Audio degradation after setting skipCheck=true option in settings.yml

[kusznerr]

We are observing different audio with skipCheck veriable set to true or false. When skipCheck is set on true , audio quality is significantly degradated (frequent cracks etc). With skipCheck set to false (default option) - audio quality is much better.

Setting done in : /usr/share/meteor/bundle/programs/server/assets/app/config/settings.yml

Option : skipCheck

Expected behavior - no difference in audio quality between both options.

Devices tested - PC and smartphone.

[ffdixon]

This is really strange! Are you able to verify this first-hand (you can hear the differences yourself)?

[kusznerr]

Hello Fred - yes , we've spent some time on this by testing multiple server configurations (GCP / AWS / baremetal etc) and it all boiled to this setting. If you want we can share some samples we recorded.

[lookas1]

@ffdixon we have client video and audio captures which we can share but I believe you can test it in your lab to see the difference. We did a multiple test runs with off and on echo test, couldn't believe the results. Do you think different profiles can be loaded with this option?

[mariogasparoni]

Which version of bbb are you running? Also, which OS and browser this happens more often.

[kusznerr]

Latest BBB version (2.2.28 and 26). Results do not vary on client side , as we are testing it across few of us.

[dariosera]

I can confirm the problem. Same results also setting the userdata-bbb_skip_check_audio parameter using join API call. I'm running BBB 2.2.28 on a DigitalOcean c-8 droplet. Initially I tested with my own voice, then I used a voice recording sent directly to the conference from audio loopback.

[ffdixon]

Thanks for the confirmation. We're going to take a closer look for the FreeSWITCH configuration for the join audio sequence browser -> echo test -> main room and browser -> main room.

[mariogasparoni]

Yes, i can reproduce this too. After listening for a while you notice mic's audio quality is better

[mariogasparoni]

Did some tests taking a look at FreeSWITCH, and it seems jitterbuffer is not being set when using skipCheck=true. There's an action on diaplan which sets it, but apparently this has no effect. Still needs more investigation, i will run a few more tests and update it here

[mariogasparoni]

@ffdixon we have client video and audio captures which we can share but I believe you can test it in your lab to see the difference. We did a multiple test runs with off and on echo test, couldn't believe the results. Do you think different profiles can be loaded with this option?

By changing skipCheck to true, you will hit a different dialplan. Although this is a different dialplan , this is intended to set same audio call parameters when skipCheck is false. Both skipCheck true/false use the same profile: external

[mariogasparoni]

Just took a closer look on FreeSWITCH. jitterbuffer is being correctly set for both dialplans (those used when skipCheck is true or false). And after running more tests, i can't really feel much difference, which makes me think if i can really reproduce this problem. I am running Chrome 86 on Linux (Ubuntu focal) and Safari 13.1.2 on macOS for my tests

@lookas1 can you share your video/audio showing the problem ? So i can see if i am able to reproduce it. you can email me: mariogasparoni at gmail

[lookas1]

@mariogasparoni, @ffdixon apologies it took some time but we decided to make new samples :-) Please check them here: https://drive.google.com/drive/folders/1snb3u84amrPf7E0UAb473ZY7SaY2wema?usp=sharing

I think names are self-explanatory.

I will not put any infrastructure details (for now at least) as we observe it on VPS, cloud VMs and bare metal servers.

Samples were taken with XBox Game Bar on Win10.

[kusznerr]

@mariogasparoni - wanted to check if you were able to validate samples we provided ?

[mariogasparoni]

@lookas1 @kusznerr , thanks for the high quality samples you guys provided. Indeed, when echo test is off i can hear some audio 'cracks' during the talk. I took a closer look on the audio process and i couldn't find any difference in dialplans/profiles that could decrease quality when echo test is off.

I am working on general audio improvements, and there's a chance i find a solution for this (can't promise though). In the meantime, can you run a similar test on a different server/setup ? Email me so i can give you access to a private server (running 2.2.29). You can run a test on it (no need to record this time, and i will make sure echo test is off) and then we eliminate the possibility there's something different in your hardware/SO/BBB-install that causes this issue.

[lookas1]

@mariogasparoni can you send details to our email [email protected]? Not sure how I can get your email (is it possible through github anyway? ;-) ) We are happy to test it on your setup.

[kusznerr]

Hello - wanted to check if there was any luck in finding what the issue may be ...

[kiakeykhosravi]

hello - we have exactly the same problem i even added it in google dev group today its really starting to create problems for us.is there any chance you guys found whats the problem?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.