I have a high voltage power supply for an electron beam system. The voltage regulator is an old analog system that works via a simple comparator and reference signal from a POT. My problem is the feedback voltage from the high voltage tank (tank meaning large steel oil filled tank with all HV components inside) is upward of 124 volts DC and is floating in respect to ground. The feedback signal comes from a 50k resistor which is part of a large resistance network inside the tank. Part of that network also runs to an analog 0-1mA meter that is a kV meter so I don't want to mess with the resistance.

I am trying to replace the analog controller with a microcontroller (ATmega) to replace the closed loop analog control with digital control. I don't want to mess with the resistance network inside the tank so this has to piggyback existing system and not unbalance the resistance.

Below is an overly simplified schematic of the power supply. As you can see R1 is the feedback resistor, two wires are taken from each of its terminals and run to the voltage regulator. R3 represents beam current measurement and overload protection (different circuit). The ground is chassis/earth ground. R2 represents the numerous resistors inside the tank from the negative side of the power supply. R4 and R5 is my planned divider to a difference amp (ignore the TL801 part number). My idea is to use a high impedance divider using 11 megs of resistance and then feeding that scaled down signal into a unity gain difference amplifier or high impedance instrumentation amplifier. The amplifier would be powered from an isolating DC-DC power supply and the signal isolated via an isolation amplifier (TI ISO124). From there I can scale it further using a POT to the uC ADC. Any thoughts?


simulate this circuit – Schematic created using CircuitLab

Notes: If anyone is wondering, since the feedback signal is within a few volts of ground, there is no need to worry about high voltage isolation, only ground isolation. The high voltage power supply normal operating range is 0-150kV and at 150kV, will produce a feedback voltage of 100V. however, there is a setting to allow to power supply to output its maximum of ~183kV for what is knows as corona clearing of the gun. That gives you a little over 123V.


I would be inclined to split R4 into two resistors, one on each side of R5, feeding the two inputs of the amplifier. Not only will this help eliminate input bias offsets, it will help limit fault currents should a problem occur.

Other than that, what you propose sounds fine.


The existing analogue circuit may have capacitance across R1. You need to investigate this.

For example if R2 has 1pF self-capacitance, R1 should have or require, (to balance the potential divider across all frequencies) a capacitance of over 1nF. Be very careful that you look into this when replacing the analogue control method.

Without the +1nF you will not be measuring stuff correctly and your digital control-loop may easily go unstable (as would the analogue loop).

Additionally, if you don't have enough capacitance you could easily destroy your front-end and maybe even cause a high-voltage risk when the HT is ramping up. I don't know what speed your HT will ramp up at and if the leakage capacitance is a lot higher than 1pF this could cause a few mA to flow through an unsuspecting finger.

Are you sure you want to mess with this? It's very easy to get things wrong in a control loop and have you looked at the impact of this? If you are happy about this then I'd also consider bolstering up the protection to your feedback op-amp and this can introduce errors so be aware.

  • \$\begingroup\$ Thank you for the tip. I checked the original schematic and there is no cap across the input. Though oddly enough there are two 20k resistors in series with the negative input, one of which has a 1uf cap across it that feeds the op-amp. The positive side of the feedback goes to the dual rail power supply ground in the regulator. \$\endgroup\$ – Mister Tea May 14 '13 at 18:58
  • \$\begingroup\$ This is the problem with high voltage measurment - the critical performance of the "system" may depend on components that are either obvious (stuff on the inputs) or not so obvious (stuff somewhere else in the circuit) AND, you have to decide if anything looks a little non-ordinary - we can't help you in this without seeing the circuit of what you have at the moment - this circuit may have something that "accomodates" the nuances of the pot divider but not at some obvious place like the input amp. \$\endgroup\$ – Andy aka May 14 '13 at 19:02

Not the answer you're looking for? Browse other questions tagged or ask your own question.