Help: Phase 90 working but distorting

[Le québécois]

Hi Phase lover, My Phaser90 distort my sound and I don't like that. It is not over the top distrotion but it's enough to discredit my clean tone. I don't know if something is wrong with my circuit or if it's natural from that FX? here are the debugging info.

Thanx for your help.

1.What does it do, not do, and sound like? It phase nicely (at least to my taste) and there is no volume drop or boost but it unfortunately distort my sound and it's stronger on the treble side. The distortion is unpleasant and I would like to eliminate it. Sound like this: http://dl.dropbox.com/u/43158392/rythme_01_06.wav
2.Name of the circuit = MXR Phase 90 (PEZ90 from tonepad)
3.Source of the circuit (URL of schematic or project) = http://www.tonepad.com/getFileInfo.asp?id=42
4.Any modifications to the circuit? Y
5.Any parts substitutions? It is actually on breadboard. The 15 uF in the LFO was changed to 10, 22 or 10 + 4.7 uf. Actually using 10 uF. This modification don't change my distortion problem. The FET (2N5457) are matched using GEOFEX improved matcher but I only succeed to find VGSon of -0.58v.... Far from -1.8 to -2.0 volt. I have try with 2sk30a-y : -1.007, -1.003, -1.002, -0.999 Volt and have the same issue... almost stronger in fact. My Zener is a 1N4733 but somehow it give 4.2 volt, not 5.1 volt so I put a 4.7k resistor in serie with the zener to have a Vref of 5.5 or so. No 1N914, If I put one (with the good orientation.... the band facing the 9V and the non-band to ground), it blow and I really don't understand why!
6.Positive ground to negative ground conversion? N
7. What is the out of circuit battery voltage? => My breadboard is power with DC adaptor. 9.28 Volt.

Voltage at the circuit board end of the red battery lead = NA
Voltage at the circuit board end of the black battery lead = NA

Now, using the original schematic as a reference for which part is which (that is, which transistor is Q1, Q2, etc. and which IC is IC1, IC2, C1, and so on) measure and list the voltage on each pin of every transistor and IC. Just keep the black lead on ground, and touch the pointed end of the red probe to each one in turn. Report the voltages as follows:

U1
1- 5.51
2- 5.51
3- 5.23
4- 0
5- 4.45 - 5.84
6- 4.65 - 5.86
7- 1.31 - 7.18
8- 9.28

U2
1- 5.51
2- 5.51
3- 5.51
4- 0
5- 5.51
6- 5.51
7- 5.51
8- 9.28

U3
1- 5.51
2- 5.51
3- 5.51
4- 0
5- 5.51
6- 5.51
7- 5.51
8- 9.28

Q1 to Q4 (2N5457) identical reading (below) and matched VGson of-0.58 to - 0.6v
D = 5.51
S = 5.51
G = 4.05 - 4.31

Q5 (2N4125 - PNP)
E= 5.33
B= 4.90
C= 3.37

D1 (1N914)--- Not there
A (anode, the non-band end) = NA
K (cathode, the banded end) = NA

Z1 (zener)
A = 0
K = 4.05v

[PRR]

It doesn't make sense for Z1 to be 4V when all other pins are showing 5.5V. "I put a 4.7k resistor in serie" -- Ah, OK, take out the 4.7K and replace the 10K feeding the Zener with 1K to "sock" the Zener with enough current to maybe bring it up to 5V.

You'd expect Q5 Emitter to also be 5.5V, not 5.3V, since it's on the common Vref line also.

[Le québécois]

Ok, I removed the 4.7k in serie with the zener and changed the 10k to 1k. Vref is now at 4.88v. It improved a bit the distortion problem. Below, the new voltage.

Question, how crucial it is to have a VGs of -1.8 or so? Can this be the problem? I feel that the distortion come from the 4 phasing stages. It feel like at a certain point, the gain or frequency (high end) is to much amplified. I don't undestand how the Drain to source resistance of the FET (sweeping between 22k and 0 ohm?) interact with the frequency but I think it's form a frequency dependent filter somehow with the 0.05 µF capacitor of each stages. At one extreme of the sweep, it probably allow or amplified to much high frequency?

Finally, with my DMM, I can easily measure the resistance of the 22K combine with the Drain-source resistance of each FET. That is the thing which need to be match right? Despite similar VGs with the GEofex Matcher here are the variation of resistance :
FET A = 2.14 - 10.0 k     VGs = -0.539
FET B = 1.57 - 8.41 k     VGs = -0.593
FET C = 1.91 - 9.14 k     VGs = -0.562
FET D = 1.55 - 6.68 k     VGs = -0.615


U1
1- 4.88
2- 4.88
3- 4.66
4- 0
5- 3.99 - 5.37
6- 4.15 - 5.31
7- 1.31 - 7.18
8- 9.28

U2
1- 4.88
2- 4.88
3- 4.88
4- 0
5- 4.88
6- 4.88
7- 4.88
8- 9.28

U3
1- 4.88
2- 4.88
3- 4.88
4- 0
5- 4.88
6- 4.88
7- 4.88
8- 9.28

Q1 to Q4 (2N5457) identical reading (below) and matched VGson of-0.58 to - 0.6v
D = 4.88
S = 4.88
G = 3.45 - 3.71

Q5 (2N4125 - PNP)
E= 4.88
B= 4.28
C= 2.81

D1 (1N914)--- Not there
A (anode, the non-band end) = NA
K (cathode, the banded end) = NA

Z1 (zener)
A = 0
K = 4.88v

[Seljer]

Try removing the 22k feedback resistor if you included it. My guitar has relatively high output pickups and with the feedback included it was just too much in my Phase 90.

That or use an audio probe (capacitor + alligator clips hooked up to a 1/4" jack) to listen to your signal through the circuit to see where it starts distortion.

[PRR]

> I only succeed to find VGSon of -0.58v

Sorry, I didn't really think about this.

When the peak signal gets close to the FET turn-off voltage, it distorts.

The exact relation depends on the circuit and the modulation.

But hot guitar can be 0.5V, and 0.58V is awful close to that.

In principle, you could knock-down the signal in front, phase it, then boost it up. But that puts the signal closer to the universal hiss, and requires multiple mods.

Can you get different FETs?

[R O Tiree]

RG's FET matching circuit matches them at 10k, so they're not fully off. From my own experience of this circuit, you need Vgs@10k of about -1.3 to -1.8V.

I think the schem calls for 2N5952?  If you cannot get them, you might try 2N5485?

[Le québécois]

> Can you get diffrent FET?

not much without special ordering but,
I go to the store and was lucky enough since they allow me to scan the entire 2n5457 bins (± 200) with the fet matcher and my DMM. None of them was above -1.0 VGs. I end up with -0.738, -0.737. I also try the 2sK30 (also suppose to be good for this circuit). They only had 17 of them so my matching was not much better and the VGs only increase to -1 or so. They don't have the 5485 or 5952.... The thing is, when I try the 2SK30 a-y, the distortion is stronger so I go back to the 2N5457 at -0.738.   

>remove the 22 k in the feedback
I did'nt mention it but its gone since the beginning. Unfortunately since I prefer the phasing with that resistor in place.


My guitar signal is not that hot. I might consider taming the beast or I will order matched FET from a store (with better VGs).
Any other suggestion are still welcome.

have a nice soldering weekend