Cranky Op-Amp

Post a schematic to alt.binaries.schematics.electronic or post a link here.

...Jim Thompson

The capacitance of 20 ft of audio cable is *minute*. That isn't a problem.

So ?

Op-amps don't have hysteresis.

Rubbish.

Sounds like you have instability. Do you have any power supply 'decoupling' ?

Graham

The first thing to fix is the capacitative load. 20 feet of cable is

- that 100pF is very much a rule of thumb).

Regular op amps oscillate when loaded with more than 100pF. There is a standard circuit to fix this - see page 9 of the LM318 data sheet

The oscillation is usually at a pretty low level at the output, but the swings inside the amplifier are apparently larger, and mess up the amplifier performance.

The LM318 is actually a pretty horrible amplifier - useful back in the early 1970's where you wanted cheap gain-bandwidth, but a swine to keep stable.

The LF356 hasn't got quite as much bandwidth, but it is much better behaved, and it can drive a 5nF load. If you are working with a gain of five or more, you could use the decompensated version - the LF357 - which does seem to be about as fast as the LM318.

There are undoubtedly more modern parts that do even better, but the LF356 and LF357 have been around since the mid-1970's and got designed into lots of stuff, so you can often get them out of university and trade school lab stocks.

As for the rest of your circuit, see if you can put it up on a web-site we can look at it, or post the circuit diagram to alt.binaries.schematics.electronic as Jim Thompson suggested.

I see. What's different is the 100 ohm R directly connected and between the capacitive output load. I'll try it.

Hmm, I've been looking for a replacement. I don't need high frequency-- < 20KHz. Although most op-amps that NTE sells are > 4MHz bandwidth. I'm wondering if there's any particular parameters to look for in a datasheet to identify if the chip is prone to unwanted oscillations and such?

I use google groups to post. It seems google doesn't support that one as it goes from alt.biology.* followed by alt.birdlung.* I'll search for a free news group, but haven't seen any that allow alt.binaries.schematics.electronic. Perhaps I can post the image at my yahoo geocities site.

Paul

the general phrase you want is "unity gain stable" or "compensated"

Those cads and bounders at National discontinued the LF157 in '04--it was the last decompensated family member. It's gone to join the TL014 and MRF9331 in Chip Purgatory.

That nice 2 pF input capacitance was a pretty nice feature for high-Z stuff. Ah well--it's the price of getting old: all your friends start dying off.

Cheers,

Phil Hobbs

My rough estimate for the capacitance of 20 feet (6 metres) of audio cable is 600pF.

Since regular op amps start oscillatong at about 100pF, this isn't minute. I first met it as a problem around 1974 (and my slightly unconventioanl solution used to be embedded in the control panels of a couple of British power stations) and I've run into it in a surprsingly large number of legacy circuits since then. The real idiots "cured" the problem by loading the output with a 100nF ceramic disk to ground, which reduces the amplitude to a few millivolts - the electronic equivalent of sweeping the dust under the carpet.

1st, you never drive a cable with an ordinary OpAmp. There are special cable drivers for this. Have a look at the LT1794 or similar.

Rene

It does depend upon the gain, and the OP is running 58dB if his circuit description is to believed. The problem is more due to positive feedback pickup in the input coil/antenna. A cable induced oscillation would rarely be intermittent. The first fix would be to make the overall gain inverting, and then R isolate the output to cable a bit.

Wait a minute-- the first op-amp seems to be set up for a gain of

Same thing with the second amplifier, which is cascaded off the first one.

So all you need is for one part per million of the output of #2 to get back to the coil and you're sunk!

That's an isolation of 120dB which is really quite a lot. You're unlikely to be able to get that much isolation at anything above low audio frequencies without some really serious ground-planes, advanced decoupling, and substantial shielding, and double-shielded coax.

Also you're going to have a heck of a problem with DC offset and drift. Do you really need response down to DC?

And how much gain do you REALLY need, and how much bandwidth?

You might be able to tame this beast if you roll-off the frequency response with a small capacitor or two across each 39K resistor. Also have separate power supply decoupling, with like 100 ohm series resistors, 10uf capacitors, and 0.01uF capacitors for Hf. And a ground-plane would help too.

I've measured some cheap cables with 100 pf per. ft.

Just throw it in the trash...

Very interesting. I'll try it. I know that an R across output helps a little.

Paul

Phil, I found a op-amp with 1.4 pf. NE5532

What do you think of that chip? Probably more bandwidth than I need, but it has high input current, low noise, and very low input impedance.

Paul

I think it's 83, but you're right in that I would eventually like 1000 for each stage. BTW, the device I'm designing is a PI metal detector. Normally I won't need a 20 ft cable, but for now I have to hook it up to a computer. The computer has too much metal and generates noise. So

BTW, was I correct in building a two stage op-amp circuit? Lets say that I need a gain of 200,000? Is it better to have a one stage 200,000 gain op-amp or two stage op-amp circuit each having 447 gain? Or possibly a low 1st stage gain of say 20 and a high gain 2nd stage of

Yes, lol. Would an Opto-Isolators work well enough or would I need to go all out and use a 3 foot optic cables? I'm thinking that 3 feet for final build is max as most metal detectors are about 3 feet long.

Oh you have no idea what a cranky op-amp I have. It's due to my lack of knowledge, but I still keep getting visions of that 10-pound sludge hammer nicely tucked away in the garage. You nailed it. Half of the time I can't DC balance the output. I'll trim the pot say 1/2 turn and then suddenly the amp avalanches and changes output polarity. Then I slowly rotate the pot the other direction and nothing happens until about 1/2 turn and the polarity instantly changes again. That histeresis must be coming from positive feedback, right? Is that caused by the output line capacitance or somehow the 2nd stage output is feeding back to the 1st stage input? I gather that's one advantage of a two-stage op-amp; i.e., less feedback from 2nd stage output to 1st stage input.

Lost of potential gain-- more the merrier. I'd like extremely high gain up to 20 KHz, but can live with only a few KHz.

A ground place near the metal detector coil is a no-no. A few IC's doesn't seem to bother it that much though. I think that somehow I need to move most of the circuit 3 feet away from the ground coil, perhaps by an optic cable. Perhaps a preamp or 1st stage op-amp could feed a

Paul

These are both inverting, right? If the output frequency in very low, try driving the output with resistors on both sides. I guess you want DC coupling. Ground loops problems occur with separated devices. I suppose the coil can also pickup RF interference, but I'm not certain of the design.

I tend to build lots of things with high impedance and not much gain, such as transimpedance amplifiers.

What I'm usually worried about with input capacitance is the nasty noise peak caused by the Cin*Rf rolloff, so the figure of merit I usually look at is v_n*C_in--as long as v_n is low enough to start with. The 5532 is okay--I've never used one, because there are lots better choices in bipolars. I used to use the OP-37 and HA5147 a lot, for instance. Joerg could do it with a couple of MMBT3904s and some nose grease, but we do what we can. ;-)

The 1-uA maximum input bias current pretty well rules the 5532 out for high-Z things, as you point out.

Cheers,

Phil Hobbs

OP37A looks like what I need. Must not be my day as NTE has no cross-reference chip. ;-( ... The search continues.

Paul

Re Gm amps (TIA) - I design these as well and use the 2SK170 JFet. This part has a voltage noise of 1nV in the audio band.7 There is a circuit which I first heard of at Bell Labs called the gm multiplier. In this method, a bipolar NPN is stacked on top of the JFet. 90% of the JFet drain current is shunted away to the positive supply leaving 10% in the bipolar collector current. This permits the NPN collector load resistor to be 10x larger and the resulting voltage gain to be 10x larger as well, up to 100V/V or higher. What I do is to use a small feedback capacitor from the NPN collector to the input, so the gain is set by the ratio of the source capacitance to the feedback capacitor. The amplifier input capacitance may be reduced if the NPN base is bootstrapped from the JFet source, causing the NPN emitter to back-drive the input through the JFet Cdg.

Charles Gilbert