How do you adapt low speed, precision operational amplifier circuit to the high speed realm? And, more importantly, how do you interpret inconsistencies that you may encounter? In this post, I will focus on one specific circuit: the difference amplifier circuit and how the device architecture will affect performance. See figure 1.
Figure 1: Difference amplifier circuit
The difference amplifier is used to reject common-mode signal or to realize the conversion from differential to single-ended signal. When considering ideal operational amplifier, where the positive node is exactly equal to the inverting node, the common mode rejection ratio [CMRR] is the well known figure of 
where is the accuracy in % of the resistors selected for the design.
This, of course, is at DC, or for an ideal amplifier. If the amplifier is non-ideal, you will have an error voltage between the inverting and non-inverting input of the amplifier. Let’s call this error 
for voltage feedback amplifiers (VFB) and 
for current feedback amplifiers (CFB). Note that I selected the same name for the resistor accuracy 
and the error voltage 
for VFB amplifier across its input. I was careful not to select the same symbol as to keep those two terms distinguishable in the discussion.
, mentioned earlier, for the CFB is the gain of the buffer located between the non-inverting input and the inverting input. 
is normally equal to 0.98V/V for a typical CFB.
The amplifier voltage error, mentioned earlier, tends to be a lot smaller with a voltage feedback amplifier (VFB) as it is corrected by a higher in magnitude correction factor, the open-loop gain than CFB. See figure 2a,b for the OPA835 (36Mhz, 250uA quiescent current) CMRR and Aol plots vs. frequency.
Figure 2: a) OPA835 CMRR & PSRR vs. Frequency
b) OPA835 Aol & open-loop phase vs. frequency
So although you will have an excellent CMRR at DC, the voltage feedback architecture will not support very good common-mode rejection at higher frequencies. To achieve better high frequency CMRR, the CFB architecture will often prove to be a better choice.
CFB amplifiers, on the other hand, will have poor CMRR at low frequencies, see figure 2 for the OPA695 (1.4GHz, 12.5mA quiescent current) CMRR and Zol plots vs. frequency
Figure 3: a) OPA695 CMRR & PSRR vs. Frequency
b) OPA695 Zol & Open-loop Phase vs. Frequency
That said, how do we improve the CMRR at high frequency? There are several solutions that come to mind. The first one is to use a composite amplifier, combining the best of the precision world and the high-speed world. This will work to moderately high frequencies. To achieve high CMRR beyond 100MHz, the only solution left is to cascade several stages until the adequate CMRR target at the frequency of interest is met.
A specific application of the above was implemented for a high-impedance differential probe circuit. The circuit is shown in figure 4 below.
The OPA659 stage will not provide any CMRR rejection, but does provide a high-input impedance usually associated with probe. The OPA2695 circuit, who’s CMRR will depend on both the accuracy of the resistors and the CMRR of the input buffer. Note that the CMRR of the input buffer will be the limiting factor. Looking at the CMRR measurement, figure 5, you will see that the OPA2695 achieves only 28dB CMRR. This is consistent with . Note that 1% resistor, used here, will have -34dB CMRR in this circuit with an ideal amplifier. Adding both errors linearly will result will results in the observed -28dB.
Figure 4: Fully differential probe using CFB as difference amplifiers.
The second stage of difference amplifier, built using the OPA2695, almost doubles the previous -28dB to -52dB, bring the combined circuit CMRR to -50dB at 200MHz.
The last stage is a buffer stage to boost the gain as required.
The complete results and measurement of the CMRR after each stage is shown in figure 5.
Figure 5: CMRR measurement cumulative after each stage.
Here are a few of my other blogs that also might help in your design challenges:
- High-current pulsed-source needed
- Wanted: Stable oscillator
- This amplifier doesn't exist...now what!?
- This amplifier doesn't exist...now what?! - Part 2
- Current feedback amplifier...how do I make it work for me?
- Reducing amplifier power consumption for SAR ADC drivers
- Correcting DC errors in high-speed amplifier circuits
- High-gain, high-bandwidth....putting it all together
- High-gain, high-bandwidth…why is this circuit oscillating?
- High-gain, high-bandwidth… how can I get it all?