1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * SpanDSP - a series of DSP components for telephony
4 *
5 * echo.c - A line echo canceller.  This code is being developed
6 *          against and partially complies with G168.
7 *
8 * Written by Steve Underwood <steveu@coppice.org>
9 *         and David Rowe <david_at_rowetel_dot_com>
10 *
11 * Copyright (C) 2001 Steve Underwood and 2007 David Rowe
12 *
13 * All rights reserved.
14 */
15
16#ifndef __ECHO_H
17#define __ECHO_H
18
19/*
20Line echo cancellation for voice
21
22What does it do?
23
24This module aims to provide G.168-2002 compliant echo cancellation, to remove
25electrical echoes (e.g. from 2-4 wire hybrids) from voice calls.
26
27How does it work?
28
29The heart of the echo cancellor is FIR filter. This is adapted to match the
30echo impulse response of the telephone line. It must be long enough to
31adequately cover the duration of that impulse response. The signal transmitted
32to the telephone line is passed through the FIR filter. Once the FIR is
33properly adapted, the resulting output is an estimate of the echo signal
34received from the line. This is subtracted from the received signal. The result
35is an estimate of the signal which originated at the far end of the line, free
36from echos of our own transmitted signal.
37
38The least mean squares (LMS) algorithm is attributed to Widrow and Hoff, and
39was introduced in 1960. It is the commonest form of filter adaption used in
40things like modem line equalisers and line echo cancellers. There it works very
41well.  However, it only works well for signals of constant amplitude. It works
42very poorly for things like speech echo cancellation, where the signal level
43varies widely.  This is quite easy to fix. If the signal level is normalised -
44similar to applying AGC - LMS can work as well for a signal of varying
45amplitude as it does for a modem signal. This normalised least mean squares
46(NLMS) algorithm is the commonest one used for speech echo cancellation. Many
47other algorithms exist - e.g. RLS (essentially the same as Kalman filtering),
48FAP, etc. Some perform significantly better than NLMS.  However, factors such
49as computational complexity and patents favour the use of NLMS.
50
51A simple refinement to NLMS can improve its performance with speech. NLMS tends
52to adapt best to the strongest parts of a signal. If the signal is white noise,
53the NLMS algorithm works very well. However, speech has more low frequency than
54high frequency content. Pre-whitening (i.e. filtering the signal to flatten its
55spectrum) the echo signal improves the adapt rate for speech, and ensures the
56final residual signal is not heavily biased towards high frequencies. A very
57low complexity filter is adequate for this, so pre-whitening adds little to the
58compute requirements of the echo canceller.
59
60An FIR filter adapted using pre-whitened NLMS performs well, provided certain
61conditions are met:
62
63    - The transmitted signal has poor self-correlation.
64    - There is no signal being generated within the environment being
65      cancelled.
66
67The difficulty is that neither of these can be guaranteed.
68
69If the adaption is performed while transmitting noise (or something fairly
70noise like, such as voice) the adaption works very well. If the adaption is
71performed while transmitting something highly correlative (typically narrow
72band energy such as signalling tones or DTMF), the adaption can go seriously
73wrong. The reason is there is only one solution for the adaption on a near
74random signal - the impulse response of the line. For a repetitive signal,
75there are any number of solutions which converge the adaption, and nothing
76guides the adaption to choose the generalised one. Allowing an untrained
77canceller to converge on this kind of narrowband energy probably a good thing,
78since at least it cancels the tones. Allowing a well converged canceller to
79continue converging on such energy is just a way to ruin its generalised
80adaption. A narrowband detector is needed, so adapation can be suspended at
81appropriate times.
82
83The adaption process is based on trying to eliminate the received signal. When
84there is any signal from within the environment being cancelled it may upset
85the adaption process. Similarly, if the signal we are transmitting is small,
86noise may dominate and disturb the adaption process. If we can ensure that the
87adaption is only performed when we are transmitting a significant signal level,
88and the environment is not, things will be OK. Clearly, it is easy to tell when
89we are sending a significant signal. Telling, if the environment is generating
90a significant signal, and doing it with sufficient speed that the adaption will
91not have diverged too much more we stop it, is a little harder.
92
93The key problem in detecting when the environment is sourcing significant
94energy is that we must do this very quickly. Given a reasonably long sample of
95the received signal, there are a number of strategies which may be used to
96assess whether that signal contains a strong far end component. However, by the
97time that assessment is complete the far end signal will have already caused
98major mis-convergence in the adaption process. An assessment algorithm is
99needed which produces a fairly accurate result from a very short burst of far
100end energy.
101
102How do I use it?
103
104The echo cancellor processes both the transmit and receive streams sample by
105sample. The processing function is not declared inline. Unfortunately,
106cancellation requires many operations per sample, so the call overhead is only
107a minor burden.
108*/
109
110#include "fir.h"
111#include "oslec.h"
112
113/*
114    G.168 echo canceller descriptor. This defines the working state for a line
115    echo canceller.
116*/
117struct oslec_state {
118	int16_t tx;
119	int16_t rx;
120	int16_t clean;
121	int16_t clean_nlp;
122
123	int nonupdate_dwell;
124	int curr_pos;
125	int taps;
126	int log2taps;
127	int adaption_mode;
128
129	int cond_met;
130	int32_t pstates;
131	int16_t adapt;
132	int32_t factor;
133	int16_t shift;
134
135	/* Average levels and averaging filter states */
136	int ltxacc;
137	int lrxacc;
138	int lcleanacc;
139	int lclean_bgacc;
140	int ltx;
141	int lrx;
142	int lclean;
143	int lclean_bg;
144	int lbgn;
145	int lbgn_acc;
146	int lbgn_upper;
147	int lbgn_upper_acc;
148
149	/* foreground and background filter states */
150	struct fir16_state_t fir_state;
151	struct fir16_state_t fir_state_bg;
152	int16_t *fir_taps16[2];
153
154	/* DC blocking filter states */
155	int tx_1;
156	int tx_2;
157	int rx_1;
158	int rx_2;
159
160	/* optional High Pass Filter states */
161	int32_t xvtx[5];
162	int32_t yvtx[5];
163	int32_t xvrx[5];
164	int32_t yvrx[5];
165
166	/* Parameters for the optional Hoth noise generator */
167	int cng_level;
168	int cng_rndnum;
169	int cng_filter;
170
171	/* snapshot sample of coeffs used for development */
172	int16_t *snapshot;
173};
174
175#endif /* __ECHO_H */
176