TW455800B - Device, system, and method for modem communication utilizing two-step mapping - Google Patents

Abstract

Modem technology is implemented in a system including a personal computer (PC) to enable communication over a PSTN using communication software which includes a mapping algorithm for mapping data into u-law signal points, which are divided into segments with different distances.

Description

455800 Printed by the Central Standards Bureau of the Ministry of Economic Affairs and Consumer Cooperatives A7 B7 V. Description of the invention (1) 范围. Scope of the invention-The present invention relates generally to a modem device, a modem-based system, and a device for The method of signal communication is specifically about high-speed modem communication using two-step mapping technology. 2. Background of the Invention Several known conventional modem devices, systems, and methods are transmitted between personal computers' via a public switched telephone network (PSTN). To avoid this type of communication, information frame mapping technology has been used previously and is currently in use. However, due to the improvement of the performance of the modem, the need for various information frame mapping techniques to reduce the average power of a set form to a minimum is increasing. SUMMARY OF THE INVENTION According to the present invention '-a PC transmitting and receiving signals via a PSTN uses a two-step mapping algorithm to reduce or minimize the average power of a set shape benefit. The PC includes a modem that communicates with a device, such as another PC, a mini computer, a host computer, a facsimile machine, etc., via the PSTN. The modem includes a two-step mapping device that uses a first-order generation function to minimize (or reduce) the use of signal points with a minimum distance and to reduce the error rate. Brief Description of the Drawings' FIG. 1A is a block diagram of a PC communication system using a public switched telephone network in the present invention. Figure 1B is a pc block diagram in the present invention. FIG. 1C is a transmission signal spectrum in the present invention. -4- This paper size is applicable to China National Graduation Standard (CNS) Α4 size (210 X 297 gram) (please read the precautions on the back before filling this page) Crack 455800 A7 B7 V. Description of the invention (2) Figure 2 is a block diagram of a 56k baud rate modem communication system in the present invention. Fig. 3 is a block diagram of a codec of a 56k baud rate modem communication system according to the present invention. Fig. 4 is a block diagram of an uncoded decoder of a 56k baud rate modem communication system according to the present invention. Fig. 5 is a block diagram of the encoder with spectrum encoding and resolution of the 56] baud rate modem communication system of the present invention. Fig. 6 is a block diagram of an unencoder with a spectrum of a 56] baud rate modem communication system of the present invention. Fig. 7 is a block diagram of an analog half-duplex modem receiver according to the present invention. Fig. 8 is a block diagram of a possible specific embodiment of an analog half-duplex modem receiver of the present invention. FIG. 9 is a block diagram of an analog full-duplex modem receiver according to the present invention. FIG. 10 is a block diagram of a specific alternative embodiment of an analog full-duplex modem receiver of the present invention. Fig. 11 is a block diagram of a possible specific embodiment of an analog full-duplex modem receiver according to the present invention. For a detailed description of a preferred embodiment, refer to FIG. 1A, which shows a configuration according to the present invention, in which a PC 100 includes a modem 101 'and communicates with a device 102 via PSTN 103, such as another PC, Mini computers, host computers, fax machines, etc. The modem 101 can operate at a digital signal transmission / reception rate of 14,400 (14.4k), 18,800 (28.8k), 33,60 (33.6k), 56,000 (56k) or its signal rate (such as baud rate). The modem also includes a two-step mapping device 104, which uses -5- this paper size to apply Chinese National Standard Soap (CNS > A4 specification (21〇χ297 Male Thin)) (Please read the precautions on the back before you fill out (This page)

45 5 8 0 0 A7 B7 V. Description of the invention (3) Generate the function once or twice to minimize (or reduce) the use of signal points with the smallest distance and reduce the error rate. Referring to FIG. 1B, the PC 100 in the figure includes a 107tPC system software 105 and a data processor software 106 connected to a central processing unit (CPU) 107 via a conventional data and instruction bus 108. The CPU 107 operates the modem software 106 to coordinate the operation of the modem controller (not shown), and directs the transmission of the signal via the modem 060. During the operation of the modem software 106 by the CPU 107, the digital signals are generated and sequenced in the message box of the modem 101. If the data signals are transmitted, the digital signals can be applied to a screen image, or The situation of the transmission of sound signals' These signals may correspond to a block of speech. Each message box contains information combined with the N symbol and is defined by the κ binary. The bins are ordered ' within a digital signal stream and are divided into sections. Referring to FIG. 2, a 56k baud modem communication system digital modem 101 is shown connected to an analog modem 201 connected to a device 102 via a PSTN 103. Its sample rate is 8 kHz. In each sample, a coder (not shown) operates on a set of input bits. An encoder (not shown) is printed by the Zhengong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economics (please read the precautions on the back before filling this page). The output of the coder is used to generate output signal samples. The samples are sent to #Law d / A 205 via digital network 203 and converted into analog signals. # 法律 D / A 205 uses a payout number format with a total of 255 digits that can be entered. The 255 values are divided into 16 segments: +1, +2, ..., +8 '-1, -2, ..., _8. The distance between two adjacent points in the i-th segment is. For simplicity and without limiting this value, let c = i. The minimum distance between the +1 and _ 丨 sections is 1 ′ 〇 and the two sections are shared. These two minimum distances -6- This paper size applies to the Chinese National Standard (CNS) A4 specifications (21 {5 × 297 · ^ ~ || · Printed by the Bayer Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs 455800 A7 ____B7 V. Description of the invention ( 4) Segments with a distance of 1 may not be counted because there is no significant loss of transmission quality or reception quality, so there are 224 points left. By analogy, a pair of small points can be ignored, so increase the minimum Distance ◦ Every time 32 points are removed, the minimum distance is doubled, and the benefit of 6dB noise margin is obtained. Among the points that can be used for signal transmission, these external points have the maximum distance 'away' and become available. Good point. In order to use these larger points more, a traditional shell mapping algorithm may be modified and applied. For example, a traditional v. 34 data machine is used to divide the two-dimensional signal constellation into M Ring, each μ ring has the same number of points. These M rings are sorted according to the average power of the ring. The shell mapping algorithm more often selects these inner rings to reduce the total average power. In the present invention , These one-dimensional points are divided into 2m sections: ± (9-m) ···, ± 7, ± 8 'm = 1, 3, ..., 7. The + i and _i sections are combined into a section with a section index 8-i. The section index ranges from the maximum energy to The minimum energy is sorted a. Refer to Figure 3, where the encoder 3101 is connected to a digital network 203 and receives multiple inputs. The encoder 301 includes a section mapper 303 and an interleaving encoder 304, both of which provide Input signal to bitmap 305. After the encoder encodes these transmission signals, it immediately transmits them from bitmap 305 to digital network 203. Now more specifically select one of the M available points and transmit N symbols K bits in the information frame, the data rate is 8000 K / N bits / second = the M points are divided into 0,1, ..., m−1 m = M / 32 sections. Note that regardless of K / N or K for each symbol bit must be an integer. In the case of non-integer K, the information frame exchange technology may be applied. However, to simplify the discussion, it can be assumed that K is an integer. If China is applied to a paper scale Standard (CNS) A4 size (210X297 mm) (Please read the note on the back before filling in this page)-Order _ 455800 Central Standard of the Ministry of Economic Affairs Printed by the Bureau of Work and Consumer Cooperatives A7 Β7 V. Description of the Invention () Selecting a point in the section requires 5 bits, you can use k == K-5N bits to do "section mapping". The maximum possible value of k It can be obtained from the following inequality: 2k ^. For example, if 192 points are used and the size of the mapping box is 8, then m = 6 and k S 20.68. If k = 20, each symbol transmits 7.5 bits and the The data rate is (k + 5N) * 8000 / N = 60 kbps. To send signal information, 'traditionally' (bits) are used every 6 symbols, but the old codec uses 1 for each symbol. Bit. Therefore, if one bit is taken every 6 symbols', the remaining LSB bits are used as the intercepted signal bits, and there are M / 2 points available for use. However, it should be noted that the present invention can be modified to meet the requirements of older codecs. With M / 2 possible points, 4 bits can be used to select an even point. The data rate is reduced by 8000/6 = 1333.33 bps. For some older codecs, the truncated bits are used for each symbol ’, so the data rate is reduced by 8 kbps. It is not necessary to use the largest possible k. If a small k is used, then

Because of the free use of external points and the "distance gain", D, if N = 8 'is selected, the segment mapping algorithm used may be the same as the shell mapping algorithm used in traditional V.34 modems. In the shell mapping algorithm, the 'cost function approaches the linear function'. This linear function is very suitable for estimating how to minimize the average power, but not for the average distance. If a criterion is used to minimize the average distance, the ideal cost function should be inversely proportional to the distance. The distance is an exponential function. For example, if you use 7 segments and a linear cost function, choose 2 The cost of each segment is $ points, 1 segment is 4 points, and 1 segment is 6 points. The cost is the same, and the average distance is 128 and (256 + 64) / 2 = 160. Therefore, the rating is -8. This paper size is applicable to China National Standard (CNS) Α4 specification (210 × 297 mm) {Please read the precautions on the back before filling this page]

4 5 58 0 0 A7 B7 5. The method of the invention description (6) should be to modify the mapping algorithm 'to obtain the maximum-average distance. —Another way of the shell-mapping algorithm is to assume that there are 7 rings and the following generation functions can be achieved: g 1 = 1 + x + x2 + x3 + x4 + x5 + x6 two g2 = gl * gl = l + 2x + 3x2 + 4x3 + 5x4 + 6x5 + 7x6 + 0X7 + 5X8 + 4X9 + 3X10 + 2Xu + xt2 g4 = g2 * g2 = l + ... x24; g8 = g4 * g4 = l + ... x ', s 〆 'The size of the far G4 table is 4m-4 · and the size of the G8 table is 8η-8, where m is the number of rings. -If k = 7, modify the cost function to 0,1,2,4,8,16,32, and the generation function becomes:

gl = l + x + x2 + x4H g2 ^ 1 * gl = l + Z \ ^ 2 \ ^ + 2x5 + 2x6 + 2xs + 2x9 + 2XI0 + 2X12 + 2X164- .. —ΙΪ 1- -I-II =- ιν _- ^ I-li I. {Please read the note i on the back before filling this page) g4 = g2 * g2 = R .. :: g8 = g4 * g4 = K ... x: 128 .256 Ministry of Economy Although printed by the Central Standards Bureau member H Consumer Cooperative, the size of the G4 table is l + 2k; the size of the G8 table is l + 2k + 1. Although the same mapping algorithm can be used, these tables are much larger and more complex. On the other hand, a larger average distance may not be the only criterion. Another criterion is to minimize the frequency of the smallest segment. Since the error seems to occur more often in the smallest segment ', this criterion may improve the results. In order to obtain this result, the paper size is in accordance with China National Standards (CNS) A4 (210 × 297 mm) 45 58 0 0 A7 B7 V. Description of the invention (7 The printed fruits of the Consumer Cooperatives of the Central Government Bureau of the Ministry of Economic Affairs, The generating function can be: gl = l + X + X2 + X3 + x4 + X54-X41. In this case, the cost of any 8-point bundling with at least one point in the minimum section is lower than that of all kinds without any point. 8-point combination. In other words, the section is selected only after using various combinations that do not have the point. A special case is that if the minimum frequency of the minimum distance is a difficult criterion, Bayes can use a simpler The generation function is as follows: 'gl = 6 + X In this case, its cost function is a segment that must not be used. Another feasible criterion is to reduce the usage of the sub-section to obtain the generation function: gl = 5 + x + x9 here In this case, the 2 segments whose cost function is 〇, 〇〇, 〇, 〇, i9 should not be used. The table size is appropriate, and its performance is slightly better in the other embodiment described above. If you use _ Of the cost function, then the cost does not just refer to the segment. For example, if Using the final cost function, cost 0 means segment to 4 'cost i means segment 5 and cost 9 means segment 6. The segment that should be used in this case can be obtained by using the second generation function. Decision. Similar to the above shell "mapping expression" search the zs table, a ,, and "pass the total cost of 8 bluffs. D If the total cost of the child is 0, you may need to use the second generation function such as: hl = l + x + x2 + x3 + x4 h2 = hl * hl and only the minimum This rule can be borrowed and only the minimum f Please read the notes on the back before filling in this page j " binding. Order '-10- this paper Kill μ sword t ϋ n standard 210χ297 ^ 7 A7 B7 4 5 58〇 〇 --- 5. Description of the invention (8 h4 = h2 * h2 Then use the same mapping algorithm, you can get the range from 0 8 segment indicators from 4 to 4. If the total cost of the 8 symbols is not 0, then use the material table to divide the 8 'cost into two costs of 4 symbols. If one of the four symbol costs is 〇, then go to the dagger table to find the four segments of 0 to 4. Refer to the same. 'If the cost of the four symbols is not 0, then continue to use the § 2 table to cost the four symbols Two costs into two symbols. If the cost of one of the two break costs is 0, use the M table to obtain two segment indicators. Note that the h table is very small. 8 segment indicators ( From 0 to m_ 丨) is derived from the section map,. The index 0 means the section between + 8 and _8, which has a total of 32 points. After that, pick one of these 32 points with 5 bits For the sixth symbol, because the LSE is a truncated signal bit, only four bits are used to select one of the 16 points. Weaving code is an effective technique that can improve performance. When an interleaved encoder 304 is used, a 4D code may be used in a modem such as v.34. The symbol is one-dimensional and 4 symbols can be combined into a 4D signal to be applied to the 4D code. The cost is 1 bit ' per 4D, i.e. 25 bits per symbol, or 2 kbpS. In every 4 symbols, input 3 encoding bits and generate 4 bits for selecting the secondary set. It is possible that 1 bit is used as the LSB of each symbol for selection between even and odd points. In this case, each mapping information frame has k + 27 bits. In the above example, k = 15 and the data rate is 56.67 kbps. Because the truncation bit is also on the LSB, the LSB is selected between even and odd points, so special consideration must be given to the truncation symbol bit in each 6th symbol. The paper scale applies to the Chinese national rate (CNS > Μ specifications (2! 〇X 297 mm) -------- install --- ^ ---- order ----- 1. Line (please read the precautions on the back before filling this page) The Consumers' Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs, Printed Together 455800 A7 —B7 9 V. Description of the Invention ((Please read the precautions on the back before filling this page) How to deal with this parent weaving code " ~ A possible way is to use this The coding bit is used as the first bit of these symbols with dynamic bits, that is, for these special symbols, 'the signal distance is doubled. The Viterbi decoder will cut the chirp signal for this symbol in various ways. Min .. In this case, there is a loss of 1333 bps in the data rate. 'Wang Yi, the intercepted signal bits affect the selection of points within the segment, but do not affect the segment mapping itself. So the mapping box The size does not need to be 6. For example, the above-mentioned shell mapping algorithm can use a similar mapping box of size 8. Use The parser for 1/6 intercepting the foot signal bits is as follows. Define a large message box with% symbols, that is, 3 mapping boxes with 8 symbols each, or 6 coding boxes with 4 symbols each. In the large frame, there are segment mapping bits, 18 coded bits' and 92 uncoded bits.. {Si, 〇, i, Sit〇, 2, " .Si, 0, k}, {Ili'〇J2i-〇5l3i > 〇} > {QiA〇-IiW i.0,2,2,-· Qi, 0,2,4}, {Qi.0,3,1.Qi .0,3,2, · Qi, 〇, 3j4} > {IluJ2uJ3, J) {Qi > 1AlAl, 0,2, ..; Qi) li〇4}) {Qiu U, 2.2, ". Qu, 2,4}, {Qi, l, 3, l, Qi, l, 3,2, ·-.Qi, i, 3,4)-. R · Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs

Qu. ,, .. Such as 12- This paper size applies Chinese National Standard (CNS) A4 specification (2 丨 0X297mm) 4 5 5 8 0 0 A7, B7 V. Description of invention (1Q)

i, 2A2 »· • Qi, 2,2.4}» {Qi, 2,3, l > QiA3,2 > Q〇, 33} J

{lii, 3.12 ^ 3,13; ^}, {Qji3i〇, i, Qi > 3, 〇) 2, ... Qi > 3i〇i4} > {Qi.3, iiiJQi, 3.1,2i --- Qi13 > lt4} 3 {Qi, 3,2,1 > Q

{Si, 2, l, S, " .3 to}, I {II i, 4,12 ^ 4,13 ^ 4), {Qi (4 (〇 > bQi, 4,0,2 > ' --Qi, 4,0,4}, {Qu.l.wQiAl, 2j --- Qi, 4, I.3} ^ Qi.4.2, l »Q! I, 4,2,2, ... Qi, 4,2,4), {Qi, 4,3, i, QiA3,2, ... Qi, 4,3,4). {I li, 5, Kj ^ JSi sJ ^ Qj ^ hQi, 5, 〇.2 > '** Qi, 5,0,4} > {Qi, 5, U > Qi.5, l, 2) · i, 5,2,2, ... Qi, 5,2,4}, {Qi.SAwQiMAQiWj} 0 l / 6 The data rate of the intercepted signal bits is (k * KB6,67). The data rates for various M values are shown in the table below. (Please read the precautions on the back first and then fill in the horse clothing training. ”Load Μ m Distance k Data rate (kbps) Fully intercepted data rate 224 7 2 22 58.67 52 192 6 4 20 56.67 50 160 5 6 18 54.67 48 128 4 16 16 52.67 46 96 3 32 12 48.67 42 64 2 64 8 44.67 38 32 1 128 0 36.67 30 Today, the Central Standards Bureau, Ministry of Economic Affairs, Consumers Cooperatives printed the table 1. The data rate and the value of M are redundantly intercepted data rate In this case, the profiler is relatively simple. Its period is _ a mapping box. {SuA2, ... Si, k} >

川 i, 〇'I2 丨,. A ,. }, {Q;,. . Mountain QiA〇 > 2, Qw W

Ql〇, 2,3}, {Qi.oAhQi.mQm]},, 〇 ;; Q; > 1〇3}, {Qi, 1,1,1, Qi, 1,1,2sQi (11 3) s {Q; j 2 lsQ,: ^ -13- This paper size applies Chinese National Standard (CNS) A4 (210x297 mm) '--------- 4 5 58 0 0 A7 —B7 V. Description of the Invention (11, printed by J Industrial Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs, each 8-symbol mapping box has k sector mapping bits, 6 coding bits, and 24 uncoding bits The data rate is listed in the table above. The data rate is listed in the table above. From the table, it can be seen that the data rate supporting a 56k baud rate requires only I ”points, and its minimum distance is 4. Therefore To achieve proper performance, an SNR of about 40 dB is required. If the mapping gain and coding gain of the segment are considered, it is not difficult to meet this requirement. If the circuit does not support this requirement, then it will Μ is reduced to 160. After that, a data rate of 54.67 kbps can be operated with a 30dB · ^ SNR using a button, for example. ': From this table, it is also known that if it is also reduced !, it is usually reduced by 2 And the data rate of the dragon minus 2kbpSDmW can provide -6 noise margin. Because Therefore, use power minus 1 instead of using coding to obtain a similar rate with better noise tolerance, and make the encoder 401 easier to achieve the work, as shown in Figure 4. The data rate when not coding is shown below. SNR 46 40 34 28 22 16 10 Table 2 · The data of the uncoded case must be noted that the conventional convolutional code is invalid —------------------ --------------- ~ dou ', the reason is partly due to the uneven distribution of the signal. In the external section, the signals such as μ, μ, etc., are separated due to too much separation. No coding is required * Only the smallest segment is required — • 14- — Effective steps M m Distance k Data rate (kbps) 224 7 2 22 60.67 192 6 4 20 58.67 160 5 8 18 56.67 128 4 16 16 54.67 96 3 32 12 50.67 64 2 64 8 46.67 32 1 128 0 38.67 This paper size applies to China National Standard (CNS) Α4 size (210X297 male thin) (Please read the note on the back before filling this page) Order 455800

A7 SI V. Description of the invention (12. 骡 Only need to add redundancy in the smallest section. Hybrid calendar: the device has a spectrum failure phenomenon at DC, but the random data signal has a flat spectrum from 0 to 4 kHz. Therefore, the transmission signal should not have dc and the frequency of the component must be extremely low. In order to eliminate the above-mentioned DC component 'encoded and unencoded encoders, another specific embodiment of 601 is shown in Fig. 5 and Fig. 6, including forming a frequency spectrum DC suppressor (spectrum shaper) 503. The first 3 symbols of every 4 symbols are generated as usual. The running sum of all previously sent signals remains unchanged. If the running sum is negative, then the 4th symbol Positive values: vice versa. In this way, the DC signal and extremely low frequency components are suppressed. In short, for random input data, the probability of the fourth symbol being positive or negative is the same. Design 'the data rate table is changed as follows. (Please read the note on the back before filling this page) " booking. Order printed by the Ministry of Economic Affairs Central Standard Bureau employee consumer cooperative Μ n 1 distance k data rate (uncoded) 224 7 2 22 58.67 192 6 4 20 56.67 160 5 8 18 54.6.7-· 128 4 16 16 52.67 〆 96 3 32 12 48.67 64 2 64 8 44.67 32 1 128 0 36.67 Table 3 • Data rate with DC suppressor Note that '56 kbps is still possible- 15- 56.67 54.67 52.67 50.67 46.67 42.67 34.67 The size of this paper is suitable for financial standards (CNS) A4 specifications (210; 297 mm) — 455800 A7 --- -M _____ 5. Description of the invention (13) When keeping the same minimum distance M may increase by 16. In particular, every other point can be used for the smallest segment ^ For example, 16 points can be used for the second segment 'while IN points are used for the third to eighth segments and the minimum distance is maintained 4 β In order to use this mapping algorithm, the number of points in each group should be the same. Therefore, divide the total number of M points into m = groups, that is, divide a normal section into 2 groups, and the smallest section with only half points Is 1 group. Then the range of rn is from 1 to 15. 'We use 4 bits to select a point from each group. By using more points, the data rate may increase by 1 or 2 kbps " In the method, the generation function of the odd number m can be: gi = (m-3.) + 3x; hl = l + x + x2 + · · + xm "4; or gi = (m-5) + 2x + 3X9; hl = l + x + x2 + ... + xm'6; The performance rate table is as follows: (Please read the precautions on the back before filling this page} Load M m Distance k Data rate (uncoded) Data rate (coded) 240 15 2 31 59.67. 57.67 224 14 2 30 58.67 56.67 · 208 13 4 29 57.67 55.67 192 12 4 28 56.67 — 54.67 176 11 8 27 55.67 53.67 160 10 8 26 54.67 52.67 144 9 16 25 53.67 51.67 128 8 16 24 52.67 50.67 112 7 32 22 50.67. 48.67 -16- This paper size applies to the Chinese national standard (CNS) A4 size (210X297 mm) 455 800 A7 £ 7 Table 4. Using the data rate of reduced points and referring to FIG. 2 again, the 64 kb / s digital signal passes through the digital network 203, and then is converted into an analog signal by the // law D / A converter 205. The analog signal passes through the analog. Area loop to the 56k modem receiver 201. Printed by the Consumer Standards Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs V. Invention Description (14) 96 6 32 20 48.67 46.67 80 5 64 18 46.67 44.67 64 4 64 16 44.67 42.67 48 3 128 12 40.67 38.67 32 2 128 8 36.67 34,67 16 1 256 0 28.67 26.67 (Please read the notes on the back before filling out this page) Refer to Figure 7 'which shows a half-duplex 56k analog modem receiver 701. The analog signal from the loop in this area is converted to digital by a linear A / D converter. Since it is not easy to properly control the input analog signal level and the sampling timer, a slave law d / a has not been completed at this position. Instead, after the DC, AGC, interpolator, and equalizer are removed, a linear A / D converter is used and the law of martingale is incorporated into the linear converter. At the center of the area, the output of these equalizers should match the linear quantified signal output with the law D / a. The verbalization output is converted from # law to linear conversion to obtain the digital signal sent by the network. After the decoder (the coded version of the Viterbi decoder and the uncoded version of the simple cutter) decodes the received No. 5 D, the segment mapping decoder operates immediately to obtain the k mapping bits. yuan. The intercepted signal bits and spectrum forming bits are removed, and the descrambler operates to obtain user data. -17- This paper size applies to Chinese national standards (CNS > A4 size (210X297 mm) 4 5 5 8 0 0 Α7 ------ ^ ___ V. Description of the invention (15) 涊 Equalizer to 8k The sampling rate works, so for a span of 40 symbols, about 96 taps are needed. Since there is only one regional loop to be equalized, the span can be reduced. However, because the signal spectrum is wider And the channel fails at Dc; and there is attenuation at near 4 kHz, so a relatively long span may be required. To avoid this linear equalizer to enhance noise, a decision feedback equalizer can be used, especially when the When encoding is used, the decoding delay is not eliminated. Figure 8 shows the decision feedback equalizer. Consider Figure 9 'which shows a full-duplex 56k modem 801. The aforementioned 56k transmission foot direction is only from digital to analog. In the other direction, a small basic operation is required. Printed by the Consumers Cooperative of the Central Sample Bureau of the Ministry of Economic Affairs (please read the note on the back before filling out this page) Note that because there is no recent response, you only need— Far Echo Canceller. Cancels this echo if spectrum is used The input of the amplifier is a transmitted signal that does not have a flat spectrum. This may affect the initial convergence speed. The input and input decimation of the echo canceller are both 8k. Because the transmitted signal is only at 8 The bit time is true, so we can use one word for the two samples of the total delay line of the far echo. Therefore, if a round-trip delay is 1.2 seconds, the delay line requires 4.8 words. A similar time span of the V.34 distant echo canceller may require 96 taps. Note that the input delay line also requires 96 words. Automatic gain control (AGO) and demodulation are performed after the echo cancellation. Inside this Before the interpolator, a sampler of 2 is used to increase the sampling rate to 16 kHz. A fundamental frequency 3 sine interpolator can be used for 16k to 3χ converters and the time response at the same time. On the interpolator The subsequent devices and operating methods are -18. This paper wave scale is applicable to the Chinese National Standard (CNS) M specifications (210x297 mm). Printed by the Bayer Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economy. 455800 A7 _B7_ V. Description of the invention ( 16) systemic nature-this One of the disadvantages of the architecture is that demodulation is performed at an 8 kHz sampling rate. This makes the sine / cosine table used for this demodulation very long. To avoid this problem, Figure 10 shows another specific embodiment. After AGC A 1: 3 upsampler is used to increase the sampling rate to 24k. Then a 3x demodulator is used to perform the pass sampling rate conversion and timing recovery, so the sine / cosine table required is shorter. Refer to Figure 11 showing the duplex modem 1101. Note that since the received 56k signal has a sampling timing of 8 kHz and its spectrum is as high as 4 kHz, the output rate of the I. downsampler should be 2 8 kHz . Except for 2400 baud rate, 4x rate is used, 3x rate is acceptable. However, at 2400 baud, fewer taps can be used per eliminator. For example, in the other baud rates, each time the eliminator can use 40 taps, and in the 2400 baud rate, only 30 taps are used, so the total number of taps is still 120. After the echo cancellation and AGC, an up-sampler is used to obtain the mnx sampling rate. The up-sampler is needed because the 8k signal received for this sample has a spectrum that is -4 kHz to 4 kHz wider than the baseband 3x V.34 signal. For the 3rd sine interpolator, m = 3 seems appropriate. A fundamental third sequence sine interpolator can be used for both the mnx to 8k conversion and the timing response. Referring again to Figures 1, 2, and 56k, the symbol rates used by the modems 101, 102 are 8 kHz. The encoder takes the mixed input binary data to generate the output digital symbols. These symbols are sent to the #Law D / A on the other side via a digital network and converted into analog signals. -19- This paper size applies to Chinese National Standard (CNS) A4 (210X297 mm) (Please read the precautions on the back before filling this page)

455800 A7 B7 V. Description of the invention (17) The β-law D / A uses 16 signal segments divided into + 1 + 2, ^ and a total of 255 possible signal levels. The distance between two adjacent points in the first section is 0 * 2 丨 ’丨. For simplicity, let c = i. Zones +1 and -1 have a minimum distance of 1, and 0 is shared by both zones. Furthermore, many encoders / decoders perform poorly at these small points. So 'we often cull these two segments and use the remaining 224 points. However, these small points are still available and will be discussed later. We can also eliminate more small points in order to increase the minimum distance. For every 32 points removed, the minimum distance is doubled, so a noise margin of 6dB_ is obtained. During the correction, the five receivers measure the affection of the channel, such as the SNR and encoder / decoder characteristics, decide how many points should be used, and notify the remote transmitter. Because these points are not powers of two, to make the data rate extremely large, we can use this information frame mapping technique, an example of which is the shell-type mapping formula used in V 34. In a V.34 modem, this shell mapping algorithm maps binary data to one-dimensional QAM signal points. In a 56k modem, all signal points are _dimensional. We can group all the points we want to use into m groups. In the private law encoder / decoder, the one-dimensional point of the distant temple can be divided into 16 sections. Assume that we use these consumer co-operatives printed by the Central Bureau of Standards of the Ministry of Economic Affairs (please read the notes on the back before filling this page) 2 «1 of the 16 sections: ± (9-111), ... 7, ± 8, 111 = 1,2,. '. 7. A natural way to divide these points is to divide the + i and -1 sections into a group with a group index. That is, the outer segment has a smaller group index. Of all the points we have available, it is more commonly used because it has a larger distance from the outer point. This is in contrast to the V.34 shell mapping. In the shell mapping, we often use small dots in order to minimize the average power to obtain several morphological gains. -20- This paper size is applicable to China National Standards (CNS) A4 specifications (2 丨 0X297 mm) 455800 A7 V. Description of the invention (18) ^ " ~~-To achieve this goal, this _ is available at V.34 uses a shell mapping algorithm, but its cost function must be modified. (Please read the notes on the back before filling out this page) The mapping design of the 56k modem is detailed later. In V.34, 'we rank the loops in order from minimum energy to maximum energy 歹 a!' T as a cost function. At this time, we refer to these ring fingers as "sequentially from the largest to the smallest energy. If we can use μ points, 'and send out k-bits in the information frame of N symbols, then the data rate is face to face. The number of points per second is divided into groups with indexes ^ '㈣ and the like to add groups. Wang Yi, K / N representing the number of bits per symbol is not necessarily an integer, even κ. A similar V. 34 The design of the transformation of the information box can be used in the case of non-integer Gen. For simplicity, we can assume that Gen is an integer. We need 5 points in order to select-points in the group, so we use k = K_5N bits for this group. Group mapping,, similar to V.34 shell mapping. The maximum possible value of k can be obtained from the following inequality:

2kSmN For example, if we use 192 points and the size of the mapping box is 8 ′ then m = 6 ’and k 20.68. Suppose we use k = 2〇, we will send 7.5 bits 7G per symbol, and its data rate will be (k + 5N) * 800 / N = 600kbps. The policy of Zhengong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs is similar to the shell mapping of V.34. We do not need to use the maximum possible value of k. In other words, we do not need to use the smallest possible value of m for the same k. In the V.34 example, if we use a larger M, we get several morphological gains. Here, for larger m, we must use several smaller points' to reduce the minimum distance. However, since these minimum points are extremely important points', a larger m is generally not used. -21-1 This paper size is applicable to China National Building Standard (CNS M4 specification (2I0X297)) 455800 A7 B7 19. V. Description of the invention (If N = 8 is assumed, the segment mapping algorithm can be similar to the fork " The key is to choose a suitable cost function so that you can read the “Notes” and fill it in again. ”This page allows the performance to reach the optimal level, while maintaining the memory and complexity. Lower degree. In the shell mapping algorithm used in V.34, the selected cost function is a ramp function, that is, the cost is equal to the ring index. This cost function can be expressed by a production function : Gl = l + X + x2 + x3 + X4 + x5 + ... xn'-I „Ordered by the Consumer Affairs Cooperative of the Standards Bureau of the Ministry of Economic Affairs, because the average power in a ring is approximately proportional to the index of the ring, So this is the best cost function that can minimize the average power. If we use the same cost function here, we will get more external groups than internal groups, but the performance is not the most good, For example, for the sake of simplicity, we set N = 2 and m = 7, and the cost of group 丨 is I, i = 0; 1, ... 6. Then use one symbol in group 0 and another in group 6 Sign, it will produce a lower total cost (0 + 6 = 6) than the total cost (4 + 4 = 8) when two symbols are used in group 4. Since the former does not use the smallest point, and The smallest point is prone to errors, so the former is used more often. The slope cost function can minimize the average group index, but it cannot minimize the error rate, because the error probability of different groups is very different. Total error The probability P is P = P〇f〇 + P! Fl + ... + Pm.1fm.l where 'Pi is the probability of error of the i-th group point and fi is the relative frequency (probability) of the i-th group i = 0 , Ls .., ml. Because ρ; (ΐ = ί), 1, ... π-2) is smaller than Pm_i, the total error probability is approximately equal to l-a better criterion here is to set the most The frequency of the group is minimized. This can be achieved by using the following cost function-22- This paper size applies to the Chinese National Standard (CNS) A4 specification (2ΙΟχ: 297 male thin) 4 5 5 80 0 A7 B7 Five invention descriptions (2 Item 0: (0, l, 2, '.., m-2,8 (m-2) + l) The generation function of m = 7 is: ί Please read the notes on the back before filling in this page ] Gl = l + X + X2 + X3 + X4 + x5 + x41. In this example, any one in its smallest group is at least / 胥 —point eight

The cost is higher than that of all octets without any points. Therefore, the order of octets in this cost function is as follows. Text (〇SG), (lSG), (2SG), ... (8SG) (commit: the smallest group) Because 2k is usually less than mN, so we usually do not m. Smaller groups can be used to eliminate this error rate. ^ H is reduced and the disadvantage of the cost function is that the ordering of the generating function is too high, its table and high complexity. To avoid these problems ’suggestion == a two-step mapping algorithm. When using this new algorithm, the complexity is similar to the mapping algorithm based on the slope cost function, but it can achieve the same effect, as described above. Printed by the Central Standards Bureau of the Ministry of Economic Affairs of the Bayer Consumer Cooperative, because the minimum frequency of these smallest points is a rule, we can use a very simple generation function to express it: gi = (ml) + x In this example +, the cost The function is 〇, 〇A ..., M, except for this minimum group, we don't have to worry about using that group. Using this cost function, its cost is not D θu p, t & let + /, 疋 the index of the group. For example, cost 0 means group 0 to m-ι, and cost t means group 1 with minimum distance. The problem is at cost 0 ’how can we say to use that group. To solve -23 "this paper size _ 中国 iii 准 (CNS) Α4 · _ (―210 ^^ · 4 5 5 80 0 Α7 Β7 V. Description of the invention (21) To solve this problem, we can generate the function in one order: g 'l = l + X + X2 + ... + xm-2 g, 2 = g, l * g'l g, 4 = g'2 * g'2 ~ Z, 4 g'8 = g'4 * g'4—Z, 8 is similar to the V.34 shell mapping algorithm. 'We first search through the first table that is currently only 8 degrees long in order to find the total cost of the 8 symbols. If the cost is 0, go to the secondary generation function, search through the Z, 8 table, and perform other mappings in the same way as the V.34 shell mapping to obtain from 0 to m- 8 group indicators of 2. Printed by the Shellfish Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs (please read the precautions on the back before filling this page). If the total cost of the 8 symbols is not 0, use the following table to The cost of splitting the 8 symbols becomes the cost of 2 4 symbols. If one of the 4 symbols' cost is 0, go to the secondary generation function and use the rest to find out the ZU (or simply use the table) 4 symbols of the secondary Cost 'and finally find 4 group indicators from 0 to m-2. Similarly, if one of the costs of the 4 symbols is 0, go to the secondary generation function to calculate the secondary The cost of two symbols, and finally calculated the two group indicators from 0 to m-2. Note that the tables of the first group and the second group are small ^ The size of the table is about the cost of the slope of V.34 type The function table is the same β and its complexity is also similar. We 従 group map to get 8 group indicators (from 0 to m-1). The indicator i means the substantial section + (8-i) and-(δ- i), after a total of 32 points, use 5 bits to pick one of the 32 points. The data rate is (k + 40) kbps. The data rates of various M are listed in the following table. -24- This paper scale GM China National Standard (CNS > Α4 size (210 × 297 mm) 4 Μ Printed by the Bayer Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs 5 5 80 0 A7-____ £ 7 V. Description of the invention (22) M m k data rate (kbps) fm. 1 (new method) in-] (slope) fm.! (new method) U slope) 224 7 2 22 62 0.0831 0.105 0.791 192 6 4 20 60 Γ 0,0824 0.1123 0.734 160 5 8 18 1 58 1 0.125 0.1436 0.87 128 4 16 16 56 '0.25 0.25. 1.0 96 3 32 12 52 0.234 0.2448 0.956 64 2 64 8 48 0.5' 0.5 1.0 32 1 128 0 40 1.0 1.0 1.0 Table 5. Data rate Use the new algorithm and the original V 34 mapping with the slope cost function for the relative frequency of the smallest group in the M value table. The proportions are also listed in the table. We know that for practically useful m-values (relatively large), the heart is reduced by more than 20% (and the error rate is the same). For small m, the new algorithm can only / have any other method to help because we have to run out of its full number 'or almost all available combinations of eight symbols. But these examples do have some minor benefits. From the table, we know that if m decreases by i, k decreases by 2 and the data rate decreases by 2 kbps. A decrease of m means that the minimum distance is doubled, giving us a noise margin of about 6dB or more. Therefore, if any coding technology is used to improve the performance, a coding gain of more than 6dB can be obtained at the cost of 1/4 or less bits / symbol. "When maintaining the same minimum distance, you can also use] VI Increase by 16. In particular, we can use each of the groups smaller than the group we use separately — -25-Applicable Standard (CNS) (210X297 mm) for this paper size " ------- (Please first Please read the notes on the back of the page and fill in this page again.] • Crack. -Ι1Τ 45 580 0 A7 ______J37 V. Description of the invention (23) points. For example, we can use the 16 points of the 6th group to use the 0th to 5th groups at the same time. 192 points, and maintain a minimum distance of 4. In order to use the group mapping algorithm, the number of points in each group must be the same. Therefore, we can divide the total number of M points. Group, and M is from 1 to 15. We use 4 bits to select a point from each button. With more points: we can increase the data rate, generally 1 or 2 ribs 1 ^. In this gorge In the projection algorithm, the generation function of the odd number m is: gi = (m-3) + 3x The secondary generation function of cost 0 is: .g, = l + x + x2 + ". + Xm · 4 the cost of 1 The generation function is: g " I = l + X + X2 For even number M, we can use g1 = (m_2) + 2x = 2 [(m / 2-l) + x], and the result is the same as our previous use. The same. 'The data rate is (k +32) kbps, as listed in the table below. (Please read the notes on the back, and then fill in ^ this ear.) Central Bureau of Standards, Ministry of Economic Affairs, Industrial and Consumer Cooperation Du printed -26- This paper size applies Chinese national standards ( CNS) A4 specification (210X297) 4 5 5 8 0 〇 Α7 _β7 24. V. Description of the invention (m) Minimum distance k Data rate (kbps) fmin new fmin slope fmin new / fmin slope 240 15 2 31 63 0.1560 0.1676 0.9308 224 14 2 30 62 0.0831 0.1055 0.7877 208 13 4 29 61 0.1476 0.1668 0.8849 192 12 4 28 60 0.0824 0.1125 0.7324 176 11 8 27 59 0.1729 0.1964 0.8803 160 10 8 26 58 0.1250 0.1442 0.8669 144 9 16 25 57 _0.2685 0.2783 0.9648 1? 8 8 16 24 56 0.25 0.25 1 112 7 32 22 54 0.3485 0.3575 0.9748 Q6 6 32 20 52 0.2340 0.2470 0.9474 80 5 64 18 50 0.5061 0.5261 0.9731 '64 4 64 __ 16 48 0.5 0.5 1 48 3 128 12 44 1 1 1 32 9 128 8 40 1 1 1 16 1 256 0 32 1 1 1 ---- 'J Pack-(Please read the precautions on the back before filling this page)-Ding Staff of Central Standards Bureau, Ministry of Economic Affairs Consumption cooperatives print m Table 6. Data rate including fmin This table also lists fmin, has The probability of each point of the ramp function and the cost function cost V.34 shell mapping for a small distance of the proposed use of, and the proportions thereof. It can be seen that when 16 points are added ', although the minimum distance has not changed, the number of such minimum points has increased by 50%', so the probability of using these minimum points also generally increases, as does the probability of errors. Therefore, the increase of the data rate is not entirely without the cost of performance. -27- This paper size applies to China National Standard {CNS) A4 specification (210X297 mm > * 1εξ- ~ £ · β 455800 A7 JB7 V. Description of the invention (25) Symbol description of graphic components 100 Personal Computer (PC) 101 Modem 102 Device 103 Public Switched Telephone Network (PSTN) 104 Two-step Mapping Device-105 System Software 106 Modem Software 107 Central Processing Unit (CPU) 108 Instruction Bus 201 Analog Modem 203 Digital Network 205 / z Law Digital-to-Analog Converter (D / A) 301 Encoder 303 Section Mapper 304 Interleaved Encoder 305 Bit Mapping 401 Encoder Central Standard of the Ministry of Economy Please fill in this page again) 501 encoder 503 spectrum generator 601 encoder 701 56k analog modem receiver 901 56k modem 1101 duplex modem-28- This paper music standard is applicable to China National Standard (CNS) A4 specification (210X297 male) 4) 5 5 8 0 0 A7-B7 V. Description of the invention (26. In addition, due to the large m and k values, although the edges are reasonable, the size and complexity of the table also increase. The calculation of the data rates listed below does not take into account the reduction due to other reasons, such as the intercepted signal bits. These bits consume 1-33 kbps of the data rate (please read the precautions in the back first) (This page)-Binding Central Standard of the Ministry of Economics, printed by employee Xiaoxian Cooperative Co., Ltd. This paper size applies to Chinese National Standard (CNS) A4 (2 丨 〇 X 297 mm)

Claims (1)

4 Printed by the Employees' Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs 5 80 0 as C8 D8 6. Scope of patent application: 1. A modem device for transmitting and receiving data via a network, including: A two-step mapper with a set shape gain average power. 2. A data communication system for data communication via a personal computer and using a network as a medium, comprising: at least two network media devices connected in spaced positions; one of the two devices includes a personal computer, the computer With ...-a digital modem for digital data communication via the network medium, the digital modem includes a two-step mapper that must reduce a set morphological gain average power, and the other device of the two devices has- An analog modem used to receive analog data from the network medium. 3. — A method for transmitting digital data through a network, including the following steps: Map digital data by comparing distances, and skip these data points that are close to 0, and reduce—set the average of the form gain power. -30- This paper is _ ^? ^ Degree universal Chinese national kneading rate (€ yang). 8 4 ^ grid (210 乂 297mm) (please read the precautions on the back before filling this page) Order.