# Floating above ground…

It is not at all uncommon to get a call from someone claiming that the parts shipped to them are out of tolerance on one or more parameters. This caller had just such a claim. The offset voltage of his OPA336 was several millivolts when it should be less than 0.125mV. As is the norm, I requested a schematic of the circuit he was using. That circuit is drawn here. My caller was just attempting to measure the input-referred voltage offset on the bench. By his calculation the output voltage from this circuit divided by the non-inverting gain

should be this elusive value. However, the situation was confused on two fronts. First, the voltage was far greater than the value given in the specification table. Second, the value did not change with changes in the gain resistors.

He had measured the input offset of op amps for other projects with no problem. This technique had worked wonderfully. What was wrong?

I carefully reminded my caller that this would require the output of a perfect zero offset op amp to swing all the way to ground. He had been looking at the output swing limitation of the amplifier.

The cure for this problem is to generate a voltage something above ground that will be the reference or zero for the circuit. In the circuit shown here a voltage as low as 0.1V or even 50mV would work. You can assign a descriptive name to this voltage such as signal ground or analog ground. This is the circuit signal-zero-voltage.

There are several ways to derive this voltage ranging from a basic resistive voltage divider to a stable voltage reference. Since this is the zero point for the signal it can be critical to the system accuracy. A first approach might be to drop in a high accuracy reference, however, depending on the system topology this may be a waste of resources. The importance of the absolute accuracy of the analog zero depends on the reference point of input signal and how it is referenced at the next stage.

If the input signal is referenced to ground then any variations in the analog zero will appear as a signal, however, if the input signal can be referred to this analog zero then a simple voltage divider may be acceptable. The second half of the decision depends on how the output is used. It may be possible to use the reference voltage from the A/D converter as the analog zero. In this case any variations in the reference are also changes in the analog zero and the errors can be self-canceling.

An alternate approach to the analog zero stability issue would be to select an A/D converter with a differential input stage. In this case the analog zero would be connected to the negative input of the A/D converter, which would reference the signal to a less than perfect analog ground.