I’m thinking about strategies for reducing feedback/bleed in a corpus-based musicking system and would love to hear if anyone has developed solutions for this. I’m using an air mic on my cello and analyzing that signal to trigger samples with flucoma/sp.tools. I’m performing this system in a club where the monitors will bleed back into the cello mic.
First idea is to use a mel-scale filterbank to analyze the corpus sampler output, and invert those melband values to apply gain reduction to the audio input before the rest of the analysis.
Second idea is to use inverse convolution where an impulse response is created from the sampler output and applied to the input . I’m aware of the HISS tools which seem to be a good place to start, but I don’t have a good sense of how to use them for this task. I’m also not sure if the latency/computational cost will be prohibitive for my otherwise low-latency system.
Any thoughts, opinions or guidance is welcome. Thanks!
If you’re open to it, I think the easiest way is to change the mic you’re using so that it’s less sensitive in noisy environments. How nasty would an SM57 or a pickup contact mic be?
For sure. I am planning to get a cello pickup which should help significantly, and my current condenser is pretty directional. Nevertheless the cello body picks up a lot of the sound from the loudspeakers so I’d be interesting in exploring some kind of attenuation
I’d definitely start with mic choice here, particularly since you can use separate mics for analysis and amplification. So a (hyper)directional mic and/or a contact mic (if you have one) could be used to feed the analysis stuff, then use the mic you normally do for vanilla amplification.
Latency/computation shouldn’t be an issue here as convolution is quite cheap and adds no latency, so it’s more about having/taking the time to create appropriate IRs for each part of the chain, in situ.
If this is something you want to do much more, you can look at better amplification options. These work really well with cello strings and can be quite installed without much difficulty (perhaps some 3d printing):
@alicee has used them in her Feedback Cello project:
I have done miracles with good mic (range, pattern, position). If you can afford trying them, Schoeps figures-of-8 with the null point aiming at the speaker (or in my case at the other instrument I wanted to reject) gave me good results.