Phase 90 R28 mod idea

[Mugshot]

i had a friend build for me a phase 90 copy based off tonepad's project file. browsing DIY for possible mods to get rid of the distortion in the circuit (just a wee bit, but there are times the distortion is not needed), i read about the R28 mod. long story short, i decided to pull the feedback resistor out, kinda like the smoother phasing that way.

consequently, being the curious nut that i am, i tried installing a diode into the r28 spot. depending on the orientation, it can get really feedback-y and squeal like a pig!  i junked this effort afterwards seeing this may not be practical.

later this week, an idea struck me. why not install a diode again to have a one-way signal regeneration in series with a 10K resistor and a 1Meg pot to control the one-way feedback?

watchathink guys?

[Mark Hammer]

In fact, that's similar to how it is usually done when there is a control for feedback amount.  The MXR units simply went with a fixed feedback amount, rather than the extra production cost of a pot, a knob, and extra machining.

[Mugshot]

dont worry mark, mine already has a hole for a new pot anyway 

what about the diode idea? do you think it's feasible? what im thinking is that it will allow only a one-way signal traffic rather than a "bleed-through" when using a resistor. idk if i have explained it well enough.

[Mark Hammer]

It is quite likely you will have no feedback whatsoever, because the normal signal fed to a phaser is well below the half volt you would need to get it to conduct and the phaser adds no gain to increase the odds of the feedback signal being a large enough amplitude to cause the diode to conduct.

[Mugshot]

It is quite likely you will have no feedback whatsoever, because the normal signal fed to a phaser is well below the half volt you would need to get it to conduct and the phaser adds no gain to increase the odds of the feedback signal being a large enough amplitude to cause the diode to conduct.

hmm, this gets me thinking. i forgot about this fact 

anyway, i'll play with Ge diodes tonight and a lower pot value (say, like 10k) and report back to you. im thinking the 0.3V needed for the Ge Diode to conduct may allow for this to happen 

[Mark Hammer]

Well, you could get it to work if you either fed the unit a hotter signal or simply added some gain to the first stage.  That would tend to introduce clipping in the FETs, though.

[Mugshot]

Well, you could get it to work if you either fed the unit a hotter signal or simply added some gain to the first stage.  That would tend to introduce clipping in the FETs, though.

yes a hotter signal is what i have in mind Mark! mine sits after the delay, so maybe the extra push from the boosted signal is enough to make the diode work as it envisions it  idk if this is really feasible, but i will play with the circuit and report back tomorrow if anything.

[Mark Hammer]

Well, there is what is feasible, and there is what is desirable.  I'm not expecting anything particularly pleasing from this.   If it "works", I expect the result to be something that appeals to fans of Metasonix modules ( http://www.metasonix.com/index.php?option=com_content&task=view&id=27&Itemid=53 )....an "acquired taste". 

Any diode in series with the signal will produce crossover distortion.  In your instance, with a single diode in series, you will get one half cycle passing the diode, and the half-cycle that does pass will have its sides chopped.  That will make the frequency content in the feedback path largely devoid of most bass, but NOT exactly the same as simply using a cap to eliminate most bass from the feedback path.

What you might want to consider is using a small (6k8) "default" feedback resistor, in series with a variable resistance (50k pot should be more than enough), and running a back-to-back pair of germanium diodes to ground from the junction of these two.

Looking at the Tonepad schematic for the P90 - http://www.tonepad.com/getFileInfo.asp?id=42 - you can see that the default feedback resistor is 22k.  Values smaller than about 11k are likely to result in oscillation, however if the diodes clip the feedback signal, that will not likely occur.

[Mugshot]

hmm, it had me thinking 

last night i tried a Ge Diode in series with a 27k resistor, there was a subtle effect but not remarkable. resistor values any lower than 15Kresulted to distortion in the circuit. actually even without the diode the circuit distorts (possible cause was the FETs used, original 22K was removed due to this). hmmm, i will try your suggestion tonight Mark!

[Mark Hammer]

Phase shift stages are inverting, so if you want the feedback to accentuate the notches, it has to recirculate around a loop of odd-numbered stages (3, 5, 7, etc.).  When flangers recirculate the delay signal, obviously they don't do it through stages like a phaser does, but they still have the challenge of keeping the input+recirculated signal within the limits of what the delay chip can handle.  So, what you will often see is an op-amp mixer stage just at the start of the delay path, where the feedback signal comes back to.  That op-amp stage, which is illustrated here in the Boss BF-2 ( http://www.freeinfosociety.com/electronics/schemview.php?id=117 ), will have a pair of diodes in the feedback loop so that if the input+feedback signal exceeds some threshold, it will be clipped by the diodes.  That clipping may add some harmonic content, but it also serves as a crude limiter.

An additional mixer stage like that, situated before the phase-shift stages could work the same way.  That is, one could set the feedback amount to "stun", but the diodes would keep it in check.  For it to work/sound like a phaser normally does, though, the entire loop would need to be inverting.  That means one of the following:
a) you have 5 phase-shift stages
b) you have 4 phase shift stages but the mixer stage where the signal goes back to is inverting (as it is in the BF-2),
c) you use 4 phase-shift stages and the mixer is non-inverting, but you stick a unity-gain inverting stage in the feedback path so that the feedback is inverted before it gets mixed in.

[mac]

In mine (tonepad layout) I solder to wires instead of r28. wires connect to a 15k and a 50k linear pot with on/off switch. This way I can switch r28 on and off and vary its resistance.
Less than 15k and well, you know

mac