Buffer underrun about chatterpi

[ViennaMike]

What type of pi are you using?

…

On Fri, Sep 8, 2023, 7:30 AM diveblue ***@***.***> wrote: Using 5 files of 6-10MB each, PIR trigger. function well for about 2 hours, then I get several buffer underruns reported and playback / trigger sensing stops ALSA lib pcm.c:8545:(snd_pcm_recover) underrun occurred — Reply to this email directly, view it on GitHub <#14>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHXZ3XHNA5WRE4HRWEIE6DXZM2XTANCNFSM6AAAAAA4QRVDHQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[arwpc]

Pi Zero W or a Pi 4. Thanks! I love the software! On Fri, Sep 8, 2023 at 8:33 PM Mike McGurrin ***@***.***> wrote:

…

What type of pi are you using? On Fri, Sep 8, 2023, 7:30 AM diveblue ***@***.***> wrote: > Using 5 files of 6-10MB each, PIR trigger. function well for about 2 > hours, then I get several > buffer underruns reported and playback / trigger sensing stops > > ALSA lib pcm.c:8545:(snd_pcm_recover) underrun occurred > > — > Reply to this email directly, view it on GitHub > <#14>, or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AAHXZ3XHNA5WRE4HRWEIE6DXZM2XTANCNFSM6AAAAAA4QRVDHQ> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > — Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALVQ63PAP6HKURH6K2BIXKTXZPBOFANCNFSM6AAAAAA4QRVDHQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[diveblue]

Pi zero w. Incidentally, the image did not work so I had to do the manual install of the prerequisites

…

On Fri, Sep 8, 2023, 9:33 PM Mike McGurrin ***@***.***> wrote: What type of pi are you using? On Fri, Sep 8, 2023, 7:30 AM diveblue ***@***.***> wrote: > Using 5 files of 6-10MB each, PIR trigger. function well for about 2 > hours, then I get several > buffer underruns reported and playback / trigger sensing stops > > ALSA lib pcm.c:8545:(snd_pcm_recover) underrun occurred > > — > Reply to this email directly, view it on GitHub > <#14>, or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AAHXZ3XHNA5WRE4HRWEIE6DXZM2XTANCNFSM6AAAAAA4QRVDHQ> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> > — Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHRTTKWS5O33RBGZB2NFNZLXZPBODANCNFSM6AAAAAA4QRVDHQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[arwpc]

Ditto that. I’m trying to keep the cost down from the Pi4 because I’d like to network several together. ;)

…

On Mon, Sep 11, 2023 at 5:49 PM diveblue ***@***.***> wrote: Pi zero w. Incidentally, the image did not work so I had to do the manual install of the prerequisites On Fri, Sep 8, 2023, 9:33 PM Mike McGurrin ***@***.***> wrote: > What type of pi are you using? > > On Fri, Sep 8, 2023, 7:30 AM diveblue ***@***.***> wrote: > > > Using 5 files of 6-10MB each, PIR trigger. function well for about 2 > > hours, then I get several > > buffer underruns reported and playback / trigger sensing stops > > > > ALSA lib pcm.c:8545:(snd_pcm_recover) underrun occurred > > > > — > > Reply to this email directly, view it on GitHub > > <#14>, or unsubscribe > > < > https://github.com/notifications/unsubscribe-auth/AAHXZ3XHNA5WRE4HRWEIE6DXZM2XTANCNFSM6AAAAAA4QRVDHQ> > > > . > > You are receiving this because you are subscribed to this thread.Message > > ID: ***@***.***> > > > > — > Reply to this email directly, view it on GitHub > < #14 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AHRTTKWS5O33RBGZB2NFNZLXZPBODANCNFSM6AAAAAA4QRVDHQ> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> > — Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALVQ63PSADKIDQKZVVJ4K5DXZ6IPFANCNFSM6AAAAAA4QRVDHQ> . You are receiving this because you commented.Message ID: ***@***.***>

[ViennaMike]

OK, this one will be a tough one. There's no doubt that ChatterPi really pushes a Pi to its limits. The issue is clearly related to taking too long in the audiopy callback function that determines the jaw movements, but I have not found a way to make that run any faster. Any suggestions for further optimizing that code welcomed.

Try increasing the BUFFER_SIZE parameter to 16384. In the past, this seemed, in my setup, to elliminate the problem on the Pi4.
Changing the CPU governor to performance mode may also help, but it wasn't sufficient by itself. The downside of a larger buffer size is that it introduces delay in the output.

Hope this helps! I started a separate issue regarding the image file not working.

[ViennaMike]

I'm testing a fix for the buffer underrun problem. Rather than calling the heavy duty math processing for determining jaw movements at the same rate that chunks are fed to produce audio, only calling for it 50x per second. So far, this seems to be keeping buffer underruns away, at least on my Pi 4. This can probably be made less frequent if there's still a problem on slower machines like the Pi Zero, but dropping it to only 10x per second had a visibly detrimental effect on jaw movement.

[ViennaMike]

Ran repeatedly. Sometimes the audio cut out while the jaw kept moving properly, but that seems to be a different problem. Did not see a single "buffer underrun" error. Fixed code uploaded. Declaring this bug fixed. If anyone sees buffer underrun problems with this new code, please start a new issue.