Comb filter fans, check this out

[DiyFreaque]

Ken Stone has come out with this, the CGS57 Bi-N-Tic filter:

http://www.cgs.synth.net/modules/cgs57_bintic.html

I know it's a synth module, but here is another thing that might sound wild in a stomp box.  

Culled from the Synth DIY list is a description of Grant Richter's (creator of the Wiard Synth) for a version he built back in '99:

"The circuit produces the sharpest comb filters that I have heard.
The sound is very similar to a flanger or phase shifter.
While the original design used an ultrasonic clock,
by far the most interesting sounds are produced by
clocking it in the audio range. It sounds really good
with a rhythm box driving a sequencer controlling the
clock frequency. "

Ken's design can filter and also produce signals.  Note pin 8 of the 4051's should be tied to 0V instead of -7V (Ken mentioned that on the CGS list.  The boards are correct, though).

Cheers,
Scott

[ExpAnonColin]

Very intriguing...  How is the comb filter's spacing determined?  A filter that is controlled by a clock frequency... very interesting.

-Colin

[puretube]

reminds me of an old "Elektor" project from the early 70`s
(which I never got working properly...) pre-BBD-chip era,
which had 40 100n caps switched to ground sequentially with 7400 (?) NANDs... (or some similar opencollector gates - i forgot).
[S/N ratio close to 6dB]
:lol:

[Johnny G]

would you believe i was actually looking at this last night. had a look around the site while looking for a simple and fun filter project that i could build out of mostlly discrete transistors (me no have any op-amps )

[Transmogrifox]

Very intriguing...  How is the comb filter's spacing determined?  A filter that is controlled by a clock frequency... very interesting.

-Colin

If we can indeed assume that his theory of why it works is correct, then the filter (center frequency) spacing is determined by the RC time constants set by the array of capacitors:  1/x, as we are talking about a 1/(2piRC) function, and each paralell cap adds an integer multiple of one cap.  That's his claim.

A problem here, is that we have a transient analysis situation here.  I certainly think it would be interesting to sweep the circuit on an spectrum analyzer and see if the overall response isn't approximately comb filter at those frequencies as he claims.  There's so much other stuff going on here:  You have an input variable that changes with time, and filter response defined by the time variable, that is, a time-varying impedance.  You have "1/RC" as a time constant, but when "R' is a function of time, this assumption breaks down, as the transform that is used to derive this simple algebraic expression is computed assuming R and C are CONSTANTS.  If they become functions of time, then the transform equation becomes significantly more complex than "1/RC", and likely does not break down to "1/RC(t)", such that skipping "C(t)" around at ultrasonic frequencies may not necessarily result in a comb filter.  This filter is a basic state variable biquadrature filter, so we might be tempted to think "what if I just had a bunch of these in parallel and added the outputs together, then I'd have a comb filter...or I could change its center frequency so fast that it appears to have a center frequency at all those frequencies at the same time"....the former is the theory.  Unfortunately, any filter response derivation usually sets the initial condition to 0 and assumes the transient response has passed by the time we're concerned about what the filter is doing.  when you're switching filter parameters faster than the transient response of the filter, you generate a weird sequenced nonlinear distortion...and filtering of some sort.  I'm sure as hell not going to do the math to figure that one out--but I bet it sounds pretty cool, who cares what it's ACTUALLY doing as long as it sounds cool.

No doubt it's a weird filter, and a very interesting one at that.  I think I'll build one sometime--and I imagine it does generate some very radical sounds, not to mention the modulation cross-terms.  Changing the clock frequency probably does some of the most radical stuff--This is how we know "comb filter" is kind of a lark.  It's a very nonlinear filter that adds as much as it takes away.  Very very cool.

[DiyFreaque]

This is how we know "comb filter" is kind of a lark.

Hmmmm....

I have never known Ken to state that any design of his does something that it does not do.  He is more prone to understatement, if anything.  His work is excellent and incredibly prolific - google on "Ken Stone" Synth and take a look at what pops up.  

I can't say that I even begin to understand this filter (my understanding ultimately will come from the breadboard, and just try googling Biquart Filter    ), but I would not label the comb filter description as a lark without giving it a go.  I say this based simply on Ken Stone's work and the origins of the filter in Electronotes, both of which carry some heavy-duty credentials.

Cheers,
Scott

[ExpAnonColin]

As my understanding level rates about at the same as Scott's or, probably lower, I have to say that even if it's not a comb filter per se it sounds quite interesting, it definitely falls into the realm of "new effects" if anyone stompboxicizes it.

That being said, I wonder how much of the controls could be omitted...  as a stompbox probably wouldn't need the sync or FM controls.  

Oh, and I meant to ask, how does one "thermally couple" a transistor to a resistor?

-Colin

[puretube]

thermo-goop & clamp;
or both mounted onto a heatsink close together + thermal compound...

[Mark Hammer]

I wonder how much of the controls could be omitted...  as a stompbox probably wouldn't need the sync or FM controls.  

Synth modules tend to be a bit "overcontrollable", from a guitarist's perspective.  Of course, one needs to understand that it's not so much because of any desire for control overkill, but rather because the production of more natural sounds requires modules to adapt in a more flexible way.  So, it is common for filters to have control optins for not only an envelope generator CV, but to have the initial frequency track the pitch of the note being played, to permit some modulation of the cutoff/centre frequency, and other things.  If you plug a guitar into an envelope controlled filter, there are tons of things you can vary with your picking hand alone, which require no adjustments to the filter other than sensitivity in order to produce a broad range of tones.  Synths need a little more help in that regard.  Of course, if you feed such a filter a complex and easily modifiable input signal, then many of those control inputs CAN be omitted with sacrificing  much of your tone palette.

[StephenGiles]

I would say that this circuit is well beyond 98% of posters here, myself included. I would like to hear a sounclip just to hear if it is worth any time - being absolutely honest. But for the 2% - an unveiled challenge indeed!  
Stephen

[LP Hovercraft]

Thanks a bunch for the link.  There's a lot of great stuff there that I'll have to try.  Synth comb filtration sounds incredible with cymbals (or White Noise).  I'm ordering a PCB right away!  

[Tim Escobedo]

Don Lancaster has a fairly easy to understand description of the switched cap comb filter here:
http://www.tinaja.com/glib/muse91.pdf

[DiyFreaque]

Don Lancaster has a fairly easy to understand description of the switched cap comb filter here:

Oooh!  Thanks for the link.

Cheers,
Scott

[Transmogrifox]

This is how we know "comb filter" is kind of a lark.

Hmmmm....

I have never known Ken to state that any design of his does something that it does not do.  He is more prone to understatement, if anything...

Cheers,
Scott

Don Lancaster has a fairly easy to understand description of the switched cap comb filter here:
http://www.tinaja.com/glib/muse91.pdf

Thanks for putting me in my place.  It turns out my interpretation of what the circuit does was off.  I read Don Lancaster's article and things began to click.

This filter is not switching caps in parallel, rather, that chip is an analog multiplexor, thus it is sampling the input in each capacitor, and the stored charge is discharged at a later time.  This, then, DOES seem a more intuitively reasonable method for creating a comb filter.

I would contend that if the circuit operated how I initially thought it did, then all of the above I said is correct.  It was my bad in misinterpreting the circuit in the first place and trying to be the typical engineer that has to find something wrong with the other guy's idea  :evil:  :twisted:  :wink:

So I over-did it on the emoticons.  Thanks very much for that link to Lancaster's article.  It was good reading.  Now I really want to adapt this thing to guitar.  It would fit into my current envelope filter module very well.

[puretube]

probably based on this: http://www.pat2pdf.org/patents/pat3526858.pdf...

and usable for this adaptive filter: http://www.pat2pdf.org/patents/pat3445685.pdf