FET Matching advice

[Beo]

My purchase of 100 2n5952 transistors from Mouser seems to have yielded a lot of well matched sets. I'm trying to decide how greedy to be, as I think I have a few ranges where I could have as many as 18 transistors in a matched set. Here's my data. Please give me advice on whether these are good sets to work with for phaser applications. Thanks!
Travis

Set of 18: VGS from -1.518 to -1.552, max/min difference of 2.2%
(I can make a set of 16 out of these with a max/min difference of 2.0%, or a set of 12 with a max/min difference of 1.2%)

Set of 12: VGS from -1.620 to -1.641, max/min difference of 1.3%

Three Sets of 8:
-1.842 to -1.873, max/min of 1.7%
-1.669 to -1.694, max/min of 1.5%
-1.489 to -1.504, max/min of 1.0%

(In these cases, do I want the set that matches closest, or do I want the set with biggest or smallest VGS?)

A bunch of sets of 4 with max/min diffs of 1.3% or lower.

Here's a graph.

[haveyouseenhim]

.

[Beo]

Sorry I cant help, but can you point me in the direction of some clear instructions on how to match fets? I'm having trouble understanding how it works.

I wish someone would make a video tutorial. Some things I just need spelled out to understand.

As usual, geofex has the answer:

http://www.geofex.com/Article_Folders/fetmatch/fetmatch.htm

[PRR]

You want Vgs to be much more than signal voltage.

Here is a thread where Vgs of 0.6V or 1.0V apparently would "distort my sound ... enough to discredit my clean tone."
http://www.diystompboxes.com/smfforum/index.php?topic=99100.0

Vgs of 4V or more, it may be impossible to get full swing in 9V systems. (You don't have this problem.)

I would set-aside your 1.3V group. They may be great amplifiers/buffers, but might be marginal as phase-modulators on guitar-level signals.

[haveyouseenhim]

.

[anchovie]

You mean the Geofex schematic? That's a single opamp. The inputs labelled + and - on the left of the triangle are the non-inverting and inverting inputs. You'll notice that the line coming out of the top side of the triangle goes to 9v and the one out of the bottom goes to ground - that's the power supply!

[haveyouseenhim]

.

[Beo]

You want Vgs to be much more than signal voltage.

Here is a thread where Vgs of 0.6V or 1.0V apparently would "distort my sound ... enough to discredit my clean tone."
http://www.diystompboxes.com/smfforum/index.php?topic=99100.0

Vgs of 4V or more, it may be impossible to get full swing in 9V systems. (You don't have this problem.)

I would set-aside your 1.3V group. They may be great amplifiers/buffers, but might be marginal as phase-modulators on guitar-level signals.

Guys, please try not to hijack the thread too severely.

PRR, I appreciate the advice, but I don't really get it. The 2n5952 datasheet specs VGSoff between -1.3 and -3.5V. My testing is right in this range. The 2n5952 is a standard part for the P90 type phasers. I'm pretty sure I measured VGS correctly using the test circuit. I would think my -1.6V FETs will be perfect. I guess the proof will be on the breadboard.

Travis

[haveyouseenhim]

Sorry beo. totally didn't realize i was threadjacking till now 

[R.G.]

PRR, I appreciate the advice, but I don't really get it. The 2n5952 datasheet specs VGSoff between -1.3 and -3.5V. My testing is right in this range. The 2n5952 is a standard part for the P90 type phasers. I'm pretty sure I measured VGS correctly using the test circuit. I would think my -1.6V FETs will be perfect. I guess the proof will be on the breadboard.

The proof is always on the breadboard. But Paul is correct, you really can't use raw JFETs with signal peaks near the Vgsoff as variable resistors and expect low distortion unless you worry about and provide for the signal swings some way. If you think about it, if the gate is sitting at -1.6V with respect to the source, there's -2.6V on the drain, and the channel is expected to be a resistor, there's some point on the channel where the gate, which overlays the whole channel, is going to be forward biased and conduct into the channel, causing distortion.

In the bad old days when the P90 was designed, it was common to simply expect guitar signals to be "small", in the 100mV range and not worry about it. Today, it's unusual to have an effect run with just a single coil guitar as an input. Even if the guitar is not loaded with mega-output humbuckers, it's common for the first few effects to be boosters or distortion pedals with big outputs, in the one to four volt peak range. If the signal is already distorted, a little JFET distortion doesn't get noticed much, but big guitar signals and boosted signals will let it be obvious.

You're both right, in that the 1.6V FETs will work OK for "small" guitar signals, such as raw single coil outputs. It will also have some signal peak distortions on bigger signals. And signals are limited practically to about 4V peak in a nominally 9V system, so if you can get to 3-4V Vgsoff, so much the better. This means the LFO has to have a bigger signal, too, to modulate the bigger Vgsoff, and the bias point for the sources gets to be critical, and ... well, you see how the complexity builds. The P90 was a good, quickie design for a simpler time that we just don't look at too closely.

Many people prefer the phasing sound of JFETs with drain feedback to the gate to linearize the JFET as a voltage-variable resistor, as in the P45. It has its own set of complexities and design challenges, but it lowers the JFET distortion.

My advice? Put in any one of your matched sets and listen. You'll get a good P90 sound with any of them, but least distortion on peaks with the biggest magnitude Vgsoff sets - if that matters to you. If you don't like it, go for bigger Vgsoff sets, or make the leap to the feedback-enhanced JFET circuit, which was never put in a production P90 to my knowledge, but is a trivial change to the overall circuit.

[Beo]

My advice? Put in any one of your matched sets and listen. You'll get a good P90 sound with any of them, but least distortion on peaks with the biggest magnitude Vgsoff sets - if that matters to you. If you don't like it, go for bigger Vgsoff sets, or make the leap to the feedback-enhanced JFET circuit, which was never put in a production P90 to my knowledge, but is a trivial change to the overall circuit.

Thanks RG. I never really considered that the P45 architecture was in some way correcting for a limitation in the P90 design... I thought they were just two different approaches to achieving phasing. I'll Breadboard both to compare. I assume that the P45 FET architecture can be expanded for as many stages as desired, similar to your P180 extension of the P90 design?

Also, using the P90 FET architecture, might it be helpful to put a signal limiter in front of the circuit and a gain recovery after? That way we could limit the signal level going through the FETs.

Finally, I hear you in that I should use my strongest FETs (ie. biggest absolute VGSoff). What do you think of my correlation... is 2% or less good for a matched set?

[Gurner]

Finally, I hear you in that I should use my strongest FETs (ie. biggest absolute VGSoff). What do you think of my correlation... is 2% or less good for a matched set?

What tolerances are the caps & resistors in your allpass stages?

[Beo]

What tolerances are the caps & resistors in your allpass stages?

2% metal resistors, 5% or 10% film caps.

[Gurner]

2% metal resistors, 5% or 10% film caps.

then I'd have said that 2% FET matching would likely be fine (but there are some on here who are phaser fanatics & can likely give you a definitive asnwer!)

[R.G.]

I assume that the P45 FET architecture can be expanded for as many stages as desired, similar to your P180 extension of the P90 design?

Yes.

Also, using the P90 FET architecture, might it be helpful to put a signal limiter in front of the circuit and a gain recovery after? That way we could limit the signal level going through the FETs.

If you're going to that much trouble, use a compander instead. SE571 is a standard answer to dynamic range issues in many audio setups.

Finally, I hear you in that I should use my strongest FETs (ie. biggest absolute VGSoff).

It all depends on what signal you feed it. If you're under a peak signal of a volt or so, don't sweat it. If you're feeding it boosted signals, or humbuckers, go big. If you don't know what it will be getting, as is mostly the case, worry and think about the issue a lot. That's what I do, anyway.  

What do you think of my correlation... is 2% or less good for a matched set?

Considering that a lot of people can't be bothered to match at all and post about good results with out-of-the-box sets, 2% is a masterful matching result.

As has been belabored by others, Vgsoff matching, which my matcher doesn't do exactly anyway, is only an approximation to what you really want, which is resistance matching for a range of Vgs. What you really want is for the JFETs to be at the same rds for the same Vgs. What you can settle for is them being close, and one or more JFETs not either bottoming out or maxing out while the others are still in the good rds region. I have always taken that set of reasoning to lead to the approach of getting them close and not worrying too much about perfection. Getting perfect matching won't make the sound of the phaser magical.

Probably...  

then I'd have said that 2% FET matching would likely be fine (but there are some on here who are phaser fanatics & can likely give you a definitive asnwer!)

The fanatics you have with you always.   

[Beo]

Thanks RG. Great help!

And of course, now I have something else to research and learn about... companders! This hobby never stops... with each questions answered, new questions pop up! I am so down the rabbit hole!

[R.G.]

This hobby never stops... with each questions answered, new questions pop up! I am so down the rabbit hole!

Yep. I've been at it for over 40 years now. It's - intriguing.

[PRR]

-

[Beo]

> Please give me advice on whether these are good sets to work with for phaser applications

I am sorry my advice offended you.

Geez no, PRR, I was not offended. Sorry my comment came across strongly. I meant only to clarify one of my questions: whether the correlation of my data was in the range that is typically sought when matching FETs. I think I have confirmed that 1-2% is great correlation.

The info I learned from you and RG is immensely helpful. On the breadboard, I will test to see if my FETs cannot handle my signal raw. If not, I will have options I can explore, including reducing my signal (compander), trying different feedback designs (e.g. P45) or trying a different set of FETs (e.g. 2n5458 with -2.5 avg VGSoff per datasheet). See... I do listen to advice that is given

I've also read that the biasing can be important to avoid distortion on the top or bottom of the sweep, and that adjusting the resistor divider from center can help. Lots of ideas to experiment with now. Thanks guys as always. I'm still amazed that after all these years, the vets are still willing to help the newbs.