Nulling a long-tailed pair

[Spehro Pefhany]

Okay, I have a long tailed pair (actually a Gilbert cell differential
output), presumably diffused resistors to Vdd, and there's a DC offset
and tempco of that offset. Current is about 1mA on each side.

I tried nulling the differential output voltage by adding a high
resistance to the lower side to Vdd, and the tempco seems about the
same magnitude, but the sign flips.

Is this to be expected? Lucky chance? Experimental error?

 

[Spehro Pefhany]

You could measure the collector currents currents of the two outputs
(with low-z milliammeters or low value, low TC resistors) and see what
the currents are doing. That would help separate some variables.

The resistors and collectors are connected on-chip so I can't get to
them individually. I suppose I could make a small signal measurement
of the resistors by assuming the collectors are high-Z and putting a
small AC signal on the collectors but it might be hard to get that
number accurately.

Diffused resistors have rotten TCs, so maybe your external resistor
messed up their matching.

The drift appears to be roughly PTAT relative to the offset.

 

[whit3rd]

Okay, I have a long tailed pair (actually a Gilbert cell differential
output), presumably diffused resistors to Vdd, and there's a DC offset
and tempco of that offset. Current is about 1mA on each side.

I tried nulling the differential output voltage by adding a high
resistance to the lower side to Vdd, and the tempco seems about the
same magnitude, but the sign flips.

The offset voltage of a differential op amp can be trimmed by imbalancing the
currents in the two input transistors. The temperature coefficient of
an op amp can be trimmed by balancing the collector currents
(Hillbiber, an old Fairchild engineer, had a scheme for doing this
on a uA741 through the trim pins), but the offset voltage suffers.

Somehow, the uA725 input circuit was designed so that zero tempco
occurred at zero offset voltage. It was the go-to op amp for many
years, for microvolt (like, thermocouple) applications.

 

[Spehro Pefhany]

The offset voltage of a differential op amp can be trimmed by imbalancing the
currents in the two input transistors. The temperature coefficient of
an op amp can be trimmed by balancing the collector currents
(Hillbiber, an old Fairchild engineer, had a scheme for doing this
on a uA741 through the trim pins), but the offset voltage suffers.

Somehow, the uA725 input circuit was designed so that zero tempco
occurred at zero offset voltage. It was the go-to op amp for many
years, for microvolt (like, thermocouple) applications.

Yes, I remember that pretty well. In this case, I'd be willing to add
a couple of trims if I can reduce the TCVos by an order of magnitude,
provided it doesn't require days of measurement in the thermal
chamber.