volume: zipper noise with fast changes

[thesofproject]

volume: zipper noise with fast changes

[tlauda]

@plbossart HiFi3 or generic implementation?

[plbossart]

This was on APL so I guess HiFi3.

…

On 07/13/2018 02:22 PM, Tomasz Lauda wrote: @plbossart <https://github.com/plbossart> HiFi3 or generic implementation? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#103 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGIa7jjsiTl_qjvF5j7i9bMosvOFiBvks5uGPNhgaJpZM4VPQJF>.

[lgirdwood]

@singalsu this may need ZC or a smaller step size.

[singalsu]

@lgirdwood Sure, I will check some options after doing the updates to my UP2. Also as 3rd option logarithmic step could cope with larger step length. Zero crossings detection based volume update could be best but it would need largest amount of work.

[singalsu]

I can now repeat the issue with my UP2+HiFiBerry after some struggle with fresh code. It's a bit hard to hear the zipper with music but with sine (997 Hz) it's quite audible.

[singalsu]

The effect is also a bit different when pressing cursor keys in alsamixer vs. running successive commands like
amixer sset 'PGA1.0 Master Playback Volume 1',0 10 10
and
amixer sset 'PGA1.0 Master Playback Volume 1',0 32 32
so there may be two different issues. For the cursor keys case I wonder if the previous ramp has completed before the new target gain is received. The zipper noise is stronger in the keys press case. Recording the playback should confirm what happens.

Note: I'm testing with gcc compiled FW. But it looks like the HiFi3 version just uses different copy function versions while the volwork code is shared so the issue should be the same.

[singalsu]

I will submit a pull request with 250 ms long linear volume ramp (was 64 ms) and with gain value update rate increased to every 1 ms (was 2 ms). I can still hear very silent zipper noise with sine wave but no more with music. If it's considered still too noisy maybe there's still room for some ramp length increase. Or consider changing ramp to logarithmic (still a small change) or even use ZC detection (major change, includes HiFi3 optimized code).