High Pass Filter Node introducing unwanted offsets

It seems that opening and closing the low pass filter introduces an offset to the signal.
HighPass%20DC%20offset%20bug
Top display is the signal in, bottom display is the output of the highpass node.

This is the patch in the gif: High Pass Bug.audulus (11.4 KB)

At first I thought you were talking about a phase shift which is unavoidable, but now I see that you are talking about a DC offset in the signal. I believe that this is a consequence of the filter algorithm used for the built in nodes. I believe they are exponentially weighted, moving average, first order filters. When you rapidly change the filter cutoff frequency, the filter takes a while to settle because of the delay inherent in the moving average. This effect is more pronounced the closer you get to an alpha of 1. Assuming that alpha is <1 the offset will eventually decay. I don’t believe that this is something that can be fixed without changing the type of filter. Bi-quads have some similar problems which is why they often don’t modulate well.

1 Like

Thanks for the insight. I added some slew to the input signal to the high pass node and the problem self-corrects more quickly, but I had to make sure it was also VERY little slew because the of asymptotic decay of the signal going through the low pass filter was attenuating the LFO signal.
44%20PM

1 Like

Lots of filters end up having DC offsets, including analog ones (my SEM one had a -3v dc offset when in HP mode even with nothing going through it).

1 Like

Well! Today I learned…

1 Like

The offset in this one is only short term. I doubt you would be able to detect it at audio frequencies unless you were modulating the cutout frequency at audio rates as well. In that case I guess you would be introducing an additional low frequency sub-harmonic. The offset is only noticeable when the modulation frequency is similar to the cutoff frequency. I did get some interesting waveforms while I was playing with @robertsyrett’s setup:

1 Like