So what happens if you don’t match JFET’s on a Phaser?

[Rambozo96]

I been reading up on phasers and how it's recommended for the JFET's to be matched and even seen in an old MXR advert for the phase 90 about how they match JFETS.

Thanks in advance

[PRR]

In a 4-stage phase shifter, if the four stages are not at the *same* frequency, the effect is reduced.

OK, let's ask the Idiot. I simmed a perfect (static) phase shifter with matched Rs, and another with Rs scattered at wild values.

The phase curve is much less steep.

When combined with the straight input, the dips are further apart (or the peaks are wider). In fact the "wild" amplitude plot could cover everything a guitarist might play in one passage, "no effect"; the "matched" curve is more sure to nick some of the sound.

[Myampgoesto12]

I have been curious about this as well. This helps me a good bit.

[Mark Hammer]

I take a slightly different tack.

JFETs provide a means for having a varying resistance.  Each JFET has a range of operation where the drain-source resistance can be made to vary by applying current to the gate.  When the inevitable trimmer is set, you are moving/biasing all the JFETs to a point where they can now all be "nudged" into changing that drain-source resistance.  The "movement" of the phaser requires that all JFETs involved, across all stages, provide a steadily accumulating or decreasing amount of total phase shift above some minima as the LFO feeds all their gates. 

The effect you hear is the result of the total phase-shift across all stages.  Because of tolerances of the components involved, including the caps, the JFETs, and whatever resistors may be in parallel with the JFETs, the frequencies where the minimum and maximum phase shift occur can be different for each stage.  In a P90, there is a .05uf (really .047uf) cap feeding the non-inverting pin, and a JFET+22k from the non-inverting pin to Vref.  When the JFET is at its maximum off resistance (in the megohms), .05uf and 22k gives 90 degrees of phase shift for everything above 144hz.  Of course, between the cap, resistor, and JFET tolerances, that maximum phase shift might be happening at 140hz in one stage 158hz in another, and so on.  All within spitting distance of each other, but not identical. 

As the LFO applied to the JFET gates starts to produce a drop in their drain-source resistance, the specific resistance produced will not be identical.  What matters is that, for every stage, the frequency where they apply maximum phase shift is now moving upwards.  If, theoretically JFET+22k now = 10k, then max phase shift would be occurring at 318hz.  Across four stages, that would be 360 degrees of total phase shift at 318hz.

What happens if their ranges do not line up?  Suppose at some point, out of 4 JFETs - A, B, C, and D - A, B, and D are still changing their resistance, corresponding to the LFO output, but D has reached the end of its range and cannot change any more.  That could be at the "top" of its sweep, or at the "bottom" of its sweep.  The other stages are still contributing ever more, or ever less (depending on direction of sweep) phase shift, but one has crapped out.  The result is that the sweep seems less smooth.  I liken it to the "turnaround" when you watch competitive swimmers race end to end in an Olympic pool.  What matters considerably, and gives some swimmers a competitive edge over others, is the smoothness and fluidity of their turnaround.  There can be nothing halting or jerky about it or else they lose momentum.

Similarly, if the range of change of all JFETs involved do not align, such that some are changing drain-source resistance when others are not, the "turnaround" in the sweep just feels less musical.  My phaser sensai, Mike Irwin (call me if you're out there, Mike), taught me that one of the many reasons (and there are several) why photocells make for such great phasers is because they will always yield resistance changes throughout the full range of whatever illumination source is being used.  Indeed, the photocells are not at all matched for their specific individual resistance ranges.  What matters is that they all move together, with none of them "standing still".

So, although we do absolutely NO testing for the specific drain-source resistance of the JFETs used, we always try to select those whose range is fairly similar such that they are all yielding progressively greater or lesser drain-source resistance as the gate current changes, with none of them reaching the end of its range before any of the others.

I hope this makes sense.

[Paul Marossy]

Like 15 years ago I built a Phase 45 with J201s, which is not supposed to work, and I didn't match them either. It works pretty well, and I get a pretty pronounced phasing effect with distortion. Maybe I got lucky?

From what I remember, it's most important to match FETs on the four stage Phase 90.

[Kevin Mitchell]

Wow. PRR came in armed with images. Thank you!
I recall someone working on a jfet phaser that had a trimmer on each stage as a workaround to even things out. I'm not sure what became of it but the idea did catch my interest.

[pinkjimiphoton]

back in the day ya could actually get some fets outta a batch that would be really close and sometimes get away with out matching.

jc maillot has a great phaser project currently awaiting my attention that doesn't match them.

but in general, if ya don't match them, the stages don't kick off n on the same, and it will seriously limit or destroy the ability to phase.

rullywow has a jfet test product, it's rg's design and works great

[EBK]

Careful, Jimi. You just mentioned J.C. and R.G. in the same post.   

[Mark Hammer]

A number of commercial phasers, including the MXR Phase 100, and the Boss PH-2, include several fixed phase-shift stages.  As owners of those pedals can attest, you still get great phasing out of them, even though some of the stages aren't "moving".  HOWEVER, they start out that way, and stay that way.  What you hear is the cumulative phase-shift of the fixed stages and the swept stages.  The less musical-sounding effect occurs when one or more of the stages that are supposed to sweep, and DID sweep, stop sweeping before they reach the top or the bottom of the sweep range.

I think it is fair to note that, in designing the Phase 90, MXR went with a single sweep width, intended to be good enough for the fastest and slowest speeds of the pedal.  Where the sweep width is modest (and the LFO output current-limiting equivalent is a higher value, like 3M9), there's a pretty good chance that any two or four JFETs of the same type will work fairly well.  Or rather, there is a smaller chance that one or more are going to crap out over such a limited sweep range, once a common bias is set.  If the width of the sweep range is extended, however (and some designs will permit the sweep to be ultra-wide for very slow speeds), I suspect that matching is probably a wise strategy to employ; especially if using more than 4 stages.

Some time back, Steve Daniels (Small Bear) had sent me 16 pre-measured 2N5457 JFETs to see if they could be suitable candidates for phasing.  They were all spec'd for ldss, transconductance, and Vp (like I know what those mean  ), and could be ranked from most to least on each of those.  I had a Phase 90 board ready with sockets for the JFETs, and found that any 4 "consecutive" JFETs in the 16 sounded just fine, once the bias was adjusted.  For instance, #3 thru #6 have ldss specs of 3.45 to 3.71ma, and Vp specs of 1.62 to 1.65 (#16 had ldss=2.91 and Vp=1.4, while #2 had ldss=3.72 and Vp=1.66).  They aren't "matched", just close enough to do the required job.

And to repeat, as Mike Irwin had conveyed, and demonstrated, LDRs are rarely matched (and seldom able to be matched), yet sound great in phasers, because it is a very rare instance where one or more yield no more resistance change at some point in the sweep.  The over-arching objective is to have ability-to-change for whatever sweep range is going to be covered.  As long as a given JFET can do that, it's good enough.  Matching can assure that, but in many contexts may not be absolutely necessary to achieve it.

[amptramp]

There is a version of the Phase 90 using PWM switching with series and parallel resistances and CD4066 switches to replace the variable resistance of the FET stages.  Since the off resistance is quite high and the on resistance is quite low, the resistance is simulated quite well with switches driven by the same PWM waveform.  It is attributed to Trolltech ASA.  This should work perfectly well with no matching required.

[Mark Hammer]

In the late '70 and early '80s, MXR produced a number of products that used PWM of CMOS switches to great effect.  Among them was the Envelope Filter, the Commande Series phaser, and the Analog Delay.  In all instances, PWM of a solid-state switch was used to sweep or adjust a filter, by simulating a varying resistance.

The advantage this had over JFETs was that they did not require any matching or component selection.  The advantage that this had over OTAs (like the Small Stone) was that it had better headroom and tolerance for hotter input signals.  The advantage it had over LDRs was that it took up less space, used less current, and required no calibration.

[PRR]

> PRR came in armed with images.

Yes; and I got the R and C in the wrong places. We'd normally put all the R "to ground" (actually a bypassed bias voltage) so they would take the same Gate voltages.

No real effect otherwise. The phase curve would slant the other way, but we only hear the difference not the slant.

[StephenGiles]

Mark has hit the bullseye with his replies here! Interestingly, I played with phasers based on the ETI phaser which used a 4049 cmos chip instead of fets. Presumably all of its inverters are matched,  but the phasing effect was a muddle!!