v17rx.h

Go to the documentation of this file.
00001 /*
00002  * SpanDSP - a series of DSP components for telephony
00003  *
00004  * v17rx.h - ITU V.17 modem receive part
00005  *
00006  * Written by Steve Underwood <steveu@coppice.org>
00007  *
00008  * Copyright (C) 2003 Steve Underwood
00009  *
00010  * All rights reserved.
00011  *
00012  * This program is free software; you can redistribute it and/or modify
00013  * it under the terms of the GNU Lesser General Public License version 2.1,
00014  * as published by the Free Software Foundation.
00015  *
00016  * This program is distributed in the hope that it will be useful,
00017  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00018  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
00019  * GNU Lesser General Public License for more details.
00020  *
00021  * You should have received a copy of the GNU Lesser General Public
00022  * License along with this program; if not, write to the Free Software
00023  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
00024  *
00025  * $Id: v17rx.h,v 1.63 2009/04/12 04:20:01 steveu Exp $
00026  */
00027 
00028 /*! \file */
00029 
00030 #if !defined(_SPANDSP_V17RX_H_)
00031 #define _SPANDSP_V17RX_H_
00032 
00033 /*! \page v17rx_page The V.17 receiver
00034 \section v17rx_page_sec_1 What does it do?
00035 The V.17 receiver implements the receive side of a V.17 modem. This can operate
00036 at data rates of 14400, 12000, 9600 and 7200 bits/second. The audio input is a stream
00037 of 16 bit samples, at 8000 samples/second. The transmit and receive side of V.17
00038 modems operate independantly. V.17 is mostly used for FAX transmission over PSTN
00039 lines, where it provides the standard 14400 bits/second rate. 
00040 
00041 \section v17rx_page_sec_2 How does it work?
00042 V.17 uses QAM modulation, at 2400 baud, and trellis coding. Constellations with
00043 16, 32, 64, and 128 points are defined. After one bit per baud is absorbed by the
00044 trellis coding, this gives usable bit rates of 7200, 9600, 12000, and 14400 per
00045 second.
00046 
00047 V.17 specifies a training sequence at the start of transmission, which makes the
00048 design of a V.17 receiver relatively straightforward. The first stage of the
00049 training sequence consists of 256
00050 symbols, alternating between two constellation positions. The receiver monitors
00051 the signal power, to sense the possible presence of a valid carrier. When the
00052 alternating signal begins, the power rising above a minimum threshold (-43dBm0)
00053 causes the main receiver computation to begin. The initial measured power is
00054 used to quickly set the gain of the receiver. After this initial settling, the
00055 front end gain is locked, and the adaptive equalizer tracks any subsequent
00056 signal level variation. The signal is oversampled to 24000 samples/second (i.e.
00057 signal, zero, zero, signal, zero, zero, ...) and fed to a complex root raised
00058 cosine pulse shaping filter. This filter has been modified from the conventional
00059 root raised cosine filter, by shifting it up the band, to be centred at the nominal
00060 carrier frequency. This filter interpolates the samples, pulse shapes, and performs
00061 a fractional sample delay at the same time. 192 sets of filter coefficients are used
00062 to achieve a set of finely spaces fractional sample delays, between zero and
00063 one sample. By choosing every fifth sample, and the appropriate set of filter
00064 coefficients, the properly tuned symbol tracker can select data samples at 4800
00065 samples/second from points within 0.28 degrees of the centre and mid-points of
00066 each symbol. The output of the filter is multiplied by a complex carrier, generated
00067 by a DDS. The result is a baseband signal, requiring no further filtering, apart from
00068 an adaptive equalizer. The baseband signal is fed to a T/2 adaptive equalizer.
00069 A band edge component maximisation algorithm is used to tune the sampling, so the samples
00070 fed to the equalizer are close to the mid point and edges of each symbol. Initially
00071 the algorithm is very lightly damped, to ensure the symbol alignment pulls in
00072 quickly. Because the sampling rate will not be precisely the same as the
00073 transmitter's (the spec. says the symbol timing should be within 0.01%), the
00074 receiver constantly evaluates and corrects this sampling throughout its
00075 operation. During the symbol timing maintainence phase, the algorithm uses
00076 a heavier damping.
00077 
00078 The carrier is specified as 1800Hz +- 1Hz at the transmitter, and 1800 +-7Hz at
00079 the receiver. The receive carrier would only be this inaccurate if the link
00080 includes FDM sections. These are being phased out, but the design must still
00081 allow for the worst case. Using an initial 1800Hz signal for demodulation gives
00082 a worst case rotation rate for the constellation of about one degree per symbol.
00083 Once the symbol timing synchronisation algorithm has been given time to lock to the
00084 symbol timing of the initial alternating pattern, the phase of the demodulated signal
00085 is recorded on two successive symbols - once for each of the constellation positions.
00086 The receiver then tracks the symbol alternations, until a large phase jump occurs.
00087 This signifies the start of the next phase of the training sequence. At this
00088 point the total phase shift between the original recorded symbol phase, and the
00089 symbol phase just before the phase jump occurred is used to provide a coarse
00090 estimation of the rotation rate of the constellation, and it current absolute
00091 angle of rotation. These are used to update the current carrier phase and phase
00092 update rate in the carrier DDS. The working data already in the pulse shaping
00093 filter and equalizer buffers is given a similar step rotation to pull it all
00094 into line. From this point on, a heavily damped integrate and dump approach,
00095 based on the angular difference between each received constellation position and
00096 its expected position, is sufficient to track the carrier, and maintain phase
00097 alignment. A fast rough approximator for the arc-tangent function is adequate
00098 for the estimation of the angular error. 
00099 
00100 The next phase of the training sequence is a scrambled sequence of two
00101 particular symbols. We train the T/2 adaptive equalizer using this sequence. The
00102 scrambling makes the signal sufficiently diverse to ensure the equalizer
00103 converges to the proper generalised solution. At the end of this sequence, the
00104 equalizer should be sufficiently well adapted that is can correctly resolve the
00105 full QAM constellation. However, the equalizer continues to adapt throughout
00106 operation of the modem, fine tuning on the more complex data patterns of the
00107 full QAM constellation. 
00108 
00109 In the last phase of the training sequence, the modem enters normal data
00110 operation, with a short defined period of all ones as data. As in most high
00111 speed modems, data in a V.17 modem passes through a scrambler, to whiten the
00112 spectrum of the signal. The transmitter should initialise its data scrambler,
00113 and pass the ones through it. At the end of the ones, real data begins to pass
00114 through the scrambler, and the transmit modem is in normal operation. The
00115 receiver tests that ones are really received, in order to verify the modem
00116 trained correctly. If all is well, the data following the ones is fed to the
00117 application, and the receive modem is up and running. Unfortunately, some
00118 transmit side of some real V.17 modems fail to initialise their scrambler before
00119 sending the ones. This means the first 23 received bits (the length of the
00120 scrambler register) cannot be trusted for the test. The receive modem,
00121 therefore, only tests that bits starting at bit 24 are really ones.
00122 
00123 The V.17 signal is trellis coded. Two bits of each symbol are convolutionally coded
00124 to form a 3 bit trellis code - the two original bits, plus an extra redundant bit. It
00125 is possible to ignore the trellis coding, and just decode the non-redundant bits.
00126 However, the noise performance of the receiver would suffer. Using a proper
00127 trellis decoder adds several dB to the noise tolerance to the receiving modem. Trellis
00128 coding seems quite complex at first sight, but is fairly straightforward once you
00129 get to grips with it.
00130 
00131 Trellis decoding tracks the data in terms of the possible states of the convolutional
00132 coder at the transmitter. There are 8 possible states of the V.17 coder. The first
00133 step in trellis decoding is to find the best candidate constellation point
00134 for each of these 8 states. One of thse will be our final answer. The constellation
00135 has been designed so groups of 8 are spread fairly evenly across it. Locating them
00136 is achieved is a reasonably fast manner, by looking up the answers in a set of space
00137 map tables. The disadvantage is the tables are potentially large enough to affect
00138 cache performance. The trellis decoder works over 16 successive symbols. The result
00139 of decoding is not known until 16 symbols after the data enters the decoder. The
00140 minimum total accumulated mismatch between each received point and the actual
00141 constellation (termed the distance) is assessed for each of the 8 states. A little
00142 analysis of the coder shows that each of the 8 current states could be arrived at
00143 from 4 different previous states, through 4 different constellation bit patterns.
00144 For each new state, the running total distance is arrived at by inspecting a previous
00145 total plus a new distance for the appropriate 4 previous states. The minimum of the 4
00146 values becomes the new distance for the state. Clearly, a mechanism is needed to stop
00147 this distance from growing indefinitely. A sliding window, and several other schemes
00148 are possible. However, a simple single pole IIR is very simple, and provides adequate
00149 results.
00150 
00151 For each new state we store the constellation bit pattern, or path, to that state, and
00152 the number of the previous state. We find the minimum distance amongst the 8 new
00153 states for each new symbol. We then trace back through the states, until we reach the
00154 one 16 states ago which leads to the current minimum distance. The bit pattern stored
00155 there is the error corrected bit pattern for that symbol.
00156 
00157 So, what does Trellis coding actually achieve? TCM is easier to understand by looking
00158 at the V.23bis modem spec. The V.32bis spec. is very similar to V.17, except that it
00159 is a full duplex modem and has non-TCM options, as well as the TCM ones in V.17.
00160 
00161 V32bis defines two options for pumping 9600 bits per second down a phone line - one
00162 with and one without TCM. Both run at 2400 baud. The non-TCM one uses simple 16 point
00163 QAM on the raw data. The other takes two out of every four raw bits, and convolutionally
00164 encodes them to 3. Now we have 5 bits per symbol, and we need 32 point QAM to send the
00165 data.
00166 
00167 The raw error rate from simple decoding of the 32 point QAM is horrible compared to
00168 decoding the 16 point QAM. If a point decoded from the 32 point QAM is wrong, the likely
00169 correct choice should be one of the adjacent ones. It is unlikely to have been one that
00170 is far away across the constellation, unless there was a huge noise spike, interference,
00171 or something equally nasty. Now, the 32 point symbols do not exist in isolation. There
00172 was a kind of temporal smearing in the convolutional coding. It created a well defined
00173 dependency between successive symbols. If we knew for sure what the last few symbols
00174 were, they would lead us to a limited group of possible values for the current symbol,
00175 constrained by the behaviour of the convolutional coder. If you look at how the symbols
00176 were mapped to constellation points, you will see the mapping tries to spread those
00177 possible symbols as far apart as possible. This will leave only one that is pretty
00178 close to the received point, which must be the correct choice. However, this assumes
00179 we know the last few symbols for sure. Since we don't, we have a bit more work to do
00180 to achieve reliable decoding.
00181 
00182 Instead of decoding to the nearest point on the constellation, we decode to a group of
00183 likely constellation points in the neighbourhood of the received point. We record the
00184 mismatch for each - that is the distance across the constellation between the received
00185 point and the group of nearby points. To avoid square roots, recording x2 + y2 can be
00186 good enough. Symbol by symbol, we record this information. After a few symbols we can
00187 stand back and look at the recorded information.
00188 
00189 For each symbol we have a set of possible symbol values and error metric pairs. The
00190 dependency between symbols, created by the convolutional coder, means some paths from
00191 symbol to symbol are possible and some are not. It we trace back through the possible
00192 symbol to symbol paths, and total up the error metric through those paths, we end up
00193 with a set of figures of merit (or more accurately figures of demerit, since
00194 larger == worse) for the likelihood of each path being the correct one. The path with
00195 the lowest total metric is the most likely, and gives us our final choice for what we
00196 think the current symbol really is.
00197 
00198 That was hard work. It takes considerable computation to do this selection and traceback,
00199 symbol by symbol. We need to get quite a lot from this. It needs to drive the error rate
00200 down so far that is compensates for the much higher error rate due to the larger
00201 constellation, and then buys us some actual benefit. Well in the example we are looking
00202 at - V.32bis at 9600bps - it works out the error rate from the TCM option is like using
00203 the non-TCM option with several dB more signal to noise ratio. That's nice. The non-TCM
00204 option is pretty reasonable on most phone lines, but a better error rate is always a
00205 good thing. However, V32bis includes a 14,400bps option. That uses 2400 baud, and 6 bit
00206 symbols. Convolutional encoding increases that to 7 bits per symbol, by taking 2 bits and
00207 encoding them to 3. This give a 128 point QAM constellation. Again, the difference between
00208 using this, and using just an uncoded 64 point constellation is equivalent to maybe 5dB of
00209 extra signal to noise ratio. However, in this case it is the difference between the modem
00210 working only on the most optimal lines, and being widely usable across most phone lines.
00211 TCM absolutely transformed the phone line modem business.
00212 */
00213 
00214 /* Target length for the equalizer is about 63 taps, to deal with the worst stuff
00215    in V.56bis. */
00216 /*! Samples before the target position in the equalizer buffer */
00217 #define V17_EQUALIZER_PRE_LEN       8
00218 /*! Samples after the target position in the equalizer buffer */
00219 #define V17_EQUALIZER_POST_LEN      8
00220 
00221 /*! The number of taps in the pulse shaping/bandpass filter */
00222 #define V17_RX_FILTER_STEPS         27
00223 
00224 /* We can store more trellis depth that we look back over, so that we can push out a group
00225    of symbols in one go, giving greater processing efficiency, at the expense of a bit more
00226    latency through the modem. */
00227 /* Right now we don't take advantage of this optimisation. */
00228 /*! The depth of the trellis buffer */
00229 #define V17_TRELLIS_STORAGE_DEPTH   16
00230 /*! How far we look back into history for trellis decisions */
00231 #define V17_TRELLIS_LOOKBACK_DEPTH  16
00232 
00233 /*!
00234     V.17 modem receive side descriptor. This defines the working state for a
00235     single instance of a V.17 modem receiver.
00236 */
00237 typedef struct v17_rx_state_s v17_rx_state_t;
00238 
00239 #if defined(__cplusplus)
00240 extern "C"
00241 {
00242 #endif
00243 
00244 /*! Initialise a V.17 modem receive context.
00245     \brief Initialise a V.17 modem receive context.
00246     \param s The modem context.
00247     \param bit_rate The bit rate of the modem. Valid values are 7200, 9600, 12000 and 14400.
00248     \param put_bit The callback routine used to put the received data.
00249     \param user_data An opaque pointer passed to the put_bit routine.
00250     \return A pointer to the modem context, or NULL if there was a problem. */
00251 SPAN_DECLARE(v17_rx_state_t *) v17_rx_init(v17_rx_state_t *s, int bit_rate, put_bit_func_t put_bit, void *user_data);
00252 
00253 /*! Reinitialise an existing V.17 modem receive context.
00254     \brief Reinitialise an existing V.17 modem receive context.
00255     \param s The modem context.
00256     \param bit_rate The bit rate of the modem. Valid values are 7200, 9600, 12000 and 14400.
00257     \param short_train TRUE if a short training sequence is expected.
00258     \return 0 for OK, -1 for bad parameter */
00259 SPAN_DECLARE(int) v17_rx_restart(v17_rx_state_t *s, int bit_rate, int short_train);
00260 
00261 /*! Release a V.17 modem receive context.
00262     \brief Release a V.17 modem receive context.
00263     \param s The modem context.
00264     \return 0 for OK */
00265 SPAN_DECLARE(int) v17_rx_release(v17_rx_state_t *s);
00266 
00267 /*! Free a V.17 modem receive context.
00268     \brief Free a V.17 modem receive context.
00269     \param s The modem context.
00270     \return 0 for OK */
00271 SPAN_DECLARE(int) v17_rx_free(v17_rx_state_t *s);
00272 
00273 /*! Get the logging context associated with a V.17 modem receive context.
00274     \brief Get the logging context associated with a V.17 modem receive context.
00275     \param s The modem context.
00276     \return A pointer to the logging context */
00277 SPAN_DECLARE(logging_state_t *) v17_rx_get_logging_state(v17_rx_state_t *s);
00278 
00279 /*! Change the put_bit function associated with a V.17 modem receive context.
00280     \brief Change the put_bit function associated with a V.17 modem receive context.
00281     \param s The modem context.
00282     \param put_bit The callback routine used to handle received bits.
00283     \param user_data An opaque pointer. */
00284 SPAN_DECLARE(void) v17_rx_set_put_bit(v17_rx_state_t *s, put_bit_func_t put_bit, void *user_data);
00285 
00286 /*! Change the modem status report function associated with a V.17 modem receive context.
00287     \brief Change the modem status report function associated with a V.17 modem receive context.
00288     \param s The modem context.
00289     \param handler The callback routine used to report modem status changes.
00290     \param user_data An opaque pointer. */
00291 SPAN_DECLARE(void) v17_rx_set_modem_status_handler(v17_rx_state_t *s, modem_rx_status_func_t handler, void *user_data);
00292 
00293 /*! Process a block of received V.17 modem audio samples.
00294     \brief Process a block of received V.17 modem audio samples.
00295     \param s The modem context.
00296     \param amp The audio sample buffer.
00297     \param len The number of samples in the buffer.
00298     \return The number of samples unprocessed.
00299 */
00300 SPAN_DECLARE(int) v17_rx(v17_rx_state_t *s, const int16_t amp[], int len);
00301 
00302 /*! Fake processing of a missing block of received V.17 modem audio samples.
00303     (e.g due to packet loss).
00304     \brief Fake processing of a missing block of received V.17 modem audio samples.
00305     \param s The modem context.
00306     \param len The number of samples to fake.
00307     \return The number of samples unprocessed.
00308 */
00309 SPAN_DECLARE(int) v17_rx_fillin(v17_rx_state_t *s, int len);
00310 
00311 /*! Get a snapshot of the current equalizer coefficients.
00312     \brief Get a snapshot of the current equalizer coefficients.
00313     \param s The modem context.
00314     \param coeffs The vector of complex coefficients.
00315     \return The number of coefficients in the vector. */
00316 #if defined(SPANDSP_USE_FIXED_POINTx)
00317 SPAN_DECLARE(int) v17_rx_equalizer_state(v17_rx_state_t *s, complexi_t **coeffs);
00318 #else
00319 SPAN_DECLARE(int) v17_rx_equalizer_state(v17_rx_state_t *s, complexf_t **coeffs);
00320 #endif
00321 
00322 /*! Get the current received carrier frequency.
00323     \param s The modem context.
00324     \return The frequency, in Hertz. */
00325 SPAN_DECLARE(float) v17_rx_carrier_frequency(v17_rx_state_t *s);
00326 
00327 /*! Get the current symbol timing correction since startup.
00328     \param s The modem context.
00329     \return The correction. */
00330 SPAN_DECLARE(float) v17_rx_symbol_timing_correction(v17_rx_state_t *s);
00331 
00332 /*! Get a current received signal power.
00333     \param s The modem context.
00334     \return The signal power, in dBm0. */
00335 SPAN_DECLARE(float) v17_rx_signal_power(v17_rx_state_t *s);
00336 
00337 /*! Set the power level at which the carrier detection will cut in
00338     \param s The modem context.
00339     \param cutoff The signal cutoff power, in dBm0. */
00340 SPAN_DECLARE(void) v17_rx_signal_cutoff(v17_rx_state_t *s, float cutoff);
00341 
00342 /*! Set a handler routine to process QAM status reports
00343     \param s The modem context.
00344     \param handler The handler routine.
00345     \param user_data An opaque pointer passed to the handler routine. */
00346 SPAN_DECLARE(void) v17_rx_set_qam_report_handler(v17_rx_state_t *s, qam_report_handler_t handler, void *user_data);
00347 
00348 #if defined(__cplusplus)
00349 }
00350 #endif
00351 
00352 #endif
00353 /*- End of file ------------------------------------------------------------*/

Generated by  doxygen 1.6.2