Reading a voltage which is grounded to a different potential?

[Faringdon]

Hi,
Do you know of a lower component count solution for the following?...
As you can see, there is a Diff Amp which reads a 3V source voltage…..then passes it to circuitry which is grounded to a slightly different potential….and that circuitry there must read the 3V voltage correctly.
The attached does it but is too high in component count. Do you know of a  lower component count way?

[TimFox]

Have you considered an integrated "instrumentation amplifier" chip?

[Faringdon]

Thanks yes, i have looked through some datasheets, but dont yet find any that give the diff bus type output....they all give an output referenced to their ground.......i was also wondering about isolation  op amps?

[Marco]

A non inverting summing amplifier can do it by just adding the offset, but there's a lot of resistors inserting gain errors.

A LTC1043 can do it without gain errors, but 10$ is insane.

[capt bullshot]

Hi,
Do you know of a lower component count solution for the following?...
As you can see, there is a Diff Amp which reads a 3V source voltage…..then passes it to circuitry which is grounded to a slightly different potential….and that circuitry there must read the 3V voltage correctly.
The attached does it but is too high in component count. Do you know of a  lower component count way?

Depending on the desired accuracy, source impedance and complications, you can do the whole shebang with just one OpAmp and four resistors (differential amplifier). Next bet would be a two OpAmp textbook circuit that does provide high input impedance, but less AC CMRR. And finally, that "classic" three OpAmp instrumentation amp can be done way cheaper than a lower parts count integrated solution, but that also depends on your overall requirements regarding e.g. accuracy, DC CMRR, AC CMRR, ...

[HalFoster]

Stick with the Instrumentation amplifiers - or a current sense amp if the common mode voltage is too high and just add an op-amp inverter if you need a differential output.  Texas Instruments has a whole series (INAxxx) that are very accurate, easy to use and much more affordable than the LT parts.

Hal

[Wolfram]

What is the required bandwidth, accuracy, CMRR, output impedance, source impedance, signal level, dynamic range, maximum voltage difference between the two grounds? There are tons of potential solutions, but the most suitable one depends a lot on the actual requirements.

A plain differential amplifier, sitting at the output-side potential and measuring the difference between your signal and its ground reference is the simplest, as mentioned. Making it a full instrumentation amplifier might be required if you have special requirements in regards to accuracy or CMRR, or your source impedance is high. Dedicated instrumentation amplifier chips start at around 20 cents. Then there are many more exotic options if you have special requirements, flying capacitor voltage sampling, delta-sigma analog voltage isolators, linearized optocouplers and so on.

[Faringdon]

What is the required bandwidth, accuracy, CMRR, output impedance, source impedance, signal level, dynamic range, maximum voltage difference between the two grounds?

Thanks, its not particularly special.....as follows...

........the idea is to actually make a UC3902 out of separate blocks. So we need a differential "share" bus....thats why this is needed

UC3902
https://www.ti.com/lit/ds/symlink/uc2902.pdf?ts=1664976108297&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FUC2902

The '3902 is a device which passes a "share" bus between paralleled power supplies, so that they can use that voltage to trim their output current, so that all PSUs share the total output current.....the "share" bus must be a differential bus, since at each power supply, the ground is slightly different in potential.

The actual "share" bus voltage is simply the voltage output of the current sense amplifier, of the PSU that is shipping the most current out of all PSU's.

We cant use the '3902 because its gain is fixed at 40...and it uses a fixed offset voltage of 35mV....and we may need it to be higher than this.

[Faringdon]

A-ha, after all, the UC3902 "double diff amp method" does actually work...when i tried it before, i stupidly didnt have the opamp power rails high enough in voltage and i was thus getting saturation of signal near the rails......so the attached shows the differential share bus being correctly received by all three positions.......i am a little concerned though, that the negative of the diff share bus is not actually driven...its floaty......i prefer a driven bus line , since its better for noise immunity.

Any nicely driven diff share bus options greatly appreciated.

I am still thinking this job  is made for an isolation amplifier?

[Terry Bites]

A bunch of precsion opamps and precsion resistors will always cost a lot more and perform less well than an isnt amp or two. As suggested, Inst amps are the way to go see www.analog.com/en/analog-dialogue/raqs/raq-issue-161.html  and SBOA001A from TI, easy. [ Dont try to paralllel "zero drift types, it will go badly.]
The choice of inst amp depends on cost/ accuracy trade offs you're willing to make. And the availabliity. The CMV range of the inst amp range will also matter.
For HV there are things like the INA149. Dont forget to use protection.

Check TI AD etc for "fully differential amps". They are diff in diff out.
You can get around ground this issue by sending a current over your pair rather than a voltage.  Because you are measuring a current, minor voltage drops in the line are irrelevant.  You gain a lot of noise immunity as well. A technique very popular in industrial settings for these reasons. www.bapihvac.com/application_note/the-science-of-4-to-20-ma-current-loops-application-note/  You get to power the remote end into the bargain. Precise conversion of a few mV to a few mA is easy. www.allaboutcircuits.com/technical-articles/how-to-design-a-precision-current-pump-with-op-amps/  Roll you own here, current loop signalling ics are pricey and unobtainium. You dont have to go with 4-20 of course.

Going iso is fine but you will have the extra expense of an isopower supply. 

[Marco]

Fully differential amplifiers tend to be meant for RF with signals around a common mode, they aren't necessarily very good at providing a precise signal relative to the output common mode (the common mode is not very important in a fully differential chain). Nor are they designed to work close to the rail.

[Haldor]

A summing amp is the simplest way to do this. I designed this circuit a couple of weeks ago to convert a bipolar (+/-5V or +/-10V) input signal into a 0.1 to 3.1 volt output for connection to a processor's ADC. The summing circuit only uses the 1st op-amp.

Since this is feeding an ADC I added a 2nd order, 4kHz antialiasing filter (my sampling rate is 10 kHz). The antialiasing filter should be designed for your ADC sampling rate (filter corner frequency < ADC_RATE / 2.5). If you want to use a simple 1st order filter, then increase the sampling rate to 10X of the bandwidth of interest.

The design was based on this app note from TI.

https://www.ti.com/lit/an/sloa097/sloa097.pdf?ts=1665349867438

[Terry Bites]

How does this deal with ground potential differences?
BTW I made a little tool- no macros don't worry.