0
\$\begingroup\$

Basically I will be calibrating circuits to a high degree of precision. Keeping it minimalistic, I figure all I needed for that are resistor networks, voltmeters and a few high precision voltage references. Like I said it has to be very high precision, so I figure the voltage differences would be in the microvolt range.

I've looked at different options from expensive high precision lab voltmeters to oscilloscope preamps (I suppose it could be adapted as a preamp for a voltmeter?). They are all so expensive.

Since my circuits have to work both with AC and DC signals, we could simplify everything by only calibrating against DC signals. So noise becomes a lot less of a problem, since it's DC.

So now I figure, just get a Low Noise Instrumentation Amplifier (or preamp) that will amplify by a fixed factor of 1000, to bring the microvolt to millivolt range, which my multimeter can then measure.

Is there any problem with this whole approach? If you have really good schematic designs for an Instrumentation LNA, can you guys share it with me?

\$\endgroup\$
7
  • \$\begingroup\$ To get a meaningful answer you may need to give a bit more information. What do the circuits do? What signals go in? What signals come out? What level of precision do you need, and what level of accuracy do you need? They are not the same, and the latter costs much more than the former. How big is the budget? One thing to note from your question thus far is that DC is probably not the way to go for uV level signals, as you'll hit problems with thermoelectric voltages and 1/f noise. AC may be better. \$\endgroup\$ – Jack B Dec 11 '15 at 11:18
  • \$\begingroup\$ Yes, I've read on the subject of drifts from DC op-amps that has voltage gains in the thousands. I could simply go with zero-drift op-amps (most of which I've read uses choppers). I'm not constrained to common parts as I can always order harder to find ones from Digi-Key. \$\endgroup\$ – kozner Dec 11 '15 at 12:30
  • 1
    \$\begingroup\$ A zero drift, chopper based opamp avoids thermoelectric drift from dissimilar metal junctions inside itself. It does nothing for the dissimilar metal junctions between it and the device under test. For that to work you have to chop the excitation going into the device under test, and sync that with the chopping of the amp. That basically boils down to doing an AC measurement. \$\endgroup\$ – Jack B Dec 11 '15 at 12:40
  • \$\begingroup\$ I can ground it to the PCB's large copper area and then thermally insulate it inside a good enclosure. The wires can be far enough from the op-amp IC that it doesn't "affect the package's junctions". Coz, of course, that's how I'll build it, I'll put the device to be tested directly on the pins of the package. \$\endgroup\$ – kozner Dec 11 '15 at 13:01
  • \$\begingroup\$ As for the chopping of inside the op-amp, the frequency, I don't think matters; relative to the load of it's output. It chops it up and integrates everything back in a self contained manner, I believe. \$\endgroup\$ – kozner Dec 11 '15 at 13:03
2
\$\begingroup\$

Here are the pitfalls as I see them: -

  • The connection between two dissimilar metals is called a thermocouple and this can produce a voltage of several micro volts per degree centigrade. This is why thermocouple amplifiers use cold junction compensation and they are rigorous about this. Getting an accuracy below 0.1 degC is almost unfeasible. Note that I said accuracy and not resolution.
  • The best DC spec op-amp I use (and I do use some pretty good ones) has an offset voltage of typically 5uV. This represents an error in your measurement system and there's very little you can do about it except perform very exact calibrations quite often.
  • There is also a temperature related drift with this offset too so a stable environment is important
  • Amplifier bias currents can play havoc with DC voltage measurements if the source (the thing to be measured) has a significant output resistance. For instance a strain-gauge bridge as a resistance of typically 350 ohms in a lot of applications and, if you choose an amplifier (or op-amp) that has a bias or offset current of 10 nA you get an offset error voltage of 3.5 uV. Clearly for measuring sources that have a much higher resistance you have to choose amplifiers with much lower bias currents.

So, my advice is look at what is available and set your requirements to be achievable. I'm not going to give you diagrams because they have IP value and this is a free advice site.

\$\endgroup\$
11
  • \$\begingroup\$ How about Zero-Drift op-amps?? \$\endgroup\$ – kozner Dec 11 '15 at 12:34
  • \$\begingroup\$ I use them but read the specs - they aren't actually zero drift and who, in reality, can expect them to be. Send me a link of one that you consider to be good and I'll point out the problems if you want. \$\endgroup\$ – Andy aka Dec 11 '15 at 12:36
  • \$\begingroup\$ Also, please stop blocking me from asking any more question as I already have too many accounts that I have follow, beside my public ones. \$\endgroup\$ – kozner Dec 11 '15 at 12:40
  • \$\begingroup\$ Also, when measuring DC voltages, a fairly cheap compromise is to subtract the offset. For a fixed amplification of 1000 and say an offset of 200nV, you only need to subtract 0.2mV from the output of my multimeter. Who's afraid of a little bit of arithmetic? \$\endgroup\$ – kozner Dec 11 '15 at 12:45
  • \$\begingroup\$ Also (just saying "also", because I've established it in the last 3 comments), the only problem with said arithmetic is that the offset has to be established and doesn't drift. \$\endgroup\$ – kozner Dec 11 '15 at 12:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.