Basic query about allpass filters - and I mean basic

[Mark Hammer]

A quick look at the average op-amp based phase-shift/allpass stage reveals the following:

• the input signal to the stage is "split", with one path going to the inverting input and the other going to the nininverting input
• the op-amp is set up for unity gain, with the feedback and inverting-input resistor equal in value (often 10k or 100k)
• the noninverting input is fed with a crude highpass filter, made up of a small value cap (often .01uf or close to it) and whatever it is that forms the resistance to ground (FET, LDR, fixed resistor, etc.)

The amount of phase shift created in such a topography increases with frequency, up to some maximum.  As the value of the capacitor increases, and as the resistance increases, the point at which the phase shift increases beyond zero gets lower and lower (as does the frequency where it reaches a maximum of 90 degrees for that stage).

From what I understand, however, there is no requirement that be the resistance of that pair that goes to ground.  It can still be an allpass stage if the resistance is what goes into the noninverting pin, and the capacitor goes to ground (i.e., their positions are reversed).  My understanding is that it is simply more convenient for the resistance to go to ground (especially with FETs, and that is why we normally see what we see in schematics.

Here is where my puzzlement begins.  If the configuration is flipped around, such that the cap now goes to ground, this starts to look like an RC lowpass filter going into the noninverting input, rather than the more conventional CR highpass filter.  Does this now mean that phase shift increases as one goes lower in frequency, in contrast to the CR configuration where it gets higher with frequency?  And if so, does this mean that one can create full-spectrum shifts in phase by cascading combinations of CR and RC-type allpass stages?

The reason why I ask this is some busride-thinking prompted by looking over the PH-2 schematic, as well as thinking about the Phase 100.  Both of these use a quartet of fixed phaseshift stages to increase the intensity of the effect by adding a fixed amount of phaseshift to what is created by the swept stages.  This results in some additional notches being created, but because the phase shift added by the fixed stages tends to apply mostly to higher frequency content, the additional notches tend to disappear when the phaser sweeps lower (which is pretty much where you would want the extra notches to show up).  I was thinking that if extra phase-shift could be added in a more uniform manner by means of flipped allpass stages, the advantage of additional fixed stages could be spread out more evenly.

Is this muddled thinking?

[A.S.P.]

the shift`ll go in the opposite direction (other side of zero... - don`t let Mark P.read that). 
think: lead/lag

[A.S.P.]

soon, Sean, stm, jcm, and Transmo... will post calculations of those "transformations" 
- I have to shut up, again, here - 

http://www.elixant.com/~stompbox/smfforum/index.php?topic=40841.msg294293#msg294293

[MR COFFEE]

ASP said it right.

This results in some additional notches being created, but because the phase shift added by the fixed stages tends to apply mostly to higher frequency content, the additional notches tend to disappear when the phaser sweeps lower (which is pretty much where you would want the extra notches to show up).  I was thinking that if extra phase-shift could be added in a more uniform manner by means of flipped allpass stages, the advantage of additional fixed stages could be spread out more evenly.

Is this muddled thinking?

Umm, I think maybe if I read you right. 

Notches disappearing?    Sounds like the resistance range in the phase-shift leg is not properly centered for the capacitor value selected, or too wide a sweep range so that the notches are going either subsonic or supersonic. Perhaps what you are referring to here?

The staggered-frequency phase shift network like in the UniVibe doesn't have to be limited to four stages, and if you are going for a different phase-shifter sound with a really wide sweep, it may help to stagger the cap values to get more of what you *may* be aiming for sonically. Just guessing what you mean and are after here... Twelve stages can sound really thick if you can stand the noise level.

[A.S.P.]

sorry, Bart, for having left you out in above list of people who can explain better than me... 

(Mark: this issue does have s.th. to do with what I`ve mentioned as "TSP" a while ago...)

[Mark Hammer]

Notches disappearing?    Sounds like the resistance range in the phase-shift leg is not properly centered for the capacitor value selected, or too wide a sweep range so that the notches are going either subsonic or supersonic. Perhaps what you are referring to here?

Actually, the idea was that if there is really no appreciable phase-shift added below, say, 350hz, that the intensity of the effect (and I'll temporarily equate "intensity" with number of discernible notches) woul diminish slightly when the sweep was at its lowest point, relative to what was added by the fixed stages when the sweep went higher.

(Mark: this issue does have s.th. to do with what I`ve mentioned as "TSP" a while ago...)

I was actually thiking only of the phase-shift path, but now that you mention it, I guess a fairly uniform phase shift applied across the full spectrum of the "clean" path could be used to produce a through-zero phasing.

[George Giblet]

> and the capacitor goes to ground (i.e., their positions are reversed).

Yes it still behaves as all pass.  There are actually a number of all-pass circuits around.

>  My understanding is that it is simply more convenient for the resistance to go to ground (especially with FETs, and that is why we normally see what we see in schematics.

Yes.

>Does this now mean that phase shift increases as one goes lower in frequency, in contrast to the CR configuration where it gets higher with frequency?

You have to be a little careful on this one.   On the R to gnd circuit the phase goes 180deg -> 90deg -> 0 with increasing frequency.  On the C to gnd circuit the phase goes 0 -> -90deg -> -180deg with increasing frequency.     In both cases the slope of the phase shift is negative, it just starts at a different point.  It is not as if you get positive phase slope on the C to gnd, 0 -> 90deg -> 180deg.

The most intuitive way to understand the C to ground circuit is that the behaviour is the same as the R to ground with the output inverted (gain -1).     So you won't get a through zero effect, it's more like inverted stereo.

[A.S.P.]

you might have misinterpreted me cc. "TSP", but that doesn`t matter...  <0>

[Mark Hammer]

Okay, fellas.  In which case can one approach two fixed stages the way one approaches an active crossover and use an R-to-ground and C-to-ground stage, with component values well chosen, and have pretty much close to a uniform 90-degree shift across the spectrum?

[MR COFFEE]

Hi A.S.P.

list of people who can explain better than me... 

Thanks for the vote of confidence, but not necessarily
I try 

Hi Mark,

use an R-to-ground and C-to-ground stage, with component values well chosen, and have pretty much close to a uniform 90-degree shift across the spectrum?

Nope. That's not the way it works. But you can get a pretty much close to a uniform 90-degree shift across the spectrum if that's what you are after.

If you want 90 degrees phase shift across the frequency spectrum, you use two chains of the same circuit, either multiple R-to-ground stages or C-to-ground stages, but with different frequencies in each stage in both chains. They are carefully selected to give a fairly accurate 90 degree shift across the audio spectrum. Takes at least 4-6 stages in each string and precision components or "tuning".

It's what Bob Moog used to call the DOME filter. Moog, Harald Bode and others used them in frequency shifters. Jurgen Habile has some nice info on it on his site.

Why don't you tell us what you are trying to do here and maybe we can be more helpful? I thought you wanted to make a phaser that did a deeper phasing

Bart