantizer is constructed in the time domain equivalent frequency 2 / Reduce 32 times 128-tone _SDM circuit, frequency interpolation and frequency reduction are performed before the FFT and 128 tear, respectively, as shown in the figure: two it is called 建 constructed in the time domain equivalent frequency broadcast and constructed in 牿Domain-specific frequency reduction. It can be seen from (20) that the W =] signal is subjected to frequency domain interpolation 丨 28_iFFT# / the complex subcarrier signal is rotated: = expanded '128 bits of the 128-point complex subcarrier signal】fft_ (S) can be output 32 times Equivalent to frequency interpolation such as. (4) and ^ generation 26 (24) 1330467 into (22) can get dg(n + lxm) = ^- + e(n + lxm)-e(n + (l-\)x\2S), V0<« <127,0</<31 Add the formula (( + />028) / = 0,1,...,31 to get • d,q («) = 2] ^, (« + / X 128) /-0 ]31 31 = ^Σχ^) + Σ^ + /χ128)-Κ« + (^-1)χ128) (25) /=0 /=0 =x〇7) + e( Rt + 31xl28) —e〇-128) " =x(«) + e,. («), VO < « < 127 because J〆") adds 32 NRZ signals 'so j '〆&quot ;) is a quantized signal of 〇, ±A,..., ±16Δ hierarchy. Therefore, the 128-sound-SDM modulation circuit constructed by interpolating/reducing the time-domain equivalent frequency by 32 times is a five-bit

Lu quantizer. Five-digit 128-SDM D/A and five-bit 128-SDM A/D $ are constructed using %• domain-specific frequency interpolation/reduction B=25 = 3 2 design's can be completely equivalent With frequency interpolation/reduction processing, the length of the fast inverse Fourier transform (IFFT) and the fast Fourier transform (FFT) is not increased as the frequency interpolation magnification increases, and the effect is eliminated. 1 28-SDM circuit quantifies the purpose of noise, reducing the effect of logic gate and processing time. When the amplitude of the 128_IFFT output signal exceeds the 27 1330467 half-quantity unit (0.5 Δ) of the 128-SDM circuit, the 128-SDM circuit will generate overload quantization noise, and the portion defining the 128-IFFT output signal amplitude exceeding 0.5 Δ is the cut. Noise (26) 0, for|x(«)| < 0.5Δ < 0.5Δ - x(n), for x(n) > 0.5Δ -0.5Δ - x(n), for x(« < -0.5Δ When the input signal of the 128-SDM circuit is not more than 0.5A, the quantization noise is evenly distributed. When the input signal of the 128-SDM circuit is not less than 0.5 △, the quantization noise is the clipping noise. The quantization noise of the previous 12th 8th. That is, e(n)= 〜6^(-0.5厶,0.5/\), V|x(8)| 2 0.5Δ eo/, (")= (17) + _ 128), V|x(«)| 2 0.5 Δ (27) where ~" "(- 〇·5△' 0.5Δ) ' represents e"yv〇) evenly distributed in [-0.5△' 0.5A]. When the input signal of the 128-SDM circuit is in the frequency domain Interpolating 32 times of OFDM signal' then 128-sigma-SDM circuit zero-point quantization noise is e,.(8)= < Bu(8)Ν〇.5Δ 1(8) μ〇·5Δ (28) It can be seen from (28) that frequency interpolation can be used to reduce 128-SDM 28 8 1330467 roads are all denominated to quantify noise. The smaller the △ is, the smaller the conditional probability of quantifying noise 30 is. The quantization is complicated: Erli's noise, so one...?=== Know the probability of the 128-sDM-SDM circuit to enter the signal: Difference and only: The most quantized unit a makes the zero-quantization noise i take small '! The dynasty ttSQNR κ is the most: To match the 28-side circuit The quantized unit of the input signal. Because: The 〇 draw signal tends to be normal when it is large enough ^ is = complex complex subcarrier secret IFFT loses two = number. Gaussian random variation of zero mean and variance is ~2; Is interpolated by the input signal x at the frequency of ίίΓ, times the condition 12 12 6L2

2^)Q 2σν (29) where

Qiu)

Exp (30) Day ^(10) knows that the zero-point quantization noise side M is the minimum, the early position is △, and the A is the quantization unit constructed in the time domain equivalent frequency interpolation/reduction 32 times the 128-tone - SDM circuit (5-bit quantitative crying, 128-IFFT output signal, you can get the maximum size. - Li (B) five 疋 128-SDIVI D / A conversion cry 29 9 1330467 as shown in Figure 5, both UWB-OFDM transceiver uses 128-IFFT/128-FFT modulation/demodulation of fixed-point operation to generate OFDM symbol signals, and transmits/receives UWB-OFDM signals with 5-bit D/A and A/D converters. The conversion will generate excessive quantization noise. The present invention discloses that the SQNR of the 5-bit 128-SDM D/A output UWB-OFDM signal does not decrease as the AMFFT length W becomes larger, and the highest SQNR can be obtained. Time domain equivalent frequency interpolation/reduction £ = 32 times W = 128 tone - SDM circuit's five-bit digital analog converter (do = i〇g2L = 5-bit 128-SDM D/A converter) as shown in Figure 7. As shown, the five-bit 128-SDM D/A converter includes a five-bit quantizer that is digitally constructed in a time-domain equivalent frequency interpolation/reduction 32-times 128-sound-SDM circuit. The present invention discloses UWB-OFDM. The five-bit 128-sound-SDM D/A converter of the shooter quantifies the high-precision floating-point value of the 128-IFFT output using a 128-sound-SDM circuit constructed with a time-domain equivalent frequency interpolation/reduction of 32 times. The quantized signal of 0,士A,...'± 16 A level can be expressed as a five-bit fixed-point value' where λ is the optimal quantization unit, and the quintile fixed-point value is outputted by the equivalent of the 5-bit voltage-level decoder output. Voltage, sampled by S/Η circuit, input smoothed low-pass filter output UWB-OFDM signal. 128-IFFT output floating point value is constructed by using time domain equivalent frequency interpolation/reduction 32 times 128-tone The SDM circuit quantizes the OFDM signal of the five-bit fixed-point value of the maximum SQNR, and only needs to add an additional = 3x128x32 = 12288 complex addition operations, without complex multiplication at all. As shown in Figure 4, there are 128 UWB-OFDM transmitters. -IFFT outputs 128-point OFDM signals χ(0), χ(1),...,x(i27), after twice the time spread spectrum circuit output spread-spectrum double OFDM signal x(〇), 30 (8) 丄: (1) _··· X(127) ' x(〇) ' χ(1),...,x(127), twice rounded by ^bit D/A The time spread frequency UWB-OFDM signal is constructed in a five-bit 128-sound-SDM circuit 1 of the double-time spread spectrum circuit. It is called MS 5.bit 128_than. The double-time spread spectrum 〇FDM^ is said to be two consecutive strokes. The same UWB-OFDM signal =, , = into, after the A WGN channel is synchronously received, after five bits A/D, it is twice the time of the five fixed-point value spread spectrum OFDM signal 〇 (〇) Heart (〇) + X〇) + e,〇) + «ι ' ...' x(127) + e,.(〇) + 1.7 χ(〇) + e,(0) + ^128,x(l) + e/(l) + «129> , x(i27) W127) + "255 ' where,,e,·(1),...,e,.(127) are X(〇)' , (1 ) '..' towel 27) The quantization noise, "...丨,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, The fourth (fourth) despreading signal output of the double-spreading spread spectrum circuit is the first point value of the double-time spread-spectrum OFDM signal and the first +12, so the despreading signal can be (4) the value of + y(m) = b^lig^l±5J+b(w + 128) + g,.(m + 128) + „,,,··^ (31) 2 = 4/«) + ^m) + ^±S±i28 5 Vm = 〇5 ?127 Let σ„2 be the received AWGN noise variance, and (28) know that the variance of the received A WGN noise after the despread output is

E 5il± em + 128 2 ^k2,1+4

Vjml -ση+ση 4 ~~2 (32) 1330467 From (32), it can be seen that the UWB-OFDM signal exhibits the spread spectrum twice, and the received AWGN noise variance is reduced to one of the original ones. Therefore, the signal-to-noise ratio of the five-bit D/A and A/D transmission/reception twice-time spread-spread OFDM signals can be expressed as, SNR. • TSP — 0.5σΙ (33) where σ? and < respectively The 〇FDM symbol 〇FDM symbol signal for transmission is quantized by the five-dimensional D/A and A/D. Constructed in the time domain equivalent frequency interpolation twice the five-mile dry coffee ^^ is called the coffee side circuit 2: the rate domain equivalent frequency interpolation circuit is constructed in the time domain equivalent, the rate reduction circuit replaces the map Four uw: coffee and spreader. The root can know the domain, etc. "(1)’...,χ(127> is constructed at 4 times the effect rate interpolation twice the circuit output 〇 5 〇. rush 27) and. ·5χ(.),〇5χ(1),入)邱27/乂'..., = 0 fixed-point value of the 0-character signal == —.as; rate reduction ^ by (four) structure (four) domain equivalent frequency 1330467 31 _ 31 31 <1(«) = ΣΑ(Λ + 128/) =丄艺x(«) + £e〇+ 128/)-e〇7 + 128(/ -1)) /=〇64/= 0 /=〇=0.5x(«) + e{n + 31 x 128) - e(n -128) ξ 〇.5x(n) + et (n), V« = 0,1,... , 127 (34) The second time spread symbol will be equal to the last 32 symbols. The output symbol dq nd («) = Σ ί (« +128/) = ~ constructed by reducing the time domain equivalent frequency by 32 times. Σ χ(«) + ^ e(« +128/) -e{n + \ 28(/ -1)) /=32 〇4/=32 /=32 =0.5x(«) + e(«+ 63χ 128)-e(«-31x128) ξ 0.5χ(«) + e;. {η +128), V« = 0,1,... ,127 (35) where today (") is 丨28 sound - The NRZ signal output by the SDM circuit (the value can only be ±〇_5Δ)' From (34) and (35), the first time spread symbol of the 2-TFi 5-bit dSDM circuit output 〇·5χ(〇) + e/(0),〇.5x(l) + e,(l),...,〇,5χ(127) + β/(〇) and the second time spread symbol 〇.5x(0) + palpitations 28), 〇5λ: (1) + heart (129),, 0.k(127) + e/(255) are all 〇 5五(〇),〇5χ(1),,〇5叩7) The five-dimensional quantized value (the value can only be 〇,±, ;fFl Μ1 128_SDM rounds out the first time spread symbol The yuan is not exactly the same (quantization noise is not the same as the factory time two " the first time of the shed output frequency and the second two times = the time domain equivalent frequency is reduced by twice the output, X (X (127) ) After construction in the time domain, etc. (§) 33 effect frequency interpolation / subdivision (34) and (35) can be used to 128-tone-SDM circuit output, (25) (36) = ¥ (four), + _ = And)) () + e(n + 63x128)-e(„_128), V^ = 〇5l }127 mouth ^ I/» is the nrz signal output by the 128-sDM-SDM circuit (value; connect: d' by ( 36) It can be seen that the five-bit 128-SDM A/D conversion: 槿 护 护 鼙 鼙 间 展 展 展 与 与 与 与 与 与 与 与 与 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄 狄The output de-emphasis 128_IFFT output floating-point* value of the OFDM symbol signal set value (value can only be 〇, ± A, ± 2 △, ..., ± 32 Δ), the media can be improved in the resolution - bit, Therefore; the Japanese Shouzhi equivalent frequency interpolation / reduction of twice the five-digit 丨 28 sound SDM D/A and. The A/D converter can be viewed as a six-bit 128-tone § swollen d/a and a quantized resolution of the A/D converter. Therefore, the present invention discloses that the UWB-OFDM transceiver uses the five-bit (3) tone _sdm d/a and the a/d converter to construct a circuit with a daily domain equivalent frequency interpolation/reduction twice, in addition to In place of the double-time spread/de-spread function, the SQNR of the 5-bit 丨28-SDM D/A and A/D converter transmitting/receiving UWB_〇FDM signals can also be increased with the bit 兀. The invention discloses a 6-bit w-sound D/A and A/D conversion constructed in time domain equivalent frequency interpolation/reduction B times

Benefits can also be applied to UWB-OFDM communication system with 1 time spread spectrum and # subcarriers, and obtain quantitative analysis of +1〇g/bit#音_SDM 1330467 D/A and A/D converter degree. (C) 5-bit 128-SDMA/D converter

With a five-bit A/D converter, the number of resistors and comparators increases as the bit resolution of the A/D state is increased by two, so the A/D converter circuit volume cannot be effectively reduced. With reduced power consumption. Constructed in the time domain equivalent frequency interpolation / reduction z = 32 times 128 sound _8 〇 ^ ^ circuit five-digit analog digital converter 0 = l 〇 g2 /; = 5-bit 128-SDM A / D converter As shown in Figure 9, the five-bit 128_SDM A/D contains a five-digit array of analog-time equivalents of the time domain equivalent frequency interpolation/reduction of the 32-times '_SDM circuit. As shown in Figure 9, the five-bit 28_SDM A/D Cheng = state can be divided into two parts of the oversampling 32 = 25 circuit (constructed in the time frequency interpolation circuit). 2. The analog SDM circuit includes a subtractor, and: the integrator one-bit A/D converter and the ~bit D/A conversion are reduced to $10. 3. Digital accumulator (constructed in the time domain equivalent frequency pass). The UWB gutoscope received by # is passed through the anti-aliasing low-second input S/Η circuit, and the circuit will keep the same output value G. The f# terminal's 黯 circuit completes one bit NRZ = two every ^32 seconds. After 11 and tired Jiajie 'every ^ seconds of force σ device can output a point to do a 〗 - five-bit construction of the ΐ bit 128 side A / D converter contains to compensate / reduce 128 sound - The coffee quantization circuit, dt? analog sdm circuit replaces a large number of resistors and compares the second while simplifying the circuit's volume and power consumption. Yes, some UWB-0FDM receivers use 52 35 1330467 bit A/D. If the 128-SDM A/D converter uses only a single SDM circuit unit, the 5-bit A/D converter requires an oversampling ratio L = 25 =32, making the 128-SDM A/D working clock up to 32x528MHz = 16.896GHz. In order to reduce the operating clock speed of the 128-SDM circuit, the present invention discloses that the five-bit 128-SDM A/D converter can be processed in parallel using two SDM circuit units, and the oversampling ratio can be reduced to 25 Λ'. 128-SDM A/D converter circuit with low working clock. When V = 1, it means that the five-bit 128-SDM A/D converter uses two parallel processing SDM circuits. As shown in Figure 11, the S/Η circuit sample value X will remain G = 1/528ΜΗζ = 1.894. The output sampling rate of the nanosecond 'S/Η circuit is upper /2 = 16 times of the input sampling rate'. Therefore, each sampling value X of the S/Η circuit is oversampled 16 times and is repeatedly output 16 times. The oversampled output signal X is divided into two paths, one of which is the limiter output signal xw of the X through the five-bit 128-SDM A/D converter dynamic range half amplitude (0.5AL/2 = 8Δ), when the input signal amplitude Exceeding ±8Δ, the limiter will output ±8Α; when the input signal amplitude is between ±8Α, the limiter will output equal to the input, so 8Δ, for 8Δ < λ: (37) - 8Δ, for x < -8Δ x, for |x| < 8Δ The difference between the output signals of the limiter of x and 8 △ amplitude is ~, then (S) 36 (38) 1330467 χ ~ 8Δ, for 8Δ < χ ^ =^ -^, = jx + 8A, forx <-8A 〇, for \χ\ < 8Δ - ~ and ~ respectively input parallel two sets of SDM circuits with a 16-fold circuit output constructed at equivalent frequency reduction ~ and ~ four bits Quantity: 佶Q (〇 and QQJ. From (37) and (38), if x does not exceed 6 ^ a 128-SDM A/D dynamic range (± help = ±16Δ), = • the state range is half of the original * (±0.5 help = ±8Δ). So the quantity two = / only need 丨 og2 (L / 2) = & - 丨 = 4 bit resolution to obtain the same SQNR with the five-bit resolution, so construct Time domain equivalent frequency interpolation magnification (oversampling ratio) Reduce to half of the original (24 = 1 6 /, then the operating clock rate of two parallel processed SDM circuits should be '. Half as 528MHZ = 8. Phase GHz, every 1/(16x528MHz) = 0.1184 nanoseconds The output signal is input to the post-stage accumulator, and the accumulator is completed every 1.894 nanoseconds. • The under-constructed time-domain equivalent frequency is reduced by 16 times. The output four-bit quantity ^ value 0 (especially ") and 〇 (々)'疋 0 (") - [" / 〇. 5 八] is the quantized value, where [v] is the nearest integer to V and closer to zero than V. From (37) and (38) can be obtained ^ (39) 1330467

QM + QM Q(SA) + Q{x - 8Δ), for 8Δ < x -^(8Δ) + + 8Δ), for x < —8Δ Q(x) + 0, for |jc| < 8Δ Q(x), for 8Δ < x 16 + Q(x) -16, for 8Δ < 16 + Q(x) +16, for x < -8Δ = ^ Q(x)·, for x < -8Δ = Q(x) Q(x), for x| < 8Δ Q(x), for \x\ < 8Δ From (39), we know that Q(x„) and Q(xJ will equal S/ The five-bit quantized value Q(x) of the H-circuit sampling signal 。. Originally, two accumulators are required to complete the two-stage SDM circuit output signal construction in the time-domain equivalent frequency reduction, and then the quantized values of the two accumulators are output. Adding. If the two sets of S DM circuit output signals are added first, only one accumulator is needed to complete the time domain equivalent frequency reduction, as shown in Figure 11. When V = 2, it means The five-bit 128-SDM A/D converter uses four parallel-processed SDM circuits. As shown in Figure 12, the S/Η circuit sample value jc will remain at 1.894 nanoseconds, and the S/Η circuit output sample rate is input sampled. The rate of L / 4 = 8 times, so each sample value X of the S / Η circuit will be repeated eight times after oversampling eight times. The oversampled output signal X is divided Two ways, one of which is: c passes through the limiter output signal of the three-quarter amplitude (0.5ALx3/4=12Δ) of the dynamic range of the five-bit/V-SDM A/D converter, so 12Δ, for 12Δ < x χί( = < -12Δ,forx < -12Δ (40) x, for |x| S 12Δ The second difference is the difference between the output signal of the limiter and the amplitude of 1 2 A, then 38 1330467 (41) WX -JC": Χ-12Δ, forl2A<x χ + 12Δ, 〇Γχ〇Γχ<-12Δ 0, for |χ|<12Δ From (40), if x does not exceed five-bit 128-SDM A/D converter The dynamic range (± from /2 = ±1 6Δ), then the dynamic range of X, = is one quarter of the original (±〇.5ΔΖ74 = ±4Δ), so quantifying X丨 requires only l〇g2(^/4) ) = 6 - 2 = 3 bit resolution to obtain the same SQNR as the five-bit resolution. The 12A amplitude limiter output signal; ^ continues into two paths, one of which is a five-bit 128-SDM A/D converter dynamic range half amplitude (〇.5AIx 1/2 = 8Δ) limiter output signal, the relationship between (40) and X is 8Δ, for 8Δ < xu x'u - < - 8Δ, for xu < -8Δ 8Δ, for 8Δ < x =< -8Δ, for x < -8Δ (42) xu, for |x;/1 < 8Δ x, for |x| < 8Δ • The second is: the difference X between the output signal of the limiter and the amplitude of the 8 △ amplitude, which is obtained by subtracting (40) from (42) The relationship between X and X is 4Δ, for 12Δ < X -4Δ, for X < -12Δ χΊ — x'd = xu ~xu =' χ ~~ 8Δ, for 8Δ < x < 12Δ (43 ) x + 8Δ, for - 12Δ < x < -8Δ 0, for |x| < 8Δ From (43), the dynamic range of λ:2 = is the original quarter 39 1330467 (±0·5ΔΙ/ 4 = ±4Δ), so quantizing jc2 with only three-bit resolution gives the same SQNR as the five-bit resolution. The 8 Δ amplitude limiter output signal; cV continues to be divided into two paths, one of which is a limiter of xV through five bits] 28-SDM A/D dynamic range quarter amplitude (0.5 ΔΙχ 1/4 = 4 Δ) Output signal X4, from (42), the relationship between x4 and X is 4Δ, for 4Δ < x 4Δ, for 4Δ < x X4 4Δ, for x'u < -4Δ = < — 4Δ, for x &lt ; -4Δ (44) <,f〇rx, for |x| 乞4Δ The second is the difference between X and the limiter output signal x4 of the 4A amplitude x3, which is obtained by subtracting (42) from (44) The relationship between x3 and X is x3 = Xu ~ X4 - < 4Δ, for 8Δ < x -4Δ, for x < -8Δ x - 4Δ, for 4Δ < x < 8Δ x + 4Δ, for - 8Δ <; jc < -4Δ (45) 0, for |x| < 4Δ

It can be seen from (45) and (44) that the dynamic range of x3 and x4 is one quarter of the original (±0.5AL/4 .== ±4A), so quantization x3 and x4 only need three-dimensional resolution. Obtain the same SQNR as the five-bit resolution. Therefore, the time-domain equivalent frequency interpolation magnification (super-sampling magnification) is reduced by a factor of four (23 = 8), and the working clock rate of the four parallel-processed SDM circuits is only one quarter of the original one. 8x528MHz = 4.224GHz, every 仏 = 1/(8χ528ΜΗζ) = 0.237 nanoseconds to complete the output one-bit NRZ signal input after the stage accumulator, the accumulator is completed every 1330467 1.894 nanoseconds to construct the time domain equivalent frequency reduction 8 Times, the three-dimensional quantized values QO丨), Q(x2), Q(x3), and q(X4) are rotated. From (38) (43), (45) and (44), Q{X\ ) + Q(X2 ) Qix3 ) Q(X4 ) Q(x-\ 2Δ) + ρ(4Δ) + ^(4Δ) + ρ(4Δ), for 12Δ < χ Q(x +12Δ) + Q(-4A) + 0(-4Δ) + Q(-4A), for xs ^ 12Δ ^(0) + Q(x - 8Δ + Q(4A) + Q(4A), for 8Δ < jc < 12Δ = < 0(9) + - 8Δ) + 0(4Δ) + 0(4Δ), for -12Δ S xs _8Δ = g(x) (4 7 ) Q(〇) + Q(^)+ Q(x ~ 4Δ) + Q(4A), for 4Δ < x < 8Δ Q(〇) + 0(〇) + Q(x + 4Δ) + ^(4Δ), for - 8Δ < x < ^4Δ Q(〇) + Q(〇) + 0(0) + Q(x\ for H ^ 4Δ Originally required four accumulators to complete four sets of sdm The circuit rotation signal is constructed in the time domain equivalent frequency reduction' and then the quantized values of the four accumulated p outputs are added. If the output signals of the four groups of SDM circuits are first = added, only one accumulator is needed to complete Constructed in the time domain equivalent frequency reduction, as shown in Figure 12. The five-bit 128-SDM A/D converter can be extended to use SDM circuits with thirty-two parallel processing and no longer need to be oversampled. Maintain a sampling rate of 528 MHz, as shown in Figure 13. Table 1 shows the sampling at 528 MHz. - The five-bit A/D converter uses two working clock frequencies in parallel with four sets of SDM circuit units. As shown in Table 1, '128-SDM A/D converter uses two groups and four sets of SDm circuit units in parallel. Processing architecture, the working clock of the 1 28-SDM circuit can be reduced to one-half and one-quarter respectively. Therefore, the dSDM A/D converter of the parallel processing architecture can be selected according to the circuit recipe 1330467. Appropriate oversampling rate 'by reducing the circuit speed by increasing the circuit's working speed, the best choice between circuit work and volume, and thus reduce the A/D converter circuit volume and power consumption. (D) Simulation results table The second is used for UWB-OFDM transceivers - to nine-bit 128-SDM D/A and A/D converters to transmit/receive QpSK, 16qam

The best quantization unit with the 64-QAM UWB-OFDM signal, where the most φ good quantization unit A is the signal variance of QpSK, 16-QAM and 64-QAM UWB-OFDM by 0.0078, 0.039, 0.1641 and bit resolution 6 = log2L = 0 '1' ..·, 9 is substituted into (29), and then the quantization unit A is generated such that the zero-quantization noise variance is minimized. Figure 14 is a comparison of UWB-OFDM transceivers using one to nine bit USDM D/A and A/D converter to transmit/receive qpsk UWB-OFDM signals, for different quantization units 0.7 A, 〇.8 into, 0.9A, A, 1.1A' 1.2λ, 1.3 Α SQNR. The upper limit of the SQNR of the one-to-ninth D/A and A/D in the figure is the result of the calculation according to (21). As can be seen from Fig. 14, the SQNR of the QpSK UWB-OFDM signal is maximized using one of the best quantization unit A to the nine-bit 128-SDM D/A converter. Figure 15 shows the analog UWB_OFDM transmission machine according to Table 2, using the best quantization unit of different bits 128-SDM D/A and A/D converter respectively transmit/receive] 00 groups QPSK, 16-QAM and 64-QAM SQNR for UWB-OFDM signals. Figure 15 shows that for each bit resolution of the 128-SDM D/A and A/D converter, the SQNR of the 128-SDM D/A and A/D converter transmitting/receiving UWB-OFDM signals is increased by about 5 dB. And the optimal 42 1330467 quantization unit different bits 128-SDM D/A and A/D converter transmit / receive QPSK, 16-QAM and 64-QAM UWB-OFDM signal SQNR are the same, and modulation mode Nothing.

The bit error rate (BER) of the UWB-OFDM transceiver in the time-synchronized AWGN channel is determined by SNR and SQNR. Increasing the SNR of the transmitted UWB-OFDM signal can only reduce the influence of AWGN noise and cannot improve UWB-OFDM. SQNR of the transfer machine. Figure 16 shows the 128-point complex subcarrier signals of 100 sets of QPSK, 16-QAM and 64_QAM modulated respectively. After 128-IFFT processing of floating-point numerical operations, the OFDM symbol signals of floating-point values are output. The U WB-OFDM symbol signal is transmitted to the AWGN channel through a four-bit 128-SDM D/A, five-bit 128-SDM D/A converter and an ideal D/A converter (no quantization noise)' output. 'Receive UWB-OFDM signal under time synchronization' and convert to four bits by four-bit 1 28-SDM A/D, five-bit 丨28-SDM A/D converter and ideal A/D converter Meta- and quintuple fixed-point values and floating-point numeric digital OFDM symbol signals, input 1 28-FFT demodulated into 128-point complex sub-carrier signals, and the maximum likelihood law of the nebula decoding 'generates the position of the two-bit data Meta error rate (BEr). It can be seen from FIG. 16 that the present invention discloses a 28-iFFT output QPSK OFDM signal using a floating point operation, and transmits/receives a qpSk UWB-OFDM signal through a four-bit or five-bit 128-SDM D/A and A/D conversion. The BER is already approaching the BER of the ideal D/A and A/D transmit/receive QPSK UWB-OFDM signals. Because the UWB-OFDM transceiver uses four-bit and five-bit 128-SDM D/A and A/D conversion' when the minimum coding distance between the nebula modulated signals is fixed, then the transmitted/received UWB-OFDM signal The quantization noise variance will increase with the modulation order, so as the modulation order 43 1330467 becomes higher, the BER of the four-bit and five-bit 128-SDM D/A and the A/D converter The worse the BER of the ideal D/A and A/D converters. Four

Element 128-SDM D/A and A/D converter transmit/receive 1〇〇 group 16-QAM UWB-OFDM signal BER can be reduced to zero at SNR over 23dB, five-bit 128-SDM D/A and A/ D converter transmission/reception 1 〇〇 group 64-QAM UWB-OFDM signal BER can be reduced to one ten thousandth (10 - 4) when the SNR exceeds 30dB. Therefore, the UWB-OFDM transceiver uses the present invention to disclose the SQNR of the 5-bit 128-SDM D/A and A/D converter transmitting/receiving 16-QAM and 64-QAM UWB-OFDM signals in the AWGN channel. System requirements for BER. As shown in FIG. 17, the UWB-OFDM transceiver uses the five-bit 128-SDM (referred to as 2-TFI 5-bit 128-SDM) D constructed by the present invention to be constructed in time domain equivalent frequency interpolation. /A conversion, generating the first time spread symbol and the second time spread symbol (referred to as 1 st and 2nd TS-Symbol) of the 64-QAM UWB-OFDM signal of twice the time spread spectrum, and the sampling amplitude thereof The five-bit quantized value of the 64-QAM OFDM symbol signal of the 128-IFFT output floating-point value (the value can only be 0, A, ..., ±16A). However, the 1st and 2^ TS-Symbol generated by the 2-TFI 5-bit 128-SDM D/A converter are not exactly the same, and the received lst and 〗 〖 TS-Symbol goes through the five-bit 128-SDM A/D The despread symbol output by the converter is the six-bit quantized value of the 64-QAM OFDM symbol signal of the 128-IFFT output floating-point value (the value can only be 0, ±〇.5A, bandit, ± 1.5Δ...,±16A), lst TS-Symbol and 2nd TS-Symbol average (despreading frequency) are defined as 2-TFD 5-bit 128-SDM A/D OFDM generated by time equivalent frequency reduction Signal. So the time domain equivalent frequency interpolation doubles the double time generated by the exhibition 44 1330467

The frequency of a five-bit OFDM signal is reduced by twice the average of the time-domain equivalent frequency to generate a gain' such that the five-bit 128-SDM D/A and A/D conversion are constructed by interpolating/reducing twice the time domain equivalent frequency. The SQNR is the same as the six-bit 128-SDM D/A and A/D converter. The simulation results verify that the dynamic quantized value of the two-digit OFDM signal lst and 2nd TS-Symbol generated by the double-time spread-time two-bit 128-SDM D/A converter using the time-domain equivalent frequency interpolation is between Between ±16A and the quantization unit is △' and the two OFDM signals pass through five bits] 28-SDM A/D conversion and time domain equivalent frequency reduction twice the development of the deTS-Symbol of the OFDM signal The quantized value is between ±16 Δ and the quantization unit is 〇.5 Δ', which is equivalent to an increase in one-dimensional resolution. It can be seen from FIG. 15 that the SQNR of the five-bit and six-bit 128-SDM D/A and the A/D converter are about 25 dB and 30 dB, respectively. Therefore, the present invention discloses that the UWB-OFDM transceiver uses five bits. 128-SDM D/A and A/D converter are built in time domain equivalent frequency interpolation/reduction double circuit, except that it can replace the double-time spread/de-spread function, and five bits 1 28- The SQNR of the SDM D/A and A/D converter transmitting/receiving U WB-OFDM signals can be increased by 5 dB. After the same method is promoted, the 6-bit W-SDM D/A and A/D converters constructed in the time domain equivalent frequency interpolation/reduction L times can also be applied to the B-times spread spectrum and The W subcarrier OFDM communication system generates the SQNR of the I_2L bit W-SDM D/A and the A/D converter. Figure 18 shows the 128-IFFT output floating-point value of the 128-point complex subcarrier signal generated by the floating-point value of the 128-point complex subcarrier signal generated by the analog UWB-OFDM transmitter. OFDM symbol signals, which simulate three different D/A and A/D architectures respectively: 45 1330467 (a). Use the ideal D/A output UWB-OFDM signal to transmit to the AWGN channel and receive UWB-OFDM under time synchronization conditions. Signal, input the ideal A/D conversion to floating point value of the OFDM symbol signal input 1 28-FFT demodulation to 128 point complex subcarrier signal through the maximum likelihood law of the nebula decoding back to the binary data. (b) Repeat the input of the five-bit 128-SDM D/A twice, output the UWB-OFDM signal with twice the time spread spectrum to the AWGN channel, and receive the UWB with twice the time spread spectrum under the condition of time synchronization. OFDM signal, 轮Fdm symbol signal converted into five-bit 128-SDM A/D converted to two-digit five-point fixed-point value, and the two-digit five-bit OFDM signal with twice the time spread spectrum is averaged and then input. The 128-FFT demodulation is a 128-point complex subcarrier signal that is decoded back to the binary data by the maximum likelihood law of the nebula. (c) Time domain equivalent frequency interpolation twice and five-bit 128-SDM D/A converters generate twice the time spread UWB-OFDM signals are transmitted to the AWGN channel, receiving twice under time-synchronized conditions Time-spreading UWB-OFDM signal 'input five-bit 128-SDM A/D converted to two-digit five-bit fixed-point OFDM symbol signal' is constructed to double the time-domain equivalent frequency (twice the time show The two-bit five-bit OFDM signal is added to obtain a six-bit OFDM signal, and the input 128-FFT is demodulated into a 128-point complex sub-carrier signal, which is decoded by the maximum likelihood law of the nebula to generate two-bit data.

The signal-to-noise ratio of the UWB-OFDM signal transmitted/received by the ideal D/A and A/D converter is SNR = σ〗/σ, and substituting (33) can obtain five-bit 128-SDM D/A and A/D. The converter transmits/receives the double-time spread-spectrum UWB-OFDM signal with a signal-to-noise ratio of SNRTSP= σϊ2/(〇.5σ„2+σ|) = 1 / (0.5-SNR—1 + SQNR—1). When SNR 2 0.5.SQNR, then SNRTSP 46 1330467

=1 / (0.5'SNR-1 + SQNR_I) < SNR ' is the ideal d/A and the signal-to-noise ratio (SNR) of the UWB-OFDM signal transmitted/received by the A/D converter is greater than the five-bit 128-SDM D /A and A/D converter half-quantized signal-to-noise ratio (0.5'SQNR)' then 5-bit 128-SDM D/A and A/D converter transmit/receive twice-time spread-spectrum UWB-OFDM received signal The signal-to-noise ratio (SNRTSP) is smaller than the signal-to-noise ratio (SNR) of the UWB-OFDM signal transmitted/received by the ideal D/A and A/D converter. It can be seen from Fig. 15 that the SQNR of the five-bit 28-SDM D/A and A/D transmission/reception UWB-OFDM signals is about 25 dB 'Bei 1J. When SNR > 0.5.SQNR = 22dB, use five bits. 128-SDM D/A and A/D converters transmit/receive double-time spread-spectrum UWB-OFDM signals with higher BER than UWB with no ideal time-frequency for transmitting/receiving with ideal d/A and A/D converters The BER of the OFDM signal is shown in Figure 18. The UWB-OFDM transmission machine uses the present invention to disclose a five-bit 128-SDM (referred to as 2-TFID 5-bit 128-SDM) D/A and A/D constructed twice in time domain equivalent frequency interpolation/reduction. The converter, transmitting/receiving the SQNR of the UWB-OFDM signal with twice the time spread spectrum is 30dB. Similarly, when the SNR 2 is 27dB, the 2-TFID 5-bit 128-SDM D/A and A/D converter are transmitted. The BER of a 64-QAM UWB-OFDM signal that receives twice the time spread spectrum is greater than the BER of a 64-QAM UWB-OFDM signal that uses an ideal D/A and A/D converter without time spreading. Compare the six-bit 128-SDM D/A and A/D converters of Figure 16 and Figure 18 to transmit/receive twice-time spread-spectrum 64-QAM UWB-OFDM signals and 64-QAM UWB without time spread spectrum - OFDM signal, it will be found that the BER can only be reduced to about one ten thousandth (10-4), because the D/A and A/D converter transmit/receive signals 47 1330467 SQNR will determine the BER can reach Minimum value (lower limit of BER), and SQNR does not change due to double-time spread spectrum, so use 5-bit 128-SDM D/A and A/D converter to transmit/receive 64-QAM with double time spread spectrum The BER lower limit of the UWB-OFDM signal is not improved by time spreading. The 2-TFI 5-bit 128-SDM D/A and A/D converter disclosed in the present invention can make the BER of the JWB-OFDM transceiver better than the ideal D/A and A/D converters. UWB-OFDM transceiver with twice the time spread spectrum (2_TS).

In summary, the structural features of the present invention and the various embodiments have been disclosed in detail, and the present invention can be fully demonstrated that the purpose and the efficacy of the present invention are both rich and progressive, and the value of the industry is extremely valuable. As never seen in the market, according to the spirit of the patent law, the present invention fully complies with the requirements of the invention patent.

The above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the equivalent variations and modifications made by the scope of the present invention should still belong to the present invention. Within the scope of the patent, I would like to ask your review committee to give a clear explanation and pray for it. It is the prayer. [Simple diagram of the diagram] Table: Table 1 five-bit 128-sound-SDM A/D converter uses the working clock frequency of the flat circuit and the analog SDM circuit pair \ table. Table UWB-OFDM Transmitter uses five-bit 128-tone _SD and A/D transmission/reception QPSK, i6, Qam UWB-OFDM signal optimal quantization unit β ^ ΑΜ 48 € 1330467 Illustration: Figure 1 (a Digital analog converter basic architecture block diagram Figure 1 (b) analog digital converter basic architecture block diagram Figure 2 linear Midread three-bit quantizer transfer function diagram three UWB-OFDM transfer machine basic architecture diagram Figure 4 Μ立元 A/D converter (6 = 3) system diagram Figure 5 UWB-OFDM transceiver system block diagram Figure 6 (a) TV sound-SDM circuit block diagram Figure 6 (b) / V sound -SDM circuit discrete system block diagram Figure 7 is constructed in time domain equivalent frequency interpolation / reduction 32 times 128 tone - SDM circuit five-bit digital analog converter Figure 8 UWB-OFDM transfer machine used in the time domain, etc. Effective frequency interpolation/reduction of twice the five-bit 128-sound-SDM digital analog converter and analog-to-digital converter. Figure 9 is constructed in the time domain equivalent frequency interpolation/reduction 32 times 128-six-simplified five-bit analog digital Converter diagram ten analogy SDM circuit diagram eleven constructed in time domain equivalent frequency interpolation / reduction 16 times 128 tone - SDM The five-bit analog-to-digital converter of the road uses two SDM parallel processing circuits. Figure 12 is constructed in time-domain equivalent frequency interpolation/reduction of 8 times. The five-bit analog-to-digital converter of the 28-bit-SDM circuit uses four SDM parallels. Processing Circuit Figure 13 uses 128 SDM parallel processing circuits for 128-SDM five (I, 49 1330467 bit analog-to-digital converters)

Figure 14 Comparison of U WB-OFDM transceivers using SQNR for one to nine digits of 128-tone SDM D/A and A/D

Figure 15 compares the UWB-OFDM transceiver using the best quantization unit of the five-bit 128-SDM D/A and the A/D SQNR. Figure 16 compares the four-bit 128-sound D-A, A/D , five-bit 128-sound D/A, A/D and ideal D/A, A/D transmission/reception QPSK, 16-QAM and 64-QAMUWB-OFDM signals Figure 17 is constructed in time domain equivalent frequency interpolation Complementing the five-bit 1 28-SDM D/A produces a double-time spread-spectrum 64-QAM UWB-OFDM signal

Figure 18 compares the BER and the five-bit 128 of the UWB-OFDM transceiver used in the AWGN channel to construct a time-domain equivalent frequency interpolation/reduction of twice the bit 1 28-SDM D/A and A/D. BER-SDM D/A and A/D BER diagram description:

50 1330467 Table Table 1 Analogous SDM Number (2A') Interpolation/Reduction Magnification (A = 2A'-A) Operating Clock Frequency 1 32 16.896GHz 2 16 8.448GHz 4 8 4.224GHz 8 4 2.112GHz 16 2 1.056 GHz 32 1 528MHz

Table 2 b 1 2 3 4 "5 ' 6 7 8 9 A Δ (64-QAM) 1.001 1.368 1.730 2.067 2.378 2.662 2.931 3.174 3.430 A Δ (16-QAM) 0.488 0.667 0.844 1.008 1.158 1.299 1.434 1.562 1.690 Δ (QPSK 0.218 0.298 0.378 0.451 0.518 0.582 0.640 0.691 0.768

Claims (2)

1330467 X. The scope of application for patents: ί. - A device for the five-bit 128-tone digit-to-analog and analog-to-digital digits of the orthogonal frequency division multiplexing ultra-wideband transmission machine, including::: digitization state The system includes a time domain equivalent frequency interpolation benefit and a time domain equivalent frequency reducer, and has a 128-sound-SDM circuit; A five-bit analog/digital converter consisting of the following components, constructed in a time-domain equivalent frequency interpolation/reduction 128-sigma-SDM quantization circuit, replacing a large number of resistances and comparisons with an analog SDM circuit of a parallel processing architecture The quantized device is simplified, which simplifies the electric power consumption; ' Μ borrow = component, its two-bit data is transferred to the ίίΐ wave signal by the nebula encoder, and the sub-carrier signal per m point is serially converted and paralleled and input 128· fast reverse Fourier transform is converted into a = frequency division multiplexing signal 'and then with the five-digit digit quantization of the above components, five five digits / analog converter action, five-digit US sound dream two: two analogies and five The bit-bit 128-sDM analog-to-digital device adopts a design constructed in time domain-based equivalent frequency interpolation (-: deduction (10)), and the time domain equivalent frequency interpolation/reduction times轼: It can be completely equivalent to frequency interpolation/reduction processing, and the frequency of the 3rd frequency insertion rate is increased by a factor of 5%, resulting in a fast inverse Fourier transform..., the length of the fast Fourier transform is increased to 祁% to eliminate the second tone-S DM circuit quantization noise 'according to 128• fast inverse Fourier transform and 128· fast Fourier transform, five-bit 128-SDM D/AS】 52 I33Q467 UWB-OFDM transmitter is constructed twice in time domain equivalent frequency interpolation, The 5-bit 128-SDM D/A output signal achieves the effect of generating twice the time spread-spectrum UWB-OFDM signal, while the UWB-OFDM receiver is constructed to reduce the time-domain equivalent frequency by a factor of two, and the SQNR can add 5 dB. The de-spreading orthogonal frequency division multiplexing signal input 128-fast Fourier transform is demodulated into a 128-point complex sub-carrier signal, and finally the binary data is decoded back through the nebula decoder. 2. The device for applying the five-bit 128-sound-SDM digital-to-analog and analog-to-digital digits of the orthogonal frequency division multiplexing ultra-wideband transmission machine according to the first aspect of the patent application, wherein the five-bit analogy/ The digital converter further includes the following components: an oversampling 32-fold circuit; and an analog SDM circuit including a subtractor, an integrator, a one-bit A/D converter and a one-bit D/A converter; And a digital accumulator. 3. The apparatus for applying the five-bit 128-sound-SDM digital-to-analog and analog-to-digital digits of the orthogonal frequency division multiplexing ultra-wideband transmission machine according to the first aspect of the patent application, wherein the time domain equivalent frequency The interpolator further includes: a fixed length (128 points) fast inverse Fourier transform (IFFT); a memory register; - an oversampling 32 times circuit; and a parallel/serial converter . 4. The device of the five-bit 128-SDM digital-to-analog ratio analog analogy and the analog-to-53 Ϊ330467 digital position applied to the orthogonal frequency division multiplexing ultra-wideband transmission machine according to the first application of the patent scope, wherein the time domain, etc. The effect frequency reducer further comprises: a serial/parallel converter; a memory register; a parallel/serial converter; a 128-dimensional vector accumulator circuit unit; and a parallel/serial converter And a fixed length (128 points) fast Fourier transform (FFT). For example, the five-bit π 128 tone-SDM digital to analog and analog-to-digital device of the m-crossing frequency division multiplexing ultra-wideband transmission machine described in the i-th patent application scope, the time domain equivalent frequency interpolator The compensation process is after a fixed length (128 points) fast inverse Fourier (four) change, the interpolation process is performed in the time domain and the fast inverse Fourier transform length of the orthogonal frequency division multi-band ultra-wideband transmitter * needs to follow the frequency The interpolation magnification is increased by 32 times, and the length of the fast inverse Fourier transform is increased by 4096. The reduction of the time domain equivalent frequency reducer is before the fast Fourier transform at a fixed length (10) point, and the reduction processing is performed by the positive father frequency division. The fast Fourier transform length of the multi-bandwidth ultra-wideband receiver does not need to increase by 32 times with the frequency reduction ratio. The length of the fast Fourier transform is increased to 4〇96, and the +bits of the two-element bead can be selected according to the Nebula® coding method. 64-QAM. 0 or = the luxury patent patent earning item is applied to the octave 128-sound/SDM digital-to-analog and analog-to-analog device for orthogonal frequency division multiplexing super-transfers, where the five-digit quantizer Constructed in the time domain, etc. 54 1330.467 Effective frequency interpolator and receiver are constructed in the time domain equivalent frequency reducer, which can effectively eliminate the quantization noise of 128-SDM circuit and reduce the number of logic gates and processing time. 7. The apparatus for applying the five-bit 128-sound-SDM digital-to-analog and analog-to-digital digits of the orthogonal frequency division multiplexing ultra-wideband transmission machine according to the first aspect of the patent application, wherein the five-bit 128_SDM D /A and A/D use the best quantization unit to transmit/receive QPSK, 16-QAM and 64-QAM UWB-OFDM signals have the same SQNR' five-bit 128-SDM D/A and A/D can meet the peak mean value of 16 -QA1V [Performance requirements for BER with 64-QAM UWB-OFDM communication systems. 8. The device for applying the five-bit 128-sound-SDM digital-to-analog and analog-to-digital digits of the orthogonal frequency division multiplexing ultra-wideband transmission machine as described in claim 1 of the patent scope, wherein the five-bit analogy/ The digital converter implementation can also employ a parallel processing architecture of thirty-two analog SDM circuits that do not require oversampling while maintaining a sampling rate of 528 MHz. 55