# Voltage reference voltage drop

I have two potentiometers and a single voltage reference, connecting them causes a few mV voltage drop in the simulator:

Adding an op amp solves the problem but then comes input offset voltage and temperature stability problems:

simulate this circuit – Schematic created using CircuitLab

How do I properly use a single voltage reference for two potentiometers without getting voltage drop?

• Why not use fixed R’s with a bandgap Vref , and quad OA because Zeners are pretty wide tolerance >5% and temperature sensitive. Commented May 11, 2020 at 19:58
• @TonyStewartSunnyskyguyEE75 Can you please explain a little bit more? you mean I do that in the simulator to get a "voltage reference like" behavior? because on the PCB I don't have enough room to use a quad op amp... Commented May 11, 2020 at 20:02
• The ratio of Load reg Err= source / load+source = Load regulation error. The Zener source has a lower incremental resistance dv/DI but not low enough. Whereas the OA reduces its Zout by AOL Commented May 11, 2020 at 20:05
• @TonyStewartSunnyskyguyEE75 I'm going to use LM4040 Vref, I'm not sure though is it going to behave like a Zener or not. what does "AOL" stands for? Commented May 11, 2020 at 20:11
• Are you really going to have a switch in your circuit or will the pair of potentiometers always be connected? Which grade (A/B/C/D) of LM4040 were you planning on using? Commented May 11, 2020 at 20:32

The accuracy of the LM4040 is guaranteed with a 100uA current and typically changes only 300uV for a current change from 75uA to 1mA. The guarantee is +/2mV at room temperature and +/-15mV over a wider span.

Self-heating will affect it too.

So if you use even a modest precision op-amp you won't affect the initial accuracy or drift much. The antediluvian OP-07 has guaranteed Vos of as low as 75uV (which is << 2~15mV) and a guaranteed drift of less than 1.3uV/°C compared to the +/-31uV typical drift of the best LM4040 at the optimum current.

There are much better op-amps available than the 1975-era OP-07 in 2020.

So, yes, buffering it if you expect large changes in load will help, and it will reduce self-heating drift and changes with source voltage even if the load is fairly fixed. You can also low-pass filter out some of the LM4040 noise, which is probably around 200-250uVp-p.

That will also allow you to operate it at an optimum current, with a large series resistor (perhaps from a relatively high voltage supply as in your example) so the current is closer to constant.

You should really add up all the sources of error and compare it to your target, to see what is necessary or desirable.