Re[2]: telehouse - 25 broadway

Unnamed Administration sources reported that Leo Bicknell said:

I think "limited" here is a bit misleading. Most of these engines
are continuous duity rated, which means within the bounds of
maintenance items (oil changes, gasket changes, and other routine
maintenance) they can run non-stop. Most gensets are essentially
the same components that power diesel-electric locomotives that
routinely run for weeks at a time without shutdown.

They may be continuous duty rated, but in this application?

I.e. - the ones I have worked on (to limited extents;
change filters, oil, repair FUBARs in the control systems
design [1] had two rating: N kw continuous duty; {say} 1.4N
intermittent. Intermittent may well be 24H or less; I
bet the 1.4 is a f(run time/week) or such.

Here in ice storm country every few years power is out for 3-4 days
in some areas after bad storms. Generators routinely run for 3-4
days and cause no concern over their condition, other than to move
up routine maintenance due to the additional run time.

I suspect dust is going to be the largest problem for all generators
in the area. It will reduce radiator effectiveness, wear on external
parts (fans and the like), clog air and oil cleaners. Again, these
devices are made for some fairly harsh environments so 48 hours is
no concern, I would think expecting 2 weeks of reliable power if
they are attended to even in these conditions is reasonable.

This is just as much a case of matching equipment to application
as is buying a new backbone router. Do you buy based on a "100 year
flood" peak; the "Mothers Day calling" peak or the "105% of >1hr"
model. Justify your decision the the beancounter. 25 points.

I do not know what all the Ma specs [Sean: is this p/o the FCC
regs, NEBS, or ??] call for, but I'd guess 2-3 days tops.

[1] Engines are big noisy devices. When you use high impedance
input logic (from slow crude mechanical sensors such as
oil pressure and water temp switches) that responds in a few
microseconds, and have no filtering debounce on them; duh, you get
lots of false alarms. Slow things down to a 500ms time constant,
and they work far better. Filter the power supply {Hello! The
generator just started -- yep, the line WILL spike...} well as
well, and hey, it does work after all. I hope other folks did
better design work than those guys.

Every time I've seen a standby generator sized for a computer data
center the 'continuous run' figure was used for the maximum loading.
Data center loading is generally fairly well known, if you have a
N KW UPS, it's known to draw N + M KW on input at full load, and the
generator needs to be able to supply N + M for continuous run.

Overrun amounts are _very_ limited. Typically < 30 minutes @10%
over, and < 5 minutes @20% over. I don't think I've ever seen one
with a rating for more than 20% over. Even a bean counter won't
expect a power outage to be over in 30 minutes.