[time-nuts] UPS for my time rack

Poul-Henning Kamp phk at phk.freebsd.dk
Tue Oct 13 14:54:58 EDT 2015

In message <4FD0F30EBAEF49609DF207E3EE61C15B at pc52>, "Tom Van Baak" writes:

>I used to rely on one massive UPS (along with natural gas generator)
>for my entire lab. Eventually I found it more reliable and convenient
>to have localized power backup. By local I mean backup for a single
>shelf, or even a single instrument.

The big gain is avoiding the DC->AC conversions.

AC->DC conversions are covered under EnergyStar and similar programmes,
so they're generally 90% efficient or better.

But DC->AC conversions, for instance in inverters and UPS's are not
covered, the argument going that they are only run very seldomly,
and therefore capital cost would be wasted.

In the professional segment, UPS's which require a forklift,
efficiencies are good, in some cases very good, because power costs
money at that scale.

But anyting you can fit into a rack will typically have horrible
losses, the smaller the UPS the worse.  I've personally measured
sub 50% efficiency in one case.

The argument against running stuff on 12 or 24V DC is the short
circuit currents, and the absense of an affordable standarized

For the short circuit currents the only cure is fuses and caution,
and for connectors there seems to be no hope of a standard - ever.

China forced USB through as the standard for mobile phones, but
despite several valiant attempts nobody has ever managed to get
anything above 5V/5W standardized.

Here's the website of the IEEE WG which came out with a standard
(IEEE 1823:2015) this May:


It will cost you $160 to see the full standard, which is a very
good and strong reason why adoption will be slow, and nobody
I've talked to expects it to go anywhere ever.

At the same time USB has come up with 100W power concept
which is not compatible, since IEEE uses CANbus and different

I've not heard any rumours that China man nail this one.

Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

More information about the time-nuts mailing list