Random shutdown of high bandwidth DVI displays due to poor grounding

Yeah, that’s a long title but it says it all.

I’m not an electrical engineer and what i discuss here was an interesting learning experience, at the cost of customer satisfaction – on what happened to be something due to customer’s electrical installation fault.

Comments are appreciated. It’s always awesome to learn something from the community.

A large videowall was commissioned a few months ago. It’s a very typical installation without much to it – 5×3 configuration, each display is FullHD. Issue was simple by itself: from time to time (sometimes per hour, a few times a day… depends on.. stuff?), each of the splitters would randomly lose input signal. Catch: only a few of them would not display any symptoms and nothing was at all obvious.

TL;DR: it was the UPS’ fault!

The wall is fed via a NVidia M4000 graphical card, with each DP port being converted to DVI-DL, and feeding a DVI splitter. There are 6 splitters for this videowall.

Things we did

  • Swap of one splitter + DP converter
  • Swapped some cables around
  • Your usual FW + SW upgrade on everything
  • Prayed to multiple religions and beliefs.

This was promptly escalated to our supplier and after a few weeks of site visits, lab tests et al, they did two similar setups with a very small difference: only one of them was properly grounded. The issue manifested itself almost immediately on the ungrounded system.

Sure there were tests on the customer, but only using a multimeter – not the right tool for the job, but it’s not like everyone has a ground resistance tester laying around, right?

The tool for the job was a Kyoritsu 4120A. Testing non UPS proved boring:

All readings were stable around 0.5 Ohm mark – between 0.48 and 0.51 Ohm to be exact.

But UPS circuits were rather interesting. First measure: 50 Ohm. A tad high..? Second test: 75 Ohm.

3rd test: infinity. 4th test 45 Ohm.

OL resistance = no ground is present!

Clearly something was off and looked like we had a floating ground when the UPS was under load (but not on battery power – go figure!).

Setting the UPS on bypass brought back expected values below 0.6 Ohm, and, imagine that: problem went away.

Conclusion: UPS was messing with electrical ground – probably leaking a lot of stuff in it, changing potentials over the instalation. Don’t ask why. Despite it’s a large 36KVA unit, it’s some cheap OEM with no insulation transformer.

This whole thing reminded me of this great article at hackaday testing ESD effects on a transistor – somethings aren’t binary and there are some inconsistent states that WILL lead to inconstant results. That’s why there are rules and regulations!

Leave a Reply

Your email address will not be published. Required fields are marked *