Differential amplifier question.

The difference between the two circuits is in how they handle differential and common mode noise. The half C circuit works only with differential nois e. The two cap circuit works on both differential and common mode noise. I would use all three caps. The primary pole would be the two caps to ground and a smaller cap in between the two inputs. It might not be necessary but you'll have the three footprints on the board to play around with for any i ssues that you find later.

Which is best sounds like an experiment.. and as Wandered says depends on the type of noise. Personally I'd start with the grounded cap. (That's what I do on my bridge -> differential circuits.) Grasp your ground firmly... Maybe that's like "always invert".

But if it's mostly differential noise.... \ One issue I've seen with the caps to ground is that if your signal gets too close to the RC corner freq. then the RC's never match perfectly and you get different "gains" in each channel.

George H.

At radio frequencies, your differential amplifier is unlikely to offer good common-mode rejection. So, for frequencies higher than your intended signal, interference has to be filtered. That's why a capacitor to ground is important, and you ought to consider RF bypass to the positive and negative power supplies as well. Differential gain is deliberately rolled off, is less likely to have any surprising RF sensitivity.

Opinion has nothing to do with it. The GND'ed C case does a common mode to differential conversion not present in the ungrounded C circuit. This means a common mode voltage will appear as a differential voltage to the input i nsofar as the C's mismatch. There is no such effect in the ungrounded C cas e. Then since LTC6263 CMRR is pretty good, like 80dB out to 10MHz (?), no c ommon mode rolloff is necessary. Grounding the input C will ruin this massi vely, maybe 40dB or worse.

o differential conversion not present in the ungrounded C circuit. This mea ns a common mode voltage will appear as a differential voltage to the input insofar as the C's mismatch. There is no such effect in the ungrounded C c ase. Then since LTC6263 CMRR is pretty good, like 80dB out to 10MHz (?), no common mode rolloff is necessary. Grounding the input C will ruin this mas sively, maybe 40dB or worse.

Oh I totally agree. You've gotta think about where all the corners are. One cap and 1% R's.. When AC coupling, if the corner is within a factor of ten of the frequency of interest, then start worrying... maybe way before that if the signal is small.

George H.

If the opamp rectifies gigahertz signals from cellphones, then having no (deliberate) capacitance at all to ground might also not be the best option.

I have seen systems where the main (biggest) capacitor is between the signal lines (which sets the roll-off for differential signals) and then small capacitors from each signal line to ground, for common-mode RF rejection purposes. Since the capacitors to ground are much smaller (e.g. at least 10x smaller than the differential capacitor), their mismatch degrades the CMRR less than it would if they were big. One would choose the relative sizes of the capacitors depending on the matching of the capacitors to ground that one can afford to buy, and the CMRR that one needs, and the amount of common-mode RF rejection that one needs (which is determined by the level and frequency of likely RF exposure and sensitivity of the op-amp to RF rectification).

Chris

as

to differential conversion not present in the ungrounded C circuit. This m eans a common mode voltage will appear as a differential voltage to the inp ut insofar as the C's mismatch. There is no such effect in the ungrounded C case. Then since LTC6263 CMRR is pretty good, like 80dB out to 10MHz (?), no common mode rolloff is necessary. Grounding the input C will ruin this m assively, maybe 40dB or worse.

on.

Same thing here:

-671.pdf

Great link, Fred and thanks everyone, three capacitors it is.

Cheers

uch

two

de

ts

d as

s,

de to differential conversion not present in the ungrounded C circuit. This means a common mode voltage will appear as a differential voltage to the i nput insofar as the C's mismatch. There is no such effect in the ungrounded C case. Then since LTC6263 CMRR is pretty good, like 80dB out to 10MHz (?) , no common mode rolloff is necessary. Grounding the input C will ruin this massively, maybe 40dB or worse.

tion.

AN-671.pdf

Nice... Thanks, George H.

Den torsdag den 21. april 2016 kl. 03.02.26 UTC+2 skrev George Herold:

ch

wo

as

to differential conversion not present in the ungrounded C circuit. This m eans a common mode voltage will appear as a differential voltage to the inp ut insofar as the C's mismatch. There is no such effect in the ungrounded C case. Then since LTC6263 CMRR is pretty good, like 80dB out to 10MHz (?), no common mode rolloff is necessary. Grounding the input C will ruin this m assively, maybe 40dB or worse.

s small.

I've often seen that on RTD driving with a current source one side is the high impedance current source, the other side grounded

-Lasse