Opto in amplifier screws up output

Looks like too much gain. (slew rate limit on the opamp.) Try less current in the AGC part of the circuit.

George H.

FET couplers work as controllable resistors only for very small voltages across the FET, plus minus very few tens if millivolts. After that it leaves the ohmic region and distorts. Figure 2 shows it, past +/-50mV it all flattens out.

You might be better off with an LDR and a light source. Unless it's for a product in a RoHS country, then no dice because they contain Cadmium and some armchair bureaucrats outlawed that.

Way too much javascript and tracking crudware on that site.

Was the resistor the same value as the resistance of the photoFET at your bias point? Those suckers are pretty nonlinear at high resistance.

Cheers

Phil Hobbs

Yes, I had them both around 60k, and your right way in the non-linear region. I've dropped the value down to 4k, and it's not perfect but looks a whole lot better.

thanks!

What's the time scale on those scope shots? That photofet has 50 pF of capacitance at 15 volts, probably lots more for small signals.

You might consider using an analog multiplier, or buying an official VGA.

Or the poor man's VGA, an OTA. LM13700 is still in plentiful supply (for the moment), and is more stable than the opto, anyway (which is probably saying something).

The signal level isn't any more (~60mV with predistortion diodes in use), but you get a lot more voltage at the output for that swing.

Tim

OTA chips (LM13700, NE5517 etc.) are well-behaved variable gain devices, as long as you keep the differential Vin under 20-50 mV. You get a current-source output, so voltage shift and Zoutput control are freebies.

Ditto what Joerg and Phil have said. I assume these things act like regular JFETs. My experience matches what's been said: the key to making a decent circuit out of them is to keep the voltage swing across them small.

Or, use a DUAL opto, one to send and the other as the feed back for the input stage.

You'll have isolation, if that is what's being sought?

Jamie

The OP just states that he wants a voltage-controlled amplifier.

No, that doesn't help for this problem. What you suggest will help keep the conductivity (or resistance, depending on what you choose) linearly proportional to the control input, but this is a different problem: For a fixed control input, if the device is a FET there will be a non-linear relationship between Id and Vds. To fix this one needs to keep Vds very small (which might lead to poor signal-to-noise ratio), or change to a Cadmium Sulfide LDR etc., or re-design it to use a "digital potentiometer", multiplying DAC, VGA, etc.

Chris

Chris

If you use the input diodes, the input is a current as well.

Cheers

Phil Hobbs

FET based VCAs can work very well.

Set up the JFET as an voltage divider with the JFET as the R to ground.

Keep the audio voltage across the DS very small.

Add feedback of 1/2 of the audio on the Drain to the gate control signal to cancel the second order distortion.

Google this, it is a well known technique that works very well.

Apply the DC VCA control to the gate along with the distortion cancelation signal.

Mark

Across a narrow range of gains (say 2:1), not bad. They're terrible if you need cutoff, though (more than 1:10, say). This rather limits their use -- e.g., small range calibration*, low modulation AM, audio effects (vibrato), etc.

(*But for calibration, the residual nonlinearity would probably be as big a problem as the drift..?!)

What you get is, when the gain is set low, it's operating at very low [negative] Vgs: near pinchoff. It takes only a little drain voltage to fully pinch off part of the channel, and therefore enter the linear (read: constant current) region. The feedback resistors help a little, but only by a factor, not by orders of magnitude. Hence, it's no good if you need orders of variable range.

I suppose you could try a limited range like 1:2, and subtract the input from it, so it's variable down to zero. But now you're subtracting harmonics that weren't there before, so you'll have a residual that's inevitably distorted, no matter the signal level (asymptotically speaking, that is). There's also no good way to null it; it has to be calibrated live, otherwise when you wanted "1%" you might end up with "-5%" instead!

BJTs (e.g., OTAs, internal mult. and VGA cores) are fantastic for wide range (up to, what, 7 decades or so?), drift about as much (due to Is and Vbe), and their distortion and input range don't vary with bias (because the derivative of an exponent is an exponent is an exponent! :) ).

Tim

That's a bad design. It needs a series-R before the photoFET. Ground the (+) input of the op-amp; do not connect to an input or output node.

Have you figured this one out yet?

I never did.

The wrong-direction spikes are from gate-to-source capacitance. The big long slope is drain capacitance discharging into the 5K resistor.

I designed a cool circuit recently, wrecked by mosfet g-d capacitance.

Oops, make that gate-to-drain capacitance.