Audio Skipping in Simultaneous Audio Presentation and Recording

[CarsonGMR]

I've been collecting data in a study that simultaneously collects audio using

Code:

background_audio

and presents audio using

Code:

audio

.  Things seem to be working well overall, but I happened to notice that for at least one participant, the audio stimulus seems to skip in several trials -- in this case the stimulus is also picked up in the audio recording, and I noticed that small sections of the audio stimulus repeated.
My best guess is that this is an issue with doing simultaneous audio output and input that is outside the scope of the SSG, but thought I'd check if anyone else has noticed audio skipping in other studies?
Before diving in to deep, I thought I'd also just check if there are known issues with using audio (which I think uses a standard js audio object) and the background_audio recorder (which I think is implemented here)?
Thanks FF team for putting together this great system!! I've been incredibly impressed with how flexible it has been!

[Lucy - FF Researcher Support]

Hello! We have not heard about an issue of using audio stimulus + background audio response from other researchers. Could you clarify what you mean by "the audio stimulus seems to skip in several trials"? Is the audio stimulus playing in a "skippy" fashion, or does the trial "skip" the audio stimulus (as in it doesn't play at all)?

Thanks for the kind words about FF!

[CarsonGMR]

Hello! We have not heard about an issue of using audio stimulus + background audio response from other researchers. Could you clarify what you mean by "the audio stimulus seems to skip in several trials"? Is the audio stimulus playing in a "skippy" fashion, or does the trial "skip" the audio stimulus (as in it doesn't play at all)?

Thanks for the kind words about FF!

[Ting]

Interesting! I'm inclined to think it was due to a particularly under-powered device (e.g., Kindle Fire devices, the kids version especially), since the recording and the audio stimulus go on at the same time? On some lower-end devices, sound duplex processing (outputting sounds and recording at the same time) can be a taxing task. You mentioned that you heard the audio stimulus in the recording, which points to this hypothesis as well - the tablet CPU was so busy that it couldn't keep up with echo cancellation!

Is it necessary to start recording as soon as the audio stimulus starts playing? If not, that might be the first point to optimize. For example, use the audio stimulus' "barrier" property and delay the recording till after the audio stimulus finishes playing.

We'll look into if there are more efficiency ways to play an audio file too, but I don't think there's too much free performance to be squeezed out of there, unfortunately.

BTW - thanks for the super detailed illustration!