Basic bonehead question about adding stages to a phase shifter

[Mark Hammer]

All other things being equal - i.e., no negative impact on sweep, or other change in range - does adding more stages (2, 4, etc.) to a phase shifter result in adding notches below the existing range?

I ask because attempts to use the expansion boards for the Ropez, or experiments where I've added a few fixed stages to a Phase 90, seemed to increase its propensity to "howl".  I understand the risk of howl stems from small cumulative deviations from unity gain over stages that translate into a feedback signal that is greater amplitude than the input signal.  But even with the feedback turned down, the howl is much lower than the original build - minus additional stages - would sweep.

If my hunch is correct and adding more stages moves the starting point downwards, I gather this would suggest a suitable adaptation is to reduce the value of the caps in the phase shift stages.  So, for example, adding more stages to a Phase 90 might warrant exchanging the stock .047uf caps for .039uf.

[armdnrdy]

I don't have a technical answer for you but it might help to take a look at the Jetlyzer schematic.

http://www.dirk-hendrik.com/temp/Ibanez_Jetlyzer.pdf

That 8 stage phaser incorporates the same .047uf caps and 10K resistors for each phase stage.

[Seljer]

Adding more stages of the same filter adds more notches above, while slightly lowering the lowest notch, in the same way stacking two identical first order filters to get a second order filter result in lower cutoff frequency than of the individual filters by themselves.

In a simple two stage phaser the phase response of allpass filter goes from 0° to -360° (with a notch provided where it goes through -180°)
If you add more of the same stages, you end up with slightly lower corner frequency, except now the reponse goes from 0° to -720°  (with a notch at -180° and -540°)

Heres from a simulation of the typical jfet phaser ciruit, blue is the signal after two stages, green is after four stages, which is then mixed with the non-filtered signal resulting in the red response

A two stage phaser would have one notch at about 10.5kHz

[Paul Marossy]

I don't have a technical answer for you but it might help to take a look at the Jetlyzer schematic.

http://www.dirk-hendrik.com/temp/Ibanez_Jetlyzer.pdf

That 8 stage phaser incorporates the same .047uf caps and 10K resistors for each phase stage.

Wow, that's crazy looking. 

[Mark Hammer]

As near as I can tell (and this assumes Dirk reversed the board and drew it accurately), the "Jet Tone" is just a noise source combined with the input signal.  The sidechain, formed by Q16a/B and Q17A drives an optoisolator that brings in the noise source when the input signal is strong enough. 

Strategic...but weird nonetheless.

[Jazznoise]

Changing the all-pass's cap value will surely just change the phase shift value - ie: where you were spacing each notch an octave, you'd now be spacing it closer to a 5th. So the "texture" or "density" of the phase shifters sweep would be altered?

That is indeed a badass looking phaser.

[Mark Hammer]

Changing the all-pass's cap value will surely just change the phase shift value - ie: where you were spacing each notch an octave, you'd now be spacing it closer to a 5th. So the "texture" or "density" of the phase shifters sweep would be altered?

Nah.  The notches occur at those points in the spectrum where the cumulative phase shift across stages happens to be 180 degrees, 540 degrees, 900 degrees, etc., at that particular point in time.  Identical-value caps, whatever the value happens to be, tend to yield deeper notches than those instances where the caps are different values.  That's why Uni-vibes sound less "focussed" than phasers do.  Shifting the values of all caps upward or downward by the same amount has the same net effect as adding or subtracting an LFO offset voltage.

I did hear a phaser once that used a different order of all-pass filter and appeared to produce more closely spaced notches, bt that may have been an auditory illusion.

[Devius]

Ok, real dumb question but please humour me...

Notches are where the frequency stops and reverses direction?

[Mark Hammer]

The notches occur where the total amount of phase shift is enough to place the input signal perfectly "out of phase" with a phase shifted version of itself.

If you were to split an input signal, and combine an inverted copy of the signal with itself, those two would cancel out completely; like matter and anti-matter.  The reason you would not hear anything at all is because every last teeny bit of the input was placed 180 degrees out of phase, via the inversion. 

With a phase shifter, there is no complete inversion of the entire signal.  Rather, the amount of phase shift is different for different parts of the spectrum.  The point/s in the spectrum where there just happens to be enough total phase shift to produce "inversion" and cancellation keeps moving around.  What we call "notches" are those places across the spectrum where the inversion and cancellation is most complete (keeping in mind that there is still a large quantity of spectral content to the signal where there is no cancellation whatsoever).  Adjacent parts of the spectrum, where the amount of phase shift is greater or lesser than 180/540/900 degrees, results in reduced/incomplete cancellation.

So, you are sort of kind of on the right track, except that "reverses direction" is perhaps not the right terminology to use.  Perhaps the most accurate might be "where the input signal and phase-shifted signal are closest to being electronically opposite".

[Thecomedian]

mhmm.

[Mark Hammer]

Sure you're in the right thread, buddy? 

[Devius]

I dig it.