Is there a way to buffer a large DC signal accurately?
I would like to measure a voltage that can go from 0 to 50VDC, with 16 bit accuracy (so down to the mV level at the highest voltage, and µV levels at the lowest). Since this is a 10ppm measurement, I need to be very careful about not affecting the voltage that I am trying to measure.
The typical way to do this (I believe) is to use a unity gain buffer at the voltage, with something like a FET op amp.
The problem is that there aren't that many op amps that can have 50VDC inputs. Of those that can, there are fewer that can have 50VDC outputs. And finally, of those, there really aren't that many that would be very accurate at low voltage measurements. Typical offset voltages in the mV are common I believe. Finally, I'm not aware of any that can measure down to the lower rail.
Anyone have any idea how to buffer something like this? Noise is less of a concern due to the DC nature of the voltage that is being measured, but accuracy is important.
Best Answer
if you have all day (relatively speaking) to make your 'DC' measurement you could servo the supplies of an accurate low voltage (say zero-drift/chopper) amplifier to be close to the measured voltage using a cruder amplifier. Using something like a hybrid high voltage op-amp hundreds of volts would be possible.
Edit: The concept would be to build a voltage follower buffer that (say) is supplied from a 5V supply, but that 5V supply is floating (DC-DC converter, say), and ground for it is connected to the output of a crude high voltage op-amp (say an expensive Apex 400V amplifier with -10/+350V). You'd need protection on the input such as clamp diodes and a series resistor because it will see high voltage occasionally as well as on the output since it will subject whatever it is connected to to high voltage (maybe just a series resistor). Put a divider across the floating 5V supply and servo that to 0V with respect to system ground using the crude amplifier. Input protection for the low voltage amplifier, whatever follows it, and noise contributed by the crude amplifier that is too high frequency to be eliminated by the low voltage buffer are possible problems, as is slew rate. Fortunately there are easier solutions for 50V signals!
However your 50V+ requirement is not all that daunting for a 2015 design, and you can simply slap down an LTC2057HV and give it something like a 0V/+55V supply. Any output load should be to the negative rail. Offset drift is only 15nV/°C and 4uV maximum offset, so 16 bit accuracy down to about 0.25V full scale, and (with single point zero null) for +/-20°C, more like 20mV full scale typically. Chances are your connectors (thermal EMFs), reference drift, PGA and other components will add a lot more error.