Rail to rail opamp output performance measurement with an oscilloscope

[240RS]

I need to assess output offset from the plus and minus 15V rails of an opamp and have an oscilloscope (Keysight 3000 with passive probes and only a 1000V differential probe), DMM (6.5 digits Keithley) and a (dual) mV accurate voltage source (R&S NGPS 40) to help me with that.

I feed a couple of kHz square wave in the saturated opamp region to the positive input with a grounded negative input of the opamp. On the scope I measure plus and minus 15.0 V, if I add plus and minus 15.000V to the two channels of the scope it all overlaps perfectly (as intended. It is a couple dollar AD precision opamp)

I just wonder how I could “zoom in” on either the positive or negative pulse at the output and compare it to the +/-15.000v reference.

[RoGeorge]

Power the opamp from a floating power supply, or from batteries, then hook the oscilloscope GND to one of the opamp's + or - supply.

[240RS]

Great. Yes, I can do that. But: if I connect it like this:

- isolated +/- 15V to opamp
- scope ground to +15V rail
- probe to opamp output
- scope in (say) 10mV/div to measure offset from positive rail during the positive duty

….. wouldn’t I then have -30V during the negative duty? I was afraid that is a risk with the scope in 10mV/div range. Or is that no issue?

[RoGeorge]

That is only to observe how close the output of the opamp goes near that supply, while ignoring the other values.

The oscilloscope won't get any damage from 60V.  The max voltage is the same, no matter the volts/div settings.  Usually the max voltage is written on the front panel, near the BNC connector.  If not, it's written in the user manual, at the oscilloscope specs.

Same with the DMM, you can use the mV range and apply hundreds of volts by mistake, and the DMM won't get any damage.

[temperance]

Good luck trying to measure the off set from either rails at 10mv div. You will saturate the input of the scoop and nothing will be visible.

But why do you want to measure this when such things can be found in the data sheet for the op amp. If the design depends on it, it's a better idea to find an op map which does what you need out of the box?

[RoGeorge]

It works for low frequency, or with DC if the input stage of the oscilloscope can not get out of saturation fast enough.

[temperance]

Maybe, but you will see some recovery anyhow, messing up your measurement.

Bu the real question is why you want to do that. If you just want some +-15V reference square wave, I would opt for a different circuit working in class C which doesn't require any op amps. Also, if the +/-15V is that important, how about the stability of those rails.

Given that the op doesn't know how to use his scope and other equipment, I get the idea that the op is trying to do something which can be solved in a different way. But I might be very wrong of course.

[240RS]

The oscilloscope won't get any damage from 60V.  The max voltage is the same, no matter the volts/div settings.  Usually the max voltage is written on the front panel, near the BNC connector.  If not, it's written in the user manual, at the oscilloscope specs.

Same with the DMM, you can use the mV range and apply hundreds of volts by mistake, and the DMM won't get any damage.

That’s good news. Relays always click when going to lower V/divs and expected it would be an issue. So it isn’t.

Will try tomorrow. Also curious about saturation of the scope input stage. Hadn’t thought about that.

As for the “why”, it is really just an experiment in a practical situation. I had bought a “precision differential probe” for $50 on alibaba and was gutted to see that the noise alone was in the range of 20mV. I just want reasonable differential measurements. And now this case came up and I thought that there must be a different way to do a differential measurement and that is by making it a single measurement with a 15.00V DC offset. And tried to find whether the scope could do that directly. Couldn’t find it, but in a way a seemed logical to me if you could set the center at negative 30V and see the delta. Didn’t show up in the manuals though. And hearing about saturation, that is probably also a reason it wouldn’t work.

Thanks for all the feedback!

[jwet]

You might know this and if so disregard,  you can use two scope channels and and the scope will subtract them and display them as if they were a pseudo channel.  On some scopes, this is called the "math" channel, on old scopes it was a setting or was called invert and add.  As long as you stay within range of your scope, you won't hurt anything.

The diff probe that I think you said you have? is a nice way to do this. 

The difficulty with all this stuff is Common Mode rejection and the diff probe should be good in this respect.  Ideally, you could take two big voltages and subtract them leaving just a tiny residual- like 14.00001 - 14 = .00001.  What happens is the larger the numbers get, the poorer the subtraction goes.  A fair CMMR spec might be 60 db or 1/1000.  This is error in the subtraction based on how big the numbers are.  At 60 db, this means if you had 10v, you would get errors of about 10v/1000 or about 10 mV- you couldn't resolve anything much less.  Good CMMR requires precision matching of components so little errors don't creep in.   Instrumentation amps are made especially to have high CMMR, the precision resistors required are all on chip and are trimmed at the factory- CMMR it can be over 100 dB, 10 PPM or 1/100000.

Good luck with your experiments.  I don't quite understand what you're up to but at least this should give you some idea of what to expect.

[240RS]

Thanks! Good input on the CMRR.

I didn’t use math channel since I expected that to be two 8bit samples being subtracted and hence not give the accuracy I was looking for.

[jwet]

You're likely right with the math channels but it might be worth a try.  I wonder if things like add and subrtact are done in the analog domain and fancy stuff is done in the digital domain, probably not in a non lab type scope.

[2N3055]

You're likely right with the math channels but it might be worth a try.  I wonder if things like add and subrtact are done in the analog domain and fancy stuff is done in the digital domain, probably not in a non lab type scope.

Nope.
On digital scopes of today, all math is done in digital domain.