Dear JMGL,
Yes, you will have to definitely look into compensating your divider. Attached is a simple simulation of your divider with 6.8pF (C1) stray capacitance, giving you the observed 0.08 rad phase shift. Also attached is a divider with 136pF compensation capacitor, giving you a flat frequency response and no phase shift. This of course is a simple simulation with ideal components, but it still shows the importance of compensating the high voltage stray capacitance.
A few important issues:
- C1 is not a real capacitor, but symbolizes the stray capacitance (Cs). This stray capacitance varies with the placement of leads, your hands, etc.
- to limit this non-stable stray capacitance, the divider needs to be shielded, e.g. in a metallic Pomona box.
- C2 and C3 are examples used to compensate the stray capacitance. Their exact value depends on the actual stray capacitance (20x Cs) and is therefore strongly dependent on your mechanical layout. The shielding is probably reducing Cs already.
- At least part of the compensation capacitance should be adjustable to be able to compensate Cs in the circuit. Drill a small hole in the Pomona lid and if possible use a plastic tool to turn the adjustable capacitor.
- make sure that the rotor part of the adjustable capacitor is grounded.
- Components: I would use SMD 1206, thin film resistors and C0G or film capacitors for at least 250V. Non-COG capacitors vary too much with voltage and temperature.
- build a small circuit board to mount the components securely so they don't shift and change the compensation.
- use as much ground area on the circuit board as possible. Inductive voltage drops on the ground side go right into the low-voltage side and effect the phase shift.
- Keep in mind the rules for high frequency design. To have less than 0.005 rad at 1kHz sine wave, the cut-off frequency of your divider must be more than 200kHz. If your wave form is other than a pure sine, the limit frequency must be much higher than this, probably to several MHz.
- Choosing the divider resistance is always a trade-off. Too low a resistance, and the loading of the circuit as well as the self-heating of the resistors becomes an issue. Too high a resistance, and the resistors come less stable and the influence of the stray capacitance increases. E.g. with say 2x 10MOhms and 1MOhms on the low side, 6.8pF would create a phase shift of more than 31 degrees.
- The output cable of the divider is effecting the response also. On the one hand, it can act as an antenna for the stray fields if not properly shielded. On the other hand, its capacitance is in parallel to the compensation capacitance and can even overcompensate if too long.
- If your ADC doesn't have a buffer amplifier built in, all bets are off. Normal ADC's switch capacitances back and for during each measurement period. If, for such an ADC, the source impedance is too high, its response is not at all what you would expect.
- To reduce the influence of the output cable and ADC, place a voltage-follower Op-Amp directly at the divider output if possible.
A last thought: why don't you use a1:10 oscilloscope divider? A good one should be shielded and have the compensation capability build in. It expects a 1MOhms load for the 1:10 ratio, if you reduce the load (probably to 500kOhms) it could easily be brought to 1:20.
Best Regards,
Dirk