Jfet matcher simple project

[akis]

I needed to match some JFets which I have been using on a Wein Bridge oscillator to provide thermal stability. Each oscillator uses 2 pairs of JFets in the feedback loop configured as pure diodes. The better the FETs are matched the better the oscillator's signal - amplitude, temp stability, frequency stability.

After many experiments I discovered that the BF244A (not B and not C) are the FETs that work the best so I went and bought at least 60 odd of them.

I then went and looked for a jfet matcher but did not find anything nice. So I made my own. Maybe there are nice things out there and I only needed an excuse to make my own. Anyway.

The idea is that we drive two FETs in parallel and we compare the current that flows through them at various gate drives. If the currents match then we claim the FETs are a pair.

Because JFets have pinch off voltages sometimes approaching -12V, the device is powered by +/- 12V. I was at two minds what supply to use for this, a 12V to +/-12V DC DC converter chip costs around £4, and on ebay I have bought switched mode AC to 12V DC / 1A units for £3.50 - so anyway my schematic does not show where the +/- 12V comes from.

The device uses two pots. One pot is the bias which sweeps from -V to +V into the gates of the FETs. As we move from -12V up we will hit the pinch off voltage. At that point it is reasonable to start the measurements. An LED comes on when then pinch off voltage has been reached.

The second pot is the tolerance. How much difference will we allow in the two FETs before we dismiss them as not a good match? The larger the tolerance the looser the pair will be.

In practice I insert the two FETs and sweep the bias just beyond the pinch off point. At that moment I have a green LED telling me that the FETs are now switched on. I may also have another green LED to tell me they are matching or red LED to tell me they are not. I will then sweep the bias all the way to 12V and expect to see the green LED light to stay on indicating that the two FETs match. If the red LED comes on, indicating a breach of the current tolerance level, I then increase the tolerance to make this test pass.

I then remove FET number 2 and place it aside as my best match so far. I then insert another FET in its place and re-run the sweep. If the RED light comes on at any point during the sweep then this new FET is a worse match than the previous one (which I have put on the side). If the RED light stays off, I then reduce the tolerance pot until the red light *just* comes on. Since I have reduced the tolerance this new FET is a better match so it replaces the previous one I had put on the side.

This process repeats until I find the FET whith the lowest threshold pot position - that indicates that the current pair of FETs are a very good match requiring very little tolerance to keep the green light on. It is like a bubble sort process where we try to find the minimum, if you wish.

I have used this on breadboard and have sorted through dozens of JFets - it is dead easy and very quick.

The schematic is below. Q1 and Q2 are the DUTs. U3A detects the differences in drain voltage and amplifies by 10. Needs to be a FET input opamp because it is sensing voltages quite close to positive rail. D2 helps reduce this voltage somewhat to make things easier but is not needed. U2A takes the absolute value of the differential voltage since it could be positive or negative. U2B compares this voltage against the "Tolerance" setting and indicates Green if less and Red if breached. Finally U3B is a cheap way to tell when Q1 has just come on - it is not perfect but it is not easy or required to measure nA which is the typical off-current of JFets.

A voltage meter could be attached to the bias pot to show us the actual pinch-off voltage which would be the reading shown as the U3B LED comes on.



U2A

[dannyf]

Seems to be very complicated.

I would retain the bridge configuration, and simply use a coil meter between the drains to indicate inbalance.

If you want to make it into a tracer, you can use a staircase oscillator to drive the gates.

[void_error]

The first idea I had when I read this was to use a single supply with a TL431 to lift the sources above ground and differential amp with a LM393 window comparator driving an LED to indicate matching. One pot will set drive voltage while the other will set the comparator window voltage for maximum allowable imbalance between the JFETs.

[akis]

Already been there and done that.

The coil meter is not very helpful because it does not store the worst imbalance. You could throw away everything and simply plug a multimeter on the two drains, or a scope, same principle as the coil meter, I checked visually the imbalance on two traces on the scope screen, as the bias changed, and I can see the distance between the two traces, the imbalance. But then, I  need pen and paper to write down the size of the imbalance as I go along trying to find a good match, and if I get interrupted I will not remember what the imbalance actually was (unless I have been keeping notes). It's like the peak detectors that come with VU meters on mixing decks and consoles - without them you'd need pen and paper to write down the levels across different passages.

This little device does this for you, the position of the Tolerance pot is a marker for the "best match so far" and you do not need to write down anything, no need for scope or coil meter or multimeter, no need to remember anything and is very accurate as well. And in terms of complexity, well, two op-amps, 21 resistors and 5 capacitors, that is not complex is it? I am planning to put in inside a small case, 8x9 cm, to resemble a small, handheld tool.

[rob77]

i like the idea  relatively simple and very convenient to use