Phase 90 Repair Problems

[Cruton]

Hi everyone,

I'm repairing a MXR phase 90 (a block-logo reissue with through-hole components) and have gotten a bit stuck. I repaired one of these before fairly easily, but this one has me stumped. I have a hard time removing components from these double-sided PCBs without messing up the solder pads, so I thought I'd better ask for some insight here before needlessly replacing any more components.

I'm going off of the schematic posted at electrosmash here:https://www.electrosmash.com/images/tech/phase-90/mxr-phase-90-script-logo-schematic-parts.png . I know this isn't exactly the same as my version (PCB marked Rev. D), but it's been sufficient to start sorting it all out.

As often happens, someone plugged it into the wrong power supply. I replaced the polarity protection diode and it powered on, but passed no signal. The signal was dying at the input buffer, so I replaced the TL072 that serves as the input buffer and the LFO. At that point it passed signal, but no phasing happened. The audio was stopping at the output of the first phase stage, so I replaced the TL074 op amp that runs the 4 phasing stages. Now audio makes it through all stages, but alas, without any phasing. There are a few red flags that come up when looking around at the voltages, but I'm not sure where to go next. Here are the voltage readings:

Power Supply: 9.51

TL072:

Buffer      LFO
1: 4.35      8: 9.51
2: 4.35      7: 1.36-8.88 (Rate at Min), 4.47-4.61 (Rate at Max)
3: 3.95      6: 3.73-5.34 (Rate at Min), 4.52-4.54 (Rate at Max)
4: 0.00      5: 3.58 - 5.39 (Rate at Min), 4.32-4.34 (Rate at Max)

TL074:

1: 4.34 - 4.36      14: 4.34 - 4.36
2: 4.34 - 4.35      13: 4.33 - 4.35
3: 4.32 - 4.35      12: 4.32 - 4.34
4: 9.51         11: 0.00
5: 4.34 - 4.35      10: 4.32 - 4.35
6: 4.34 - 4.36       9: 4.34 - 4.36
7: 4.34 - 4.36      8: 4.34 - 4.36

Q1 (2N4126):

C: 2.55-2.57
B: 3.76-3.78
E: 4.34-4.36

Q2-Q5 (2N5952) [readings identical across the 4 FETs]:

D: 4.33- 4.35
G: 1.99-2.35 (rate min), 2.14-2.15 (rate max)
S: 4.34- 4.35


So the Vref is only ~4.3V, but I was expecting 5.1V since the Vref is set by a 5.1V Zener. The Zener doesn't look cooked, and a quick continuity check shows that it's only conducting one way as it should. I don't know if a Zener can get partially cooked and started conducting at a lower voltage though? Anyway, I pulled C8 (the 15uF power supply filtering cap) to see if that had been compromised and was allowing some of Vref to ground, but that didn't change anything. I haven't tried removing C4 (47nF across Vref and Vbias) yet, though attempting to do an all-or-nothing test with the multimeter showed no continuity from Vref to Vbias, but a low reading when measuring from Vbias to Vref. Not sure if that's indicative of a cooked cap or just a symptom of it trying to test continuity with it in the circuit?

The next question would be whether having Vref ~ 4.3V (for whatever reason) would explain the lack of phasing. I know that the JFETs take careful biasing to get the right swing of resistance in response to the LFO output, but most of the wisdom I've found on the 'net is just about tweaking the trimmer to get ~2.5V on the gate of the FETs-- in this case, they seem to be close to that, but I realize the "magic" 2.5V assumes a 5.1V Vref at the source. I'm not sure whether the difference between 5.1V vs. 4.3V Vref at the source (and 2.5V vs. 2.1V Vbias at the gate) would make the difference between typical phasing and no phasing at all.

I also read on the GGG phase 90 build documents that ~2V should be showing up at the output pin of each op amp phasing stage, and that's definitely not the case here.

Measuring the resistance across the drain and source of the FETs shows that the resistance will oscillate between ~19-23K when the rate is set low, but stays at ~23k when its faster (maybe my meter is just too slow)-- at any rate, this seems like a small change in resistance? Not sure if that's normal or part of my lack-of-whoosh problem.

Sorry for the brain dump-- I've been trying to sort this out in my spare time for over a week now and I think it's starting to get the better of me! I appreciate any insights the gurus here might have on where to go next with this little troublemaker.

[willienillie]

It's hard to get too far without the actual schematic for your version.  I'm not sure the Electrosmash schematic matches any of them exactly, to be honest.  You say someone plugged in the wrong power supply, do you mean wrong polarity?  Are there tantalum caps on the board?  Those don't tolerate reverse polarity at all.  Even if you have aluminum electros instead, those should probably be replaced.  I think your JFETs are fine, from what you report they seem like they are trying to do their job.  The non-polarized caps (like the 47n) should be okay too.

[Cruton]

Thanks for the reply. Yeah, a good schematic would be handy, but the usual searches haven't turned one up for me. I'll dig a little more, and if I can find one I'll post it here. So far most of what I've found are just value differences, although I've found at least one place where there are resistors not listed in the electro smash schematic (ie, one in series with each leg of the bias trimmer). Anyway...

As far as the power supply, yes, I believe they plugged in the wrong polarity supply, but I also recently read something about frying things by plugging in an AC PS instead of a DC PS. My guess is it was just the wrong polarity but I don't know.

There are some tantalum caps, the C8 power supply filtering cap I already mentioned, and the the cap that connects the + input of the LFO to ground (C10). I already tried removing the first to no avail, but haven't tried C10 yet. It does look like the solder for C10 has gotten hot at some point, so maybe that's a good place to start. That cap shows no continuity across its leads, so at least it's not failed short circuit, I think (though I guess that should be obvious since Vref isn't 0). So if that cap has started leaking a bit, that could draw Vref down because it's connected to the + input that should otherwise be at ~Vref since the - input is coupled to Vref?

[Mark Hammer]

The zener is there to act as a voltage regulator.  Keep in mind that the trimmer simply divides down a supply voltage to provide the bias that the JFETs need.  The P90, like a great many pedals of that era, assumed it would be powered by a battery.  And, as the battery wears down, that divided-down bias voltage will drift, requiring the user to open the pedal and tweak the trimmer - something that is beyond many end-users, and is definitely not gig-friendly.

The zener takes the potentially varying voltage (good pun, eh?) of the battery, and chops it down to something fixed, such that whatever the trimmer carves away from that will be stable.  As long as the zener voltage is somewhat greater than the needed bias applied to the JFETs, things should work.

[Cruton]

Thanks for chiming in, Mark. Just to make sure I'm following-- it sounds like my slightly-lower-than-expected Vref is unlikely to explain why I'm not getting any phasing in my case, since the Zener is producing a stable Vref high enough for me to bias the JFETs to the ~2.5V gate voltage needed in the Phase 90.

All righty then, back to the drawing board... this little pedal is driving me nuts.

[Cruton]

Ah, I just figured it out! Based on Mark's comment that the Vref difference was unlikely to be the cause, I took a step back and thought through it again... LFO is fine... Phasing stages are fine... JFETS and bias are fine... so maybe the JFETS just aren't getting the LFO signal? The big resistor between Vbias/JFET gates and the - input of the LFO (R22 on the linked schematic, 3M9 there, but 2M7 in mine) looked like it had seen some heat at the pads, but wasn't charred itself... knowing that they often fail open, I just jumpered it with a 2M2 I had on hand, and voila! Phasing!!!

I'm going to (carefully!!!) desolder the old one. Is there any reason why 2M2 would be a poorer choice here than 2M7? I've read that the goal there is just to limit the current to the JFETs since they don't need much, but maybe my understanding is overly simplistic.

Thanks again for the help. You all definitely saved me a heck of a goose chase.

[Cruton]

...and just answered my own question there. Looked at aron's schematic for the Phase 90 and saw that it has 2M1 in the R22 spot: https://aronnelson.com/gallery/main.php/v/Schematics-etc/MXR_Phase_90.gif.html?g2_imageViewsIndex=1

So I'm guessing that the value of that resistor isn't critical, so long as it's large "enough".

[duck_arse]

if you set the fets bias to "2V5", in a great many cases, due to jfet manuf [in]tolerances, it will be set wrong. someone may well have measured that voltage on a working production unit, but the trimmer is there specifically to allow the voltage adjustments needed to suit the devices soldered to that particular board.

[Cruton]

I see, that makes sense to me. What in your opinion would be the ideal way to set the "correct" bias voltage when FETs are already in the circuit like this?

[willienillie]

Set it by ear, where it sounds best is the correct setting.

[Mark Hammer]

Ah, I just figured it out! Based on Mark's comment that the Vref difference was unlikely to be the cause, I took a step back and thought through it again... LFO is fine... Phasing stages are fine... JFETS and bias are fine... so maybe the JFETS just aren't getting the LFO signal? The big resistor between Vbias/JFET gates and the - input of the LFO (R22 on the linked schematic, 3M9 there, but 2M7 in mine) looked like it had seen some heat at the pads, but wasn't charred itself... knowing that they often fail open, I just jumpered it with a 2M2 I had on hand, and voila! Phasing!!!

I'm going to (carefully!!!) desolder the old one. Is there any reason why 2M2 would be a poorer choice here than 2M7? I've read that the goal there is just to limit the current to the JFETs since they don't need much, but maybe my understanding is overly simplistic.

Thanks again for the help. You all definitely saved me a heck of a goose chase.

Great detectiving!

The P90 lacks a sweep width and manual/offset control, to set where in the spectrum the sweep happens and how wide a sweep is produced.  MXR selected defaults for those, as well as feedback/resonance, that would work for the entire range of speeds.  A "proper" phaser should have all 4 controls, plus a mix control, but the P90 came out at a time when 1 or 2 knobs was the rule.  Some diagrams will show the feedback resistor as 24k and others 22k.  Some will show the speed-resistor as 3M9, and others as 3M3.  3M9 yields an audibly narrower sweep, and 22k yields greater resonance.

The current coming off the LFO sums with the current coming from the trimmer output.  The JFETs have what I guess you might call "sweep headroom" limitations.  In other words, you can make the drain-source voltage change by applying more current to their gates, but not more than X. So whatever the LFO current-limiting output resistor value is, between it and the bias current, they shouldn't push the JFETs past the point where they won't change D-S resistance.  Personally, I wouldn't aim for much below 2M7 or above 4M3, and even those may prove too wide a range. 

I find I am fine with a 3-way toggle to set width, another to set feedback, and a variable offset pot.  The latter essentially replaces the stock 1M fixed resistor off the trimmer wiper with a 510k fixed resistor in series with a 500k linear pot, wired as variable resistor.  Ideally, you set their combination for maximum resistance, and then adjust the trimmer for the lowest gurgliest phasing attainable that has a full sweep.  As the pot resistance is reduced, the phasing should go from low and gurgly to higher and swirly, giving a different feel.  Since that is feeding the JFET gates more current, it may sum too high with the LFO current if the stock 3M9 value is set too low.

[PRR]

> Vref is only ~4.3V, but I was expecting 5.1V since the Vref is set by a 5.1V Zener

Probably right. A Zener is not a sharp drop-off like falling off a dock. It is a soft thing like my little dog falling off the soft couch. The "5.1V" is rated for a specific high current, often half the Watts rating. This P90 runs the Zener near 0.4mA or 0.020 Watts, far below the smallest Zener I have ever owned. Mostly small Zeners are tested at 5mA. 10X what the P90 hits it with.

[Cruton]

All right, so setting by ear is the way to go. It sounds pretty good now, but maybe I should adjust the trimpot to see if I can dial it in better.

Thanks for the extra explanation, Mark. To make sure I understand-- a lower-value resistor limiting the current from the output of the LFO to the FETs gates will increase the sweep width, but with a much lower value (like possibly the 2M2 I had on hand) there may be some portion of the sweep where the behavior of the FET isn't really changing because the current has already pushed it to its "sweep headroom" limit. This might sound like a sort of a "flat spot" in the phaser sweep, I imagine, as opposed to a smooth (well, kind of) sweep.

PRR, I appreciate the lesson on Zeners (and the analogy about your dog). That will undoubtedly be very helpful when troubleshooting in the future. That Vref discrepancy had me hung up for a long time when it had nothing to do with the problem!

[Mark Hammer]

All right, so setting by ear is the way to go. It sounds pretty good now, but maybe I should adjust the trimpot to see if I can dial it in better.

Thanks for the extra explanation, Mark. To make sure I understand-- a lower-value resistor limiting the current from the output of the LFO to the FETs gates will increase the sweep width, but with a much lower value (like possibly the 2M2 I had on hand) there may be some portion of the sweep where the behavior of the FET isn't really changing because the current has already pushed it to its "sweep headroom" limit. This might sound like a sort of a "flat spot" in the phaser sweep, I imagine, as opposed to a smooth (well, kind of) sweep.

PRR, I appreciate the lesson on Zeners (and the analogy about your dog). That will undoubtedly be very helpful when troubleshooting in the future. That Vref discrepancy had me hung up for a long time when it had nothing to do with the problem!

Yup.  You got it.  When it comes to FET-based phasers, you really can have "too much of a good thing".
Here's what you can do with two of these puppies.  For simplicity, I opted for 2 position push-button options.  Using pots for everything would have made it too cluttered, and to complicated to get back to specific sounds.  The sound isn't great here, but you'll get the idea.

[Cruton]

Whoa, that's a super cool build!! That "bubbling up" sound, as you described it, sounded awesome. The push button options were perfect for demonstrating what you were describing about the width, range, and feedback in a phaser, thanks for sharing that!

I was surprised at how nice the vibrato sounded-- was that just accomplished by cutting out the dry signal? That gets me way more excited about the prospect of building a phaser project than I was before.

[Mark Hammer]

Yup.  All the "vibrato" required was lifting one end of the resistor that mixes dry in with wet.  And yes, the rising phase is a neat sound.  The Phase 99 does some of the same tricks, but mine does more, like the vibrato, vibrato+phase (synced and unsynced), 2nd phaser on demand via the footswitch, and the independent offset and sweep width for each.  That said, stereo is nice and so is a smaller footprint.

The little DPDT push-button switches are nice, compact, and cheap as dirt (half the rice of cheap toggles), but are a bugger to line up with holes as well as attain the optimal height above the enclosure surface.  Too high and the caps catch on the chassis.  Too low and you can't  tell whether it is pressed down or not.  Takes practice to get it just right.

[Cruton]

I had never really looked into the p99, but it sounds pretty cool. I think the extra features of yours is well worth the added size and lack of stereo, but maybe that's just me.

Thanks for sharing all of this, and for answering my question about the vibrato mode. It has me thinking about building a phaser with toggles for width and offset like you described, plus a second stomp switch to change between phase/vibrato. The ability to switch between the two on the fly was probably my favorite feature on the Rotovibe I use to have, but alas, I sold it for a JHS unicorn v2 (there's that small footprint appeal again). Now I find myself bending over for that chorus/vibrato toggle all the time.

I can see what you mean about the trickiness of those little DPDT push-button switches. That said, the end result in your build looked awesome, and I can appreciate the appeal of not having so many potentiometers that it's easy to get lost. May have to grab me some of those and mess around with them on an old enclosure for a bit to see if I can get the hang of positioning them.

[Mark Hammer]

A useful compromise to the little pushbuttons, with much easier machining, is the short-handle SPDT 3-position toggles.  The underside still takes up the same amount of room, but the shorter bat handle does allow for closer spacing to knobs without getting in the way of either seeing the current switch position, or being able to manoeuver them.