Nulling the opamp's input voltage offset and its TC

[iMo]

When discussing the "standard" 399's 7V->10V design with a bipolar OP07 (not a chopper/AZ) we usually neglecting the need for a Vos nulling (here we talk always the nulling with the trimmer between the pins 1 and 8 ).

It has been indicated in some posts couple of times the TC of the Vos "will be zero when the Vos is nulled".

I've recently had a look into below paper from TI, which indicates the TC of the Vos actually does not depend on the level of "nulling". They even suggest an NTC thermistor in the external trim circuit.

Now, I want to trim off the Vos and its TC of my OP07 in my "standard 399" circuit.
My Qs:

1. the TI's range for the nulling Vos (OP07) with the trimmer is 4mV (DS) - it sounds to me even with a 10T trimmer (no padding resistors in series with the trimmer) to trim off say 50uV would be an uneasy exercise

2. after the nulling the TC of the Vos, like for example +1uV/C, will stay the same - is that correct??

3. how to make the nulling in the real "standard 399" circuit - I mean how to measure the input voltage offset has been nulled?

 

[Kleinstein]

With BJT based OP-amps there is usually still the correlation between TC and offset.  It is not perfect, but can be relative close.

The TI article included shows examples with FET based OP-amps. For these there is less to no correlation between offset trim and TC. It depends on the type.
Anyway, many moder OP-amp types don't have the user option to trim the offset anymore, but only the factory trim or auto zero.

For a reference an offset from the buffer would not be that bad, as long as it is reasonable constant. Consider the offset as part of the reference voltage.

The trim range may be on the large side, to make sure that one can really reach zero. One is usually free to use a high value trim pot and this may get less trim range.

[iMo]

And how to trim off the offset voltage in the actual 7V->10V circuit?
How to measure the offset has been trimmed off?
PS: Imagine I got the 399's 7->10V with an OP07, I added the 10T trimmer, say 50k or 100k wired between pin 1 and 8 (wiper to +Vcc) and I want to null the OP07's input offset voltage..

[Kleinstein]

There is usually no need to trim the offset in the 7 V or 10 V references. Many modern amplifiers like OPA207 or OPA205 or ADA4077 don't even have the offset trim option.
Also the old OP07 is good enough without a trim.  Some 0.3 to 1 µV/K is not that relevant relative to a 7 V ref. voltage at the input.

[iMo]

I know, I know, I know..  ... but again - how to do it in the above/below circuit?

For example to short the opamp's input pins together and trim the Vos voltage down till the voltage between input and output of the opamp will be 0V??
Or?

[EC8010]

Just because a 10T resistor is convenient does not mean it is accurate or stable. Quite apart from the backlash in the worm driving the gear, there's flex and slip in the wiper onto the very short track. Trimmers are fine for removing small errors, but I think for something like this I'd want to use fixed resistors to do most of the trim, then a variable. You also need a means of disconnecting and grounding the op-amp's input so that it sees 0V from a low source resistance to do the trim. Definitely not short the op-amp's input pins together as it would have infinite gain and saturate at one or other rail. All without damaging the reference...

[iMo]

Ok, let us assume the trimmer used is an "ideal" 10T one.

So I have to wire the none-inv input to ground, I have to use a negative Vee then (say -5V) for the opamp, I would assume, and then trim the output down to the 0.0V. Right?
When I revert back to the single supply none grounded input - will the offset change?

[EC8010]

Ah, I see what you mean. I was assuming symmetrical supplies, which you don't have. Sorry. However... If you short the feedback resistor you will change the gain from about 1.4 to unity and any internal offset will now be in series with the input voltage. So if you put a meter between non-inv and output, you should see the offset and be able to tweak it to zero.

[iMo]

Ok, so:
1. I have to disconnect the Rz (bootstrapping the zener) from opamp's output and wire a 10V source to it instead
2. short the FB resistor (I think this is not needed)
3. measure the voltage between none-inv input and output, trim it down to zero.

[Kleinstein]

For zeroing (if really needed) one can measure the voltage at the input side to the OP-amp, between the inverting and non inverting inputs.  To avoid possible oscillation from capacitive loading, some 10 K in series to the meter at the inverting input can be a good idea. EMI could still be an issue.

[jfet]

You may want to read the datasheet for the LT1001.  On page 6 and 7, they have some info and effects about offset nulling.

[PCB.Wiz]

2. short the FB resistor (I think this is not needed)

If you short the FB resistor, you have thrown away the Opamp gain, which can help you here.
If you short the inputs instead, as you suggested in #4, the opamp gain acts as a comparator and the output swing is easier to measure.
One item to watch would be bias current effects.

[EC8010]

Ok, so:
1. I have to disconnect the Rz (bootstrapping the zener) from opamp's output and wire a 10V source to it instead
2. short the FB resistor (I think this is not needed)
3. measure the voltage between none-inv input and output, trim it down to zero.

Yes, I think that should do it. It might not even be necessary to do (1) because you're only looking for the op-amp offset, but it might have a dependency on input voltage, so you're probably right to do that.

[David Hess]

It has been indicated in some posts couple of times the TC of the Vos "will be zero when the Vos is nulled".

I've recently had a look into below paper from TI, which indicates the TC of the Vos actually does not depend on the level of "nulling". They even suggest an NTC thermistor in the external trim circuit.

Properly designed bipolar input parts have the property of nulling their temperature coefficient of input offset voltage when their input offset voltage is nulled.  This includes the OP-07 (really the OP-05 and maybe uA725) and its decedents.  For these parts, the temperature coefficient of input offset voltage tracks the input offset voltage so well that by adjusting for a specific input offset voltage, the temperature coefficient of input offset voltage can be used as a temperature sensor.

1. the TI's range for the nulling Vos (OP07) with the trimmer is 4mV (DS) - it sounds to me even with a 10T trimmer (no padding resistors in series with the trimmer) to trim off say 50uV would be an uneasy exercise

Fixed precision resistors should be used to restrict the trim range to a minimum, or the trim can be implemented with fixed precision resistors.

2. after the nulling the TC of the Vos, like for example +1uV/C, will stay the same - is that correct??

Yes.

3. how to make the nulling in the real "standard 399" circuit - I mean how to measure the input voltage offset has been nulled?

In theory the voltage between the inverting and non-inverting inputs could be measured directly, and I have sometimes done it this way, but in practice the high impedance at these points and the low signal level make this problematical.

The usual method is to short the non-inverting input to signal ground (the bottom of R1 as shown), and rely on the operational amplifier to provide an amplified low impedance output representing its own input offset voltage.  The value of R1 might be temporarily decreased with a shunt resistance, or the value of R2 increased, to increase the closed loop gain for an easier measurement.

[CurtisSeizert]

Properly designed bipolar input parts have the property of nulling their temperature coefficient of input offset voltage when their input offset voltage is nulled.  This includes the OP-07 (really the OP-05 and maybe uA725) and its decedents.  For these parts, the temperature coefficient of input offset voltage tracks the input offset voltage so well that by adjusting for a specific input offset voltage, the temperature coefficient of input offset voltage can be used as a temperature sensor.

Does this imply that the input transistors are generally very well matched and much of the Vos comes from mismatch of the collector loads? Either way, this does make sense from the standpoint that trimmed parts with very low TC (like the OPA205) are pretty cheap, and I would think it would be very laborious to separately trim VosTC if it were not related, which would presumably drive up the cost of the parts.

[Kleinstein]

Getting low TC when the offset is small even applies to a differential amplifier with not so well matched transistors.
The difference in VBE for 2 transistos with both operated with a constant current is to a first approximation propoertional to the absolute temperature. In this approxiation one would get a drift proportional to the offset. It is only 2nd order effects (e.g. temperature dependenet stress) to still have a small TC when the offset is zero.

[David Hess]

Properly designed bipolar input parts have the property of nulling their temperature coefficient of input offset voltage when their input offset voltage is nulled.  This includes the OP-07 (really the OP-05 and maybe uA725) and its decedents.  For these parts, the temperature coefficient of input offset voltage tracks the input offset voltage so well that by adjusting for a specific input offset voltage, the temperature coefficient of input offset voltage can be used as a temperature sensor.

Does this imply that the input transistors are generally very well matched and much of the Vos comes from mismatch of the collector loads? Either way, this does make sense from the standpoint that trimmed parts with very low TC (like the OPA205) are pretty cheap, and I would think it would be very laborious to separately trim VosTC if it were not related, which would presumably drive up the cost of the parts.

Trimmed parts are relying on the same relationship between Vos and VosTC.

Some of the Vos comes from mismatched collector loads, and some comes from differences between the transistors.  I am not sure what proportion is from which.

I remember reading an application note or article which went into the details, but it was a long time ago.

... It is only 2nd order effects (e.g. temperature dependenet stress) to still have a small TC when the offset is zero.

Strain has a major effect.  Precision analog parts, and analog parts in general, use extra encapsulation steps to mechanically isolate the die from the package.

[dietert1]

Maybe not a short over the OpAmp inputs but a small resistor, so the amplifier stays in regulation. Maybe 5 or 10 Ohm. If one uses a jumper to connect and disconnect that resistor, one can see the input offset balance at the OpAmp output. The "10 V" should be (almost) the same with and without the test resistor.

Regards, Dieter

PS: With the jumper in place, you amplify TC of the offset voltage as well and easily measure it. You will be able to decide whether you want low offset voltage or low TC and how well these two adjustments match.

[iMo]

Provided the "opamp's Vos TC depends on the Vos level" we may compensate the TC of the reference with it..

[David Hess]

Provided the "opamp's Vos TC depends on the Vos level" we may compensate the TC of the reference with it..

Yes, that could work, but you would need to know the TC of the reference somehow, and precision references already have at least first order compensation of their TC so second order effects will not be linear.