Jfet Matching Questions ... Again !!!

[petemoore]

I'm using 1/2 of a TL082.
 I'm getting readings from J201's...
 1.14
 1.14
 1.15
 1.16
 1.08
 I've tested 5 and the readings are all within that range.
  Do these look like VGs off readings from Jfets? ie does it look like they're reading 'right',...they're well within the few percent difference specified with the Schematic of the matcher... / :cry: ...?

[petemoore]

I still haven't given up the goal of comprehending enough about what's going on to use the matcher yet. I just haven't found what takes to make it snap into place for me.
 Something about the voltage at which the Jfet shuts off being matched. I think I understand that much anyway.
As far as how to conect the meter [range, polarity..], what to look for, and what the 10k with the extra marking 'set' [at bottom right of the 10k's] means, what it's used for, and how to make use of it, I'm in the dark.
 I'm sure it seems simple when it 'snaps', but I've been stretching to try and figure this one out, on and off for years now.
 I took these readings from the Jfet article at Geo.
 -.05v - 4.0v These I think are the Min / Max Mfr Specs on a Q.
 -1.32v
 -2.72v
 2.5v
 2.6v
 I don't understand the negative voltages, I don't recall measureing a Neg voltage on a Jfet.
 >>>Edit, I thought it over and figure those must be Pos ground devices or something./
 Anyone whos'e tried this with J201's could you tell me if those numbers I saw are anomolous or showing likely good candidates for a phaser?

[petemoore]

6 Phase stages wires, 15 socket lugs apiece...whew...bleary eyed...enough for one night.
 Once I got started and started marking made connections [to keep my place], they went pretty easy...I just had to redo one misconnect, the left and right sides of the OA are different, easy to lose track which side, when the board gets flipped around alot like that...so I think the Stages are wired correctly.
 Tomorrow for the bottom part...the Little OA down there. This I'm planning as using for the last half of the last OA. I'm certainly hoping the OA's are 'seperate enough in one chip halves that which OA does what in the circuit doens't matter...
 So the only things left to do besides the six stages is everything...lol...
 I guess I can just stick what I have in there as far as Jfets' and hope for the best...if the phaser works I can figure I'm lucky, or the matcher also works. Any opinions on the posted VGs off voltages or whatever that Is I posted above?
 Then there are the diodes...I will need to find out what diodes are intended for this phazer.
 Anyone know what the 's' quiggly line of a diode mark [the one at the end of the arrow that's usually flat and 90 degrees to the marking that indicates a wire] indicates? Which types of diodes are used in this build ?

[RedHouse]

It would be nice if RG would chime-in, it's his circuit.
(isn't it?)

[Mark Hammer]

Mike Irwin is the guy you want.  As he tells it to me, the matching is related to sweep.  He prefers to use LDRs in his phasers, which are almost impossible to match tightly, yet the phasers sound great with terrific sweeps.  Why?  Because the LDRs have such a broad range of resistance values that they will demonstrate change in resistance across the entire range of wherever the LFO decides to go - i.e., the LFO output goes a little higher and the LDR still changes resistance.

In contrast, JFETs can have points where they decide "I think I'm gonna go to sleep now" and no further change in effective resistance value is obtainable from them for the remainder of that part of the sweep. I guess the best way to think of it is as if you had, say 8 stages sweeping as the LFO goes upward, and well before the LFO reaches maximum positive swing, one JFET craps out and becomes equivalent to a FIXED resistor, then another, and another.  So, you've now gone from 8 stages (and 720 degrees of phase shift at some frequencies) sweeping to 5 (with 3 fixed stages) for a portion of the sweep.    Where the phaser has the opportunity to sound dramatic, it bows out and says "No thanks.  I'll just stand in the background here.".  If the JFETs are all matched for where they're gonna crap out, then you can adjust the LFO appropriately so that change is produced in ALL of them over the entire sweep of the LFO.  However dramatic the effect is will be constant over the entire sweep cycle.

[petemoore]

I may have stuck 5 Jfets into the Improved matcher the right way.
 Here's the Voltages the DMM read on 'em...
 1.15
  1.15
 1.16
 1.14
 1.08v...
   Please tell me these don't look right.
  Or...
 Or tell me maybe It's possible to get  readings in this range using J201's on a properly wired, properly hooked up, Improved Jfet matcher, connected to a DMM set to the right setting, with the Jfet in there the right way.

[petemoore]

...






  BUmp

[Boofhead]

For the J201 the datasheet shows a pinch-off voltage of -0.3V to -1.5V.  

These are negative because the voltage is gate to source- the gate is always usually less than the source on an N-channel device.  The sign is just a matter of getting the multimeter around the correct way, and really the sign is irrelevant for matching purposes.

The pinch-off voltage is defined at zero drain current.  RG's circuit measures the voltage at finite current, not zero, so the measured voltages from the matcher will be slighly lower in magnitude than the specs.

Your measured values look OK considering the above, however, it's possible wrong connections could give similar voltages.  If you want to check the jig maybe try a JFET with a higher pinch-off voltage - of course you have to get the pinout right.

[R.G.]

Lemme try.

JFETs are depletion devices, which means they normally are "on", no signal required to make them conduct. Like tubes, you have to put a signal on them to turn them off.  With the gate open or shorted to the source, JFETs conduct a current called "Idss", for Current (I) in the drain, with the gate shorted to the source. That is, Vgs=0.

You turn a JFET off with a voltage on the gate negative with respect to the source. A JFET will turn really, really off - giga-ohms worth of off. The voltage between the gate and source that causes drain current to drop to zero is listed as Vgsoff, for "Voltage between gate and source that makes the drain current turn off".

At full on (Vgs=0), the JFET channel between drain and source is resistive. The resistance is also listed on the data sheet as Rdson. Common values are 10 ohms to a few hundred, with the low ones on JFETs designed for switching, the high ones for JFETs designed for RF amplification.

Changing Vgs between 0 and Vgs off changes the channel resistance from Rdson up to essentially infinity. If you want a resistor between Rds on and infinity, chances are you can get it with a JFET.

Phasers need a resistance that wiggles around in that range. Many phasers need a resistance between, say, 1K and 100K. JFETs do this nicely. But they are very non-uniform. Not only do different JFET types have different resistances per Vgs, they vary from one to another within the type.

Case in point. The J201. This is a switching JFET, but it still has a resistive region below pinchoff. (Pinchoff is the drain to source voltage that causes the channel to stay at a fixed current with higher voltages; we have to use JFETs below pinchoff to use them as variable resistors.) The J201 has a Vgsoff spec of about -0.1 to -1.0V ( if i remember right; I don't want to go chasing datasheets right now.) That means that any given J201 **may be** one that goes from 20 ohms Rds with Vgs=0 to fully off with a Vgs of -0.1V. The entire range of resistanct it can sweep is compressed to a 0.1V control signal.  A seemingly identical one **from the same manufacturer** may go from 20 ohms to infinity with a Vgs sweep from 0V to -1.0V, a control range that's ten times larger. This is for supposedly identical devices.

Other JFETs like the prototypical 2N5292 from the P90 have Vgsoff specs of -0.5 to -3.0V. That means the control range from full on to full off is contained within the range of 0 to -0.5V for some devices, 0 to -3.0V for other devices, with most of them in the middle.

My matcher does a single point match. That is, it lets you measure the Vgs voltage that makes a given example of JFET act like a 10K resistor (or whatever resistor you stuff in for Rmatch). That is, it makes the voltage and current in the JFET be equal to the voltage and current for the 10K.

This insures that if you get a quad of JFETs matched this way, you are dead certain that at least one point in the control voltage sweep, they will all be in the desired spot on the resistance curve, and you're dead certain to get some phasing around that spot.

This is how the original 2N5292's were matched for the P90, and probably others. It's not perfect, and here's the fault in it.

A JFET not only has a range of resistance from Rdson to infinity with a range of Vgs, the actual resistance per unit voltage change is not linear. It's not precisely square law, or exponential. This is like pot linearity - changing the pot rotation ( that is, Vgs) a given amount in the low end will NOT give you the same resistance change as the same amount in the high end of the 0 to Vgsoff range. The resistance change per unit control voltage change is not clearly predictable. If you plotted one JFET's Rds versus Vgs, you'd get one curve. If you then plotted a second JFET, you'd get not only another curve, but the curves would not have the same curvature or start and end points, exactly.

A single point matcher ensures that you get sets that have curves that all touch (or nearly so) at one point. What happens away from that point is not known. This is the origin of Mike Irwins' comment that JFETs drop out. Some of them do. They reach the end of their curve and quit when the others are still changing resistance.

The beauty of a single point matcher is that it turns out sets good enough for commercial work ( i.e. the P90) and that non-pros can (sometimes) understand how to do a single point match, as well as the fact that there's about a 50% to 80% yield of quads in a largish batch of same-lot parts. It's fast, it's cheap, it's (sometimes) easy, and it has gotten huge herds of beginners going on something that I still get effects asking me how to do.

If you want to do it *good*, you curve match by matching at more and more points. Here's how:
1. Pick a resistance range. 1K to 100K is a good start.
2. Pick a place in the middle, say, 50K and match a bunch of JFETs there. 3. Throw away any that are not matched well at 50K.
4. Pick another resistance point, say 25K. Match the already-50K-matched parts at 25K. Throw away any that do not match at 25K.
5. Loop back through 4 as many times as you have JFETs and patience for.

Every iteration for another matched point on the resistance-vs-Vgs curve produces a more-closely-matched set of devices. Three points is pretty doggone spectacular IMHO. If you do a good job, you can eventually come up with a set of devices that respond essentially identically. These will produce the most intense phasing sound over the whole range of Vgs that you are using for an LFO. It's about as good as JFETs can get.

And that is the rub. Even **great** JFETs are not perfect. The control voltage sweep is limited and nonlinear (worse yet, nonexponential; exponential is what you want). JFETs vary from unit to unit a lot, and phasers need multiple identical somethings.

So - the matcher matches JFETs at a point where the JFET channel is in the middle of acting like a resistor, and more to the point, acting like a resistor might act in a phaser. It's a single point test. it's good enough for making phasers that work OK-ish when used with good JFETs. J201's are many things, but IMHO, not phaser JFETs. They're too sharp-edged and touchy. The 2N5485 and 2N5292 have always sounded better to me, but as I keep harping on, that's only my ears. Your mileage may vary. People seem to insist that if they get J201s, they're going to use them for everything. Sometimes that even works.

Interestingly,  I don't know that anyone has ever called me on the real flaws in the matcher, those being that it's a DC test, not an AC test and it's at a largish voltage compared to actual signal voltage. I actually designed a MK3 that did an AC test at 100mv of signal. I never put it up because
(a) it is much harder to build
(b) the results seem to be about the same.

Did that help?

[petemoore]

Yes RG.
 That helps alot !!!
 This writing has caused the 'picture to snap into focus'.
 I had to kind of conjure an abstract, moving mental 'picture' of Jfet DGS, and related voltages/resistances to get it, then I read it a few more times, because it made me so happy   .
 This wording of Jfet's in  phaser function did the trick for me, sometimes I get mental bloxx.
 Thank you for adding some terminological definitions 'in the article !!!
 I appreciate you're having taken the time to type this up !!

[Kleber AG]

:wink:  Thanks RG!
Yeah the picture started to snap into focus
And I agree pete, the "terminological definitions" really helped and answered me a lot!

Definately a must for the FAQs...

Thanks
Kleber AG