Crossover & Mixer Stompbox

This is a neat effect that I have used in software on a few recordings, but have never seen in a pedal form until now.

The idea is this - split the sound from your theremin into high and low frequencies, and process each separately. 

For instance:

1. Dial up a waveform on your theremin with low harmonic content (i.e. a very sine-wavey sound) to reduce spillage from the low notes into the high frequency side of the pedal.

2. On the low frequency loop, first some distortion to introduce some high frequency components, then into a talking machine.

3. Sweeten the high frequency loop with chorus and reverb.

Now you have, effectively, two instruments - a vox humana bass and an ethereal soprano that switches automatically depending on what pitch you are playing. 

http://sfxsound.com/xm/

I note that it uses 4th order filters (24db/octave) - so there will be a small crossover zone where tones of a certain range of frequencies will be sent to both loops in varying ratios. Could someone who understands what 24db/octave means in practical terms kindly expand on what this means in practice (i.e. How wide in semi-tones is the "danger area" where noticable amounts of the signal is sent to both loops)?

[edit] £205 inc. postage worldwide. :-(

Could you not just split the output from the theremin in two and put each into a separate filter (such as this one here, which I have absolutely NO experience of what so ever! a Parametric EQ pedal thang)?

Definitely not as elegant as the X&M.

I'm probably missing something very basic!

Well, yup, it's a possibility. This is a strength of forums - that someone will know, and point out the basic thing. :-)

But that looks like an idea. Boss do a graphic equaliser stompbox for around seventy quid - a couple of those would give a ridiculous amount of flexibility in what goes to each channel. And as it happens my etherwave has two audio outs, thanks to Thierry, (one with an effects loop) and I have a nice little couple of boxes from Jake Rothman of theremin.co.uk that will let me combine two channels (with variable amounts of ring-mod.) So maybe all I need is the boss pedals and a bunch more patch cables. I'll have to think on that. :-)

http://www.roland.co.uk/products/productdetails.aspx?p=144

Well I'm a simple kind of guy so basic does it for me every time!

I've been thinking further on this and can see a couple of potential drawbacks on using the two pedals:

with two pedals (one per split output) you'd get double the noise in the final mix, assuming the X&M does something mystical and only gives one dose of added noise.
Might not be so bad in a live situation but for recording porpoises might be a limitation.

and

you'll need two power supplies, or use rechargables!

I'm liking the idea all the same though!

This is facinating stuff, Gordon -

I came to the conclusion following the comments in this thread, that the best direction was an external "processor" box which took audio from the theremin and added vocal formants that moved with incoming pitch.

The original (and at trhe time I thought easier) was of doing this was to take the theremin HF oscillators and derive the audio frequency directly from these at high speed, to avoid latency - then to use this data to control the movement of the formants and larynx simulation, so one got a change along the lines of "a vox humana bass and an ethereal soprano that switches automatically depending on what pitch you are playing. " except that these would morph into each other rateher than "switch".

The above cannot be done with audio from the theremin - or at least it cant be done on the basis of a "control" signal..

But it can be done, as you describe, by splitting the audio into "tuned" paths and processing each seperately.

There is a lovely circuit designed for tone control - its the Ambler tilt equalizer - its really cheap and simple.. The standard circuit has one potentiometer, when in the centre, incoming signal gets to the output flat - no tone / harmonic change.. At one extreme (say CCW) you get maximum attenuation of bass (about -12db @ 20Hz) and maximum boost of treble (about +12db @ 16kHz) - the bulk of the change occurs between about 100Hz and 5kHz (-8db @100Hz, +8db @ 5kHz)  imagine drawing a line from -12db to +12db on log graph paper.. With the potentiometer fully CW, you "tilt" this line so that theres +12db @ 20Hz and -12db @16kHz (-8db @5kHz, +8db @ 100Hz) ... The exact frequencies depend on selected components - usually the tilt point (the frequency at which 0db is maintained - imagine the slope as being a lever, the fulcrum being this frequency - the potentiometer tilts the lever) is set to about 1kHz. It is possible to make this fulcrum adjustable, but the circuit ceases to be an "Ambler" and becomes a lot more complex.

The Ambler is a lot simpler than a graphic EQ - "deeper" Amblers can be constructed by wiring them in series with ganged potentiometers (I actually use voltage controlled electronic "potentiometers")

In simulations I have fed audio into two seperate "processors" each preceded by a "super Ambler" - these processors have adjustable distortion stages followed by adjustable formant filters, and these are again followed by seperate Amblers with "tube distortion" before being mixed for output.

The schematic for a basic Ambler is shown below, almost any opamp will work, and any supply (+/- with respect to ground) above about +/- 5V.

This Ambler is centred at about 1kHz (1kHz will always be output at the same level as its input, ie, 0db) . Range (-8db @100Hz, +8db @ 5kHz) with wiper fully to the right in schematic below, (-8db @5kHz, +8db @ 100Hz) with wiper fully to the left, and flat (0db) for all frequencies with wiper in the centre. C1 determines the top end roll-off, and can be increased to act as a filter for HF shit. The "centre" point frequency can be moved by changing (increasing) R2 and R3, and/or changing tha values of C2 and C3, but R2 must always be the same value as R3, and C2 must always be the same value as C3 - Also, care must be taken as performance is impaired if one just changes these components without making other changes to the circuit.

The circuit below was not designed by me - it is designed by Douglas Self (THE guru on audio circuits) and comes from his book Small Signal Audio Design - this link takes you to a page discussing the Ambler.

Fred.

That remembers me the tone control of my excellent old British Quad 44 preamplifier which I had unfortunately to sell ways below its value in dark times about 25 years ago...

Not quite sure how the Ambler fits into this, Fred. Sounds to me like it would be very good for morphing as it affects the whole spectrum, if I understood correctly. I guess that you wandered off on a tangent? 

So I think a pair of graphic equalisers is the way for me to go. That will allow me to have a very small transitional region if I want (the more bands, the smaller the morphing region I can obtain) or dial up a smooth slope on each of them to graduate continuously from one sound to another. 

Does that make sense?

NB. Roy - noise is not so much of an issue. Most of my effects are digital, so I have a nice low noise floor to start with, and I have some good tools in Amadeus Pro for denoising that could handle it.

"So I think a pair of graphic equalisers is the way for me to go. That will allow me to have a very small transitional region if I want (the more bands, the smaller the morphing region I can obtain) or dial up a smooth slope on each of them to graduate continuously from one sound to another. " - GordonC

Sorry, I misunderstood - My thinking is on the lines of dynamically changing the timbre as the pitch change, not switching from one voice to another..

Sure, a pair of graphic equalizers will give you better control / switchover points.. If thats what you want - Quite tricky to get it so that a sharp switchover wont sound strange on the transition frequency (even if you had a 48db/octave "switch-over" there will be a band covering several semitones which will sound completely different to the sound below and above this point)

If you really want switch-over, a sharp state-variable filter with LP and HP outputs would probably serve you best - Actually, a digital filter for this function could be the best bet - its far easier to create a sharp tunable filter without overshoot (peaking at the cut-off frequency) using a DSP than it is to implement this in analogue, and it should be reasonabely simple to have the cut-off sharpness and frequency adjustable to allow morphing or switching.

I suppose it comes down almost entirely to what one is trying to achieve.. Im simply trying to achieve / create / enhance the vocal qualities present in classic theremin sound, and add a little bit extra in terms of morphing formants dynamically - In fact, tha last thing I want is for the sound to suddenly switch from an "ooohh" to an "aaahh" as one crosses some particular pitch - I am not trying to create a "talking machine", simply trying to have an adjustable vocal "presence" which doesnt make the theremin sound like a vocalist (I actually think the TM is too realistic, and because its realistic the flaws in its vocal rendition become irritating) but at maximum is just below the level where one could be fooled into thinking its a human voice, and can be adjusted to a level where the effect is just subliminal (or off).

And for the above, the Ambler (or double Ambler, giving +/- 16db tilt) is working well (in simulation) and has an advantage of being extremely simple and low cost compared to multi-band graphic equalizers... I also think there is an advantage of keeping the number of user controls low - Most users dont want 48+ controls to set up the frequency response.. I understand that you need the versatility for the kind of stuff you do, but you are a rare kind of thereminist I think!

;-)

Fred.

"I note that it uses 4th order filters (24db/octave) - so there will be a small crossover zone where tones of a certain range of frequencies will be sent to both loops in varying ratios. Could someone who understands what 24db/octave means in practical terms kindly expand on what this means in practice (i.e. How wide in semi-tones is the "danger area" where noticable amounts of the signal is sent to both loops)?" - Gordon

24db/octave is not particularly sharp - its the same as you get from a Moog VCF - you will hear -24db, in fact you will hear -48db.. So you will hear signals an octave higher or lower.. If you think in terms of frequency controls on a graphic eq, these are usually +/- 12db, and you can certainly hear a frequency that has been fully cut.. even if you crank all the others up to +12db (you can play with this using FFT or graphic EQ applets available on-line)

db's are a log factor - so 6db will give a doubling of voltage and 3db will give a doubling of power and a pecieved doubling of loudness.. +/- 3db was deemed reasonably 'flat' in terms of HiFi back in the '60s.. Its real difficult to give qualitative "meaning" to the effect of 24db attenuation of a specific band, because what you actually hear depends a lot on content - attenuation of a harmonic by 24db can sound highly significant because the brain treats harmonics differently - the effect of a 24db sweeping filter is a demonstration of this - but a seperate sound at -24db can be percieved as a lot louder than a harmonic attenuated by 24db.

If you want sharp switch-over you are, IMO, going to need more than 24db/octave.

Fred.

I have sent you a .wav sweep, split into LF and HF @ 2kHz with 24db/octave cross over... Its easy to produce this in Adobe Audition 3.0 and to play with other orders etc.

Edit -> The first file I sent wasnt 24db/octave, it was based on a 30 band graphic equalizer with 1.6kHz and everything below set to -24db, and 2kHz and everything above set at 0db.. this produced the HP output, these were reversed to produce a LP output..  this is more like a 48db/octave response (except the cut only goes to -24db not -48db).

The 2nd file I sent was through a chebychev 4th order filter with cut-off at 1kHz, LPF on one chan, HPF on the other.

IMO, its really worth playing with some DAW and simulating stuff like this - you can get a really good idea about what hardware will sound like if you feed the specifications into the DAW and mess about with simulated input signals. Adobe is particularly good for this, as (I think) it comes with loads of VST plug-ins (I may have had some of these from other DAWs) and has a great selection of filters - including "scientific" filters based on classic filter topologies.