Maestro Phase Shifter Project: Which FET transistors ?

[rantony]

Hello everybody,

I would like to build a Maestro Phase Shifter stompbox based on the following schematic:

http://files.muziq.be/schematics/oberheim_ps-1.pdf

The transistors marked with a " * " must be FET N-Channel and they all must be matched. Does anybody know which transistor number I should go for and what gain I should target ?

Thanks for your help !

[Jered]

  Well one of the N FETs is labeled 2N4302 so that would be my guess. Other similar phasers I think have used 2N5952. For matching, RG has a great "how to" over at  http://www.geofex.com

[Gus]

There is good information written on the schematic.   FWIW good design with a pot will have the circuits desired operation with the pot at center.  Work out the math using ohms law for the -DC voltage at the FETs gates.

[rantony]

Thanks guys for the replies,

Sorry if I did not follow my own thread more closely. That's because I had put my project on 'hold'.

I think I'll go with Jered suggestion and try to find matched 2N4302's. I'm scared that it will be pretty difficult because I need 6 FETs. I'm just unsure about if all 6 FETs must be matched together or if they must simply be matched by pair. Any opinion on that ?

[StephenGiles]

Since 1978 or so I have built many phasers using the FET method and never bothered with matching - they still sounded good!

[R.G.]

You were lucky Stephen.

Any JFET will have a channel resistance of between infinity (when it's turned completely off) and Rds, the full channel-on resistance which is in the sub-1K range for most JFETs and can be in the sub-10 ohm range for some of them. That's a huge range.

JFET phasers work based on the time constant (R*C) of the JFET and the phase cap. It takes two stages to make enough phase shift to make a notch in the frequency response, which is mostly what we hear as phasing. If one of the time constants is sub audio and the other is ultrasonic, then they are not contributing equally and the notch is not deep enough for good phasing. There was a lot of work on phasers back in the 1970s, and some very scholarly research turned up that the most intense phasing sound results from all phase stages tuned identically.

Obviously non-identical stages work, as witness the univibe, where the phase caps are deliberately staggered. However, the univibe has another agenda. It is trying to produce a vibrato as well. For vibrato, you want some phase shift but distributed all over the audio spectrum so that all parts of the spectrum are phase delayed a variable amount and get shifted around. You still get notches, but they are not as deep. So the phasing suffers a bit to get vibrato to be better. Notice that the univibe uses nominally equal/matched LDRs and different cap values.

Bottom line - for phasing, you want them identical; for vibrato, you want them staggered, ideally equally staggered. Given that the caps are staggered in a vibe circuit, you still want the resistances to be nearly identical, although it matters somewhat less there.

The Neovibe sounds to my ear even better if you put all-same caps in. Different, but a better phaser, which is how most people use it. The vibrato effect suffers, but I don't use that much. A good mod would be to put in solid state switches to switch in different cap values to switch the 'vibe caps from all the same for phasing to staggered for vibrato. I may have to do an article on that, I guess.

But match your JFETs for best phasing sound. If you don't, it may still sound phase-y, but until you match, you don't know whether you'd like that better.

oops, almost forgot. 2N5485 makes a good phaser JFET, as well as 2N5952 and 2SK30A.

[Mark Hammer]

My sense is that non-matching will still produce phasing, but matching produces a better "turnaround".  That is, the manner in which the sweep approaches the extreme of the sweep and then starts its return in the other direction, just like competitive swimmers.  I recently got a Phase 90 board working that I had made with a thimbleful of 2N5952s I had bought from Smallbear.  I think I bought 6, and just grabbed 4 to stuff in the board, without any pretext of matching.  The phaser works, and produces all the usual expected phenomena in the middle 60% of the sweep, but at the upper 20% and lower 20% of sweep, it is simply less "musical", and less graceful in the way it makes the turnaround - a result of prematurely reaching the point where one or more JFETs refuse to change resistance any more.  When that happens, the point/s of maximal (and minimal phase shift) no longer correspond linearly to the shape of the LFO waveform.  The waveform may be purely sinusoidal but that actual change in phase may be something like a flattened sine at one or both peaks.

I concur with RG's advice to at least attempt matching and see how the result compares with unmatched JFETs.

[rantony]

It's fun to have experts replying to my thread. Thank you guys !

Yeah, I will "attempt" matching the JFETs...

some very scholarly research turned up that the most intense phasing sound results from all phase stages tuned
identically

So, all six JFETs must be matched equally ? That means I will need to test many dozens of transistors !!! 

[Thomeeque]

oops, almost forgot. 2N5485 makes a good phaser JFET, as well as 2N5952 and 2SK30A.

Which of these three would you prefer (for P90 clone)? I'm planning to order from banzaieffects.com, they offer them all (but 2SK30A is bit pricey - I have to match six as well, so I plan to buy hundred).

Btw. I've heard EVH P90 and liked it (it had significantly sweeter sweep than my /matched/ BF245B based clone), does anybody know which FETs are used there?

Thanks, T.

[Mark Hammer]

To the best of our collective knowledge, MXR/Dunlop is using 2N5952s by the boatload, regardless of P90 version.  Diffeences in the versions are with respect to other factors, not the FETs.

[R.G.]

So, all six JFETs must be matched equally ? That means I will need to test many dozens of transistors !!! 

You have that correct. However, that's not too much of a problem. Today's price on 2N5485 at Mouser is one for $0.11, 25 for $2.38, and 100 for $8.30.

The process is as follows.
1. Get a batch of those little sticky colored paper dots at an office supply place. Take a whole sheet of them and mark the dots with ballpoint pen, numbering from one to more than you have transistors.
2. Stick on one dot per transistor, perhaps as you test them.
3. Run them through my JFET tester, being careful to handle them with tweezers, pliers, etc. to keep hand heat from changing them and throwing the results off.
4. Type the transistor number and test reading into a text file or spreadsheet on your computer as you go.
5. When you've tested them all, use your computer's spreadsheet to sort them by test reading. Now you can simply pick out groups which are as close as you'll get. Matching to 5% or 10% of the reading is probably good enough. They will naturally fall into clumps of several transistors near each other. Those are your matched sets.

[Thomeeque]

To the best of our collective knowledge, MXR/Dunlop is using 2N5952s by the boatload, regardless of P90 version.  Diffeences in the versions are with respect to other factors, not the FETs.

Thanks!

[rantony]

Thanks R.G.

I'll probably go with the 2N5485. As you say it's pretty cheap. But should I expect less desirable sound from 2N5485's over 2N5952's or 2N4302's, or do the final result really depends on the precision of the matched set ?