Phase 100

[RickL]

Okay, so I'm going to build this, using the GGG layout. This is the schematic http://www.generalguitargadgets.com/pdf/ggg_p100_sc.pdf.

I want to go all out on this one, with as many controls as I can stuff into. I thought I would pass this by everyone to see if I'm missing something or if I'm getting off the rails. These are my plans:

Vibrato switch - lift R18 where it joins C7/R19. spdt centre-off switch to select stock (6k2 resistor), vibrato (no connection) and a pot (20k? + series 4k7) to mix in the dry sound.

Stages switch - sp3t switch for 2, 4 or 6 stages of variable notches.

Intensity switch - spdt centre-off switch to select stock feedback resistor (15k), no feed back (no connection) and a pot (20k? + 4k7? series resistor) for variable feedback.

Bring the center frequency trimmer out as a pot, maybe a 10k with resistors on either side to make the whole pot usable.

I'm not sure how to add a variable intensity pot. Pot (100k?) in place of the intensity switch?

Switched diodes across the speed pot for up and down ramps on the LFO?

Does it make sense to add pots to variable the resistance of the fixed stages? Maybe replace with 10k pots in series with 10k resistors.

Any other suggestions?

[StephenGiles]

What will you use for vactrols? From memory the MXR 100 didn't have much intensity, although a guy who edited UK Elektor back in the 80s adapted it for 15v operation, which apparently made a lot of difference!

[Mark Hammer]

Hi Rick,  I imagine you're one of the few who isn't, and can't, work from home.

Personally, I've never found much use for blend/mix controls that can continuously vary wet/dry from 0% to 100%.  I like a dry-lift toggle for vibrato-like effects, and I like a wet-blend control to dial back the effect to make subtler.

The stock P100 shows icons on the enclosure for the 4 switch-settings that are the combinations of 2 feedback settings and two sweep-width settings.  The LFO has a configuration that is not immediately comprehensible to me, so I can't tell if what the switch does ONLY changes the sweep width, or does something in addition to that, like change where the sweep occurs in the spectrum, or maybe the LFO shape.

The stock switch provides for no feedback or a single fixed degree of feedback.  But with 10 allpass stages (4 of them fixed), there are a lot of options for how many stages can be included in the feedback loop.  If you're up for a fully tricked-out P100, consider combining a rotary feedback-routing switch, in conjunction with an intensity/amount control.  This not only includes some "from" and "to" choices, but also series caps to attenuate feedback at lower mids and below, to avoid any unwanted oscillation, or simply change the tone.

I fully endorse bringing the Center Freq control out to the panel.  In virtually every P90 clone I've made for myself (and I think we're up to about 6-8 by now, I've included an "offset" control of some form to move things from gurgle to swirl.  Changing where in the spectrum the notches occur can really alter the perceived character and feel.

As for stages switch, I wonder if it might be more useful to disable some of the fixed stages when you select between number of stages.  Naturally, this would all have to be planned out in a way that's compatible with the suggested Feedback-routing switch.

All allpass stages should be precisely unity gain.  The risk - especially as one adds more stages beyond two and includes a feedback path - is that even marginal gains (e.g., 1.1x) for a single stage result in risk of oscillation.  I suspect that's part of why MXR decided to keep any feedback in the stock circuit rather modest (25k to mix in with a 10k at the input of the return point).  It's also why some other commercial phasers include trimmers in the feedback path - to be able to get maximum feedback but avoid oscillation.  So 1% resistors for each allpass stage gets you closer to that unity-gain ideal; enough that you should theoretically be able to dial in high feedback without risking oscillation.

Hope that meandering reply is useful.

[RickL]

I have the VTL5C3/2s that are recommended.

All allpass stages should be precisely unity gain

I must be misunderstanding the way the fixed stages work. I thought the 20k resistors took the place of the variable resistance in the variable stages, and so, set the notch at a fixed frequency. Won't varying those resistors (in pairs) move that fixed notch for that stage?

I hadn't thought of changing the feedback path, my plan was to always feedback the maximum number of odd stages, depending on the number of variable stages I selected. I'll have to see if I can work out a way to vary the feedback path when using 6 (and maybe 4) variable stages (and the suggested caps in the feedback path).

Oh, yeah, and I am actually working from home. For fairly obvious public perception reasons we're not doing any inspections in public until we get clear work guidelines (PPE, distancing, etc.).

[Mark Hammer]

Ah...okay, I misunderstood which resistance you were referring to.

If there is no feedback from any of the allpass stages to a preceding stage, then it doesn't matter if any one or more of them are above unity-gain.  The problem arises when one introduces some degree of feedback.  Once you have feedback, and one of those stages in the loop has a bit of gain above unity, then feedback starts to risk oscillation as the signal gets amplified a bit, fed back, re-amplified more, fed back, re-amplified again, etc.  One can certainly tame that by reducing the amount of feedback, but once you've dropped the feedback enough to avoid oscillation, you start to wonder what the advantage/virtue of having feedback is at all.  So, ideally, if one wants to use feedback, the feedback resistor in each stage should be equal to the resistor feeding the inverting input. 

Alternatively, I suppose, one could always pay no attention to matching, and simply reduce the gain of the stage where the feedback signal is tapped from below unity, such that where it gets fed back to is never seeing a feedback signal greater than the input.  In other words, say we have 4 stages, whose individual gains, in order, are 1x, 1.05x, 1.05x and 1x, with the last stage feeding back to the 2nd one.  Using 5% resistors, those are all realistic possibilities.  The cumulative gain is 1.1025x.  But feed that last stage back to the 2nd one, and within a few cycles, we're easily up past 1.5x and quickly headed to oscillation territory.

If we reduced the gain of the last stage to 0.9x, then the cumulative gain around the feedback loop would be .9925x.  We might have to tweak the mixing resistors between dry and wet to get a perfect 50/50 balance for maximum notch depth, but we could dime the feedback without risking oscillation.  I have to confess, I've never tried this, but now I'm curious to.

In the case of the P100, and referring to the GGG diagram, one might add a trimmer in series with R23 to drop IC2A's gain below unity, while holding R1 at a smidgen above 10k.  That way, you'd be aiming for maximum feedback, but trying to avoid the oscillation that might come with cumulative gain.  The traditional way of addressing the oscillation challenge is to ignore the gain of each allpass stage and simply limit the maximum feedback.  Maybe what I'm suggesting could get even more resonant sounds with less problems.

The four fixed stages each add another 90 degrees of phase shift for content above roughly 800hz, and progressively less for content below that.  Reducing the effective resistance of that 20k-to-ground moves that corner frequency upwards.  I suppose that can be interesting if one has wide-bandwidth signals to process (e.g. acoustic guitar with fresh strings, cymbals, etc.) but subbing 20k for a 10k resistance yields 90 degrees of phase shift in each stage for stuff above  roughly 1600hz.  The variable stages obviously add phase shift well below that, so the result would seem to be an intensification, or obviousness of notches at the upper end of the sweep, and less as it sweep lower.  Perhaps the direction of resistance change should be reversed.  That is, maybe use a pair of DPDT on-off-on toggles to increase that 20k for each pair of fixed stages.  Lets say the middle/off position of each toggle results in 15k being added to the stock 20k, making the 90-degree point just over 450hz.  One side position straps 8k2 in parallel with the 15k, yielding an effective parallel resistance of 5k3, pushing the 90-degree point up to roughly 630hz.  Finally, the other side position of the toggle bridges the 15k to bring things back down to a 20k resistance to ground (well, Vref, actually), and the stock 800hz 90-degree point.  I offer those points not as any sort of ideal, but simply something one can easily do with common part values.

I don't doubt you CAN work from home, but folks I've seen Rick's workplace, and you must be jonesing for that machine shop!   

[RickL]

Unfortunately the machine shop is long gone.  I got kicked out several years ago and I'm back in the same building downtown, same floor, that I started in 30 years ago.

I'll try your suggestion about the fixed stages, switches are cheaper than pots. And some switched capacitors in the feedback path, again spdt to select either no cap or one of two other values, same set up as your suggestion for varying the fixed stages.

[Mark Hammer]

Sorry to hear that the shop is inaccessible now.  A pity.

I tend to go for simple add-on options that expand possibilities, provide easily contrasting changes, and are easily replicable.  So 3-position toggles strike the right balance for me.  They also don't take up as much space, so they're often easier to find a spot for in standard-sized enclosures.

One of the nice things about both fixed stages and variable stages using LDRs is that they won't clip when feedback levels get higher.  I've had issues with P90 clones when I cranked the feedback beyond stock, as the JFETs hit their headroom limits.

I'm still curious about what the sweep-width half of the 4-position switch is doing to the LFO.  Admittedly, it's hard to grasp how the LFO works when it is drawn somewhat "backwards" (op-amp pointing right, transistors pointing left), but it seems like closing the switch bypasses the effect of Q3 and the Center Frequency trimmer.  That sort of makes sense, given that restricting the width of sweep always forces you to consider where that restricted range of sweep ought to be.  If one wishes to have more control over the sweep width, so as to better complement the sweep rate selected, then there would be a need to not only have the Center Freq trimmer turned into a panel-mount pot, but also maintain its influence on what that seeming Darlington pair of Q1/Q2 are feeding the vactrols. In short, we need to better grasp how the sweep width can be adjusted within this circuit.

[RickL]

Okay, so I've got the board populated (I use the 4ms/Commonsound CBCB method, kind of a hybrid of perf and PC board) and I'm moving on to the tedious wiring stage.

At Mark's suggestion I'm going to add the option of series capacitors in the feedback loop with a  on-off-on switch. Suggestions for two cap options that are likely to give noticeable differences in sound? I'm thinking 0.01u and 0.001u but I really haven't a clue.