diff amp

[green]

A diff amp I've been using with IB coils for my PI. I potted the circuit with 832B. 62uv input offset. Had some drift, noticed if I gripped the input terminals with a tweezers it drifted opposite directions, one more than the other. The terminal with the most drift had the shortest wire from the terminal to the dual 3904. Need to keep the wires the same or maybe add a 10 ohm resistor from the terminal to the transistor base to keep the base temperatures the same. Connected twisted pair and foamed the input terminals to solve drift problem. First try, probably not optimized. Didn't connect the other half of the dual op amp as the inverted output because I have an inverter already for the mono coil amplifier. Will add the next try. Suggestions appreciated. Looking at the picture I see I cropped the picture a little to much, lost the common line on the right side.
.

[Skippy]

I'm not familiar with the circuit, I guess on one of the PI threads it's used and discussed. But a few points worth considering:
*I would say that any junction of two dissimilar metal will generate a voltage (thermocouple), so you should ensure there are no non-identical junctions, everything should be identical in the input stages, so if one junction warms up, so does its complement.
*Physically, the layout of the board should be symmetrical, no big tracks on one input, thin tracks on the other, long on one, short on the other. The thermal environment should be balanced.
* I'm concerned by the quiescent point your amp operates at. I would choose components so the collector of both transistors was close to zero. I've just estimated they are both at +0.65 Volts in your circuit. I've assumes the pot is not fitted, I don't see a value shown for the pot, it's likely that with the correct resistance pot, the quiescent state would be near-zero. You should've perhaps replaced the pot with two fixed resistors...( I might be misunderstanding how it's intended to work, and I'm not allowing for the one-directional input signal)

[green]

I'm not familiar with the circuit, I guess on one of the PI threads it's used and discussed. But a few points worth considering:
*I would say that any junction of two dissimilar metal will generate a voltage (thermocouple), so you should ensure there are no non-identical junctions, everything should be identical in the input stages, so if one junction warms up, so does its complement.
*Physically, the layout of the board should be symmetrical, no big tracks on one input, thin tracks on the other, long on one, short on the other. The thermal environment should be balanced.
* I'm concerned by the quiescent point your amp operates at. I would choose components so the collector of both transistors was close to zero. I've just estimated they are both at +0.65 Volts in your circuit. I've assumes the pot is not fitted, I don't see a value shown for the pot, it's likely that with the correct resistance pot, the quiescent state would be near-zero. You should've perhaps replaced the pot with two fixed resistors...( I might be misunderstanding how it's intended to work, and I'm not allowing for the one-directional input signal)

Thanks for the reply. I was thinking a 100 ohm zero pot but was surprised at the low offset. Think a 10 ohm pot would work, maybe not for the next one. I like your idea on the quiescent point. Changing the collector resistors to 5k or the emitter resistor to 1600 might be a good idea.

[Davor]

You could fix drift permanently by a servo circuit.

[Skippy]

A few other thoughts:
You say you didn't use the second opamp - what did you do with it? I hope you didn't leave it floating. It could be oscillating at Megahertz frequencies, and messing up the operation of the other amp.
It's possible your amp is unstable, I've seen similar circuits (eg. audio power amps) and a resistor in series with a capacitor were connected between the two collectors, if I remember correctly. It was a stability aid, no I can't remember any details.

Re. the biasing: I calculated collector load resistors of 5400 Ohms were about right, to obtain 0V on the collectors, if you didn't use a pot.

[green]

A few other thoughts:
You say you didn't use the second opamp - what did you do with it? I hope you didn't leave it floating. It could be oscillating at Megahertz frequencies, and messing up the operation of the other amp.
It's possible your amp is unstable, I've seen similar circuits (eg. audio power amps) and a resistor in series with a capacitor were connected between the two collectors, if I remember correctly. It was a stability aid, no I can't remember any details.

Re. the biasing: I calculated collector load resistors of 5400 Ohms were about right, to obtain 0V on the collectors, if you didn't use a pot.

The second amplifier, I grounded the + input and shorted the output to the - input

[Davor]

So you can use it to build a servo. Good.

[green]

So you can use it to build a servo. Good.

Hi Davor
I've used servo's to control force, speed, displacement, etc. What are you doing with the servo in a metal detector?

[Skippy]

By feeding back some signal from the output to the input stage, it's possible to make the amp self-balancing, no need for pots. Usually it would involve long time-constant filters, so that only the DC component passes through. So the feedback causes a servo-action controlling the output to zero, but the AC signals pass through without being affected, or without ruining the servo-control loop.

[Skippy]

Here's an example of a servo-loop on a discrete buffer amp:
http://www.edn.com/design/analog/440...ipolar-signals
Figure 3 is the one to study - see the low-pass filter stage, and the feedback connection.

[Davor]

Yes, something down that line. The idea is to subtract DC and below 1Hz from a diff amp in order to eliminate drift. Alternatively you may put a big capacitor in series with output, but this works better, and this is DC coupled throughout.
Servo is basically an integrator in a DC stabilisation loop, and it is very good at that because it "remembers" the offset and compensates for it.

[green]

Skippy, Davor thanks for the replies.