Op-amp problem

[KeepItSimpleStupid]

What size tantalum cap did you use for the bypass? They really should be ceramics.

You also have a flux on the PCB. You have to get that cleaned off.

 

[hereward]

KISS, thanks for the reply, Yes, I have found, when soldering large joints, unless the flux residue is removed while the joint is hot it will go rock hard. I sprayed and cleaned each joint with a toothbrush after soldering; not very successfully I admit. Those blue caps are Monores resin dipped ceramics 47pF and 0.1µF.

 

[simonbramble]

Can you get an oscilloscope on the output? It might be picking up small amounts of mains hum and this is manifesting itself as small deviations on your voltmeter. Try grounding the input (at the chip end, not at the end of your long test leads) and see if the random fluctuations disappear.

With this small amount of gain, I don't think the Siebeck effect will come into play. The Siebeck effect causes microvolts of error (41uV per degC for a K type thermocouple from memory).

 

[hereward]

Hi Simon,
I haven’t got an oscilloscope. I have just been replacing the mains power supplies with two 9V batteries and also the input from a voltage divider with a 9V and 1.5V battery, and having no mains within 10 yards. The meter reading was around 100mV initially rising to about 290 mV with inputs. Seems totally irrational considering there is no problem when the circuit is breadboarded. Surely rf would affect the breadboard, I had put the circuit in a metal box at one point. Over the past 5 months I have been tinkering with just about everything, putting a jump lead from input to ground would just send the input to earth if I recall right, and if that’s what you mean. These readings seem too high and too stable to be thermocouples, but then when continually blowing on the op-amp caused 19 mV reading to double. Thanks - I'm baffled.

 

[KeepItSimpleStupid]

What's the offset with the input shorted?

There is always the possibility of a bad OP amp?

Beffore you totally throw in the towell, bend out pins 2 & 3 from the IC socket and solder directly to those pins.

 

[hereward]

Simon / KISS, With two 9V batteries as power supplies and 5V mains via a voltage divider as input the initial reading was 50 point something mV (no input) this fell very slowly. With an input the initial readings increased by 99.3 mV. With or without an input, shorting the pin side of the input resistor to earth gave a meter reading of -9.2V!!
Not wishing to vouch for the accuracy of the ammeter at such small readings I have seen the Seebeck effect register at 0.3 Amps or 0.3 mV (since both are proportional), that theoretically could be amplified to 30mV. These readings seem too high and too stable to be thermocouples. That 19mV reading is +, made me think there could be a leak from ground - I put the input and output on a separate ground from the power supply (just groping in the dark) still got an offset with no amplification. Basically I have already put the wires directly to the pins with no joy. I had suspected the op-amps from the beginning but plugging them into an existing amplifier they worked fine.

 

[KeepItSimpleStupid]

I designed a I-V converter using one of those 8 lead round packages and I isolated the non-inverting input with a teflon post. I've worked with sensitive electronics before and I could detect films of acetone and photoresist stripper residue left on a piece of glass. Yea, the currents were a few pA at 100 V, but they were measureable. The flux looks really suspicious.

That's why I'm suggesting to dog-ear pins 2 & 3 and solder direct to the pins. It gets the flux out of the equation.

 

[hereward]

KISS, soldered directly to dog eared pins 2&3; the results were similar to the previous (initial reading about 100 mV and falling with around 99.3 increase with input). Anyway thanks for the suggestion.

 

[KeepItSimpleStupid]

At this point, I have no idea what's wrong. I think you even went tha battery route?

But, all of us tried.

 

[hereward]

All of this started at the beginning of the year, I had asked on the forum if I was likely to encounter any problems with a dual op-amp to give a gain of 10 and 100. I was informed that this should be OK; but neither gain would work. Concluding that the op-amp was faulty it was decided to make a single amplifier with a gain of 100 since I had made several in the past without any problems with the actual circuit. This didn’t work either; replaced the resistors that were old stock for new ones, no joy. Breadboarded a circuit it worked perfectly. Tried a ten gain since this did not require the same nursing that a 100 gain had, that didn’t work on a soldered circuit. Someone who didn’t know any better suggested the solder must be at fault. Swapping Sn/Pb for Sn/Cu didn’t help, neither did swapping the electric iron for a gas one.
The suggestion that the breadboards capacitance was working to benefit the circuit was good but adding capacitors either made no difference or made things worse. Plugging the op-amps into an existing working circuit found them to be OK. Putting the circuit into a tin box didn’t help; neither did separating the input and output grounds from the supply grounds. Likewise using three different power supplies and batteries. I did at one point get the circuit working but the next day applying it to a shunt sent us back to square one. The best results, an initial reading of 1.2 mV rising and falling was obtained by soldering the resistors and wires directly to the IC.
It’s not the end of the world but the frustration is killing me. There is a reason for everything, including why this would work in the past but not now. If anyone makes this circuit successfully in the future please let me know.

 

[unclejed613]

looking at the data sheet, and at your circuit, i see that pin 4 (-Vcc) of the op amp is left floating (it's not even soldered to the pcb trace). without a negative rail, the op amp isn't going to work. your output will be a conglomeration of internal leakage currents.

 

[hereward]

Unclejed - It's the way the picture is orientated, if the board is turned over so the bottom edge becomes the top edge you will see pins 2, 3 and 4 are soldered.

 

[unclejed613]

https://www.electro-tech-online.com/custompdfs/2013/09/1050fb.pdf

pay particular attention to page 6, as it describes some techniques for minimizing offsets and drift that you may not be aware of. also, you mentioned you were using battery supplies? what is the actual rail to rail voltage? there are charts that show several things that can have an effect on offset and drift, and their relation to supply voltage. other factors that may be in play here are humidity (your sidebar doesn't tell us where you are from). extremes of humidity can cause unforeseen errors, such as electrochemically induced errors in high humidity, to electrostatic induction in extremely dry conditions.... also, the data sheet mentions thermally induced errors from the package leads and the use of various types of resistors. eliminate one possible error source at a time, instead of using a shotgun approach, that way you will figure out which error source you eliminated. when you sprayed solvent on the connections with the circuit live, you probably changed several things at once, temperature of the pins, electrical conductivity and dielectric properties of the pcb surface, dielectric properties between the IC pins, etc... of course it went haywire.... also, check out the part of the data sheet about guarding vs shielding of the inputs.

 

[KeepItSimpleStupid]

I still don't see a wire to pin 4. Just 2 caps and pin #4.

 

[hereward]

KISS - The wire to pin 4 is green/red (-5V or -9V) it is under the red input wire. Pin 1 and 8 are at the bottom of the board and unsoldered, pins 4 and 5 are closest to the grounds (1 to 8 anticlockwise around the IC, looking from the top). The amplification is working OK, if not bang on 100; the problem is the initial contamination of the reading that must be coming from the power supply.
Unclejed - That original amplifier was working until the track broke. Using the same board and the same components just soldered in a different place on the board caused the circuit to fail, to me this is just bizarre. Some of those precautions are for pV intrusions; unless they find their way into the 1/10 mV column they will not concern me. I will have a good read at the sheet, I suppose the problem must be in there somewhere. A max of +/- 75mv if that's what you mean by rail to rail voltage.

 

[hereward]

Hi Folks,
Bear with me a second – when I was at work they had a 100 Amp. PSU with a brass/manganin ammeter shunt (200 Amps. 200mV). When the power was switched off the meter would read up to 0.3 A. this would fall to zero when the shunt cooled. The more positive side of the shunt would be the hottest; if the heat sink was cooled with compressed air the meter would fall rapidly to zero, and if continued the reading would change polarity and go the other way as the other side became the hottest. This movement of energy/electrons could not form part of the main current because this would violate Kirchhoff’s Law; all it did was contaminate the meter reading. This was without doubt the Seebeck effect. This makes it hard for me to grasp why it affects the op-amp circuit, although I know it does because I have experienced the phenomena. Unclejed suggested reading the data sheet; the only thing there, that may help, was the insertion of unnecessary joints, strange since originally I had to reduce the joints to a minimum. Having made six joints (three pair) thinking they could swap electrons; this didn’t make a scrap of difference. I came across a board in my junk box with copper strips made up of three hole segments. If an IC socket is placed on it, only one hole is left and if this is used for the supply wires it might isolate the supply from the rest of the board, suspecting that these high initial readings are from the power supply, how I haven’t a clue. At the input and output this one hole is not enough to accommodate wires and resistors, so a link was put in to next segment, this made 7 joints at the input, counting adjacent holes soldered together as two joints.
The initial reading was then 1.2 mV, this is where I came in. Adding another joint to the left of the strip to #2 pin caused the reading to go to 350mV. So the joint didn’t draw the electrons from the op-amp so I removed it.
THE INITIAL READING WAS THEN 0.1 mV; I can live with that!! I have been switching it on and off periodically this initial reading seems to be sticking and amplifying to 99.5 mV. So I think the board and the joints may have done the trick. I am not yet opening the champagne or taking any credit because without you guys I would have still been thinking:” Well it worked before.”
See what tomorrow brings before handing out the electronic forum equivalent of an Oscar. - Thanks