# HG changeset patch
# User Mychaela Falconia <falcon@freecalypso.org>
# Date 1627549725 0
# Node ID 0f9606d1f8b861b769d7ffe912054d7288231a97
# Parent  c7514c8b5b41f396f9e7ab2e59343d2a03a00fcb
pirelli/aec: results of investigation into Pirelli's AEC config

diff -r c7514c8b5b41 -r 0f9606d1f8b8 pirelli/aec
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/pirelli/aec	Thu Jul 29 09:08:45 2021 +0000
@@ -0,0 +1,37 @@
+Reverse eng question: what does Pirelli's fw do with the AEC feature
+of Calypso DSP?
+
+Seeking an answer to this question, I sent an aur 12 command to Pirelli's fw
+via fc-tmsh.  The response came back with 48 bytes of data, consistent
+with our reference TCS211 code with L1_NEW_AEC enabled.  Here is the raw
+response in non-loudspeaker-call state:
+
+aur 12: 48 bytes
+offset 00:  02 00 01 00 14 00 CC 0C  AD 59 00 40 32 00 00 10
+offset 10:  00 10 00 00 04 00 00 00  9E FE 43 06 E5 16 34 36
+offset 20:  E5 16 43 06 9E FE B4 FC  BA FE 0F 01 40 00 C7 FF
+
+In an active loudspeaker call, two settings change:
+
+aur 12: 48 bytes
+offset 00:  02 00 01 00 14 00 CC 0C  FF 0F 00 40 32 00 00 10
+offset 10:  00 10 00 00 04 00 60 00  9E FE 43 06 E5 16 34 36
+offset 20:  E5 16 43 06 9E FE B4 FC  BA FE 0F 01 40 00 C7 FF
+
+Due to a misdesign in ETM (apparently unchanged between reference TCS211 and
+Pirelli's fw), the 48 bytes that come back are actually just 24 bytes of real
+data (new AEC config is 12 16-bit words) followed by 24 bytes of garbage
+padding.  Decoding the returned 12 words of new AEC config, we get:
+
+aec_enable = 0x0002
+continuous_filtering = 0x0001
+granularity_attenuation = 0x0014
+smoothing_coefficient = 0x0CCC
+max_echo_suppression_level = 0x59AD (handheld mode) or 0x0FFF (loudspeaker)
+vad_factor = 0x4000
+absolute_threshold = 0x0032
+factor_asd_filtering = 0x1000
+factor_asd_muting = 0x1000
+aec_visibility = 0
+noise_suppression_enable = 0x0004
+noise_suppression_level = 0 (handheld mode) or 0x0060 (loudspeaker)