Let's talk formants.

[Kipper4]

From reading the geofex

http://www.geofex.com/article_folders/wahpedl/wahped.htm#multfb

I sort of get that "more than one wah at a time" will help make a more vocal sound.

Here's the geofex on formants

http://www.geofex.com/article_folders/wahpedl/voicewah.htm

I've cludged together an envelope detector. Which is still undergoing development and testing

http://www.diystompboxes.com/smfforum/index.php?topic=113709.msg1053377#msg1053377

I've seen the talking pedal filters. Two mfb filters one fed a clipped signal the other a moxed clipped/clean signal.

I tried breadboarding the talking pedal filters along side the envelope detector (Reply 14 from the later thread) with some tweeks so it's lower gain etc.
Needless to say it needs some work.

I can't say I was ever so impressed with the filters performance with a vactrol set up.
I tried Dead Astronaughts light wah filter which performed marginally better. I tried this one because I know it will work with vactrol. But it's got a long ways to go to bode well in this application.

I have also considered using Twin T transistor filters.
So I'm looking for suggestions for which filter might best suit such an application.
Some values might help too.

I'm kinda pissing in the wind with the filters.
From looking at the geofex formants page I think I should be looking for filters
One in the region of 300hz to 1khz
And a second in the region of 1khz to 2khz

Right ?

Thanks for any help provided guys.

[anotherjim]

Even more vocal than Freppo's Sentient Machine?
I did breadboard 2 Q&D filters working in opposite sweeps. Worked it from an expression pedal, thinking that would help to better tune the effect to what's being played. Vowel was produced, but only during fast sweeps, I was hoping for a more consistent effect. Got on with something else instead.

[Mark Hammer]

Formants don't:
a) each cover the same  amount of spectral content
b) begin and end at the exact same moment
c) "move" the same distance
d) have the exact same amplitude

That's not to say that any attempt to use multiple bands that does not conform to these properties will sound lousy.  If the Bass Balls and Bi-Filter follower are any indication, it may sound very musical.  But if one's goal is to produce sounds that more closely approximate the human voice, those are the qualities you gotta pay attention to.

[Transmogrifox]

The state variable filter is probably the best for experimentation.  You can easily tap into the high pass, band pass and low pass outputs and even combine them for different filter responses.  You can change the Q independently from the resonant frequency (this could be very useful when experimenting with vocal sounds).

Using high frequency pulse-width modulation on an analog switch is a good way to go when you want to control multiple parameters with more precision.  Take a look at the MXR envelope filter for an idea of how the state variable filter was implemented with CMOS inverters and the analog switches.

What would be nice is if you could start your efforts by making a PCB with a state variable filter with a CV input for each parameter:
LP output mix + switch for phase
HP output mix + switch for phase
BP output mix + switch for phase
Resonant Frequency
Q

Once you have such a board, make several identical modules to combine an modulate in different interesting ways.

One possible help is look up Paul Nasca and ZynAddSubFX -- it's a software synthesizer that implements formant filters.  This effect was done with digitized state variable filters.  The code very roughly followed the model of an analog formant filter implementation.

Anyway, somewhere in the code I seem to recall a reference to a white paper or article on this topic in which you could get more ideas on how to modulate the formant filters to make specific vocal sounds.

Going to the extent of creating an analog voice synthesizer is probably far beyond where you want to go, but that general area of technology and study has some valuable ideas that can be simplified and used for less sophisticated guitar effects.

I would guess that a generic bank of 4 state variable filters would give you enough hardware to do some really interesting things.

As Mark suggested there are a lot more variables you may want to move around in order to create formants.  That's why having output mix control (amplitude control) will be useful:  you can also modulate the amplitude while you're also modulating other filter parameters.

The complexity of what you have to do all depends on whether you want a musical effect that sounds "vocal" or whether you want to emulate specific vowel or consonant sounds.

At some point a vocoder makes more sense -- that would be a bank of 8 state variable bandpass filters staggered like an EQ, but with amplitude and Q moduluation -- which requires another 8 band pass filters and 8 envelope detectors...I implemented one in software -- lots of fun:
https://www.youtube.com/watch?v=geJQpabZGZQ
About 2:18 is where you start hearing something

[Kipper4]

@Jim
Thanks dude. It's defo something to bare in mind.

@Mark

Yer I appreciate that. I don't expect it to take the place of a talk box just to be more vocal than you're average auto wah.

I breadboarded a 9v single supply version of the Bi filter follower with your mods. Thanks for that.
With a few mods of my own.
It's a very good starting point for developing what I would like to experiment with.
So that would be for the filters to an extent to sweep in opposing directions and see if that will produce more vocal sounds if you will.
I don't know if it will work or not but I'm time rich, money poor so it would be an achievement for a country boy Like me with just a few years in pedal building experience to make something useful.

Thanks for the help.
Feel free to chime in guys.

Rich

[Kipper4]

Good point Transmogrifox.
I did see some of the cmos inverter circuits an yer the state variable might well be a great option too.
As far as achieving some variability in the way the filters respond to a single envelope detector goes I was kind hoping that using two home made vactrols with there inherent manufacturers tolerances differences and mismatched type and size of leds.
Although this kinda works quite well in the modded bi filter I breadboarded.
I did notice it was more vocal when the sweep knob was used as well. Maybe that's an option to make that a vactrol too. Just thinking aloud.
I do realise that I might get more consistent results from a much more circuit and to be honest I'm at this stage not against the idea.
The double envelope detector or maybe even quad is something I considered.
Maybe some more research into the formant subject is required too on my part.
Understanding the perameters needed for making specific vocal sounds would at least show what might be achievable in the analogue domain.

Vocoder you say hmmm.
Don't tell D.A. He will have a duck fit.

Rich

[Kipper4]

For reference here's the bi filter mods by Mark

http://www.lynx.bc.ca/~jc/biFilterHammerMods.gif

[Kipper4]

Ok so I did a bit more faffing around on the breadboard, listened to a couple of dipthonizer vids on the tube, read a bit more and I think the wow sound is one of my favourites.

I'm still at a bit of a loss as to how to make it work consistently.
Is there a wow sounding wah that I could plaguerise the filter from?

I've been drawing a svf in eagle too and a couple of envelope ideas.
Some guidance perhaps from those that have built such a thing.

Thanks Ryan for the mod and board ideas.
I couldn't find the white papers but I'm sure some will turn up.

[Transmogrifox]

I was able to get a wow sound from a crazy envelope filter I built years ago based on a SVF.  The trick was using low pass output and lower Q than a typical wah.  The filter sweeps lower (like down to 250 Hz) and doesn't go as high (not above 700 Hz).  I think you would be on the right track using any type of low pass filter that you can tune with a resistor.

One such filter is my SKwah, which PRR thinks is better described as a gyrator than a Sallen-Key...but either way it's a 2nd order low pass that you can sweep with a single resistor.  The experiment was to make a non-inductor circuit that emulates the inductor wah (Vox, Dunlop...).  The transistor version is good for getting Q up to something comparable to the typical Crybaby.  More Q can be had by using an op amp instead of discrete transistor design as presented.
[url]http://www.cackleberrypines.net/transmogrifox/TransmogriNotes/skwah/index.html[url]