Frequency-Dependant Low Pass Filter

[liquids]

I'm trying to wrap my mind around a concept that would probably be pretty useful for 'tracking' devices, though it will undoubtedly be a complex circuit if feasible.

I'd trying to envision a way to use a guitar note/frequency (fundamental) to simultaneously control a filter - that is, a way that a frequency can be 'read' in short time to feed something like an OTA to determine a (not so resonant) low pass filter (or maybe band pass, ideally) that 'reads' the voltage/frequency and peaks/cuts off just above said frequency in response, to focus on the fundamental.

The Roland GR-300 seem to do something of the sort in a general way that I'm failing to completely understand, even when staring at the schematic for a while....

Anyhow, any ways to turn a note into a frequency-dependant voltage/current is useful...I guess this is in concept, a frequency to voltage converter?

[R.G.]

You're already there. You clean up the incoming signal so it only crosses zero once per lowest frequency, perhaps run that into a comparator to clean it up further, then into a frequency to voltage converter. The National Semiconductor LM2917 is a single-chip F-V converter that does a good job.

The problems:
1. Guitar signal is hard to clean up. It has lots of harmonics that can confuse an F-V, and varies in level a lot during the note.
2. F-Vs are not perfect, and will have ripple, just like envelope detectors.
3. It's a slow process. Short notes are comparable to the time it takes the F-V to react well.
4. 2 and 3 are opposing requirements. Good ripple removal makes for slow response.

Microcontrollers can do a decent job if carefully programmed, but their errors when they miss it are gross ones.

[liquids]

You're already there. You clean up the incoming signal so it only crosses zero once per lowest frequency, perhaps run that into a comparator to clean it up further, then into a frequency to voltage converter. The National Semiconductor LM2917 is a single-chip F-V converter that does a good job.

I've screwed with comparators (LM311) and F-V (LM2907, or is that different?) lately, but not run one into the other.  The sound out of the LM2907 sounded like a note...run through a comparator...  

The problems:
1. Guitar signal is hard to clean up. It has lots of harmonics that can confuse an F-V, and varies in level a lot during the note.

That's the funny thing.  It's always that second harmonic jumping forward on the low fretted notes...as the GR-300 seemed to address thoroughly.

With a quick Sample & Hold though, I think that at the attack of the note-which moreoften seems to trigger correctly-feeding the F-V, I might be able to get some consistency, so the filtering would stay constant per the initial 'sample' note being held and feeding the F-V until a new note 'triggers' the S&H etc?

I know I'm not the first to want this kind of thing so it makes me think 'don't bother trying.'  Maybe it's just a mental exercise.  But set-frequency low pass filtering just clearly doesn't work PERFECTLY with 6 string guitar unless you accept/work around it's limitations...and I teeter between acceptance and denial on a breadboard by breadboard basis.     And switching to flatwound strings....

[Mark Hammer]

One of the nice things about such tracking filters for guitar is that they don't need to be anywhere near as precise as P->V for oscillator purposes.  You can also tolerate a little more lag in the filter tracking than you can in the pitch tracking.

[merlinb]

That's the funny thing.  It's always that second harmonic jumping forward on the low fretted notes...as the GR-300 seemed to address thoroughly.

The GR-300 had the advantage of the hex pickup, don't forget.
GR-300 service manual: http://manuals.fdiskc.com/flat/Roland%20GR-300%20Service%20Manual.pdf

Stompboxology mentions adaptive filters in the sub-harmonic article:
http://www.moosapotamus.net/IDEAS/stompboxology/FreqDiv_SubharmSynth.htm

[R.G.]

I know I'm not the first to want this kind of thing so it makes me think 'don't bother trying.'  Maybe it's just a mental exercise.  But set-frequency low pass filtering just clearly doesn't work PERFECTLY with 6 string guitar unless you accept/work around it's limitations...and I teeter between acceptance and denial on a breadboard by breadboard basis.     And switching to flatwound strings....

What's funny about this is that a good frequency dependent lowpass filter is exactly what you need to clean up guitar signal for frequency-to-voltage sensing.

Back in the sticks where I grew up, we'd call that a self-eating watermelon. 

[liquids]

What's funny about this is that a good frequency dependent lowpass filter is exactly what you need to clean up guitar signal for frequency-to-voltage sensing.

Back in the sticks where I grew up, we'd call that a self-eating watermelon. 

I know!  It hurts my brain and/or irritates me to no end!      

'Give up, amateur.' I'm hoping that a simple static filter will be sufficient at the start of a note (as it usually is) to get the ball rolling (watermelon chomping). 

All the above has been helpful to me, at least for the mental exercise, though successful implementation may yield low (zero) cost-to-benefit ratio results.

[amptramp]

You may need to determine the period between zero axis crossings to make something that works fast enough.  Trying to determine frequency by counting cycles in a given time will be too slow.  Then you can tie this to a switched-capacitor filter.  It might be fast enough, but short notes are going to be a problem with any circuit.

[R.G.]

You may need to determine the period between zero axis crossings to make something that works fast enough.  Trying to determine frequency by counting cycles in a given time will be too slow.  Then you can tie this to a switched-capacitor filter.  It might be fast enough, but short notes are going to be a problem with any circuit.

Yep.
The cussed thing is that strong harmonics give more than one zero crossing per cycle, making it think that the lowest frequency is higher than it actually is. It's that self eating watermelon again.

I've actually been playing with something that might get around the issue of needing to already know what a note's doing to be able to figure out what the note is doing.

With moderately good single chip digital delays, one can delay a signal by a time that's short by human standards but long enough to do something with - maybe. I was thinking of inserting a fixed delay of a small fraction of a second in a signal path to give the circuits a head start on figuring things out. The delay might well be small enough that the human player unconsciously compensated for it by hitting notes a little earlier than the beat.

It's a raw, untested theory. Could be worthless. But with computers I can make more mistakes faster than I can by manual means alone. 

[WhenBoredomPeaks]

This LM13700/13600 application guide talks about a Dolby like noise reduction filter which adjust the filter dynamically to the bandwith of the music. I don't think this will be useful but maybe by some tweaking you can get the filter to detect only the fundamental as the upper limit of bandwith and filter out the rest.

Where i can see this failing is that generally the db levels of music and noise have a much bigger difference than the db levels of the fundamental and the harmonics i think.

http://www.idea2ic.com/LM13600/SUPPLEMENTAL%20LM13700%20APPLICATION%20EXAMPLES.html

[liquids]

Is pursuing this worthwhile?

http://electronotes.netfirms.com/AN354.pdf

[merlinb]

Is pursuing this worthwhile?
http://electronotes.netfirms.com/AN354.pdf

That is functionally identical to the one I linked to earlier (which is the same one used in the Octave Multiplexer and Microsynth)

[liquids]

That is functionally identical to the one I liked to earlier (which is the same one used in the Octave Multiplexer and Microsynth)

Sorry.  I did read (and save) the whole article (which I'd seen at one point previously).

Not to be nitpicky, but do you mean the same one used in the EHX DELUXE octave multiplexer (different circuit than the Octave Multiplexer) and the EHX rackmount EH-8000 Guitar Synthesizer (Far different from the 'Microsynth')?

http://img.photobucket.com/albums/v317/StephenGiles/EH_sine.gif

[merlinb]

Not to be nitpicky, but do you mean the same one used in the EHX DELUXE octave multiplexer (different circuit than the Octave Multiplexer) and the EHX rackmount EH-8000 Guitar Synthesizer (Far different from the 'Microsynth')?
http://img.photobucket.com/albums/v317/StephenGiles/EH_sine.gif

Yes, that's the one I meant, sorry for the confusion! (Know where I can find a schem of the 'non deluxe' octave multiplexer?)

[liquids]

(Know where I can find a schem of the 'non deluxe' octave multiplexer?)

Thanks for clarifying, I thought I knew what you meant, but wanted to know I knew what you meant, if you know what I mean.  

And yes, here is the 'simple' version schematic: http://topopiccione.atspace.com/pjimages/EHOctaveMultiplexer.sch.gif

If you have the time, is there any chance you can give me a 'play by play' of the 'EH sine' drawing linked above?  I have some of it in LTspice but...well, I like to understand, not just build; though the former often takes longer.  

[merlinb]

And yes, here is the 'simple' version schematic: http://topopiccione.atspace.com/pjimages/EHOctaveMultiplexer.sch.gif

Hmm, I'm getting a 404 error for that link...

If you have the time, is there any chance you can give me a 'play by play' of the 'EH sine' drawing linked above?  I have some of it in LTspice but...well, I like to understand, not just build; though the former often takes longer.  

I can try...

Audio comes to the first opamp and gets amplified, then it goes to the sensitivity control. Everything else on the top half of the diagram is simply a compressor- you can see the bridge rectifier that does the rectifying (well,...duh), it feeds a balanced amplifier, and the 1uF cap holds the envelope voltage that controls the OTA.

The lower haf of the digram is the adaptive filter or tracking filter. (The very last stage is just a comparator). Things get a bit more confusing here. There are two OTAs, one is a voltage controlled high-pass filter, the other a voltge controlled low-pass filter. They are cascaded and the output is fed back to create a state-variable filter (bascially a filter with three outputs, HP, LP and BP).

The outputs from the HP and LP filters are each sent to a precision rectifier, one positive, one negative. The outputs from these rectifiers are summed at the 741. If the HP and LP filters are in tune then the magnitudes of their output signals will be the same, so the positive and negative outputs from the rectifiers will cancel and the 741 does nothing. If they are not in tune then one of their outputs will be larger than the other, creating an error voltage at the 741. The 741 is an integrator so its output will ramp up or down, depending on the error, which drives the HP or LP filter in the right direction. You might say that the HP filter is simply being used to 'check' that the LP filter is doing its job properly, since we don't actually need to HP-filter the audio, only LP-filter it.

Finally, the audio is taken from the LP output and turned into a square wave by the comparator (well, it's a Schmitt trigger really, not just a comparator).

[Lurco]

http://liveweb.archive.org/http://topopiccione.atspace.com/pjimages/EHOctaveMultiplexer.sch.gif

[liquids]

Thanks!   

[PRR]

> I'm getting a 404 error for that link...

The topopiccione server declines to serve direct "outside" links.

Copy the URL and paste into a new browser window/tab; it works.

[Rodgre]

I love this thread. I wanted to throw this into the mix: is anyone familiar with the pitch-to-voltage converter circuit in the Korg MS-20/X-911 guitar synth? Does that circuit lend any ideas to you geniuses?

I know that I will probably never have the time or patience to build my teenage dream mono analog guitar synth (and I already have a GR-300) but the idea of having a box to do at least marginally reliable P-V conversion gets me giddy!