Repository navigation
Feedback from RT1011 - channels are swapped at every session #1838
Description
On CML platform, we have RT1011 smart amp providing feedback. we have 4 CH feedback ( from 4 individual speakers) enabled with attached Topology ( m4 is attached too)
Test procedure:
Play a known audio sample ( preferably sine tone) on Speakers
Record 4 ch data on Feedback PCM ( hw:0,10) for few seconds and make multiple recordings.
Each file shows feedback to be different - ideally they should be same for same input file.
TEST ENV:
CML device with RT1011 Amp.
SOF v0.8 backport release FW and topology changes attached on top.
Activity
sathyap-chrome commentedon Feb 28, 2020
@keqiaozhang Please reproduce one time and then we can see how to proceed debug. Codec vendor informs that the codec side, there is no channel swap.
cc @keyonjie @aiChaoSONG
keyonjie commentedon Mar 2, 2020
@sathyap-chrome can you help clarify:
you are playing stereo(2 channels) sine wav output to speakers,
you are capturing 4 channels(actually only channel 0/1 has data input) which expected the channel 0/1 getting the exact output of the playback channel 0/1 and 0s at channel 2/3.
is it possible to probe the output/input on the SSP pins on your side?
Hi @aiChaoSONG @keqiaozhang let's try if we can reproduce it on our side.
aiChaoSONG commentedon Mar 2, 2020
@keyonjie could you please have a look at thesofproject/sof#2430, and see if this merged PR has any influence on the abnormal behavior mentioned in this issue?
sathyap-chrome commentedon Mar 2, 2020
@keyonjie
you are playing stereo(2 channels) sine wav output to speakers
you are capturing 4 channels(actually only channel 0/1 has data input) which expected the channel 0/1 getting the exact output of the playback channel 0/1 and 0s at channel 2/3
I can try. But i wanted to know if we can get a dump or some s/w means to easily debug without opening up the closed chassis machine.
Also easy for codec vendors to duplicate any debug points.
@aiChaoSONG I guess not. This issue was seen with SOFv0.7 as well as SOF v0.8.
keyonjie commentedon Mar 2, 2020
the PR is to divide playback/capture channel numbers, I believe it won't fix channel swapping issue.
keyonjie commentedon Mar 2, 2020
@sathyap-chrome can you help upload the playback and captured .wav files?
sathyap-chrome commentedon Mar 2, 2020
This has 2 captured files. playback for this case was just normal music playback.
008_aec.zip
If you need sinewave for testing, here it is -
left1KHz_right777Hz.zip
I'll send out the captures from this file in a few minutes
aiChaoSONG commentedon Mar 2, 2020
On sof-dev + master, seems reocrd nothing. @sathyap-chrome @keyonjie
sathyap-chrome commentedon Mar 2, 2020
@aiChaoSONG did you make the topology modification ?
I've shared the tplg and also m4 for reference . You need to test only on the Helios device.
Can you try the SOF 0.8 backport on chrome first ?
aiChaoSONG commentedon Mar 2, 2020
@keyonjie @sathyap-chrome result on 008 backport(Chrome OS) is below.
@sathyap-chrome @keyonjie Now, I understand what the issue is. the wave seems on different channel slot at each recording.
sathyap-chrome commentedon Mar 2, 2020
@aiChaoSONG please do multiple recordings (during same playback) and check the snaphot.. above is only 1 recording, we need to compare 2 recording.
keyonjie commentedon Mar 2, 2020
@aiChaoSONG did you play the left1KHz_right777Hz file Sathya shared?
aiChaoSONG commentedon Mar 2, 2020
@keyonjie The problem is here: we have two channel with sound, and there are 4 recording channels, the two with sound distribute randomly on the 4 recording channels. sometimes on ch1&ch2, sometimes on ch1&ch4. you can see the picture I attached above. file used for playback is the one Sathya shared.
sathyap-chrome commentedon Mar 2, 2020
@aiChaoSONG this is correct 👍 #1838 (comment)
The feedback should actually get data from SSP Rx and its mapped to codec.
there are 4 individual channels on feedback. its not a playback loopback. if its behaving the way you mentioned, its actually incorrect.
keyonjie commentedon Mar 2, 2020
@sathyap-chrome @aiChaoSONG let's try to verify if the issue exist with hardware loopback mode(codec playback output jack connected to capture input jack).
29 remaining items
aiChaoSONG commentedon May 18, 2020
@keyonjie Yes, daily and PR tests run on this platform everyday,
sathyap-chrome commentedon Jul 15, 2020
Hi @keyonjie @aiChaoSONG due to other priorities i couldnt do any debug on this one.
I have some info from my test using CML 008 release :
I'm attaching the TPLG, m4 file the feedback wav snapshot and asoc graph here.
008_aec_ww29.zip
Note that the feedback is swapped as per the original issue. But interestingly i see that the pipeline for SSP1.IN is not active during my record session in both sessions.
Can you help replicate this issue on latest master ? this can help debug this faster.
any suggestions on what can be next debug step.
aiChaoSONG commentedon Sep 4, 2020
@keyonjie @sathyap-chrome Hi, Sathya, I tried the topology attached with the latest kernel and FW, there is nothing in the recorded feedback wave.
sathyap-chrome commentedon Sep 4, 2020
Hi @aiChaoSONG this is what i also observed . Hence we are not able to use the dump to check the SSP Rx data. Note that with CML v0.8, the feedback has data.
keyonjie commentedon Sep 14, 2020
Hi @sathyap-chrome Thanks for the feedback, so how urgent is this now?
keyonjie commentedon Sep 27, 2020
Hi, @sathyap-chrome can you help to try if PR thesofproject/sof#3476 help for this issue?
plbossart commentedon Jan 22, 2021
@sathyap-chrome @keyonjie @aiChaoSONG @kv2019i can we close this one?
keyonjie commentedon Jan 22, 2021
since thesofproject/sof#3476 already merged, so if either @sathyap-chrome or @aiChaoSONG re-check it with master firmware to confirm it, we can close it.
aiChaoSONG commentedon Jan 22, 2021
@keyonjie I remembered that I have tried to check this issue on master with the topology provided, no data recorded via feedback pipeline.
plbossart commentedon Jun 10, 2021
It's unclear if this bug is current or not. I will close, if this is an error @sathyap-chrome @aiChaoSONG @keyonjie please reopen with a status.