[sdiy] DC 'bounce' in second order filters

[Harry Bissell]

I'm trying to recover the fundamental of a filtered guitar waveform.  It works pretty well for the most part, but just after
the signal starts for the first time, the filter output bounces to where it has gone entirely positive (or negative) and my conversion
method relies on a signal peak followed by a zero crossing. There is a small gap while the circuit settles. The small gap is the width of a couple
of guitar cycles, clearly audible as a gap in the recovered waveform. Distortion I can live with, but a hole in the sound I can't.

I tried (out of curiosity) using a four pole filter in the past, but the combination of delay through the filter and the even worse
settling time made that less workable.

part of what I'm seeing may well be from the spice simulation I'm running, which predictably starts at "zero" time. In real life you
would have at least some history, even a few milliseconds.

(and I'm intrigued by your new music technology company. Too bad I'm not 'bay area'...)

H^) harry


----- Original Message -----
From: Donald Tillman <don at till.com>
To: Harry Bissell <harrybissell at wowway.com>
Cc: synth-diy at dropmix.xs4all.nl
Sent: Wed, 23 Nov 2011 13:37:03 -0500 (EST)
Subject: Re: [sdiy] DC 'bounce' in second order filters

On Nov 23, 2011, at 6:06 AM, Harry Bissell wrote:

> I have a second order VCF  When I first apply a signal to it (starting from a zero condition) the filter
> have a DC bounce (of one polarity or another) which prevents the signal output from crossing zero for a couple of cycles.
> 
> Is the bounce an unavoidable artifact of a second (or higher) order filter, or can I do anything to reduce it. In my application missing a zero cross is an error condition I'd like to avoid

Harry,

If I'm understanding your description correctly...

Yes, the DC bounce is an artifact of the signal suddenly being turned on at the zero crossing point.  That turn-on, effectively multiplying the audio signal by a step function, does create a DC Fourier component.

For example, if the audio signal was a 1.0 V rms sine wave, and we turned it on for exactly a single positive half cycle, that would be a 0.5 V DC value.  For a full cycle: 0 V DC.  For 3 half cycles: 0.33 V DC.  4 half cycles: 0 V DC.  5 half cycles: 0.2 V DC.  And so on.

I suppose you could compensate for it.  I'm intrigued about the sort of application where the DC bounce would be an issue.

  -- Don

--
Don Tillman
Palo Alto, California
don at till.com
http://www.till.com





-- 
Harry Bissell & Nora Abdullah 4eva