feedback in a 2-stage phaser

[thunderaxe]

hi all,

i'm in the process of designing a phaser with switchable stages and from most of the circuit designs i've seen it seems like people generally try to avoid feedback for 2-stage phasers -- the MXR phase 95 and pedalPCB XC phase just put the feedback control after the 2/4 stage switch so that it's only in the 4-stage path, and another thread i read about building a phaser circuit starting with 2 stages and then expanding to 4 and 6 said, "now that we have 4 stages we can add a feedback control"

which makes me wonder why -- is it just not useful in 2 stages? does it not work? does it cause problems?

one possible answer i've come across was a comment from mark hammer saying that for different phase reasons it's best to send feedback to odd numbered stages, which is how the phase 90/95 does it. but my two favourite phaser circuits, the boss PH-1R and mu-tron phasor ii, send it back to the beginning of the circuit, which would be 4 and 6 stages respectively, but also before the input buffer op amp, so while it's 4 and 6 phase stages, it's 5 and 7 op amp stages each. does this solve whatever problems there might otherwise be with feedback in 2-stage phasers specifically, or even numbers of phase stages more generally?

thanks!!

[jorg777]

I'm aware of at least one Eurorack phaser (NextPhase by AJH) which has independently switchable number of stages (up to 12) and which stage the feedback is taken from.  I believe the feedback always goes to the first stage.  Polarity of the feedback will be negative if tapped at an odd stage, or positive if tapped at an even stage.  I notice the minimum feedback tap is #4, while the minimum output stage is #2.  Also, there is a 2nd output, for stereo phasing; this is tied to the feedback tap.

[Mark Hammer]

There are two general approaches to use of feedback in phasers.  As noted, in one of them, the feedback path is generally to N-minus-one stages.  So, if there are 4 stages overall, feedback goes to stage 2, such that the feedback is recirculated through 3 stages (stages 2, 3, and 4)

A second approach is to feed the mixed signal (dry + wet) back to the input stage, rather than the 2nd phase-shift stage.  That is, the "result" travels all the way back to "the beginning".  Somewhat different sound.

[thunderaxe]

There are two general approaches to use of feedback in phasers.  As noted, in one of them, the feedback path is generally to N-minus-one stages.  So, if there are 4 stages overall, feedback goes to stage 2, such that the feedback is recirculated through 3 stages (stages 2, 3, and 4)

A second approach is to feed the mixed signal (dry + wet) back to the input stage, rather than the 2nd phase-shift stage.  That is, the "result" travels all the way back to "the beginning".  Somewhat different sound.

here is the PH-1R circuit, with the feedback path highlighted:



and the mu-tron phasor ii:



maybe it's just my "circuit flows from left to right" reading bias but it seems to me like in both circuits the feedback is taken from before the mixing stage, and brought back to the op-amp stage before the first phase-shifting stage. depending on how the resistors are directing the flow of traffic i can see how there might be some back flow from the dry signal to the feedback path, but in both cases that seems to me like the path that the wet signal takes to get to the mixing stage, and i don't know how much commingling of the signals travelling in opposite directions there can be?

[Mark Hammer]

In that first schematic, the feedback needs to come from the summed wet+dry output, at pin 7 of IC1B.  You've shown it as the wet-only signal going back to the start.

[RickL]

Except it doesn't, at least not according to Boss' service manual: https://aionfx.com/app/files/docs/boss-ph-1r-phaser-service-manual.pdf

which matches the posted schematic. The only difference is the switching FET in the service manual schematic. Does that FET reverse the phase of the feedback signal?

[Mark Hammer]

Wasn't aware of that.  Normally, one wants the feedback to result in something that is opposite phase to the total product of the phase-shift stages.  Those are all inverting, so an even number of them gets you a result at the end that is in-phase.  The feedback path in the PH-1R goes to the inverting pin of that initial op-amp input/splitter stage, such that the result coming from just after Q6 IS "phase-flipped", as would be the case if it was being fed to the input of IC3, a la Phase 90.

If the feedback signal is the entire effect, however (wet+dry), it is my understanding that this could be fed back to the very beginning.  I'll need to do that experiment.  If someone else has the time and interest to do it before me, I'm happy to allocate my time elsewhere. 

[thunderaxe]

Except it doesn't, at least not according to Boss' service manual: https://aionfx.com/app/files/docs/boss-ph-1r-phaser-service-manual.pdf

which matches the posted schematic. The only difference is the switching FET in the service manual schematic. Does that FET reverse the phase of the feedback signal?

yeah neither of these schematics are my own designs, they're the aion and pedalPCB versions, respectively, of the PH-1R and mu-tron phasor ii, which i imagine are as close to the original designs as possible with currently available components

[Rob Strand]

Here's the typical responses for the basic 2-stage and 4-stage phasers.

The all-pass networks follow the JFET type which inverts the low frequencies.

The plots show feedback cases:
  green:  kfb = -0.5  ; inverted feedback
  blue:    kfb = 0.0
  red:    kfb = +0.5

Feedback sums at the input and sums with a selected tap.
The later phase 90's sum in the middle which doesn't change the results and are equivalent to a 4-stage phaser with inverted feedback from stage 2.
The output sums with the input signal.  All pass and dry paths summed equally with unity gain.
[As a side note: the Boss has a dedicated feedback mixer whereas the MXR tacks on a mixing point on the all pass.   The way MXR does is not an ideal mixer the tack-on mixing point changes the way the all-pass filter *circuit* behaves.  I posted plots of the effect some years. back.
https://www.diystompboxes.com/smfforum/index.php?topic=122944.msg1161142#msg1161142
]

Block Diagram:  (2-stage takes the output from stage 2 and mixes it with the dry signal)



2 stage, feedback stage 1:


2 stage, feedback stage 2:


4 stage, feedback stage 1:


4 stage, feedback stage 2:


4 stage, feedback stage 3:


4 stage, feedback stage 4:


Conclusions:
- Clearly not all feedback cases are useful.
- 2 stage phaser probably isn't going to be that great with feedback.
- feedback from stage 1 isn't that great.  similar effect for 2-stage and 4-stage 2 phasers
- Note the difference between feedback from stage 2 with a 4-stage and a 2-stage phaser.
  The 4 stage essentially embeds to two stage.  It's the overall response that counts.
- From the plots we might expect a 4-stage phaser with feedback from stage 4 would
  be best with positive feedback.  As used on the Boss PH1R.
- In each case the sign of the feedback works better one way than the other.

We can't apply these results to a 6-stage.  There is a pattern for the best feedback sign for different numbers of stages.

[Mark Hammer]

Insomuch as many folks like to use a 2-stager as an ersatz Univibe, albeit with a single moving dip, I suppose feedback would have no point, even if made feasible.  When was the last time you saw a Univibe/workalike with a feedback control?