dukesrg/logue-fx

Crusher knob A and B have weird behaviour in NTS-1

aframires opened this issue · 4 comments

aframires commented

Hello dukesrg,

I'd like to thank you first for the effort and great work you have put into making these effects open-source.

I have installed the release version of the Crusher on my Korg NTS-1 and it seems that the knobs do not behave appropriately. I have installed all the versions (mod, delay and reverb) and it seems like the behaviour is similar:

  • Knob A seems to not affect the sound until from 0%-50% position, and then from 50%-100% it severely impacts the sound. I think it would be great if we could have more resolution in the second part, while less resolution on the first.

  • Knob B works well from 0-50% but, whenever it goes a bit past 50%, all the synth goes to complete silence.

Thank you for the great work. I'm currently looking at the code and trying to compile it myself and try to fix the behaviour. If I fix it I will send you a Pull Request.

dukesrg commented

@aframires
Knob A : you just won't not notice sample rate reduction on a source signal with no high frequency (the whole knob scale is 48KHz...480Hz)
Knob B: Just normalize the input signal to the maximum - it is not deithered and you're getting silence because the input signal level is below the quantization threshold. Build-in fractional bit-reduction lookup is used with rounding to avoid nasty noise on low-bit settings.

aframires commented

@dukesrg

Knob A: Do you think it would be nice for it to have less linear behaviour in the frequency domain, and more like a logarithmic behaviour?

Knob B: How would you do this normalization? I'm using as an input signal the built-in oscillators of the NTS1.

I'm almost compiling the projects. When I succeed I'll try to make these changes.

dukesrg commented

@aframires
Actualy sampling rate reduction factor looks like this http://yotx.ru/#!1/3_h/sH@1v7Rgzhf23/aP9gf@vgYN@IIfyv7W9tb2xt7u3uH@yTaNiNnVPG4@kW43Hr8mJ3f2sfAw==
With linear scale high frequency cutoff is almost instanteneous, so I used this kind of slope.

Using monophonic oscillator output does not create the full spectre so you just can't notice changes sometimes where it is nothing to change. Just check with the music from external audio input with both cases and it will be clear why it works like that.

dukesrg commented

@aframires BTW I hope you're testing with good enough headphones, not with the build-in twitter!