It could have a large effect: it forms a part of low-pass filter, with corner frequency dependent on range (lower voltages -> less bandwidth):
Old 10pF: about 8 MHz
New 0.5pF: about 160 MHz (ignoring parasitic capacitances)
So it improves the bandwidth, but on the other hand can cause aliasing problems. But because the old antialias only worked for > 16 MSPS anyway, it’s not a big loss.
I doubt that the 0.5pF has much effect, the 74HC4052 already has 3.5 pF of input capacitance. One could probably just desolder C9-C12 from HW 2.6 and have it work like HW 2.7, for the better or worse.
I’ve removed C9-C12 and C73-C74 and now with spec_an the -3 dB cutoff has moved to about 10 MHz. Still better than before.
Regarding the phase, I noticed a bit of skew on upper frequencies, but that could be related to wrong probe compensation (I recalibrated that later). Tomorrow I’ll do other tests.
Anyway, I managed to display the 16 MHz clock of an Arduino using the x10 probe, so maybe the situation really improved…
