Improved method of instrumentation amplifier input noise and RFI filtering

Question

The following circuit diagrams are taken out of Niel Albaugh's The Instrumentation Amplifier Handbook

---

I understand how the RC (tolerances) mismatches in Figure 4-16 (left circuit) cause the two time constants to not perfectly match and result in differential signal, and the error gets amplified to the output. OK. Then the author suggested a fix to the circuit by adding C3, and that the circuit on the right (Figure 4-20) is an

(page. 52) ... an improved method of instrumentation amplifier input noise and RFI filtering ... even though the R1C1 and R2C2 mismatch remain the same, there is less common mode voltage to differential voltage conversion with this filter.

I don't get this. If the frequency is low (no RFI), C3 is no more than an open circuit, and the circuit on the right is identical to that on the left. If frequency is high enough (RFI), C3 will appear to be a short circuit - there won't be any voltage potential across the two inputs. I don't see how C3 filters RFI other than zeroing out the inputs whenever the signal is influenced by a high frequency RFI.

Would someone explain how Figure 4-20 improved noise and RFI filtering? (I hope this is not a typo or misprint from an aged old text)

Answer 1

If the EMI is at a frequency within the amplifier's bandwidth, then it will process it like a normal signal, and it will end up in the output. You might have other unpleasant surprises, like CMRR degrading at HF so common mode noise is not rejected that well, or intermodulation, slew rate problems, etc.

However if the EMI is at a much higher frequency than what the amplifier can process, then it rectified and detected by semiconductor junctions and turned into a noise signal that looks like the envelope of the RF signal. That's why you hear the famous cellphone beeps in the loudspeaker before the phone rings, if the amp does not have EMI countermeasures.

The circuit you show works differently depending on frequency. I don't want to type "at a higher frequency" at the beginning of each line, o, from low to high frequency:

• the caps don't do anything.

• C3 which has higher value than C1 and C2 begins to filter differential HF noise

• C1 and C2 begin to act, and their mismatch does convert some of the common mode into differential mode, but that is neutralized by C3 which has much higher value

• Then finally at very high frequency, C1 and C2 short the RF to ground.

So the key is that C3 has higher capacitance than C1 and C2, so when C1 and C2 mismatch converts CM noise into DM noise, it is removed by C3.

EMI Rejection Ratio of Operational Amplifiers

AN-1698 A Specification for EMI Hardened Operational Amplifiers

There is another circuit which is very clever, from THAT1200 series of differential receivers:

Here C1 and C2 play the same role as your C3 cap, they filter differential noise. Their common node is connected to ground via one cap, C3, and this cap is bootstrapped, at low frequency the chip drives it via R3 and Cb so the current that flows through C3 does not come from C1 and C2, it comes from the driver instead. This means the common node between C1 and C2 is always kept at the common mode voltage, which means mismatch between C1 and C2 does not cause common mode to differential conversion, because no common mode current flows in C1 and C2.

Answer 2

I don't get this. If the frequency is low (no RFI), C3 is no more than an open circuit, and the circuit on the right is identical to that on the left.

Correct

If frequency is high enough (RFI), C3 will appear to be a short circuit - there won't be any voltage potential across the two inputs. I don't see how C3 filters RFI other than zeroing out the inputs whenever the signal is influenced by a high frequency RFI.

Incorrect. If the interfering frequency is high (or high enough) then C3 attenuates only that interfering frequency and (broadly speaking) leaves the lower frequency wanted signal intact.

Answer 3

I think the one thing you've missed is the principle of superposition, which applies to linear systems - and we assume here the opamp is linear.

According to this, you can apply a high frequency signal and a low frequency signal simultaneously, and (within the linear range) they won't interfere with each other.

(If the two signals, summed, exceed the supply rails, that's no longer true because the amp ceases to be linear. So assume both small signals or now).

Now the low frequency signal sees C3 as a high impedance (the open circuit you talk of) and is amplified as you expect with the designed gain.

But the high frequency (RF) signal sees C3 as a short, and is attenuated to 0, and there is no RF output - just as you say.

And the difference in gains for these two signals is precisely the "improved RFI filtering" you want.