Improve high frequency CMRR of InstAmp with extra L?

[max_torque]

Is putting a common mode choke into the input path of an instrumentation amplifier a sensible way of boosting the high frequency common mode rejection ratio of that amplifier?  Or is it impractical, due to matching or tollerance issues?

If i wanted a ~1Mhz differential bandwidth with the highest possible CMR is this approach practical?

[max_torque]

Anyone??   

[tszaboo]

Technically, I believe it would. But INA-s have very high input impedance. So you can place RC filters before them, without any drawbacks. Since the resistor can be a high value, you can use NP0 caps, therefore accurate and stable. And cheaper.

[danadak]

You can do a quick matlab evaluation of CMRR with unbalanced input
Z, and I think you will find small mismatches really hose the performance.
They way around that is to do a production line trim either with a variable
L or C in the signal path.

What CMR are you trying to actually achieve at 1 Mhz ?

Here is a CMR analysis I did for just DC, it was an eye opener. Attached, MATLAB.

And some more info that might be of use.


https://www.dropbox.com/s/plck7e95v7pw33c/CMR%20Analysis%20IA.pdf?dl=0


Regards, Dana.

[max_torque]

Hows about 150 dB Common Mode rejection at 1 MHz.......   I know, tall order eh!   

Low volume device, so there is the opportunity to trim after build, so if sensible variable L's or C's could be included that may work...

[f5r5e5d]

shielded transformer isolation? - but limited bandwidth, probably only 3 decades - so kHz to MHz

[Marco]

Hows about 150 dB Common Mode rejection at 1 MHz.......   I know, tall order eh!   

Are you trying to detect 10s of nanovolts with volt level common mode or 10s of microvolt with kV level common mode?

[max_torque]

10s of uV with kV level common mode, and that common mode includes some serious edges, think 750v in < 1uS   

[f5r5e5d]

then diff amp, ADC and isolated data transmission to receiver, requires power on both ends

[Marco]

I haven't done the math, but I think a divide down differential probe is a pipe dream for this. I think you want an isolation amplifier floating at the common mode. Some common mode effects will remain from stray capacitance, with any significant cabling before the amp you probably still want balanced inputs, but far more manageable.

Don't think there is an out of the box solution for MHz bandwidth. So you'd have to cobble something together, maybe a low frequency IC for DC with HCPL-4562 and some custom circuitry for higher frequency? Or an ADC/DAC solution with an optical link in between.

[danadak]

You are looking at ~ 25 bits of resolution, basically a lab grade instrument performance.

Ignoring temperature, load/line/EMI effects.

Do an error budget end to end, I think you will find doing 12 bits to 1/2 LSB over time/temp,
at DC, is a challenge. Include noise analysis.

Just the drift in passives to do the AC trim will be a killer. You might look into a simultaneous sampling
technique to get the CMR you are trying to achieve, but 1 Mhz I am thinking a staggering goal.

An AD7760 can hit 2.5 MSPS @ 24 bits, maybe a starting point. Maybe separate the DC from AC path
and recombine as an architecture.


Good luck.


Regards, Dana.

[David Hess]

Hows about 150 dB Common Mode rejection at 1 MHz.......   I know, tall order eh!   

10s of uV with kV level common mode, and that common mode includes some serious edges, think 750v in < 1uS

Even excluding the loss of common mode rejection ratio do to input dividers, very few instrumentation amplifiers achieve 150dB and none of them do so at high frequencies where high input impedance is not available.

Floating either the source or the instrumentation amplifier might work but if you have a kilovolt level common mode signal, then floating the source is dubious.  Floating the amplifier amounts to the same thing as processing the signal at the source and then sending it through an isolation barrier.  Gain before the isolation barrier directly improves the common mode rejection and as much as possible should be used.

So, amplify the signal at the source lowering the common mode rejection ratio requirements and then use an instrumentation amplifier.