Hey everyone, hope it's ok for me to revive this thread instead of opening a new one (as i've found the info in this one very useful so far) - i basically want to build something similar to what jeffd75 has built, but with several objects rather than one gong.
Each object would be in a closed feedback loop with a surface transducer and a piezo pickup connected to it, the Bela would sit in the middle applying gain scaling as well as filtering of each feedback loop's signals. Each feedback loop will also be monitored by a separate computer running some signal analysis in supercollider and sending control messages over OSC to the bela.
The aim of this setup is to figure out what would happen to objects in an acoustic feedback loop if you filter out their main resonant frequency but keep increasing the gain - I'm assuming that they will quickly settle on a different resonant frequency, and given enough notch filters maybe you'd have end up with a "scale" of possible resonant modes for each object - afterwards it would be possible to write a score for the system or perform it live.
I initially wanted to do this just using a standard audio interface with lots of I/O. However the digital mixer I was looking into (Behringer XR18) has an I/O delay of 0.8 milliseconds, which would result in a constant 1250Hz comb resonance on all channels, and this is true of pretty much all digital mixing desks across the board.
So i'm considering the Bela - hoping to squeeze out enough nanoseconds so that I can move the comb resonance high enough and filter it out via an analog lowpass filter in the feedback loop. At first i was looking into FPGA's, but besides being an entirely new can of worms the cost just to get a board with 8 channels of parallel Audio I/O would already be well over 1000€...
According to Andrew's paper it's possible to get 0.08ms i/o latency when using adc channel 7 and dac channel 0, which would be one order of magnitude better than the Behringer XR18! Although I'm aware that this won't be possible for all I/O pairs, even 0.2ms would already be acceptable for me.
So, to get to my question: I was wondering whether the bela's processing power is enough for at least 2 notch filters + gain scaling of up to 8 input/output pairs at the lowest possible latency using the audio expander board? Or if someone could give me a rough estimate of how many notch filters I could squeeze out of the thing (: