Buffer 64 vs 128 samples experience

Thanks for sharing your individual blind test results, stamkorg!

It just lacks information about the objective existence of a difference in your test-signals and its degree. If something went wrong with your setup, maybe the latency is unchanged/identical and results have to be near 50:50. Another disturbance variable can be latency-jitter or a high, absolute latency on your system (I guess not but who knows?)

A quick test is to mouseclick PTQ's virtualkeyboard and to record the sound from the speaker and the clicknoise with a microphone (most every laptop or smartphone has a mic built in). In an audioeditor (e.g. Audacity) the latency between those two signals can be measured.

You're right, I did not objectively prove that latency really changed. But several times I asked my son to put the buffer at 2048 and I was able to verify on the keyboard that the latency increased then.
Except to consider a defect of my system there is no reason to think that the latency was not affected by the manipulations of my son. But in absolute terms you are right. Checking this point each time would have made the experience much heavier

This is subtle so tough to design a robust testing scheme.

I would expect that it would be easier to discern differences with longer playing samples (say 60 seconds vs 15 seconds) which allow your musical mind & internal clock to settle in. How long was each test roughly?

You could test this by setting several test lengths of say 15, 30, 45, 60 seconds. Then see how the stratified stats look.

A few other thoughts

- If you are playing the same piece, over and over, maybe you get more aware. Then become bored, and less aware.

- if you are playing different material, then maybe some material is more timing sensitive,

- It might be better if you just had to identify changes in buffer size without prompt from your son. But there you might not have any data lol.

- etc.

This could make a difference maybe because of ratio of changes.

For example, on my system, I measure "total latency" of ~6ms from the noise of finger striking key to headphone output (buffer of 64 measured via Audacity). If I were running with speakers, the latency increases 3ms as sound travels say 1 metre more, for total of 9ms.

So if changing buffer from 64 to 128 increased latency by say 2ms, the ratio changes based on headphone or speaker use.

2ms / 6ms ~ 0.3333
2ms / 9ms ~ 0.2222

If your keyboard controller or computer is less efficient and your "total latency" were say 20ms, then changing buffer from 64 to 128 might be less obvious

2ms / 20ms ~0.100

That could be an issue with faster playing and bigger chords as MIDI information "cues up" at the digital piano to go down the slow DIN midi cable (or maybe USB based on how raw sensor data is handled inside the digital piano). Somewhat less meaningful as all "tests" have same issue.

Not really the point of the post but sorry for wasting your time.