PSRR match in dual opamp

Are you expecting power supply noise, as in broadband noise, to be correlated and thus cancel. Is that what all that prose you wrote says?

You are far better off finding an op amp with suitable PSRR than depending on PSSR tracking. Also use a LDO for the input stage with good PSRR.

Obviously the matching between op amps will be good, but of course it won't be tested. If you are building one device, no problem. If you are making a million devices, you don't want to depend on anything that isn't tested.

Tooling wears out. Plates get damaged. Wafer vendors change. Every chip you buy is a moving target, except that it will pass the electrical specifications. Not the typical specifications, but test parameters with limits.

Thanks for the comments. I was hoping to avoid the LDO and filter on the power supply, but it isn't a great imposition to retain them.

I should know that typical or untested performance figures aren't guaranteed - I've reminded others of it in the past.

Regards, Allan

If I had to guess, the PSRR matching is probably maybe 10-20dB better than one op-amp alone, if the bias conditions are identical for both op-amps, but probably degrade a lot when you apply a signal that changes the bias on the PN junctions such that they are not identical between the op-amps. In any case you can't rely on it if it is not specified, unless you test each unit yourself.

If there is no cancellation between the CMRR of the two sides, what power supply noise spectral density spec do you come up with?

Why not use a differential amplifier configuration (ideally an integrated one), at least for the differential measurement? This will normally have a guaranteed CMRR, amongst several other benefits. If the other measurement channel is one of the single-ended signals, or the common-mode, you may be able to extract that reading also from various points on the diff amp.

Do you need DC-coupling or will AC-coupling be used?

If it is DC coupled, or to a lesser extent if it is AC coupled, you may have a problem with noise from whatever is giving you the 1-2V bias, since that bias voltage (or a filtered version of it in the AC coupled case) will get amplified by the different gains of your two single ended amplifiers. The matching of the resistors in your two separate amplifiers is unlikely to be anything like as good as the CMRR of a decent op-amp at low frequencies. Therefore it may be a bigger source of noise in the output data than the problem that you were worrying about.

Without chopping techniques, MOS opamps are rarely the lowest noise sort below 50kHz. What is your source impedance? If it is high impedance (which perhaps led you to choose MOS input), then JFET input is likely to be quieter, especially if you have plenty of supply current available.

We could be more helpful if you could provide more information about the problem you are trying to solve. I anticipate that you will give the standard response, that it is a secret project, in which case I can't be of any further help.

Chris

Its probably about the worst matched parameter of them all. PSRR various a huge amount over all process corners and temperature, e.g. at low frequencies could be 30dB variation. So, small variations on the chip have a much bigger impact on PSRR than say, BW. The saving grace is that its usually quite high at LF, say > 80dB. Random offsets are not matched at all. That's the meaning of offset.

Not a great idea. A high CMRR in analog is a lot easier. Use a decent differential amp with gain for the differential signal. Might need to be a chopper for uV dc. uV dc is non trivial, dissimilar wire connection can give mv thermal signals.

Kevin Aylward B.Sc.

They should be close but maybe not exact. A good power supply filter would help a lot. An RC with a big polymer aluminum cap, or a capacitance multiplier.

I've never found opamp PS noise to be that much of a problem. (but add a bit of RC filtering.. or a cap. multiplier if you are worried.)

Where does the 1-2 V bias come from? That will need to be clean.

(In general I think the negative PSRR is worse than the positive rail... but only of concern when you get near the GBW of the opamp.)

George H.

Use a capacitance multiplier on the supplies. It's squishy down at DC, where the op amp PSR is high, but stiff and flat at higher frequencies where you (usually) care about that. It'll cost you 0.7 to 1 volt, but it's well worth it for a sensitive front end.

Cheers

Phil Hobbs

Stiff and quiet, I mean--like a nanovolt or two per root hertz.

Cheers

Phil Hobbs

You can look at die photos of dual op amps to see if the layout is mirror imaged or "step and repeat". I'm guessing a mirror image layout has better op amp to op amp matching.