Splitting a Guitar Signal

[cps]

Hi Folks,

I'd like to experiment with applying different guitar effects to different parts of the frequency spectrum of my electric guitar's output - more simply, I'd like to use different distortion (and other) effects for low and mid/high frequencies. To do this I need to split the guitar's output signal. Would something like this http://sound.whsites.net/project148.htm be suitable?

Also, if for example, I use a distortion on the mid/highs but nothing on the bass will I run into phase problems when I recombine the two signals?

Cheers,

Chris

[moid]

I don't know enough to tell you if that circuit will be suitable, but you will run into trouble recombining the audio paths if one of them has gone through an effect with an odd number of transistors in it because that will invert the phase of the signal. An effect that doesn't use transistors or only uses even numbers of them (2, 4, 6, 8 etc) should be fine. You may also find on recombining that one path is much much louder than the other, so you may need to consider a way to limit the volume of the audio coming in when you re-combine. Good luck  - and probably a good idea to breadboard this first.

[wavley]

Luckily this has already been solved years ago for you by Craig Anderton with the quadrafuzz.  http://www.paia.com/ProdArticles/quadrafz-design.htm

I don't know enough to tell you if that circuit will be suitable, but you will run into trouble recombining the audio paths if one of them has gone through an effect with an odd number of transistors in it because that will invert the phase of the signal. An effect that doesn't use transistors or only uses even numbers of them (2, 4, 6, 8 etc) should be fine. You may also find on recombining that one path is much much louder than the other, so you may need to consider a way to limit the volume of the audio coming in when you re-combine. Good luck  - and probably a good idea to breadboard this first.

Basically, as Moid points out, pay attention to phase when mixing.  If you put a volume control on each path and passively sum those signals you should be fine, maybe you'll need some make up gain after summing, but that's about it.

[antonis]

It doesn't matter the actual number of transistors but the number of phase reversals..
(e.g. a buffer's out is always in phase with the signal..)

In general, you have to count the phace reversals (Collectors of CE amps and/or Outputs of inverting Op-Amps) so you have 50% possibilities not to face phase problem..
(for the rest 50% just add a unity gain amp - BJT/FET or inverting Op-Amp)

[swever]

Following this thread.

As for phase – in my understanding any filters offset the phase of the signal. To make things worse, it's not the same degree of rotation for different frequencies. Each filter position would have its frequency response (in relation to amplitude) as well as phase response (in relation to frequency). That is probably a layman's and uncientific explanation but I'm pretty sure it's not too far from reality.

In practice that means that you are always messing with phase when you do any filtering, including crossovers. The question is how well the individual bands will combine back at mixdown.

As for the processing you do inside the bands, it's the same priciple. Anything that just flips phase is relatively easy to fix. Non linear (in relation to frequency) phase manipulations, like filters, may result in unpredictable tone changes when mixed down with other bands.

Unfortunately I can not givea any practical solutions or advice other than to try and see if it works or not. Sometimes the "weird" phase things result in a desirable tonality.

*disclamer* This is just the stuff that I learned from my amateur sound engeneering and production experience and may not be correct or not translate to electronic engeneering directly.

[R.G.]

As for phase – in my understanding any filters offset the phase of the signal. To make things worse, it's not the same degree of rotation for different frequencies. Each filter position would have its frequency response (in relation to amplitude) as well as phase response (in relation to frequency). That is probably a layman's and uncientific explanation but I'm pretty sure it's not too far from reality.

No, it's not far from reality at all. In fact, it's a good first approximation to reality, and a useful rule of thumb for people who don't want to dig through the math.

The next step toward the math at the bottom is to realize that for each reactive part - capacitor or inductor - you get a maximum of 90 degrees of phase shift. For the pure component, current and voltage are always at 90 degrees to one another, but we don't use these parts that way. They're nearly always hooked to resistors in a way that lets the resistor dominate the frequency response at frequencies where the resistance is much bigger than the reactive impedance.

So for a capacitor, the cap is always a DC open circuit. As you feed it AC, increasing the AC frequency, the cap has an impedance that declines with increasing frequency; double the frequency, half the impedance. So for frequencies "far away" from the place where the cap impedance is equal to the resistor impedance, either the resistor forces the phase shift to be almost perfectly to zero degrees, or the cap forces the phase shift to be almost perfectly to 90 degrees. The frequency where the capacitive impedance equals the resistance is taken as the corner or rolloff frequency of this simple R-C filter. Not surprisingly, the phase is 45 degrees right there.

For all >>analog<< filters, you can't have any change in signal size/magnitude with frequency without some phase shift. How much phase shift depends on how close the frequency is to the "rolloff" frequency of that filter. It is possible to concoct filters that have phase shift, but not magnitude change, as in the "all-pass" filters we use in phase shifters. But phase shift you have with you always in a filter.

[cps]

Luckily this has already been solved years ago for you by Craig Anderton with the quadrafuzz.  http://www.paia.com/ProdArticles/quadrafz-design.htm

Thanks but I want to experiment with more than just distortion.

[amptramp]

There is always the option of having separate outputs from the guitar for the pickups at the bridge and the neck.  The one nearest the bridge will get predominantly high frequencies whereas the one at the neck will boost lower frequencies.  You still have to pay attention to the phase.  There would be different signal levels but you can compensate for levels in the outboard electronics.  You might not need (or want) frequency filtering since it also affects the phase of the signal.

[MaxPower]

Luckily this has already been solved years ago for you by Craig Anderton with the quadrafuzz.  http://www.paia.com/ProdArticles/quadrafz-design.htm

Thanks but I want to experiment with more than just distortion.

I think the idea is that you can replace the four fuzzes with whatever you want.

[GibsonGM]

This will do exactly what you want, CPS...

http://runoffgroove.com/splitter-blend.html

[cps]

This will do exactly what you want, CPS...

http://runoffgroove.com/splitter-blend.html

Doesn't that split the signal into two identical signals - that's not really what I want. What I'm after is a low and mid/high signal.

[bool]

You will come up with an effective band-splitting circuit (that works for you) eventually; it's not rocket science by any means and there's lots of workable circuit snippets around.

What others have mentioned, I suggest that you look into phase/polarity switching on the "receive" side of your splitter. That can be used for signal-integrity "troubleshooting" - and also for purely "sound-design"; tone mangling purposes.

[GibsonGM]

You can start with 2 of the 'same signal', then filter each one independently, R-C.   

I think you might be over-thinking the phase shift 'problems' - should be pretty easy to breadboard what you want to do to assure yourself that you're not running into any issues...and the splitter-blend allows you to reverse phase if needed.

[antonis]

You can place Low Pass Filter on (say Green) axis and Mid/High pass filter on (say again, RED) axis to the circuit proposed by Sir Mike.. 
(in the way like in Quadrafuzz..)


In you don't decide (or tell us..) how many and what effects you want to involve, a general purpose "passive" splitter may fit to your need..



edit: You actually aren't faster Sir, 'cause I had to draw the above schem.. 

[wavley]

Luckily this has already been solved years ago for you by Craig Anderton with the quadrafuzz.  http://www.paia.com/ProdArticles/quadrafz-design.htm

Thanks but I want to experiment with more than just distortion.

If you remove the clipping diodes, it does exactly what you want... split into four frequency ranges with an opamp for gain to drive whatever you want and then mix back into a mono signal OR four outputs. 

[marcelomd]

VFE's Triumvirate splits the signal in three bands, and applies a different distortion to each one before mixing again. There's a schematic available on the site.

[PRR]

> you might be over-thinking the phase shift 'problems'

The classic Fender tonestack has a deep phaseshift "problem". At extreme settings it nearly nulls 700Hz.

You can do great violence to musical signals, as long as *you* have control of where the dents land.