Scanning speed and velocity mapping of Cybrid: Page 7

RIP · 2022-01-08T14:48:32+00:00

Context: I'm trying to understand deeper the frequency requirements of my ADC, per discussion in https://pianoclack.com/forum/d/243-diy-piano-sensors and particularly https://pianoclack.com/forum/d/243-diy-piano-sensors/29 -- and related one in https://forum.modartt.com/viewtopic.php?id=8956 Back-of-the-envelope: at ff piano level, hammers move 3m/s, the let off is about 1mm from strings (sensor), so the hammers move from let off to string (sensor) hit in about 300us. I know you are using shank sensors, hence both the distance and the speed are different, but the time is the same. So we have to measure a time around 300us and translate it to a MIDI velocity of about 120. Now we want the system to be able to produce velocities of 121 or 119 (assuming MIDI v1 and ignoring MIDI v2). So we want around 1% resolution in the MIDI output. If the MIDI output is simply multiplicative or divisive of the measured time, that means we need 1% resolution of the measured time, i.e. 3us, i.e. 300kHz per channel. Question: you mentioned a while back (I think it was on PW) that instead you needed an intense "mapping" to provide appropriate MIDI velocities from the measured times. Looking at your code a https://github.com/ekumanov/cybrid/blob/master/src/teensy/main/My_Hybrid_Piano_V1.2.ino#L220-L245 I think shows what you were talking about. I think I understand and can reverse engineer it, but I think a discussion is better. So the question is: how are you mapping from time to MIDI velocity, do you have a plot of the curve you are using with a range of parameter that you tried (and accepted or discarded) -- basically anything that is not already evident from the code section I linked and even anything that you'd like to add to explain about that code even if it's evident. Thanks!

RIP

It might not buy you much, but if you prefer I posted the original RAW data (binary, not MIDI, simple python suffice to extract it) at https://github.com/davidedelvento/Mybrid/tree/main/data/RPiPico%2BEAITRCA6

There is also the script to do extract the text (or plot these figures) if you so prefer.

Note: the speed of writing to the SD is not as good when also sampling the ADC, for doing both at full throttle I should use one of more of DMA, interrupts or multicore. Each one requires a bit more work than I am willing to put in for just a test, so I am calling it good for now. Open to revisit that decision should someone prove (or even just speculate) that a further increase in time resolution is useful for this analysis.

RIP

In case anybody is still interested in this discussion, I've resumed it with a slightly different focus in a new thread:

https://pianoclack.com/forum/d/916-performance-of-diy-controller-prototype/4