Phase 90 FET matching

[mdh]

There's been a lot of discussion here and in the build reports on Tonepad (http://tonepad.com/project.asp?id=7) about the matching of FETs in the Phase 90. I just wanted to add my observations to the mix and keep the discussion going. I built a Phase 90 a couple of weekends ago, and I didn't have a lot of FETs around, so I just popped in 4 random J201s, and it seemed to work fine. Of course, I had never really heard this circuit before, and don't have any other phasers, so you can take that with a grain of salt. Since those 4 J201s constituted 66% of my J201 inventory, I didn't bother with building R.G.'s matcher, I was just happy to have a cool sounding effect. The other day I got 20 more J201s from Aron, so I set about looking for matched quartets. I came up with a couple of tolerably well-matched sets, and I plugged one of those sets in today, after playing awhile with the original set. I also kept track of where the original 4 were plugged in on the board, and measured them after the fact.

The matched set measured as follows with R.G.'s tester (part designations correspond to Tonepad schematic):
Q1: -0.247
Q2: -0.250
Q3: -0.262
Q4: -0.265

Unmatched set:
Q1: -0.267
Q2: -0.166
Q3: -0.307
Q4: -0.142

How did each of these sets sound? Well, when I first plugged in the matched set, I felt like maybe I was getting a more perceptible sweep at slower speeds than I had with the unmatched set. I also had noticed some distortion with the unmatched set, and this seemed to be reduced with the matched set. What really surprised me, though, was that I couldn't get as much intensity in the effect with the matched set. I tried setting the trimmer to the position that yielded a little too much phasing with the unmatched set, and I got LFO ticking (ticking isn't exactly the right word, it's more like gated swoosh or something). In fact, even at what would have been the sweet spot on the trimmer, I still got LFO noise for slower rates. Any thoughts on why this would be a problem with the matched set but not the unmatched set?

The other thing I've been wondering about is choice of FETs. I don't have any FETs other than J201s in sufficient numbers for matching, so I'll probably order a pile of 2N5952s in my next Mouser order. I seem to recall reading someone (probably R.G.) saying that FETs with Vgsoff around -1.5 to -3 were ideal for the Phase 90, and of course the J201s fall far outside of that range. Still, I'm really curious as to why the unmatched J201s seem to perform so well, and the matched ones so poorly. I would be interested in hearing about other folks' experience with unmatched J201s versus matched 2N5952s or similar. Cmon, don't be shy!

[R.G.]

It's entirely possible that the extra distortion made any phasing more audible.

[jeff g]

Like you (and many others it seems), I used two random J201s for my Phase 45 - beautiful swooshy sound that to me is more than subtle. 

Then, thinking I should probably use the recommended 2N5952s, I built the Jfet tester.  When I put it two perfectly matched 2N5952s I had really minimal phasing - couldn't fine tune the trimmer to bring in the nice lush phase

I've since put the J201s back in and really like the sound. 

I'm planning on building the Phase 90 at some point and we'll see if it works out the same.

[3/4 North]

I'm still working on one.
I ordered twenty BF244A's and made the tester. The best I could find is a sequence 1.26 - 1.27 - 1.28 - 1.29

I'm using a byoc kit - three of the 2N5952's with the kit were closely matched and the other was different, off more than .10 points.


I'm also using carbon film resistors and I matched all the 10k's - 22k's and 150k's

[ildar]

I'm beginning to wonder if the idea of matched fets is becoming another example of "mojo"?
I did two Phase 45 builds, 1st with matched 5952s, and I got so frustrated when it didn't phase that I de-populated the board and abandoned the project. Recently, I did a 2nd build, matching the fets again, and still no phasing. I then popped in 2 unmatched J201s, and boom-works beautifully. I'd take the time to match the 201s, but why bother?

Indeed, why bother matching fets at all? Maybe J201s are magic.   

[Mark Hammer]

Matching IS an issue, and not mojo.  Like so many things, though, it will seem to be mojo if the circumstances where it matters most are not the circumstances you normally depend on.  There ARE things that are pure unadulterated mojo, like $200 knobs intended to improve tone, but at the other end of the spectrum are a lot of things that can matter very much...except that they only matter that much in special sorts of contexts.  For instance, we've had many discussions over the years about capacitor type.  In the audiophile world, where a capacitor is expected to pass an extremely complex audio signal consisting of many different component sources (percussion, voices, strings, keyboards, etc.) pooled into one very wide bandwidth signal where maintaining phase coherence between fundamentals and harmonics is a bit like keeping every sow with her piglets in the midst of a barn fire, capacitor type can matter VERY much.  But that's not reflective of what matters with a single electric guitar in a circuit dedicated to that one medium-bandwidth signal alone.

FETs are somewhat unique among control elements in phasers because they can come to a point where changes in the LFO voltage being fed to the FET's gate do not translate into resistance changes in the FET.  Unless your light source is a halogen headlight on high-beam, this will not likely happen when using LDRs, and will not likely happen using switched resistors or using OTAs.  The nope-this-is-as-far-as-I'm-going-and-I-don't-care-if-the-rest-are willing-to-change phenomenon that can happen with FETS just doesn't seem to occur with other types of control elements.

This particular phenomenon dictates that "best" performance from a phaser circuit will be gotten when the FETs used can: a) sweep over a wide enough range of resistance values that the notches can be moved around a fair bit, and b) all the FETs move together at each point, even though they don't have identical drain-source resistance values.  Of course, that definition of "best" assumes that what the user wants is a very pronounced wide sweep that sounds intense, and the fact is that not everybody uses their phaser that way.  Some folks think "bubbly fake fast-Leslie" when they think phaser, where others think medium-speed throb, and in both these instances, especially when the FETs are biased in a certain way, the weak spots in the sweep of mismatched FETs that matter so much in slow ultra-wide sweeps don't matter at all.  After all, the rule of thumb is that we are sensitive to rate of change, and when the modulation speed goes up, we prefer to turn the sweep width down.

So, the way to think about FET matching is that if you want the phaser to be able to do EVERYTHING that a phaser circuit is theoretically capable of doing, then matching is advised.  If, on the other hand, you personally only see use in some of those capabilities, and the phaser sounds good enough for your ears, then those FETs are just fine.

Ultimately, what is missing from all this discourse are samples of phasers where the FETs are poorly matched and where they are well-matched.  I'm not sure many folks here can imagine the difference in the aesthetic quality of "turnaround" at the ends of the sweep when one or more stages get obstinate and refuse to budge or participate in the entire sweep.  If any of you folks have a working circuit where the FETs are socketed, maybe you could stick in one or two randomly selected units (making sure to label the matched ones you started with) and record a sample or two.  Doesn't have to be your own playing - could be simply a sample of someone else processed through the phaser.

Alternatively, if you have a trimpot to adjust bias voltage to the FET gates, you can mimic the effects of no-more-sweep-than-this-buddy by biasing the phaser upwards or downwards from the optimum.  You will hear that when the FETs don't follow the LFO for the entire sweep cycle, the phaser loses a great deal of its intensity.

[mdh]

R.G., yes I see how the distortion might enhance the perception of phasing. As far as my immediate phaser needs go, distortion isn't a problem, luckily.

Mark, point taken about sound samples. The lack of sound samples in these discussions has been a source of frustration for me as well. I actually did record myself playing with the two sets of transistors mentioned in my first post, but haven't had a chance to listen to them carefully and pick out comparable clips. I'll try to do that in the next few days, and link the clips in this thread. I'm also inclined to pop for the big pile of 2N5952s sooner rather than later, and maybe I can start to gather some more useful data.

I'm still kind of curious about why the matched FETs seem to exacerbate the LFO noise. Maybe I need to deal with the source of that, though, rather than blaming the FETs. I remember coming across a circuit recently that called for a (obsolete?) TL022 in the LFO, and when looking to sub it, came across threads that discussed LFO noise. I'll have a look at some data sheets and see if I have any op-amps that may have better properties than the TL072 at IC1 in the Phase 90.

[Mark Hammer]

1) I, for one, would love to be able to have a graph showing typical drain-source-resistance by gate-voltage functions for each of a vaiety of FETs.

2) The TL022 would appear to be commonly used in LFO circuits because of its low current drain.  The standard 2-opamp comparator/integrator LFO generates a square wave, then smooths out that square wave by adding enough lag to turn it into a triangle.  It is the sudden current requirement of the square-wave component (which swings almost rail-to-rail) that adds noise to the power lines and audio path.  You can either: a) take pains to decouple the power lines supplying each subcircuit, b) add a slight amount of lag to the square-wave generating portion so it produces a trapezoidal waveform that reduces the instantaneous aspect of the voltage swing and spreads the consequent current drain over time, and/or c) use a low-power opamp that reduces whatever current drain is needed in generating a square-wave so that there is less impact on other chips/subcircuits.  The TL022 is part of the third strategy.

[3/4 North]

Got mine working

first - this byoc build uses six TL071's unlike the three TL072's on the tonepad site.
It has an added depth control mod.
I left off the 22k to keep it script specs.

I tried closely matched BF244A's (reversed with the source and gate pins switched)
The "matched" 2N5952's that came with the kit
and four J201's (reversed)  - all that I had sitting around

All the sets had good deep phasing once the 250k trimpot was adjusted.
The BF244A's did seem thinner but more melodic, no difference between the others.

I had a 500k depth control mod which seemed to be very subtle, but even more so with the BF244A's.

Sorry, I don't have sound clips either.

But it's a great sound! I recommend building one and if you don't match don't worry.

[mdh]

Got mine working

...

But it's a great sound! I recommend building one and if you don't match don't worry.

I wholeheartedly agree. The fact that so many people are getting satisfying results with unmatched J201s indicates that this project is a lot more accessible than it would seem at first. I still want to investigate the details a little more carefully, though, to satisfy my own curiousity.

1) I, for one, would love to be able to have a graph showing typical drain-source-resistance by gate-voltage functions for each of a vaiety of FETs.

Pardon my ignorance, but what would the test circuit for making these measurements look like? I looked at the Siliconix app note "FETs as Voltage-Controlled Resistors," but found only application circuits, and no obvious test circuit.

2) The TL022 would appear to be commonly used in LFO circuits because of its low current drain.  The standard 2-opamp comparator/integrator LFO generates a square wave, then smooths out that square wave by adding enough lag to turn it into a triangle.  It is the sudden current requirement of the square-wave component (which swings almost rail-to-rail) that adds noise to the power lines and audio path.  You can either: a) take pains to decouple the power lines supplying each subcircuit, b) add a slight amount of lag to the square-wave generating portion so it produces a trapezoidal waveform that reduces the instantaneous aspect of the voltage swing and spreads the consequent current drain over time, and/or c) use a low-power opamp that reduces whatever current drain is needed in generating a square-wave so that there is less impact on other chips/subcircuits.  The TL022 is part of the third strategy.

I haven't had a chance to try this, but the TL062 data sheet indicates that its current consumption is similar to the TL022, so I figured I'd sub that in and see if it addresses the LFO noise.