I asked the following question on the eevblog forum (Hipot interference - Page 1) but no one has responded so I'm trying here.
For most of this week, I spent troubleshooting a torque testing machine that I developed. The system includes a "microcontroller board", a stepper motor controller and motor, a torque meter, and a computer. I quickly fixed all the small problems, but I kept having an intermittent problem where the microcontroller board (Arduino Uno) would stop responding to the computer. The two are connected using a USB serial connection. I found out that the Hipot testing that was happening a few feet from me was causing the problem. Whenever they abruptly shut off the high voltage, the USB connection would stop working.
I have attached a picture of the bench with the equipment on it and a screenshot of the electrical impulse that is being created by the Hipot tester. Can someone explain what is going on? The magnitude of the impulse seems only to be affected by how close or far the oscilloscope probe is from the high voltage lead. The scope isn't part of the setup. I'm only using it to see the affect the Hipot machine is having on the local environment.
The High Voltage lead radiates a tremendous (relative) electrostatic field and the O'scope has a fairly high input impedance the end of the probe represents an area or aperture of co incidence of that field and if the impedances are relatively close then the probe tip will intercept and couple' energy from the electrostatic field into the O'scope input... In other words the tip of the probe is an antenna receiving the energy radiated by the H.V. lead the same as any other transmitting antenna. Unfortunately so is every other I/O pin on the Uno that isn't used should be terminated either high or low. If the termination might possibly cause an over current condition a 470R resistor will limit the current to 10 mA. Max. There have been many reports of Arduino boards behaving strangely at times due to "Stray" fields coupled by the hands and while I don't know if it's an issue for your problem I am certain that it should be looked into. It occurs to me that any "telemetry" or I/O connected should also see close scrutiny as well. IMO
I was thinking the same thing but I couldn't believe there would be enough current flowing back into the tester (due to the internal shunt) which could induce enough current in the probe to develop such a high voltage.
As another test, I took one of the leads of my Fluke 87V multimeter and wrapped it around the high voltage lead of the tester, but without the conductors touching, and grounded (earth) both the meter and the tester grounds to the same point then charged the tester to 1kV and discharged. I had the meter setup to record the peak DC voltage, and it recorded 73 volts. The input impedance of the meter is nominally 10M?, so 73V / 10M? is 7.3uA does that seem right?
You cannot deduce current from measured voltage and meter input impedance since there is no direct current path. Apart from that the meter read period and update time doesn't necessarily coincide with the pulse time. What you are looking at is electrostatic transmission, probably in the RF frequency range - in effect a spark transmitter of 1900s vintage. It's possible that the RF pulse is interfering directly with the internals of the processor chip so even tying the I/O to either GND or VCC might not prevent the problem. Apart from metallic shielding, distance is the solution - signal strength reduces as the cube of the distance.
Wiring, proximity, power routing, and then add to that an Electro-Magnetic Pulse. You need to consider some shielding (possibly putting you circuits an top of a good ground plane) and cleaning up your wiring so as to reduce the currents induced in your circuits or getting some more physical seperation from the HiPot machine. At some point the HiPot machine is pulling some fairly high currents at lower voltages and those current are radiated out to other devices. The closer they are to the HiPot machine the bigger those induces currents will be in other devices.