So the output voltage is equal to the input voltage divided by the ratio 2X / (2X + 2Y)
Now the CRO probe typically has an impedance of 1Mohm.
(some have a x10 switch which alters that to 10Mohms)
So the impedance presented to the test point is 1M // total impedance of the attenuator.
For testing plate circuits, of high-ish impedance, typically hundreds Kohms, we need the probe impedance to be as high as possible, so as not to overly load the circuit-under-test, which would reduce it's voltage signal as well as increase it's distortion.
That means *high* Y resistors.
However, we want the impedance fed to the audio interface balanced input to relatively low, in the tens K, otherwise we begin to run into impedance mismatch issues, which upsets our freq response sweeps.
The M-Audio Profire balanced 'line inputs' has a spec of ' > 20Kohms', so a sensible maximum source impedance would be 20Kohms.
That means X resistors of 10K each, for a total of 20K 'balanced'.
[ the junction of the X resistors is connected to XLR ground ... which is the M-Audio interface's ground ]
[one can have a 'ground lift' switch which makes/breaks connection of this junction to the box chassis]
OK then, the problem is that in order to maintain the X = 10Kohms and Y= *high* Kohms conditions, our attenuation ratio becomes quite high.
It gets difficult to achieve the lower attenuation ratios without resulting in a low probe impedance which then excessively loads the circuit under test.
That makes, for lower amplitude signals, say 50Vpp max, achieving a decent 'signal to noise ratio' into REW an issue - noise can begin to dominate proceedings as well as distortion increasing because of A/D resolution issues.
The higher attenuation ratios, say for signals of 100Vpp and greater is not too difficult and can be achieved without undue loading effects.
In my box, I have a rotary switch which varies the two Y resistors and the X resistors are constant.
Varying the X resistors (impedance seen by interface input) complicates things some more
I have six available settings - so far I have used
i) Y = 51K , X= 10K ; Idealised atten ratio div 6 ; Probe impedance approx 1M // 122K -> 108K
ii) Y = 180K, X=10K ; Idealised atten ratio div 10 ; Probe impedance approx 1M // 380K -> 275K
ii) Y = 470K, X=10K ; Idealised atten ratio div 25 ; Probe impedance approx 1M // 960K -> 490K
In practice, for plate circuits, setting i) has around 24% loading effect and setting iii) around 6%
It's a bit approximate in practice - but I aim for the highest signal I can get into REW, around -20dbFS as a reasonable minimum without loading the circuit under test more than around 10%.
Checking freq response with sweeps shows no undue rolloffs 20-20KHz bandwidths and frequently 10 .. 44KHz BW in typical plate circuits.
...
One can use more elaborate attenuation schemes, perhaps with both X and Y being variable on the switch to improve things .. it does get more complicted fairly quickly