Differentiator doesn’t seem to be working

[Puppeteer]

I’ve set up the following patch to test the differentiator (was trying to use it in a new algorithm) and the LFO meter swings to the LFO sine wave, but the LFO’ doesn’t move and stays on 0 regardless of what I set the time constant to. I need to detect the slope of an LFO (whether positive or negative) and the differential of it should do the job.  Am I doing something wrong here, or is differentiator not working as it should?

I’m on VSIG 3.3.3

[Puppeteer]

Integrator seems a bit sqiffy as well.  It pins at -256 with a sine wave input, with a little flick to something higher when it reaches 1 on the sine. Am I correct in assuming that it samples at audio rate?  Even with that, I’d expect it to spend 50% above 0 and 50% below 0.

Are there any examples you can point to that uses these blocks that I can study?

I’ve fixed my current problem using a tuned allpass to give me a 90 degree phase offset, but would definately like to understand the differentiator and integrator implementation in the H9000.

[John Baylies]

When debugging audio-rate or mod-rate (1/4 audio-rate) modules, it’s best to plug their outputs into the adc-outs and record them in your DAW so you can really see what they’re doing. The user interface modules may not update fast enough to show everything.

The attached sigfile works with Differentiator’s time_constant = 0, and LFO’s polarity = 0.

https://etide.io/4IY6z

[John Baylies]

Differentiator is used in 1980s Chorus.

Integrator is used in none of the H9000’s factory algorithms.

 

[Puppeteer]

Thanks, I’ll have a look at 1980’s chorus.

Regarding the integrator, as the block is, it’s not particularly useful. It acts as an accumulator from a reset event. I could see potential for use from a peak detection perspective for things like transient v steady state manipulation, similar to SPL’s transient designer, or maybe as some kind of triggered envelope, but struggling to see other applications.

In industrial automation, integrators usually have a time factor (or reset time) that defines a window over which the integrator acts.  Essentially the integrator integrates a time or number of samples, and discards the oldest ones while filling the buffer with new values allowing the integrator to smooth a signal, or to oppose fast change (in a similar way to a shock absorber, which is essentially a physical integrator), or most commonly to accumulate an offset signal (used in PI and PID controllers).

Single sample integrators act as back step operators or shift registers, but that is probably more easily achieved using a single sample delay within VSIG, for really low level modelling.

Can I make a request for an additional integrator block with a sample buffer over which the integration happens, with a user definable time constant (either in ms or samples).  This would be very useful for audio, for example, it would be useful for smoothing signals to meters, or for calculating short term levels for slow acting compressors or for gain automation.

[Author]

Posts