Thermal drift compensation

[throbscottle]

Ok next question...

I'm down to a few opamp choices but the killer is thermal drift - the amps that are ok in other respects have drift that is just too high, even though it is in the order of a few uV/C. So I have 2 choices. Put a heater/cooler in the box to control the temperature to within a couple of degrees C, or find a way adjust the offset compensation according to temperature. Or possibly both.

So, what does a person do this circumstance?

Thanks in advance

 

[crutschow]

The best you can do without compensation is to use a chopper stabilized op amp such as one of these. They have drift on the order of 0.01uV/C. If you must use a non-chopper amp, then I think putting it in an oven box is the best solution.

 

[NorthGuy]

Have you considered LMP2021/LMP2022?

 

[alec_t]

Haven't tried this, but could you tap off a tiny fraction of the voltage Vd across a diode and add/subtract that as an input offset? Vd is, of course, temperature dependent. The diode would need to be thermally coupled to the opamp to track its temperature.

 

[throbscottle]

Thanks for the replies. I may well have to go the chopper route, though on previous looks through the parametric tables I've ended up excluding them because of other factors, however this has been limited by my understanding of various things so it's worth another look. Building a little oven might well be the best solution since I think I'm going to need one for my voltage reference anyway. With an overall amp input drift of 0.05uV/C I would get 10C room within my ideal error budget of 500nV, 20C max.

Take the LMP2014 - I got excited when I saw that until I saw "differential input resistance: 9M" I later realised this isn't so bad since the input is entirely common mode, but what is the common mode input resistance? The datasheet is strangely quiet about that. Also input offset current is 6pA compared to bias current of -3pa. As far as I understand, offset current is the difference between bias currents, so this just doesn't look right.

Re LMP2021, Ib is too high! I need better than ~10pA. I've looked at a few of TI's chopper and auto-zero amps, they've been a case of nearly, but not quite. I need to spend more time looking at Linear's amps though I think, however a lot of those have Ib compensation so the current although tiny is +/-, which is no good for me unless it's really tiny. Same goes for Intersil and Microchip amps, both offer amps ideal in every other respect. Bit of a bugger, really. (Eg, ISL28107, Ib is 15pA typical, but the min value is -300 and the max value is +300. Too much chance involved)

The diode idea is where I had started going but I hit a kind of mental brick wall there. If I can find one with a suitable TC I wonder if it would be adequate to just put it in series or parallel with the offset null pot? There's already a constant current planned for that so it's part-way there. Doubt if good enough by itself but might complement the oven idea so I can use an amp with more average sort of temp drift?

Have to get ready for work now. Sunday afternoon too. Boooooooooooo.....

 

[NorthGuy]

10pA is only 0.01uV accross a 1k resistor (compared to 0.o3uV of rms noise that the 1k resistor will inherently produce at 50Hz sampling frequency). Why do you need such accuracy? I'm sorry, I'm new here, so I don't know, what do you measure?

 

[crutschow]

You are at the level where thermocouple voltages, due to junctions between any two different metals (such as copper and solder which is about 3µV/°C), can be much greater than the voltages you mentioned. Have you taken that into account?

 

[dr pepper]

Not an op amp answer, but you could consider using a crystal oven.

If your measureing really small voltages have you thought of a lock in amp circuit.

 

[crutschow]

.........................

If your measureing really small voltages have you thought of a lock in amp circuit.

I believe lock-in-amps are for small AC signals (they lock in or synchronize to the AC frequency), not DC.

 

[dr pepper]

I assumed they were for dc as my milliohmeter uses the technique, but maye it generates its own ac signal to measure with.

 

[crutschow]

I assumed they were for dc as my milliohmeter uses the technique, but maye it generates its own ac signal to measure with.

I think that's a variation of a chopper amp.

 

[Mike odom]

I use the opa241 for low drift/low supply current/low offset voltage.

 

[NorthGuy]

Deleted by poster. Inappropriate.

 

[throbscottle]

Well, the stupid thing is I don't need this level of accuracy, I just happen to have a 24 bit ADC and I want to use it! A 6 digit dmm is easily within its capability, and it just so happens that with a low scale of 999.999mV FSD, the LSD resolves 1uV.

However it turns out that the lowest 10uV are something of a war zone between noise, offsets, leakage, thermal drift, thermocouple voltages, emi and god knows what else, as you all know, and these are things I've learnt about as I've gone along over the last 18 months trying to work out the details of the design. I'm thoroughly enjoying trying to sort out the conflict here and it would be really great if it works.

I know I need to balance up thermocouple voltages and keep the temperature as even as possible. Again it's something I'm getting to grips with.

I did occur to me that with a dual op-amp package, the drift of amp 2 should be similar enough to the drift of amp 1 that it can be used to correct amp 1. But all considered I guess my best bet is a chopper amp after all.

 

[MrAl]

Hi,

You are right about the drift of one compensating the other when they are two amps on the same chip. I'd have to look for the way to do this though as it has been a long time since i've had to look at this kind of thing. Another way rather than compensate directly might be to use two or three to build a precision differential op amp.

But about this 24 bit ADC. Could you specify the part number so we could take a quick look? 24 bit means over 16 million combinations not just 1 million, unless you plan to average out some of the least bits. What i'd like to see is what are the full specs of this chip. It could just be that you are going too far if the spec's are not that good to begin with. But it would not hurt to look at this. Course if you have a data sheet even better.

 

[NorthGuy]

Well, the stupid thing is I don't need this level of accuracy, I just happen to have a 24 bit ADC and I want to use it! A 6 digit dmm is easily within its capability, and it just so happens that with a low scale of 999.999mV FSD, the LSD resolves 1uV.

The number of digits doesn't matter much. To begin with, the ADC measures voltage against a known voltage reference. It cannot be more accurate than the voltage reference it uses. This is probably one of the biggest factors limiting accuracy. To resolve 1uV on 1V scale you probably need voltage reference accurate to 0.1-0.5 ppm. How accurate your voltage reference is?

 

[crutschow]

...................................

I did occur to me that with a dual op-amp package, the drift of amp 2 should be similar enough to the drift of amp 1 that it can be used to correct amp 1. But all considered I guess my best bet is a chopper amp after all.

I'm not sure that's true. The offset drift in precision op amps can be due to slight differences between the two input transistors and also temperature gradients across the chip, and I'm don't think that necessarily tracks from one amp to another, even on the same chip. I agree that a chopper is likely you best bet.

 

[throbscottle]

Ok, probably some of you have looked at this datasheet before, when I first started the project, but here it is again. You can see from the sheet it's capable of resolving 100nV, however with a 5v reference the step size is 298nV, still tiny.

Have to walk dogs and go to bed - only quarter past 9, too. Sucks. Don't have time to think.

 

[NorthGuy]

Ok, probably some of you have looked at this datasheet before, when I first started the project, but here it is again. You can see from the sheet it's capable of resolving 100nV, however with a 5v reference the step size is 298nV, still tiny.

This thing measures voltage relative to the voltage applied between Vref+ and Vref- called Vref. If you get a count right in the middle this means that your voltage is 0.5*Vref (with very high precision)

If you now connect Vref+ to Vcc and Vref- to 0V, the middle count will be 0.5*Vcc (with very high precision).

Typical supply produces Vcc with 3% error and with a ripple, which means that it'll be wandering somehow between 4.85V and 5.15V. So, your measurement may mean 0.5*4.85V or 0.5*5.15V or anything in between, depending on where Vcc is at the moment. In such situation, is it worth distingushing between "0.50000001*(unknown value anywhere between 4.85V and 5.15V)" and "0.50000002*(unknown value anywhere between 4.85V and 5.15V)"?

Makes sense?

 

[throbscottle]

Thanks for the input NorthGuy, but I have worked all this out previously. I originally looked at a +/- 1mV (initial accuracy) reference from Intersil (ISL21009BFB850 which I previously used to build a calibrator for my existing meter) but I eventually realised I don't have to worry about initial accuracy, just noise and drift, so I'll be looking around again in that department to see if it's actually the best choice. The actual reference voltage isn't important so long as it's stable and the FS input can be scaled to match, then the reference can be trimmed to allow for gain error. Then it's just a case of scaling the counts to the display using a micro.

As to how accurate the voltage reference is, well, I know what characteristics to look for now, and it can go in a controlled environment too.

I had been planning to throw away the last couple of bits, but averaging sounds a better idea, I'll look into that.

I had originally worked everything out for a Vref of 5v and a scale based on a MSD of 4, but I'm re-visiting that since the reason for using a scale of 4 (the ohms converter, which I've completely re-designed, now fully differential) is no longer a limitation.

Oh, seems to be bed time again. Boooooo...