[TowerTalk] OT: Generators and UPS's

Jim Lux jimlux at earthlink.net
Fri Aug 25 11:25:09 EDT 2006

At 06:55 PM 8/24/2006, Doug Rehman wrote:
>I realize this is off topic, but I can't find an answer elsewhere and
>TowerTalk has the best technical resources I can think of. (OK, I guess I
>can tie it to towers, because the generator will run the tower winch in the
>event of a power failure and the computer in the shack has a UPS and will
>eventually control the rotator...)
>I have a generator with an automatic transfer switch. In the event of grid
>failure, I have more than enough capacity to run the house, shack, and
>office (all co-located).

>Yesterday, the grid went down for about an hour. The generator came on after
>about 30 seconds; the UPS's carried the computers during this interval. Some
>of the UPS's began to switch on and off. The generator was running at around
>40% average load; the voltage fluctuation couldn't have been very much, as
>the line conditioners were not reacting at all. While I didn't monitor the
>frequency, in the past I haven't seen it change much.

We used to encounter this all the time running computers on location sets 
off a small generator. (Honda 6kW unit) We'd have UPSes, and they would 
spontaneously switch on and off.. (several brands, but mostly APC)  We had 
ONLY our computers and UPSes and some incandescent lights on the circuits.

My theory is that they use a "the line power has failed" detector to switch 
over (which according to spec has to be within 1 half cycle, 8 
milliseconds) that is a bit picky and sensitive.  One of the UPSes had a 
diagnostic port, and it kept kicking out "frequency out of range" messages, 
even when the frequency wasn't changing (i.e. constant load).  In fact, 
even when there were load transients (which does cause a frequency shift, 
as the governor on the genny readjusts the throttle) we *didn't* get trips.

One thing we did notice that's a bit weird, but this is anecdotal, and I 
never confirmed the explanation with further testing.  We had several 
switching power supplies running off the 120V, but not with the UPS (they 
were loads that could safely be interrupted if the generator shut down or 
someone kicked the plug out), suppling a few hundred watts of load (12 and 
24VDC, as I recall) at most.  We fould that if we put 200 ft of extension 
cord between the switchers and the generator distro panel, the frequency of 
UPS trips on the computers (which were at the end of another 100ft or so 
cord run from the distro panel) dropped dramatically.  My suspicion is tha 
the switchers were putting out a fair amount of hash at just the wrong 
point in the cycle (since their current waveform, lightly loaded, was just 
a little blip at the peak of the halfcycle).. that transient was enough to 
cause a slight voltage sag at that point, making the UPSes think that there 
was an incipient power failure.  (maybe the UPS just used a simple 
rectifier/capacitor scheme to measure Line In.. remember, it has to be fast 
enough so that the transfer switch picks up within 8 milliseconds.)  I 
never got around to hooking an oscilloscope up to see.

I would suspect that modern switching supplies might not cause this 
problem.  The last 10 years has seen a LOT of emphasis on good power factor 
on the inputs, because of the harmonic content rules for EU, and the 
"neutral wire heating" problem here in the US.  You just don't see raw off 
line bridge to capacitor input filter sorts of inputs any more, except in 
surplus stuff (because they can't sell it in a product any more, there's a 
huge number of old switchers available cheap in the surplus market) or in 
some very price sensitive stuff where they're hoping nobody looks too 
closely at the certifications.  Hams, being thrifty sorts, and tending to 
accumulate lots of stuff from surplus and hamfests for "future use", are 
probably more likely to run into these problems than others.

More information about the TowerTalk mailing list