I would like to begin the discussion with some reassurances.
All the actions I will propose will not modify the frequency response or the calibration of the probe in any way.
In any case, for those interested, at the end of the discussion I will illustrate step by step the complete calibration procedure of the device, also explaining in detail the reason (not exactly intuitive...) of the presence of two capacitive compensation trimmers for each channel...
As you know, the CMRR of a differential amplifier depends both on the performance of the operational amplifier in use and on the accuracy and symmetry of the polarization network external to it.
To simplify the analysis, let's assume we use an ideal op amp. Well, in this case the CMRR will depend only on the network external to it!
Using the classic differential structure schematized in fig. 1, as you know, two are the conditions/constraints to obtain an infinite cmrr:
Zi+ = Zi- and Zp+ = Zp- (1)
if these conditions are both satisfied the cmrr will be infinite, at any frequency.
If we reduce the biasing network to simple resistors (and this probe uses only simple resistors...), the basic conditions to be respected will become:
Ri+ = Ri- and Rp+ = Rp- (2)
where once again, Rp+ and Rp- represent the parallel of the resistors that relate to the non-inverting and inverting input respectively.
In other words, the above conditions summarize the "secret" to obtaining an infinite cmrr using a linear polarization network: to have total and absolute equality between the non-inverting and inverting amplification factors respectively. Just to fix the ideas with simple numbers, a 1% of asymmetry between the two channels degrades the CMRR to -40dB. To obtain a CMRR of -100dB the unbalance must be less than 10ppm!
What it happens in the real world?
In this case the CMRR will depend on the performance of the op-amp and on the frequency behavior of the polarization network which, even if made up of simple resistors, as the frequency increases will be affected by the effects of some parasitic parameters (e.g. traces and components inductance, mutual inductive and capacitive coupling between components, traces and components coupling with the pcb reference layers and so on...).
On many differential probes that make use of this technology (DP10007 included), the above balancing (2) is generally achieved with two complementary measures:
1) The use of precision resistors
2) The insertion of a calibration resistive trimmer whose purpose is to finely adjust the amplification factor of one of the two channels
(note that this last adjustment can be either internal to the circuit under examination or external to it!).
As it is easy to imagine, since there is no high frequency compensation (as we will see, for DP10007 this statement is not entirely true...), these two measures will guarantee optimal CMRR at low frequencies only (note that the amplification factor of the balancing channel is usually calibrated at 50Hz...)
But then, what happens to CMRR at high frequencies?
As already mentioned, at high frequencies this parameter is substantially governed by the multitude of parasitic parameters affecting the entire circuit.
Having said this, you will surely have understood how I intend to proceed!
We are very close to the conclusion of the whole story!
Soon I will propose two solutions: the first very simple and immediate, intended for those who want to improve just a little the disastrous high freq CMRR of the x100 range (which, as you will see, is fundamentally due to a design error by Micsig... ), the second, more complete, performing and flexible, aimed at global performance optimization.