Protecting 5V on Computer from Spark / High Voltage

[Zilenc3]

I have a system which has a spark plug that gets generated from a transformer (12Vdc -> xx kV). Each time the spark hits the ground of the DUT (that has the HV transformer) the earth ground 'shakes'.Our monitor flickers/freezes and our USB devices fail and need to be reinserted into the PC to come back alive.

We've reduced this problem by adding varistors from DUT PSU to earth ground. This has helped significantly, but still we haven't solved the issue. The spike on the unit with a 47V varistor. 47V was picked because the 6528 we use is good for 60V. However we're seeing spikes on the 5V of the PC, which powers some of our virtual COM sensors, keyboard, mouse, etc. 

The spark is connected as follows:

HV transformer ---> Spark plug ---> DUT Ground

*EARTH  ---> Varistor ---> DUT Ground

*EARTH ---> Varistor ---> 12V

*Note that this connection is doubled, one at the terminal block by the DUT and another at the PSU supply side.

Have any of you had to deal with such an issue? If so how did you reduce or eliminate the problem? Ultimately we need to stop the surge from spreading, but I'm not sure how. We are going to buy a smaller varistor to pull down to 15V, but we're not certain that will solve it.

Current Varistor : https://www.digikey.ca/product-detail/en/littelfuse-inc/V20P30P/F11293-ND/4285788

Thanks.

[AnalogKid2DigitalMan]

Sounds like fun (but not fun at the same time).

I believe Varistors have a limited lifetime due to repeated shuntings of high voltage, perhaps high speed clamping diodes would be a better protection means.

I would also suggest getting an opto-isolated USB hub that is powered by its own wall wart to at least protect the PC from transients. Though that will not protect any USB devices downstream from it.

Probably not wise to float the DUT from earth ground either for safety reasons.

Note that indirect conduction of voltage to the other equipment due to electromagnetic inductive or capacitive coupling could also be at play here. I would try to shield the DUT and the measurement equipment separately as best as feasibly possible.

A good trouble shooting aide would be an oscilloscope to judge the effectiveness of each measure of protection.

Been a long time since I have worked with high voltage and discharges, I am sure Henrik will follow up with some good advice- after all:

“ground” is a convenient fantasy.

-AK2DM

[Zilenc3]

I forgot to mention that this PSU is connected to the PC by GPIB, which when disconnected proves to help, but not solve the issue. After I made this post I measured a +/-15V spike on USB power from the PC, even after that USB sensor is disconnected. I'm not sure if an isolated hub would do the trick, but we did consider that too. 

We have air gaps as much as possible with the spark travelling through cables rated for 30kV. We do notice more issues when we couple against USB or other DUT wires. So we moved it as much as we can. 

The system does run for ~3h or more before failing. 

I will ground the DUT to earth and see what happens. As of now when we measure with the varistor it's keeping the +12V (DUT PSU) limited to +/- 47V when it sparks. 

[AnalogKid2DigitalMan]

Here is a GPIB Isolator unit from NI:

http://sine.ni.com/nips/cds/view/p/lang/en/nid/203221

Good luck and be careful!

-AK2DM

[Henrik_Volkers]

Really sounds like fun 🙂

I wonder .. if there is a lab with such a serious HV PSU there should be people around that know how to handle such pulses and problems.

For high dI/dt (and dU/dt, just what HV pulses provide) a low impedance  current path needs to be provided.  I assume the HV PSU is a good one... (type? link to manual?) and should provide a good decoupling to the GPIB.  (but all of no earthly use if no good grounding (low impedance current path) is provided) 

A fast (analog?) scope and a shoke (current) probe (can be DIY, ferrite clamp on with some windings and a 50Ohm) can help 😉 if you measure voltages, I would always start at least with a 10:1 probe... 

Spoiler

I love my TEK 7844 with a P6015A probe

• low impedance GND and EARTH to the HV-PSU and DUT... keep the loop(s) as small as possible!
• some ferrite with some windings for the GPIB (such a bigger ferrite can be harvested from an old CRT monitor ) 
• Some more line filtering for the PSU ?
• shield the DUT (may incl. the PSU)  for the RF
• measure differential, add ferrite shokes ... BTW what sensors?

[Zilenc3]

We're using a LED sensor: http://www.optomisticproducts.com/solutions/universal-lightprobe-sensors/spectra-usb/

Keyence PSU: E3632A 

USB Adapter: https://www.digikey.ca/product-detail/en/3628/1528-2460-ND/7931507/?itemSeq=291526646

TVS: https://www.digikey.ca/product-detail/en/SA5.0A/SA5.0ALFCT-ND/407681/?itemSeq=291527380

Varistor for 12V to EARTH: https://www.digikey.ca/product-detail/en/V18MA1A/F2271-ND/485483/?itemSeq=291540872

Choke: B82144F2222K000‎, B78108E1101M000‎, or 5250-RC‎

I may just move the analog measurement version to save myself some trouble. We've been using the clip-on ferrite beads wherever the DUT wires cross non-DUT wires. Today I moved the beads to the USB line to and from the USB hub (that hosts the device). The other devices on the hub seem to be more sturdy than this LED sensor and don't crash whenever this does. I'm going to add a choke in-line to the DUT power and see how that goes. I'm also going to add a smaller TVS on the PC USB port using an adapter.

Earth is 16AWG while spark is 18AWG, and it measures pretty well.

The GPIB cable is a double shielded X5 cable. If you think a ferrite bead would help I'll add one.

Unfortunately we can't shield the DUT as the system was made without that in mind. Could you clarify though why you think we should shield the PSU/DUT? 

P.S. We've shorted DUT to earth, but no luck there.

[Henrik_Volkers]

Sorry I missed the HV transformer in my answer ... and you didn't noted it in your type list.

[Zilenc3]

So I tried adding an inductor in-line and that messed up my test sequence in terms of timing so I had to remove it.

I tried adding the 18V 40A varistor with no useful gains (pulled the spike down the 18V, but still the USB hub was crashing). I added to +12V to earth and shorted earth to 12V_GND. 

What seemed to solve it was a powered USB Hub, specifically: http://ca.dlink.com/products/connect/4-port-usb-2-0-hub/

I tried finding the power adapter for my existing Tripp-Lite hub, but couldn't find it. I decided I wanted to see if a powered HUB would do the trick and I ran overnight and my mouse and keyboard was in-tact. Now these were known to be the most durable USB devices in comparison to the LED sensor. I removed that for the time being because I thought it was a goner. It couldn't seem to recover even when sparking was off. 

I tried it again and it's operational so I'm going to plug it back in with the powered hub and see how it goes. 

[Zilenc3]

No luck using that device. It still drops-out. I'm going to switch to the analog measurement instead. 

Thanks for the suggestions.