Cranky Op-Amp

I had a problem a while back, with 80dB total gain, and it was PSRR that killed me - output was 4vpp into a couple of k, and the resultant psu noise, with TL064's having -10dB of PSRR at 100kHz (thats what I measured, bloody gain), was enough to kill me. then of course I had to try and find a fast opamp with low current, that was cheap. I couldnt decouple each amplifiers psu, because there were 40 of them and I couldnt spend the dosh....besides, at 100kHz, I needed a humungous cap, $$.

Cheers Terry

I agree. I liked the LM318 when it first came out, about a million years ago (erpp, over 35 years ago), but it was finicky to keep from oscillating, etc., and I was glad to see the last of them a few years later (did'ja notice the funny glitch in the curves at 300kHz?). Even tho it was a serious high-speed amplifier (many companies made their own second-source versions), especially with its fabulous miracle feedforward compensation with 150V/us slewing, or add overcompensation for the other side...

Despite all this, Paul should take your advice, Fred, and give 'em the old heave-ho. I did.

Thanks, I'll take Fred's advice. They also have a lot of latch-up. Oh well, there's always that dusty LM741 in the back office, lol. Just kidding and many thanks for the great help!

Paul

In article , wrote: [...]

Im not sure I followed the description. Is this a case of a JFETs input capacitance being reduced by bootstrapping?

Basically:

constant current ! ! ---[Z1]---- ! ! ! !---+-+--!-\\X1 ! In---->! ! >--+--+--- !---+----!+/ ! ! ! ----------

Obviously X1 isn't really an op-amp. It would have to have a couple of volts of offset voltage if it was etc. The overall feedback is from the output to the JFETs source so all of the terminals of the JFET have the same AC (within the limits of gain) voltage on them so no current flows in the device capacitance.

Much better is a series resistor. You add, say, 100 Ohms from the op-amp's output to the cable. Connect the feedback network to the far end of the resistor, so the loop stays closed with respect to what is sent out on the cable. This will certainly tame the effect of output capacitance.

That may, or may not, be the real problem, however. 58 dB is a LOT of gain, and extremely careful circuit layout practices are going to be necessary.

If you want a really GOOD diff amp, see the Analog Devices OP620. It is truly spectacular. I used it in a servo drive where there was a tens of millivolts current sense signal superimposed on a 100 V 100 KHz square wave common mode signal, and with a little matched resistor attenuator, it handled it with total grace! In another application, I was having trouble until I realized that a differential-mode remnant of the local AM broadcast station was coming through the AD620 without distortion, and being rectified by the next stage. Reducing the bandwidth of the AD620 was the cure in that case.

Jon

You don't get 4 MHz bandwidth at the 58 dB gain (from another post). It is gain-bandwidth PRODUCT. As you turn up the gain, the bandwidth necessarily is reduced.

Jon

Yeah, that was an integrated version of the even older trick of bypassing the lateral PNP level shifters by putting a cap from the inverting input to one of the compensation pins, in the LM301. I remembered this from a prehistoric NS app note (LB2).

That phase whoop-de-do was what made the settling behaviour of the 318 so crappy for its bandwidth.

Cheers,

Phil Hobbs

See Figure 17 of the MC1494 Data Sheet for how to handle the Lateral PNP compensation.

...Jim Thompson