FI108323B - Method and arrangement for implementing a self- compensating band pass filter in a digital telecommunications link - Google Patents

Abstract

This publication describes a method and an arrangement for implementing a self-compensating band pass filter in a digital telecommunications link. According to the method, a signal is band pass filtered H,z superscripts - 1, and encoded K,z superscripts -1, according to the Tomlinson principle in the transmitter in such a way that the band pass and encoding functions are implemented by use of common calculation arrangements. <IMAGE>

Description

108323

Method and Equipment for Implementing a Self-Compensating Bandwidth Filter on a Digital Telecommunication Connection

The invention relates to a method according to the preamble of claim 1 for implementing a self-compensating bandwidth filter on a digital communication link.

The invention also relates to apparatus for implementing a self-compensating band-stop filter on a digital communication link.

With modem connections providing broadband communications, the signal spectrum extends to frequencies as high as 10 to 30 MHz. At such high frequencies, some of the energy propagated by the copper cable radiates into the environment as radio waves. Some of the energy radiated into the environment falls on the frequency bands reserved for radio amateurs. This proportion must be limited in order to avoid interference with the amateur radio modem connection. In practice, the limiting can be accomplished either by selecting a band allocation that does not transmit the signal in radio amateur bands, or by filtering the output signal of the modem connection transmitter with bandpass filters so that the outgoing signal power density in radio amateur bands is low enough . Bandwidth filters can be analog, digital or both.

The benefits of a filtering solution over a bandwidth allocation solution are: filters are easy. . : 20 on / off switching and lane allocation optimization • \ *. · Without having to commit to the location of radio amateur bands. The disadvantages of the filtering solution are the complexity of the channel equalizers and the loss of transmission capacity relative to the * ... 'signal band remaining after filtering.

. ** ·. 25 In prior art solutions, there are two principles for correcting the signal distortion caused by the • bandwidth filter inserted in the transmitter. The distortion caused by the filter is compensated for by the distortion of the signal caused by the entire channel (including the bandwidth filters) by separate equalizers, which may be fixed or adaptive and which may still be located in the transmitter (TX), receiver (RX) or partially. Another way is to compensate for the distortion caused by the filters 2 108325 with customized corrections for filter distortion correction, thereby avoiding the substantial complication of other channel distortion correctors.

A disadvantage of the prior art is that hardware solutions can become complex 5, since either separate k.o.s are required to correct the distortion caused by the band-pass filters. Equipments for filters or distortion compensation capability of channel distortion correctors must be increased. The complexity of the implementation has several disadvantages: e.g. problems with processor capacity and code space sufficiency in processor implementation, power consumption and the required number of 10 logic ports in ASIC implementation.

The object of the invention is to provide a completely new type of method and apparatus by which the above-mentioned prior art problems can be solved.

The invention is based on the implementation of transmitter (TX) digital band-stop filters so that the filter itself also forms a Tomlinson-15 type pre-encoder, thereby avoiding substantial complexity of the channel equalizers. Adjusters only become more complex in that the decision feedback home (DFE) must be able to handle extended symbols. The output signal of the self-compensating bandpass filter output consists of expanded symbols so that both the linear coqain (FFE) and: · '20 decision feedback coqain (DFE) are able to operate at the same pin values as •' when that bandwidth filter is not used.

More particularly, the method according to the invention is characterized by: *** which is set forth in the characterizing part of claim 1.

• · · • · * • ·

The apparatus according to the invention, in turn, is characterized by what is stated in the characterizing part of claim 25.

• · • ♦ ·· »•« · *,. · The invention provides significant advantages.

Because the same computing machine performs both the filtering and the pre-coding functions, the prior art technology can be simplified from a circuit solution or a processor program, e.g. or a simpler and less expensive 108323 3 signal processor may be used. Lower power consumption results in e.g. increased reliability and the ability to pack the equipment to a smaller size.

The invention will now be described by way of example and with reference to the accompanying drawings.

Figure 1 is a block diagram illustrating one embodiment of the invention for implementing a self-compensating bandpass filter, which shows that the same F (z '') can be used to calculate both the bandpass filtering H (z '') and the precoding K (z ''). results.

Fig. 2 is a block diagram illustrating the solution of Fig. 1, transformed into a realizable form 10, from which, without limiting the generality, the baseline setting factor g may be omitted.

Figure 3 illustrates the solution of Figure 2 with a modified topology utilizing that subtraction and incrementing cancel each other out.

Figure 4 shows a rigid system according to the invention in a more general form.

Fig. 5 is a block diagram of one module of a prior art bandpass filter of prior art which implements one pole and one transmission zero.

Fig. 6 is a block diagram of one module of a prior art precoder consisting of successive modules which compensates for the proportion of distortion caused by one pole and one:: transmission zero by the bandpass filter.

• ·]. Figure 7 is a block diagram of a prior art solution in which: ···. the input to the precoder is summed with a modulus CN * 2m so that the output is, ···. limited.

• ·

Fig. 8 is a block diagram of one module of a pre-encoder of sequential modules of the prior art • · * ··· modified so that Fig. 7 • · · •; · 'The addition of 25 modulus CN * 2m could be moved to the output of the last block Nz.

The self-compensating band-stop filter according to the invention can be formed by placing the system illustrated in Figures 7 and 8 in the 'Pre-coder' block of Figure 4.

108323 4 The mathematical analysis below shows that the system thus formed according to the invention implements both precoder and band-stop filter functions.

Abbreviations used ASIC Application Specific Integrated Circuit 5 CAP Carrierless Amplitude and Phase Modulation DFE Decision Feedback Equalizer FFE Front Forward Equalizer 1.F Digital Filter (Finite Impulse 10 Response) IIR Infinite Impulse Digital Filter with Infinite Pulse Response

Response) PAM (Pulse Amplitude Modulation) QAM Quadrature Amplitude Modulation 15 RX Receiver; TX Transmitter TML Tomlinson-Harashima precoder (Tomlinson-Harashima precoder) • «· * * # (z ~ ') Transmission function of the bandpass filter, where z'1 is a symbol interval delay. ζ ~ ') The precursor transfer function, where ζ'1 is the symbol delay. The invention thus relates to a method and apparatus for implementing a transmitter • · ** «. ·· *. install a digital band-stop filter so that no separate I / O is required and, on the other hand, I do not have to significantly complicate the I / O to compensate for distortion caused by other parts of the channel M. such as the cable. A limitation of the bandpass filter according to the invention is that the filter sample frequency must be 108323 5 symbol frequencies and the filter input signal must be a symbol string to be transmitted, since the filter / precoding machine also performs the precoding / filtering function.

The relevant theory is discussed below.

The transmission function of the digital bandpass filter 5 using the symbol frequency sample frequency can be represented in the form: prz->) r10- ^) H (z ~ ') = —Z, = g- £ -, (1) Ö (z} Tin / = 1 where z2i (complex) is zero i in the z plane, zpt (complex) is the pole i in the z plane and g 10 (real) is the coefficient to obtain the desired gain for the filter. (z '') = 1 ie Zpj = 0 for all i. H (z "') can be represented as follows:

N, Np Γ M

ΠΟ - ^^ - ΠΟ- ^ 2'1) Zciz_i: H (z_1) = g- 1 + —--- * - = g · 1 + z -'- ^ -, (2): * ·. Π (1-ζρ | ζ- ') Π (1-ζρίζ-'); ** L i = i J L 1 = 1 «* 15 • · • 4 ·« «I # · * ♦ ·· 'where Cj is a complex number and M is greater than Nz-1 and Np-1. The last form. ^ follows that the constant term for both polynomials P (z '') and Q (z-1) is 1. Denote the last form • · * ··· * as follows: «1« 20 H (z ~ ') = g · ( l + z_lF (z '')). (3) 108323 6

The bandpass filter having that transmission function can be offset by the precoding Where the transmission function is: K (z'1) = -p-r. (4) 1 + z ~ 1F (z-1) 5 At the input of the precoder, the modulus CN * 2m must be added so that the real and imaginary parts of the precoder output remain between -m, + m. The number m is the largest value of symbol geometry + 1 and CN is a complex number whose real and imaginary parts are integers. It will be appreciated that the same result of F (z_1) as shown in Figure 1 can be used to calculate both H (z "1) and K (z '').

The system shown in Fig. 1 can be converted to a realizable form (without limiting the coefficient g can be omitted) according to Fig. 2, where the architecture is straightforward if the bandpass filter H (z_1) is such that it can be easily implemented as a decay: -1). If the bandpass filter is a multi-stage IIR filter, the implementation of the calculation of F (z '') for at least silicon may be complicated by equation (2).

The following shows a preferred embodiment of the invention in which the computing mechanism is implemented only for the precoder but not for the bandpass filter. System:. ' the output signal consists of expanded symbols whose spectrum includes the desired 'inhibitory bands. The extended symbols are the incoming symbols plus the module numbers specified in the pre-encoder «·: 20.

• ·

The system shown in Fig. 2 can be converted to the following form>> · in Fig. 3 (subtract and subtract):

It will be seen from Figure 3 that it is sufficient to calculate only the precoding function and generate * · J • · the required module numbers which are summed up in the system input. The system can be represented in a more general form as shown in Figure 4.

«T I I« I i «:" ': It can be seen from Figure 4 that the implementation architecture of the precoding function can be freely selected. Thus, it is not problematic, even if the implementation of the calculation of F (z_1) is constable.

7 108323

The following describes the implementation of the rigid system according to the invention when the bandpass filter is a multi-stage IIR filter. Consider an IIR filter with the same number of poles and zeros and poles and zeros at the same frequencies (Zpi = rjZzj and Np = Nz, where η is a real number and 0 <n <1). With such a filter, various bandpass filters can be implemented with sufficient flexibility. The filter may be implemented by an architecture consisting of sequentially placed modules, one of which is shown in Figure 5.

The transfer function of the precoder can be implemented by corresponding modules in which the roles of zero and pole are reversed as shown in Figure 6.

10 The module encoder CN * 2m must be summed at the input of the precoder with the output limited as shown in Figure 7.

In Figure 7, the addition of the modulus CN * 2m can be moved to the output of the last block Nz, but then the addition of CN * 2m must also be placed in each sub-block, e.g. as shown in Figure 8.

The self-compensating bandpass filter is as a whole according to Fig. 4, wherein the precoder block consists of the sequentially placed sub-blocks of Fig. 8. The modulo number CN is chosen such that the real and imaginary portions of the output of the last sub-block remain between -m, + m, when m is the largest * · t value of symbol geometry + 1.

20 •> • · • · · · ·

Ml • · m · • »· • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

• V

«Tl a (<1 l«

Claims (4)

A method of implementing a transmitter self-compensating bandwidth filter on a digital communication link, the method of transmitting at the transmitter 5. the signal of the bandwidth filter in a transmitter, which may be represented as H (z '') = g- (l + zAY {zx)), which precursor can be represented as K (z '*) = 1 / (1+ z' ^ z'1)) supplemented by a modulo operation, whereby 10. filtering (Η (ζ-1)) and precoding (K (z-1) )) the sample rate must be a symbol frequency and the input signal must be a transmitted symbol string, characterized in that the signal processing function F (z "') of both the bandpass filter (H (z-1)) and the precoder (K (z' ')) is implemented by a common computing mechanism. Method according to Claim 1, characterized in that the output signal containing the desired inhibitory bands 15 is formed by summing to the symbol sequence which is the input signal the modulus numbers according to the Tomlinson principle determined in the precoder. Apparatus for implementing a transmitter self-compensating bandwidth filter ♦ * · * · * * with a digital communication link comprising: ♦ · ·; ··: 20 transmitting means in the transmitter as a digital filter, which may be represented by ♦ ♦ ♦ φ. in the form H (z ') = g- (l + z' F (z ')) and tt »*« • · · - means for pre-coding the signal using a fixed Tomlinson principle. ·· *. precursor, which can be represented as K (z '') =. · * ·. 1 / (1 + z'1F (z'1)) supplemented by modulo operation 25. the sampling frequency of filtering (F ^ z'1)) and pre-coding (K (z '')) must be: the symbol frequency and the input signal must be transmitted symbol sequence, 108323, characterized in that the computing mechanism for implementing the signal processing function F (z '') of both the bandpass filter (H (z "')) and the precoder (K (z' ')) is common. Apparatus according to claim 3, characterized in that it comprises means for generating an output signal containing the inhibitory bands by summing the modulus numbers according to the Tomlinson principle defined in the precoder into a symbol sequence which is an input signal. • · · • 1 • · · • · · 1 · • ·, 0 1 08323