JFET Symmetry 1

[dau]

Hello NG,
I use two JFET MMBFJ309LT1G to switch 2 different channels to an inverting amplifier (OPA656U) with a gain of 0,6 (JFET3 closed). Furthermore the gain can be switched to 6 by opening this third JFET increasing the feedback resistor respectively.
All this worked in my previous layout with a +/-1V at each channel-input very well with a good symmetry. +/- 1V at the input resulted in -0,600V/+0,602V at the amplifier-output.Now I've changed the layout to add some features but not the basic amplifier circuit.Suddenly I get -0,600V/+0,610V with +/-1V input.I've been searching the whole day for a difference and have already changed the OPA656 but without success.Is it possible, that two different lots of JFETs have so different values?
Thanks

 

[VEBill]

A 1.3% (0.610v/0.602v) difference is pretty small.

 

[8060A]

dau,

You did not mention how you are controlling the J309. N Chan JFETs as switches will have varying on resistance from device to device and also the threshold of devices varies as well. Leakage current could also contribute to the difference. When using J309s, I usually pinch them off with at least -10V supplied by a comparator with an open drain output. To turn them on without introducing additional errors, a 1M resistor from gate to drain (or source, doesn't matter with symmetrical JFETs) usually suffices. Anyway, leakage from your control circuit is often a source of error. Accuracy can be improved by increasing your feedback resistances so that the on resistance of the J309 is negligible.

Also, if you are using a hard voltage to turn on the JFET, then you will get resistance modulation since the input voltage is changing the G to S voltage. You can make this circuit have lower errors and better matching with proper design.

 

[dau]

@8060A:
You're right, I pinch off the 309 with -9,5V through a Mosfet and switch them on via a 180k S-G-Resistor.Because also frequency is involved I added 39k in series to the Mosfet.
Your idea to increase the feedback resistor is probably the best way.Currently I use 300/200 Ohm as Rg/Rf.Perhaps I'm too obsessed with bandwidth.I think 1k/680 Ohm could work also, I'll try it.The actual Ron of the 309 isn't important because I use a potentiometer to adjust the final value.It only struck me that the +/- symmetry was almost ideal with my first board.