FreeCalypso > hg > gsm-net-reveng

annotate doc/TFO-xform/FRv1 @ 37:4ab7cc414ed2

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
doc/TFO-xform/EFR: document CN insertion
author Mychaela Falconia <falcon@freecalypso.org>
date Tue, 03 Sep 2024 21:20:47 +0000
parents 35d38348c880
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
34
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
1 Rx DTX handler situation in FRv1
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
2 ================================
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
3
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
4 Before we address the question of how one should implement TFO transform for
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
5 FRv1, let's begin with a more basic question: how does the Rx DTX handler (the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
6 "front end" part of the speech decoder in an end-terminal implementation) work
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
7 in FRv1? In both HRv1 and EFR, error-free comfort noise generation functions
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
8 of this Rx DTX handler are normative per the specs at bit-exact level, while
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
9 error handling functions are specified only as a non-normative example - and
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
10 the supplied reference C sources implement the full Rx DTX handler (both the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
11 normative part and the "example" part) as an inseparable part of the speech
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
12 decoder. But not so for FRv1: there is no reference C source and there are no
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
13 bit-exact definitions for any part of Rx DTX handler logic. All Rx DTX handler
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
14 functions are defined only in English prose (no code), and even in the most
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
15 normative parts the language used in the specs is quite loose.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
16
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
17 Based on what is specified (verbally, loosely) in GSM 06.11 and 06.12, there
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
18 are two principal ways in which an Rx-ECU-capable, Rx-DTX-capable FRv1 speech
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
19 decoder can be implemented:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
20
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
21 Fully modular approach: the basic GSM 06.10 decoder block (which is bit-exact,
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
22 but cannot handle BFIs or SID frames) remains absolutely unmodified, while the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
23 Rx DTX handler (which includes both error concealment and CN generation) is
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
24 implemented as a modular piece, with an "honest-to-god" 260-bit 06.10 frame
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
25 interface between the two blocks.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
26
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
27 Non-modular approach: the Rx DTX handler and the 06.10-based speech decoder are
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
28 integrated more tightly, and there is no possible stream of "pure" 06.10 codec
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
29 frames that would produce the same bit-exact PCM output as the actually
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
30 implemented "full decoder" with the built-in Rx DTX handler.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
31
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
32 Cursory reading of GSM 06.11 and 06.12 specs strongly suggests that they call
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
33 for the fully modular approach as defined above. However, because neither spec
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
34 includes any bit-exact definitions, there is no formal stipulation that the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
35 modular approach shall be used - it is entirely conceivable that someone could
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
36 implement a non-modular approach, and they would still be spec-compliant.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
37
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
38 Why would anyone implement the non-modular approach when the fully modular one
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
39 seems much simpler? After all, the bit-exact basic 06.10 decoder already
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
40 exists - surely it is easier to build a separate front-end to it than dig into
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
41 the guts of that pre-existing box? There is, however, one aspect that could
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
42 sway implementors toward the non-modular approach: interpolation of CN
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
43 parameter updates during prolonged DTX pauses. GSM 06.12 (or rather its latest
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
44 incarnation as 3GPP TS 46.012) says, at the very end of section 6.1:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
45
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
46 "When updating the comfort noise, the parameters above should preferably be
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
47 interpolated over a few frames to obtain smooth transitions."
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
48
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
49 This kind of CN parameter interpolation is mandatory in the newer HRv1 and EFR
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
50 codecs where the CN generator function is defined in bit-exact terms, hence it
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
51 makes sense that some implementors may have chosen to back-port the same feature
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
52 to FRv1.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
53
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
54 CN parameter interpolation: deeper analysis of the problem
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
55 ==========================================================
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
56
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
57 How does this interpolation feature affect the choice of modular or non-modular
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
58 design? As a non-expert on the subject of codec design, I am not able to say
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
59 authoritatively if it is possible to implement the feature of CN parameter
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
60 interpolation (and do it well) while staying with the fully modular design in
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
61 which the basic 06.10 decoder block remains absolutely unchanged, or if high-
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
62 quality implementation of this feature would require foregoing the modularity
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
63 and moving the CN-specific interpolation function somewhere inside that block,
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
64 e.g., between the output of GSM 06.10 section 4.2.8 and the input to section
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
65 4.2.9, as referenced from section 4.3.3 for the decoder.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
66
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
67 We can, however, look at how ETSI handled this problem in other codecs for
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
68 which they did mandate CN parameter interpolation in bit-exact form. HRv1 is
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
69 the best point of comparison in this regard because of this detail: the Rx DTX
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
70 handler front-end part of the official bit-exact HRv1 decoder (delivered as C
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
71 source this time, not just verbiage) is _almost_ modular, i.e., one could
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
72 _almost_ detach it into a modular piece whose output could be fed to the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
73 decoder as a new "cleaned up" stream of HRv1 codec frames. Where is the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
74 "almost" part? Answer: interpolation of CN parameters! When HRv1 decoder is
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
75 in CN insertion state, it dequantizes R0 and LPC parameters from SID frames
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
76 only when initial and update frames come in - but when it generates the actual
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
77 CN between those updates, it performs smooth linear interpolation on the decoded
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
78 parameters, *without* requantizing them into something that can be retransmitted
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
79 as new HRv1 codec frames representing the CN.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
80
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
81 Once again, as a non-expert on the subject of codec design, I am not able to say
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
82 authoritatively if the same approach that was prescribed by ETSI for HRv1 would
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
83 also work for FRv1, or if CN parameter interpolation for FRv1 can be done well
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
84 by requantizing the interpolated parameters for each individual CN output frame
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
85 and feeding them to a strictly unmodified 06.10 decoder block. It is the case,
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
86 however, that there is no pre-existing implementation available to us which we
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
87 can look at that does CN parameter interpolation for FRv1 - the TFO transform
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
88 in Nokia TCSM2 does _not_ interpolate - hence without a reference to look at,
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
89 this optional feature is a can of worms which we should stay away from.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
90
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
91 Front-end part of the speech decoder and TFO transform
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
92 ======================================================
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
93
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
94 If the party who implemented the regular end-decoder for FRv1 chose the fully
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
95 modular approach, either by disregarding the call for interpolation of CN
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
96 parameters (the spec language is "should preferably", rather than "shall") or
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
97 by requantizing the interpolated parameters on each CN output frame, then a
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
98 corresponding implementation of TFO transform for non-DTXd operation becomes
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
99 trivial: the modularized Rx DTX handler front-end can also serve unchanged as
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
100 the TFO transform!
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
101
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
102 This just-described situation holds for the current Themyscira Wireless
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
103 implementation of FRv1 codec, named libgsmfr2. (The 2 in the library name
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
104 refers to the major version of library API and dependency structure; the codec
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
105 it implements is still FRv1.) Specifically:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
106
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
107 * The full decoder implementation in libgsmfr2 follows the modular approach:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
108 the front-end Rx DTX handler preprocessor feeds "cleaned up" FRv1 codec frames
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
109 to an unmodified GSM 06.10 decoder.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
110
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
111 * No interpolation is done on CN parameters: as soon as each SID update comes
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
112 in, the new parameters are used immediately for all generated CN frames.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
113
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
114 The preprocessor part of libgsmfr2 is thus already suitable to serve as a TFO
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
115 transform for FRv1. However, before formally adopting it as such, I have had a
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
116 long-standing desire to see how this function was implemented by other vendors;
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
117 particularly, how it's been implemented in real historical TRAUs.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
118
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
119 Nokia TCSM2 TRAU implementation
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
120 ===============================
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
121
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
122 As of 2024-08, we finally have a working bank-of-TRAUs apparatus in our lab:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
123 Nokia TCSM2. This TRAU implements TFO for FRv1, HRv1 and EFR, hence we finally
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
124 got the ability to see how this vendor (Nokia) implemented the elusive TFO
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
125 transform.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
126
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
127 Here are our findings:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
128
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
129 Error concealment function
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
130 --------------------------
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
131
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
132 Themyscira implementation is based on the "example solution" of TS 46.011
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
133 chapter 6; Nokia's implementation appears to be very similar, with only a few
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
134 visible differences:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
135
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
136 * When the ECU enters the state of "speech muting" (after the first speech-state
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
137 BFI for which the last good speech frame is simply repeated), instead of
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
138 decrementing each of the 4 Xmaxcr numbers by 4, it decrements them by 11,
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
139 thereby producing noticeably faster muting than what the spec calls for.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
140
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
141 * The state of emitting fixed silence frames is entered not after the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
142 algorithmically-muted frame in which the lowest Xmaxcr reached 0 (my reading
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
143 of the "example solution" in the spec), but after the state of algorithmic
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
144 muting (decrementing Xmaxcr's by 11 each time) persisted for exactly 5 frames.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
145 If the original speech frame had its highest Xmaxcr equal to 63, the last
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
146 algorithmically muted frame before fixed silence frames will have 8 in that
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
147 Xmaxcr; if all starting Xmaxcr numbers were low, there will be 5 frames with
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
148 all zeros in Xmaxcr, random Mcr and other parameters unchanged before the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
149 switch to fixed silence frames.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
150
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
151 Nokia's TFO transform exhibits additional logic whereby the first good speech
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
152 frame after prolonged BFIs has its highest Xmaxcr reduced (but not messed with
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
153 otherwise); if that good speech frame is again followed by BFIs, the ECU goes
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
154 back to silence frame output right away - or at least that's what we saw in one
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
155 experiment. This aspect has not been studied in detail.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
156
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
157 Comfort noise generation (DTXd=0)
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
158 ---------------------------------
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
159
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
160 The comfort noise output from Nokia's TFO transform generally agrees with my
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
161 reading of GSM 06.12 spec section 6.1, the section that describes CN generation.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
162 However, the following parts were surprising/unexpected:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
163
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
164 1) The TRAU reacts to SID updates with a delay of 24 frames. Suppose that frame
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
165 #20 in the input is the initial SID, frame #24 (TAF position) is the first
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
166 SID update, frame #48 is the next SID update and so forth. In the output
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
167 from Nokia's TFO transform, the updated parameters from input frame #24 will
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
168 appear in output frame #48, those from input frame #48 will appear in output
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
169 frame #72 and so forth. There is no sensible explanation for this extraneous
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
170 buffering delay; at first I thought it was an artifact of the CN parameter
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
171 interpolation mechanism, but:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
172
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
173 2) No interpolation is done! I deliberately constructed input sequences in
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
174 which each subsequent SID update has wildly different parameters from the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
175 previous, and when the changeover does happen in the DL output after the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
176 strange delay of 24 frames, the change is immediate and abrupt.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
177
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
178 CN muting after two missed SID updates (BFI received instead of SID in the TAF
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
179 position twice in a row) is done the same way as speech muting: the TRAU emits
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
180 exactly 5 frames with decreasing Xmaxcr (same decrement by 11), then switches
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
181 to emitting fixed silence frames.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
182
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
183 SID forwarding (DTXd=1)
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
184 -----------------------
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
185
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
186 When DTXd is enabled on the destination call leg and the input frame stream to
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
187 the TFO transform includes SID frames (considering only valid SID for now), the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
188 transform does not generate comfort noise - instead received SID frames are
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
189 passed through to call leg B DL, unless they are invalid SID or the muting
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
190 mechanism has to kick in because of lost SID updates.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
191
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
192 Nokia's implementation does pass valid SID frames through (I haven't tested
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
193 invalid SID yet), but it applies the same weird delay of 24 frames to the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
194 switchover point for each update as it does when generating CN for DTXd=0.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
195
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
196 However, the part where Nokia's TFO transform (at least for FRv1) is plain
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
197 broken is CN muting in the case of lost SID updates. Here is what it does: it
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
198 decrements Xmaxcr by 4 (yes, by 4, not by 11) once every 24 frames (probably in
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
199 each TAF position), such that if the level of CN was very high before channel
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
200 breakdown, it will take up to 7.68 s before this CN is fully muted at the end
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
201 receiver.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
202
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
203 GSM 06.12 section 5.4 says: "For the second lost SID frame, a muting technique
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
204 shall be used on the comfort noise that will gradually decrease the output
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
205 level, resulting in silencing of the output after a maximum of 320 ms." The
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
206 spec gives a maximum of 320 ms for total muting of CN, but with Nokia's TFO
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
207 transform in DTXd=1 case, that maximum time is 7.68 s - spec requirement
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
208 violated.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
209
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
210 Only TFO, or regular FRv1 decoder too?
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
211 --------------------------------------
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
212
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
213 How does the regular FRv1 speech decoder (the one that ultimately emits G.711)
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
214 in Nokia TCSM2 TRAU implementation compare to what we've observed with their
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
215 TFO transform? Do they use a modular design where the regular decoder is a copy
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
216 of the same TFO transform followed by a standard GSM 06.10 decoder block, or do
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
217 they do something fancier?
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
218
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
219 Unfortunately we have no realistic way to answer this question: Nokia chose to
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
220 not implement the optional in-band homing mechanism for FRv1, thus we have no
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
221 way to pass test sequences through the TRAU in the decoder direction and see if
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
222 the output matches our hypothesis as to decoder logic. Hence the TFO transform
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
223 is the only part whose detailed behaviour we can realistically study in this
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
224 TRAU.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
225
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
226 Take-away for Themyscira implementation
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
227 =======================================
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
228
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
229 My take-away points from the preceding examination of FRv1 TFO transform in
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
230 Nokia TCSM2 are:
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
231
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
232 * Our current Rx DTX handler front-end in libgsmfr2 is fine - Nokia's
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
233 implementation is not any fancier at least in the case of TFO.
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
234
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
235 * Modularity is a good thing, and so is consistency. There is nothing wrong
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
236 with using the same Rx DTX handler block both as our TFO transform and as the
35d38348c880 doc/TFO-xform/FRv1: article written
Mychaela Falconia <falcon@freecalypso.org>
parents:
diff changeset
237 front-end portion of the full decoder in end terminal operation.