Before starting with the discussion, I would like to make some considerations about the probe bandwidth.
As noted by many users of this thread, the probe bandwidth is far from the declared 100MHz!
To verify this avoiding the artifacts of the input resonances it is sufficient to remove the input leads and test the transfer function (TF) of the device in this status.
The bandwidth of the "pure" probe more or less will be near 40-50Mhz.
As shown in the below graph, the TF of the full device (i.e. provided of the input leads) is due to the sum of the "pure" probe TF with the input LC TF.
The resonance frequency of this LC network is about 110MHz but, twisting the leads, other "modes" will overlap to the main resonance complicating the situation...
My feeling is that this resonance was intentionally used to attribute the nominal 100MHZ bandwidth to the probe although, in practice, the behavior above 40MHz is highly variable and conditioned by the geometry and mutual position of the input leads.
A further fundamental aspect that few people know is that the resonance associated with the capacitance of the input stage can also substantially influence the CMRR measurement at high frequencies.
Therefore, always give due weight to the resonant frequency originating from the common mode capacitance of the two inputs (in this device is approximately 5pF) and the cumulative series inductance of the reference path (that usually coincides with the ground path of the measuring bench).
If this resonance frequency is close to the CMRR frequency you are investigating, the effect is that you'll measure a CMRR worse than the actual value.
As example, if the measurement bench is organized as in the below drawing, the resonance frequency can be estimated at around 25MHz and can strongly affect the 10MHz CMRR measurement.
To limit this error it is usually sufficient to reduce the inductance of the reference loop by making a direct ground connection between the oscilloscope and the signal generator lead and making this auxiliary connection close to the input and output leads of the probe.
This will move the resonant frequency further to the right, away from the frequency under investigation.
In practice, this operation can be easily accomplished by simply connecting a probe between the oscilloscope and the output lead of the signal generator (generally this connection is unconsciously present since the amplitude of the input signal is measured...)
Remark / Important Note: in case of significant high frequency common mode components, the above precautions must also be adopted during the usual differential measurements (this to avoid to degrade the expected high frequency CMRR of the probe).
OK, enough with the premises!
Next time I will explain how to improve the CMRR above 100KHz.
Do not miss!!