Electrician Talk banner

THD weird reading

2K views 11 replies 4 participants last post by  JRaef 
#1 ·
I was called to a sub site that has been having some issues. We believe it got hit by lightning a few weeks back that toasted the APC ups unit, a server blade, another rack mount ups and a router switch. The server blade had a arc mark from the network port to the inside case of the server blade.

As this was a "while your over there job" i did not have the fluke power quality meter but i could access the meter on the main panel that also acts as a data recorder.

main panel is 480v Y

voltages range from 486 to 494
amp are around the 125 mark

Under power quality (thd)

a = 99.99 (max reading)
b = 3.84
c = 3.88

both b and c showed distortion in the 60 hertz range with tiny readings at higher hertz. A phase showed lots of readings in lots of different hertz ranges. (not all maxed but none showed zero)

There was also a lot of recorded events for power swells and dips.

Any idea why A phase would be showing this.
 
See less See more
#2 ·
The APS UPS's may have a setting for sensitivity to power quality, and THD is a factor the APC's use for power quality.

The distortion could come from drives ... might be the first thing to look at in an industrial facility. Switching mode power supplies in the servers could be an issue. According to this article, it could even be the rectifier on a UPS, I never heard that one before, but I wouldn't put it past APC to make a UPS that spoils its own power. They suck.

https://www.kohler-ups.co.uk/understanding-full-implications-harmonic-distortion/
 
#3 ·
Thanks for the reply.

whats got me stumped is the apc is 3 phase (not sure if its 480 or 208), the drives are 3 phase so shouldn't the THD be high on more than one phase.

The only single phase 277v load i can think of would be the lights which are old school T5 fluorescent.

Hopefully i will get back there Monday as i want to get a amp reading on the neutral and record a few swells and sags on the fluke to see whats really going on.

I think there's more to this buildings problems than a high THD on A phase but its something i have to look at.

Corroded ground cables (lightning conductors), A missing ground loop (according to the prints its meant to be there) and unstable voltages should keep me busy for a while.
 
#9 ·
Could also be that your blade server power supplies in the rack, which will all be single phase, were not evenly distributed, meaning they are all between A-B and A-C, none connected B-C. So A phase is doing "double duty" with no cancellation effects. Should be simple to check.


I once helped Google with their own blade server design (they build their own in house). Their brilliant propeller head EEs had no clue about designing for balance, they were not power guys. This is exactly what they did, they thought they were saving on wire by not bringing all three phases into the distribution box of the rack. That still might have worked had they evenly distributed racks across the total system, but no... they designed ALL of them exactly the same, phases A and C only. To give you an idea of how bad that can end up, here is a photo inside of one of their server farms. Each rack is 90A at 208V (now 3 phase).




Yes, that is a Storm Trooper in the photo by the way, inside joke from Google engineers...
 
#11 ·
I managed to connect to the Eaton panel over the network from a different site today and im now 99% sure that panel is glitched and the data is wrong. It looks like the lightning strike probably took out more than we originally thought.

Hopefully have a EE out in the next few weeks that is going to do a lightning/ground study and see what recommendations he makes. Still going to hook up the fluke and see whats going on especially with the swells and sags.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top