Phase 45 & 90 analysis/modding/hacking help?

[tatou]

I'm trying to learn more about how phase/vibe type effects work (in particular), so my current project is to hack together a "minimal parts count" 2-stage phaser for a 1590a. (I built a MBP Smoothie about a year ago but would like something a tad grittier, hence this project.) Anyway, as a starting point, I've been studying the MXR and DOD circuits, and extrapolated this Phase 90 analysis by Electrosmash to the Phase 45, as well, so I can see the "diff" in these two similar circuits.

I'm attaching screenshots of these "diffs" — blue highlights are minor value changes, green are things added from the P90 to make a P45, and red are things removed from the P90 to make a P45. The main differences seem to be (aside from the obvious 4 vs. 2 stages)...

The 45 has a 10k R in the input buffer opamp loop.
What does this do? (And on 2nd thought, perhaps the other 22k R to REF voltage should really be part of the output mixer section?)

The 45 uses a 4.7v Zener as opposed to a 5.1v.
Does this contribute to the "smoothness" by running things at a slightly lower voltage?

The 45 LFO adds a 150k resistor to ground.
Does this further half the LFO voltage to 2.4v or somesuch, and temper the LFO?

Several LFO cap/resistor values are different.
What do these changes do?

The 45 adds an G/S feedback circuit to the JFETs, with a cap and resistor.
I gather from this post by RG Keen that this loop cuts down on distortion added by the JFETs. Is this correct? If so, how does that work?

The 90 has a PNP buffer in the output mixer
What purpose does this serve? Is there something particular "lost" in the 45 by its absence, or is this responsible for some of the "grit" in the 90 circuit compared to the 45?

More generally, is there a reason these extra components exist in the Phase 45 circuit other than to "smooth it out" relative to the 90? In other words, could you get away mixing-and-matching the "simplest" of these sections, which would essentially be a 2-stage Phase 90 without the PNP buffer in the output mixer?

Bonus question: if one were to experiment with a sine LFO (or some other waveform), how would one do that while keeping the parts count low.

Thanks!

[tatou]

UPDATE: So last night, I breadboarded up the proposed circuit (essentially a 2-stage Phase 90, but with a 10uF cap in the LFO and a passive Phase 45 output mixer) — and nothing. The "wet" path only does a little tone shaping regardless of where I set the trimmer (tried both 200k and 500k pots, and removed the dry path from the circuit). I know the JFETs are matched because they're leftover from my previous by-the-numbers Smoothie project (tried 3 different matched pairs).

Any ideas?

[Stomptown]

UPDATE: So last night, I breadboarded up the proposed circuit (essentially a 2-stage Phase 90, but with a 10uF cap in the LFO and a passive Phase 45 output mixer) — and nothing. The "wet" path only does a little tone shaping regardless of where I set the trimmer (tried both 200k and 500k pots, and removed the dry path from the circuit). I know the JFETs are matched because they're leftover from my previous by-the-numbers Smoothie project (tried 3 different matched pairs).

Any ideas?

I'm no expert, but....

What values did you use for R23 and R24 in the p45 output mixer? If you used 10k I would try 150k (the value from the p90 output mixer) and see what happens.

[tatou]

UPDATE: I did manage to get the attached circuit (essentially a 2-stage P90 with the passive P45's output mixer) working on a breadboard. However, it's not *there* yet.

A few observations:

1. The circuit started working after I removed C4. This cap seems to exist in schematics for both the Smoothie (C7) and NomNom (C7), but was screwing things up for me. This cap also appears to be omitted from the DOD 201 version of the P45 circuit.

2. C6 changes the LFO behavior: lower values make the max rate faster, larger values make it slower. I gather that this is because larger values have a bigger "tank" and take longer to "fill up" or "drain" and turn the opamp output into a triangle wave, if I understand correctly how it works (source).

3. The RATE pot (500kC) effects are only perceptible in the last 1/3 of the rotation or so. It's as if the first 2/3 of the turn is so slow as to be imperceptible. I tried a 100kC pot and lower/higher values for R19, which tweaked the response a little but didn't seem to fix this problem. This is weird, since you usually see this problem with linear/audio tapers, not reverse-log. 

4. The MODE switch is the "vibe mod," which cancels out the dry signal to make a faux vibrato effect. While the summed effect is a very characteristic phaser sound — and also fairly strong — the wet signal alone is not very perceptible... basically only for low notes at the higher LFO rates.

5. The whole thing seems quieter than my P45, which is surprising because there are fewer (and smaller) resistor values in the signal path.

Any guidance on where to go from here is welcome!

[midwayfair]

Okay, so here's a couple things I see:

You've used the Phase 90's method of "doesn't really matter that all the gates are directly coupled to the AC signal because they're sufficiently isolated from the drains and sources anyway" BUT you've DC coupled the gates the bias voltage. Look again in the nom nom and you'll see that there's a capacitor between Vb and the output of the LFO. I think that's just filtering out noise.

The phase 45 does something slightly different. It uses a Zener-regulated voltage for Vb instead of a PNP emitter to produce the Vb. The LFO output is still decoupled from the Vb. However, there is a SECOND bias voltage that's used to set a specific DC point at which to hold the gates of the FETs. The gate and source are both held at essentially the same potential (~4.5V) but the gate is going to have to be somewhere else because the output of the LFO might not fall exactly in the range that is required for a symmetrical waveform based on the operating points of each FET. The Phase 90 doesn't seem to need this secondary bias voltage. Maybe the LFO is more stable because of the stiffer Vb. Maybe it's just that 4 stages are intense enough without dialing in a perfect setting. Who knows -- maybe it could be "improved" ever so slightly with a bias adjustment for the gates.

I'm trying to think of some reasons your speed control might bunch up -- because that is abnormal -- but I'm not really coming up with anything if you're sure it's connected 100% properly and the pot isn't backwards or something silly like that. It's a low pass filter formed with your C6 (the 10uF). Increasing that cap does indeed slow down the oscillation by lowering its frequency. It still ought to require a reverse taper because the relationship between frequencies is not a linear one -- it's 10 DECIBELS per decade (just like in tone controls ... all dealing with the same stuff in the end, because the LFO is just using the self-noise of the device). Half the resistance doubles the frequency.

As far as it being quieter ... I would expect any of these designs to be slightly below unity, but make sure you're getting at least close to unity without any modulation at all.

[wgc]

This is a pretty good read if you haven't seen it:

http://www.electrosmash.com/mxr-phase90

[midwayfair]

This is a pretty good read if you haven't seen it:

http://www.electrosmash.com/mxr-phase90

I imagine he has since he linked to it in his OP.

[wgc]

Whoops, guess he's seen it then!