Buying FETs in pairs?

[Tobes]

I busted one of the legs in a 2n5485 when building/debugging a ggg phase 45.

I went to a regional electronic superstore (Sayal, Vaughan Ontario) and they only sell them in bagged pairs. Does this mean that each pair is likely matched?
I was hoping they were because I haven't built the geofex Jfet matching circuit.

This is only my 3rd project, but I've got a couple lined up after I fix this one. Should I go ahead and start sticking them in the circuit, or wait and build a FET matcher? I've got 3 pairs...

[Pyr0]

Yes, for a phaser the fet's have to be matched, so your going to have to replace both of them. Lucky it wasn't a Phase 90 

[Pyr0]

just to add, hopefully they might be matched if they come in pairs already. Probably best to build the jfet matcher, pretty simple build on a small piece of vero.

[R O Tiree]

Probably best to buy a batch of 50, at least, if you're going to do more work with FETs, so you can get very closely matched pairs/quads

[amptramp]

Sayal sells a number of things in more than one unit per bag.  It does not mean they are matched.  I live about 4 km south of the Sayal on Matheson just west of Dixie in Mississauga.

[chromesphere]

I used J201's in the p45. Just plugged in random pairs until i found a pair that sounded good.  Cant do that with a p90, you would probably be swapping them forever, but for the p45, i think i used maybe, 6 before i got a pair that sounded nice.  I built two p45's this way.  So yeah, you could buy 50 j201's and give that a go.
Good luck,
Paul

[R O Tiree]

I used J201's in the p45. Just plugged in random pairs until i found a pair that sounded good.  Cant do that with a p90, you would probably be swapping them forever, but for the p45, i think i used maybe, 6 before i got a pair that sounded nice.  I built two p45's this way.  So yeah, you could buy 50 j201's and give that a go.
Good luck,
Paul

The FET matching article at geofex shows a very simple circuit that enables you to measure Vgs (at an Rds of 10k) that you can breadboard in about 5 minutes, tops. I got a batch of 100, attached a little sticky label to each one, numbered 1-100 and opened a new Excel spreadsheet. Column A was filled in 1, 2, 3,...98, 99, 100 and column B had all the measured Vgs figures. Sort by Column B and I was able to pick out loads of useable quads and pairs. Only a few were "loners". It's a little time-consuming but saves you an awful lot of swapping things around randomly (which might or might not, as you say, be successful). Another neat thing about that circuit is that, if you replace the 10k from the Source of the FET to GND with, say, a 4M7 resistor, you will get a reading within a few µV of Vgsoff.

[Sacorus]

I'd also recommend building the simple jfet matching circuit.

I didn't bother using post it notes or graphing it out, just lay them down on my work bench in V order. Found 3 matching pairs for phasers after about 20 tested. My phasers sound great!

[chromesphere]

Also, although its a simple circuit, if you want to make a PCB version of it, there is a jfet matcher project over at madbeanpedals.com in the members tutorial section.
Edit: ah, yeah you will need to be logged in to see it.
Paul

[Tobes]

Thanks for all the help everyone. I suppose my options are either:
1. Build the JFET matcher (which I was hoping to do eventually)
or 2. Use J201s instead. I reckon they tend to be more consistent?

I was hoping to plug and chug a few pairs of 2n5485 from the half dozen I bought and hope it turned out okay. But I'll probably learn more with building the geofex matcher on vero, plus it seems like a nice and simple endeavour for a first-time vero user.

Thanks!

[R O Tiree]

5485s go from Vgsoff of -0.5V to -4.0V, according to the spec sheet, so the Vgs(10k) figures will vary by the same ratio (8:1). That said, I bought a job-lot the other day and most of the 20 that I tested (at random from a very large bag!) had Vgsoff numbers in the -2.85 to -2.95V range with one at -3.9V. Even a 0.1V spread is probably too high.

J201s are no more and no less consistent than any other FET. They have a 5:1 spread of figures (-0.3 to -1.5) but that's not the whole story... You've also got to ensure that you can get the swept signal from the LFO into the (for want of a better phrase) "active" region of the FETs' response. Let's take one at the high end of the range... Vgsoff of -1.5V. It is fully on at Vgs=0V and fully shut off at Vgs=-1.5V. It's no good at all if your LFO signal is from, say, -2V to -3V with respect to the Source voltage. It's sweeping, but the FET is fully shut off at all times. How about if it's from -1V to -2V? Well, it's only "working" in this circuit for half the sweep. How about a J201 that has a Vgs of only -0.3V? Even if you can get the LFO signal biased correctly, it's only altering the Rds of the FET for 30% of the LFO sweep. That said, although I've made phasers in the past I cannot recall the range of sweep for the LFOs in those circuits... it may be that they're quite small... or can be made so with the trim pot... but you still have to get that sweep into the "sweet spot" for the FET type you're using.

If you haven't got a breadboard, invest in one - If you make a mistake, you just move a wire to the correct place (beware power connections to ICs), compared with a lot of tedious mucking about on vero, soldering, debugging, etc... prove your circuit and your logic on breadboard first, then design your vero

[Mark Hammer]

1) As noted, components sold in even-number quantities are not necessarily sold as such because they are matched in any way.  I'm sure they also sell opamps in bags of 2 or 4 and those are not "matched".

2) When JFETs are used as voltage-controlled resistors - as they are in many phasers - matching helps in improving the "turnaround" and making it more musical; i.e., that last bit at the top or bottom of the sweep.  It's not the matching of the resistance, but rather the fact that the FETs are all responding to changes in what they get from the LFO circuit.  Indeed, photocells are hopelessly unmatched for resistance, but because they generally respond to the brightest and dimmest points in the light source, none of them ever gives up or stops changing resistance.  By contrast, the more limited operating range of JFETs can mean that units A, B, C, and D are all sweeping upwards (i.e., gradually decreasing resistance) together, but at a certain point, unit B simply won't change any more, until the LFO voltage comes back down again.  Going in the other direction, all four change together going downwards (gradually greater resistance), and then at a certain point, C craps out and only A, B, and D are changing.  The matching seeks to identify multiple JFETs that will all sweep together in response to the same LFO output voltage.

3) How important paying attention to the "turnaround" is may well depend on a number of situational variables.  The fewer the number of phase-shift stages, the more the unit is depending on any single one to play nicely with the others.  Having one JFET out of 8 go on siesta for the last 10% of the sweep may well be unnoticeable.  It is probably also fair to say that the obviousness of the lazy FETs will depend on sweep width and speed, as well as resonance/feedback settings.  High feedback with slower sweeps focusses our attention on "where the sweep has taken us now", in which case not only the shape of the sweep, but the consistent co-movement of the stages becomes important to how "musical" it sounds.  If there is negligible feedback, and the sweep is fast enough that your attention is directed to the instability and bubbliness rather than "where the peaks are now", I doubt that matching makes an audible difference.