OPA2210: CMRR for a Instrumentation Front End

This circuit is very sensitive to resistor tolerances. Are you replacing only the amplifier chip, or do you have two circuits with different resistors?

Hi Hao,

Hao Jie Chan said:

Can you help with this application? The main objective is to achieve high DC and AC CMRR at least until 100kHz

I am not recommending to make your own instrumentation amplifier (IA) stage. I would recommend to use INA849 IA or similar one for your application. 

When you construct a discrete IA stage, the feedback resistor matching is super critical, where the matching in tolerance, temperature drift and parasitic near the discrete PCB layout. It is one thing to understand the IA's front end stage, but it is difficult to implement in discrete IA construction and expect to get the excellent CMRR performance. 

Enclosed is the application note that talks about the challenges of the feedback resistor matchings. 

https://e2e.ti.com/blogs_/archives/b/precisionhub/posts/dealing-with-rejection-instrumentation-amplifier-psrr-and-cmrr-part-ii

https://www.tij.co.jp/jp/lit/an/slyt683/slyt683.pdf?ts=1717599220345&ref_url=https%253A%252F%252Fwww.google.com%252F

(+) What you need to know about CMRR- The instrumentation amplifier (part 2) - Analog - Technical articles - TI E2E support forums.pdf

So to obtain excellent CMRR, say > 80dB, your feedback resistor matchings need to be better than 100ppm, which is difficult to do in discrete IA construction.  

If you have additional questions, please let me know. 

Best,

Raymond

Hi,

I am only replacing the amplifier chip.

Hi Hao, 

Hao Jie Chan said:

I am only replacing the amplifier chip.

Please explain. The IA's front end is using OPA2140 and you replaced with OPA2210, and poor CMMR is observed.

Best,

Raymond

Hi Raymond, 

I was considering INA851 before this as well, I think it's in the same family as INA849? I am targeting at least a gain of 5 in this stage. From the datasheet, both of the INAs have degraded Gain error and Gain drift at gain >1, the target is <0.1% of Gain error across a 20degC of temperature change.

Yes, I understand the resistor matching is crucial. In this circuit, the resistors are part of a resistor network, 4x 10k network, with 0.02% of tracking.

But the question remains, with 2140, I can get better performance, 2210 gives worse performance, so the resistor matching should be out of the question I think?

That is correct, Raymond

Hi Hao, 

But the question remains, with 2140, I can get better performance, 2210 gives worse performance, so the resistor matching should be out of the question I think?

I would like to know how the CMRR was tested and what input Vcm was used when and how you compare the results. Two op amps have different input Vcm range, and it is not as simple as op amp replacement and compare the CMRR directly.  

This is OPA2140's input Vcm range.

This is OPA2210's input Vcm range

Best,

Raymond

The IA front end is part of a signal chain.

IA (x5)---PGA855 (x8)---PGA855(x0.25) and interfaced to an ADC.

IA is powered from 30V and -4V.

The common mode signal is a sine wave, with 6Vpp and 3V offset.

Hi Hao, 

Let me investigate the differences between OPA2210 and OPA2140. 

Hao Jie Chan said:

IA (x5)---PGA855 (x8)---PGA855(x0.25) and interfaced to an ADC.

Are you using PGA855 as the differential amplifier stage after the discrete IA's analog front end? Could you lower the feedback resistors by a factor 10 in OPA2210 setup and see if the CMRR is improved. OPA2140 is using FET topology, while OPA2210 is BJT technology. 

I assumed that you are using your setup to measure the CMRR of the analog front end. 

Best.

Raymond

Yes, PGA855 is used as the differential amplifier after the discrete IA stage.

Unfortunately because the resistors are in a Resistor Network, this might be harder to change and to source. I'll see what I can do about it.

Yes, the CMRR result is obtained with that setup.

Hi Hao, 

You may be able to double or triple the resistor network in parallel and see if there is improvement in CMRR figures (even though you may not get the exact R values for now. 

Please let me know what you are getting. I still think that you should try INA849 for your design, keep the gains > 100 V/V.  

Best,

Raymond

Raymond,

After giving it some thoughts, I don't think the Bias and Offset current would change that much with frequency right? So they should only affect DC CMRR, not the AC part.

Hi Hao,

I assumed that your CMRR measurement is a single setup and you are comparing two different op amps for the measurement. I want to rule out other variable effects per the comparisons. 

I simulated CMRR in OPA2210 and OPA2140, and OPA2210 has better CMRR than OPA2140. The simulation is same as the figures presented in the datasheet.   The Aol in OPA2210 is also slight better than OPA2140 from the measurement in the datasheet. 

CMRR is not constant over frequencies. OPA2210 has better offset than OPA2140; OPA2140 has significantly lower Ib than OPA2210. I still believe that the feedback network is not optimized in OPA2210. 

If the setup is using single package of dual op amps, I can assume that Ib of each op amp should be matched. If not, they could be another variables. 

For OPA2140, the feedback network seems to be adequate as shown in Ia, Ib and Ic.

For OPA2210, the feedback network seems to be inadequate, shown in Ic --> may need to use lower feedback resistors. If I use a specific frequency to monitor the Ic, similar conclusion seems to indicate that OPA2210's feedback network may not be optimized. 

In the IA configuration, the common mode errors can convert to differential errors, because the analog front ends are not apple to apple comparisons. So my speculation is that this may be what is going on based on your data. 

Since you are using resistor network as the feedback network, it may be easier to parallel another identical one on top of each other and see if the CMRR will be affected, 

If you have additional questions, please let me know. 

Best,

Raymond

Raymond Zhang1 said:

If I use a specific frequency to monitor the Ic, similar conclusion seems to indicate that OPA2210's feedback network may not be optimized.

Could you please elaborate on this a bit more?

Does this mean, you can see the specific frequency component also appear at Ic for OPA2210?

From your simulation result, I can see the Ic plot for OPA2210 is "thicker", I guess noisier? But both of them are in the same magnitude (210E-15 vs 158E-15 range)

Hi Hao, 

The OPA2210's feedback network needs to be higher current feedback current than OPA2140 by approx. 600X times (typical value of Ibias in DC). 

If the feedback current in an op amp is smaller than the Ib current, it may operate in incorrect operation points, and this will lead to poor CMRR, increase Vos and frequency dependent errors. 

I think that you are implementing a single CMRR setup and comparing both IA's analog front end setup, it seems that the feedback network differences may be the only potential errors that may cause the poor CMMR results. That is why I am suggesting to decrease the values of the feedback network. 

It may be difficult to see this in DC sweep, perhaps AC frequency at 10kHz  is slightly more evident.  

The simulation has limitation as well here. My guess is that you have checked out most of the dependent variables that could degrade CMRR by the comparisons (based on the information you provided so far). The only proof  is to check it out on your setup. I do not see why OPA2210 should have poor CMRR than OPA2140, if the setup is done properly (if other dependent variables are comparable). 

If you have other questions, please let me know. 

Best,

Raymond 

Hi Hao,

Please let me know if the issues are resolved. 

Best,

Raymond

Hi Raymond,

I've not had time to do the experiment, so I would like to keep this open longer.

Hi Hao,

Let us know if you have a chance to revisit the test.

Best,

Raymond