Moog String Filter

DSP, Plug-in and Host development discussion.
Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Post Tue Mar 15, 2016 7:00 pm

This effect has become something of an obsession for me. It's a Moog effect made up of 40 highly resonant bandpass filters.

https://www.flickr.com/photos/macrutan/ ... 574187729/

These things are pretty rare, from what I gather.

Anyway, I'm wondering if anyone could make an educated guess at the filters used in these things? Maybe a bandpass version of the ladder filter? Does that even make sense?

JCJR
KVRAF
2345 posts since 17 Apr, 2005 from S.E. TN

Re: Moog String Filter

Post Tue Mar 15, 2016 9:24 pm

I built something similar years ago. Think my variant had 24 filters but could be mistaken. Maybe only 12. Fairly certain it was 24.

Back in the 1970's had read a keyboard mag article about jan hammer which mentioned in passing his use of a freeman string symphonizer outfitted with a custom made moog string resonator. So I called moog on the phone asking if they would sell me the circuit.

The nice lady said no we don't sell the circuit but we will send you a copy free. So when the circuit arrived it was a couple of xerox pages of rough hand drawn schematics, and the docs were labeled "Neil Diamond's Freeman Mod". Apparently Moog made one for Neil before he made one for Jan, or whatever.

It was two banks of multiple filters, each filter a single opamp sallen-key bandpass. I got out the calculator and figured the Q but do not exactly recall the values. Purt sure the Q's were bigger than 3 or 4 but less than 10. It was a picture of one example filter and then a hand drawn table of component values for all the filters, and a "macro" illustration how the filter bank should be wired to a freeman.

The freeman as best I recall treated each of its 5 octaves with different passive filtering. Higher octaves passively filtered brighter than lower octaves.

The 24 filters were fed by the freeman's five octaves thru a resistor grid, so that low octaves were mostly fed thru the low filters and high octaves were mostly fed into the high filters, with middle octaves a variable mix between the high and low filter banks.

I built the filter bank and installed in my freeman, along with some additional channels of chaotic analog chorusing. The filter bank did not turn the freeman into the london symphony orchestra, but at the time it seemed to make the freeman a little more "realistic".

Unless it has rotted away, I still have the schematic in a file cabinet somewhere. On the long to-do list was to eventually make a novelty plugin with the same frequencies and Q.

In about the same 1970's period, was a scientific american magazine article about the science of violin building, which along with other information had a table of the many resonances of a well carved violin top plate. Maybe Moog had already thunk up his resonator from other sources, or maybe he got the idea from that scientific american article, dunno. It is hard to remember details that far back. What did I eat for lunch yesterday, anyway?

Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Re: Moog String Filter

Post Wed Mar 16, 2016 8:07 am

Thank you, JCJR. That was a fascinating read. Googling gets some interesting results. John Lord used a Moog String filter on his solo album made in the early 80s. Apparently, Wendy Carlos had one, too.

There was a gentleman who was building his own string filters a few years ago who has since passed away. There's some interesting information on his site.

The very few audio samples I've heard of these things do seem to impart a woody, cello like sound. They appear to do best with simple, single oscillator sounds rather than ensemble sounds.

At any rate, thanks again for your post. What would be great is if someone created a set of impulse responses from them (I asked about that over in the effects forum but got no takers).

JCJR
KVRAF
2345 posts since 17 Apr, 2005 from S.E. TN

Re: Moog String Filter

Post Wed Mar 16, 2016 8:44 am

Long ago examining the chart of resonant frequencies in that scientific american article, which was ONLY for violins, not violas, cellos or contra-bass--

Maybe I messed up the analysis, but the major frequencies at the time, according to that plot, appeared to be approximately 2/3 octave apart, and a primary resonance was near 300 Hz. So at the time I supposed that one might get "in the ballpark" for violin, with a standard third-octave equalizer, boosting every alternate band "starting from" 300 Hz, and cutting the bands in-between. To taste. :)

IOW, on an ISO 1/3 oct EQ, boost 315, 500, 800, etc. And cut 400, 630, 1000, etc.

I used that trick on occasion with other synths. Can't say whether it made it any more realistic or not. One's ear is easy to fool with placebo effects.

There are so many narrow resonances, and possibly the resonances vary among individual violins, or individual cellos. So I wondered whether the exact placement of the resonances would be responsible for any potential beneficial ear-candy, or alternately maybe the ear candy would merely result from having lots of resonances but the exact frequency of the resonances do not really matter so much? Dunno.

For instance people like spring reverbs or echo-chorus effects for ear candy. Look at a freq plot of a spring reverb and it is HUNDREDS of very narrow resonances. Similarly a multi-delay echo-chorus with generous delay can have MANY resonances, but I don't recall counting how many.

I don't know if a spring reverb would sound any better or worse if you invert it-- Turning all the hundreds of resonances into nulls, and turning all the hundreds of nulls into resonances? Maybe to the ear, either version would sound about the same?

Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Re: Moog String Filter

Post Wed Mar 16, 2016 6:20 pm

I've done some experimenting with equalizer plugins as a tool for more string like sounds. My very limited experience is that the boost/cut range just isn't deep enough on your typical equalizer to get the woody resonances of stringed instruments. Plus, as you noted, the peaks and valleys in the frequency response of a stringed instrument are very numerous and narrow, well beyond the reach of most equalizers.

Also, those ranges you mention imply that maybe a comb filter approach might be viable. I've experimented along these lines and I *think* (all too familiar with the placebo effect you mention) that I'm able to get a more string like sound.

I've set up a patch in Cobalt that uses the delay effect along with a sawtooth waveform. The left delay line is set to 7ms while the right is set to 23ms. The delay lines are set to cross feed each other. I have the feedback set moderately high. Additionally, I have the oscillator synched to a master oscillator. The slave (the one you hear) is transposed up two semitones. The idea is to mimic actions of the bow on the string.

With no delay:

http://www.lesliesanford.com/music/temp ... -delay.mp3

With delay:

http://www.lesliesanford.com/music/temp ... -delay.mp3

I've started writing a 'delay resonator' plugin that uses three pairs of stereo delay lines to explore this further (if one pair gets good results, three pairs should be three times better, right? :)). If I can ever get around to finishing it...

JCJR
KVRAF
2345 posts since 17 Apr, 2005 from S.E. TN

Re: Moog String Filter

Post Fri Mar 18, 2016 12:17 pm

Thanks Leslie. Your example sounds fairly "cello-like" even without the delay resonances. Tone is good, but often if the tone is "roughly in the ballpark" it sounds realistic if played like the instrument one wishes to emulate. Your short snip seems played like a cello, with good use of dynamics and vibrato. A more realistic sample, but not played like a cello, would probably sound "more fake".

To my ear on the laptop speakers the delay resonances seem to make it a "little warmer" or whatever. Maybe later will listen on phones or the studio speakers. I don't think the resonances make the phrase "sound worse" than the dry version, so seems of little risk inadvertently ruining the tone with the resonances.

Do you recall the url of the fella's web page, who had been building "modern" resonators? Sounds like an interesting read.

****

Had occasionally thought about building a resonator plugin. Had mostly thought about filter bank rather than delay resonators, though had also considered making a delay bank with similar attributes.

One way would just make a big bank of programmable bandpass filters. Perhaps offer a preset with the frequencies and Q copied from the Moog freeman mod, and a screen editor (or simply let the user make a text file list and load it into the plugin) to make any set of frequencies and Q one might want.

Or a more "automatic" way where there are controls for number of filters, center tuning of the bank, frequency spread around the center frequency, etc. Options to allow spreading up, down, or both directions from the center frequency. Options to select different spreading ratios, from logarithmic spread, up thru linear, and then up to exponential frequency spreading.

For instance with logarithmic spreading, each additional higher-frequency resonator would get closer together. With linear spreading it would be about the resonance spacing of a chorus delay unit, equal absolute frequency between each resonance. With exponential spreading, the frequencies would get farther apart as more high resonances are added.

****

Have wondered if it would really be a "big audible difference" between a bank of fairly narrow bandpass filters, versus a conventional phase shifter layout with a bank of allpass filters mixed with the original signal to make notches? Assuming that the phase shifter layout allows specification of frequency and Q for each allpass filter in the bank. It seems that the majority of phase shifter effects will pile a bunch of allpass filters at the same Fc, so that the combined phase response of the stack of allpass filters determines the spacing of the peaks and notches. A "resonator" phase shifter probably wouldn't LFO or otherwise modulate the center frequencies of the filters, except as an optional special effect.

If I recall, the way you mix an allpass bank with the original signal, you can get two results-- Either a null at DC, or a peak at DC. All the frequencies of the comb filter would be the same, but with one mixing method frequency X is a peak, and the other mixing method frequency X is a null.

Which is about the same with delay comb filters. Depending on the polarity of the mixing, given a certain delay, you either get a peak at frequency X or a null at frequency X. Positive-polarity mixing is slightly more "risky" of amplitudes getting too loud, especially if applying global feedback to make the peaks/notches narrower.

It would be similar with a bank of bandpass filters-- Mixed in-phase with the original signal you would have peaks, and mixed out-of-phase with the original signal you would have notches. Or actually, depending on the percentage of out-of-phase mixing, maybe wide-ish notches with narrower peaks at the center of the notches or whatever.

But maybe with a bandpass resonator, it would often be used 100% wet with no mixing with the original signal?

****

My main concern with a delay-based resonator, is that the peaks and notches are regularly spaced according to the delay amount. Which might not agree with the resonance spacing of real world acoustic instruments. Also, chorus is such an old and common effect, modern ears might identify the "sound" of that kind of frequency spacing and think "that is a delay-mix synth" rather than "that synth sounds a lot like a violin". ?

JCJR
KVRAF
2345 posts since 17 Apr, 2005 from S.E. TN

Re: Moog String Filter

Post Fri Mar 18, 2016 12:40 pm

I noticed on the pictures of that standalone moog resonator, he offered a Q control and several controls to select the mix of various filter banks.

I suppose the filter bank selectors would be used to pick one kind of resonances for a cello track, versus another kind of resonances for a violin track.

The variable Q control, maybe he didn't use banks of single-opamp bandpass filters in this box? Maybe he had some kind of electronically tunable filter in that standalone box?

Alternately, PERHAPS he could get good enough Q control with a bank of low-Q simple bandpass filters plus global feedback. Bigger feedback would make the filters have higher Q. But would be "touchy" with too much feedback creating self-oscillation. Never tried that, dunno if it would be a feasible way to make a variable-Q filter bank out of a bunch of simple analog bandpass filters.

The mix of banks was one problem had thought about, in making a useful resonator plugin. Installing in the freeman, he was able to send each octave into a different mix of bandpass filters. But for polyphonic synth playing into a resonator plugin, it would seem hard to split out low bright synth notes to go to one mix of filters, split out mid bright synth notes to go to another mix of filters, split out high bright synth notes to go to another mix of filters, etc. Simple crossovers wouldn't do the trick, because low sawtooth tones have high harmonics. So a simple crossover wouldn't prevent the high harmonics of low notes from being fed into the high filter bank.

So if there were any virtue of a filter bank for poly-playing, maybe the filter bank would work best integrated into a string synth plugin? That way the synth plugin would know the pitch of every note and know what mix of resonator filters to use.

Doing the strings one instrument/track at a time is probably best for realism, but I liked playing polyphonic. Such as the freeman "feebly" attempted to be an orchestra on a keyboard, the low notes sounding "kinda like" bass, then higher notes sounding "kinda like" cello, viola or violin.

Sometimes there have been sampler patches rather successful in making smooth multisamples like this. Play low and get bass fiddle, play high and get violin. Some such attempts are too gappy at transitions, but have seen some patches which seem fairly successful in smoothly blending all the strings onto a single keyboard layout.

Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Re: Moog String Filter

Post Fri Mar 18, 2016 12:47 pm

JCJR wrote:Do you recall the url of the fella's web page, who had been building "modern" resonators? Sounds like an interesting read.
http://www.jhaible.info/string_filter/j ... ilter.html

There are also a few forum posts here and there you can find through Google where his worked is discussed as well as string filters in general.

I'll respond more fully to your posts later (thanks for your kind words about my cello test) but wanted to get this link to you in case I'm not able to respond before next week.

camsr
KVRAF
6859 posts since 17 Feb, 2005

Re: Moog String Filter

Post Fri Mar 18, 2016 3:12 pm

Your delay sounds very good, can't wait to try it.
Image

JCJR
KVRAF
2345 posts since 17 Apr, 2005 from S.E. TN

Re: Moog String Filter

Post Sat Mar 19, 2016 8:58 pm

Leslie Sanford wrote:
JCJR wrote:Do you recall the url of the fella's web page, who had been building "modern" resonators? Sounds like an interesting read.
http://www.jhaible.info/string_filter/j ... ilter.html

There are also a few forum posts here and there you can find through Google where his worked is discussed as well as string filters in general.

I'll respond more fully to your posts later (thanks for your kind words about my cello test) but wanted to get this link to you in case I'm not able to respond before next week.
Thanks Leslie. That is a fun site. Doh, I forgot that there had been electronotes articles about resonators.

Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Re: Moog String Filter

Post Sat Mar 26, 2016 6:13 pm

Here's another test example:

http://www.lesliesanford.com/music/temp ... st-001.mp3

I'm running a Tetra synth through a VST prototype plugin that's using two stereo delay lines. Basically, it's four feedback comb filters. The sounds is kind of cello like, but it needs work. I think if I get a UI on this plugin, I can fine tune the delay times to get better tuning (as it is I'm hard coding the delay times which doesn't make experimentation very easy).

I find that feedback comb filters give a bit more bang for the buck than an array of bandpass filters. I set up another prototype plugin with 40 bandpass filters, then attempted to tune them using an impulse response from a violin as a guide. The results were kind of violin-like, but not as much as naively running the sound through some comb filters.

I'll keep experimenting.

User avatar
Chris-S
KVRAF
2602 posts since 10 Nov, 2013 from Germany

Re: Moog String Filter

Post Sun Mar 27, 2016 1:08 am

Leslie Sanford wrote:I find that feedback comb filters give a bit more bang for the buck than an array of bandpass filters.
The peaks and notches in the Moog string filter are distributed logarithmically, while they are linear distributed when using a Comb filter.

What comes closer to the real distribution e.g. for a violine?
Last edited by Chris-S on Sun Mar 27, 2016 1:37 pm, edited 1 time in total.

camsr
KVRAF
6859 posts since 17 Feb, 2005

Re: Moog String Filter

Post Sun Mar 27, 2016 4:26 am

Well, harmonics are linearly spaced in frequency, but does it sound better?
Image

User avatar
Chris-S
KVRAF
2602 posts since 10 Nov, 2013 from Germany

Re: Moog String Filter

Post Sun Mar 27, 2016 4:48 am

I meant the resonance peaks, not the harmonics.

Leslie Sanford
KVRAF
1595 posts since 4 Dec, 2006

Re: Moog String Filter

Post Sun Mar 27, 2016 6:34 am

Chris-S wrote:What comes closer to the a real distribution e.g. for a violine?
A typical frequency response for a violin looks something like this:

http://electro-music.com/pm_tutorial/Bow-WoodenBody.htm

I'm not sure if I can answer your question, though. With regards to my preference for comb filters, it could be that a logarithmic distribution of resonant bandpass filters would be better (the article above prefers bandpasses over combs) and that it's my implementation that's not doing the job.

Return to “DSP and Plug-in Development”