Re: Linux PC mounted in a freezer

From: Floyd L. Davidson (floyd_at_barrow.com)
Date: 03/06/05 

Date: Sun, 06 Mar 2005 12:57:57 -0900

In Sane <insane@insane.insane> wrote:
>Actually, I really was thinking of a freezer, just because a refrigerator
>would probably not be able to cool strongly enough to combat the active and
>continuous heat supply inside.

Yes. Except a freezer's thermostat control may not work at
temperatures as high as you'll need. I often have lunch with a
refrigeration fellow, and if I can remember to ask next time I
see him, I'll get some definitive answers about what you can
expect as far as freezers vs. refrigerators. Offhand I'd say
either would work, and what you want to look at is the size of
the compressor.

As for temperature controls... there are ways around that. See
below, regarding temperature monitoring and control systems.

>Somebody commented about leaving it at it's highest setting so that the hard
>drives don't get too cold. I was thinking that even at it's lowest setting,
>it might not be able to take away enough heat from the power supplies to
>actually keep it cool enough.

No. That is the same as people who warm up a cold house by
setting the thermostat 10 degrees higher than the temperature
they actually want, thinking they'll get more heat faster with a
higher setting. It doesn't work that way because the thermostat
just turns it on and off, but doesn't adjust the intensity with
which it works when on.

In this case you want to set it to turn off at the lowest
temperature your equipment can tolerate. From what you've
described that necessarily is going to be several degrees above
freezing, otherwise you will get condensation and have a mixture
of water or ice forming on your equipment. Which is fine if and
only if the equipment is designed to operate under those
conditions, and fatal if it is not. For typical computer
equipment of course it is fatal. Somebody previously mentioned
38F...

And someone mentioned early in this thread that you should
research what overclockers do with cooling. Particularly, look
at some of the vapor phase or other cooling systems that can
actually bring temperatures down below freezing. You'll soon
enough understand the significance of temperature limits!
Cooling a CPU to something close to or lower than freezing means
waterproofing the socket and motherboard area. It's not easy or
practical.

>Also, I was hoping to put this in my garage, just like an air-conditioned
>telco equipment rack, or, even better, like a Liebert "Little Glass House"
>air-conditioned rack, or a similar enclosure that I've seen for sale by
>BlackBox.
>
>The Liebert runs around $10K.

Makes an old freezer look very nice, eh?

Perhaps at this point I should mention another possibility...
Move to Alaska's North Slope!

This has actually been explored commercially (I'm not sure if
the project is still alive or totally dead either).
Specifically, for a data storage facility at Deadhorse (Prudhoe
Bay). Power is cheap, even if you want to use megawatts. There
is a fiber optic to provide high bandwidth connectivity to the
Lower-48. The big plus though comes when you look at cooling a
large disk storage system (anything from a few gigs to many
mega-tera-whatevers). The outside temperature is commonly below
freezing (yesterday it was -25F here in Barrow, but today has
warmed up to -10F, which is typical for several months of the
year) and never gets above the mid 70's even in the summer.

Which is to say, cooling a massive disk array here is pretty
easy compared to anywhere else. Instead of air conditioning
with refrigeration, a reasonable bank of fans will do the same
thing!

(I'm sure that will cause you to check the real estate market in
Deadhorse... so I should also tell you now that they won't even
let you through the gate... ;-( And bandwidth is expensive
everywhere else, so consider this only if you don't need
connectivity...)

>> >Jose Maria Lopez Hernandez <jkerouac@bgsec.com> wrote:
>> >> In Sane wrote:
>> >>> I was thinking a freezer would be needed instead of a fridge, because
>> >>> the heat output of a 400W power supply would heat up the inside of a
>> >>> fridge pretty quickly.
>> >>
>> >> That's the problem. The inside of the freezer it's not
>> >> ventilated. You would need a ventilated system anyway.
>> >
>> >That would seem somewhat surprising. A freezer has to have a pretty
>> >potent heat exchange system to keep drawing heat out of it; that would
>> >surely continue to work...

It isn't an open (external) ventilation system, and there is no
need for one. There is a need to move air inside the box
though.

However, the power supply is indeed a problem that could be
cured in a number of ways, and one would be to have a closed air
intake and outlet system, where most of the heat from the power
supply could be blown out of the box without involving the
refrigeration system.

To whatever degree one wanted to lower the operating temperature
of the power supply, that could be done by automatic temperature
controlled louvers allowing some of the refrigerated air to be
added to the air system's supply duct into the power supply,
controlled by a temperature probe in the outlet duct. Note that
carried to the extreme such is system is just a very complex way
of doing exactly what happens if it isn't there... putting the
entire heat load on the refrigeration module.

>> >The thing I'd _expect_ to be a problem would be the possibility of
>> >condensation resulting from temperature gradients. I'd definitely
>> >want a fan running inside to help keep temperatures fairly constant
>> >inside the freezer.
>>
>> BTW - he wants a refrigerator, or the refrig section of a
>> combo, not a freezer. If there's a freezer section, he needs to fill
>> it with thermal mass like glad-ware containers of water, and not have
>> any equipment in it.

Regardless of whether this device is built in a refrigerator or
a freezer, it would be a very good idea to add significant
thermal storage.

However, even if this is in a refrigerator with a freezer
section, under *no circumstance* should any part of it be
allowed to get cold enough to freeze any of that water. Which
probably means the freezer section should essentially be
dismantled, or at least kept entirely empty. Put the heat
storage in an area that will have ambient inside the box
temperatures.

The whole idea is to *avoid* anything close to 0C, and having
ice in any part of the unit is definitely going to cause air
temperatures close to 0C (because of the high number of calories
required to convert ice to water compared to that required to
raise the temperature of the surrounding air).

>> He wants to set the stat on the fridge to it's highest (
>> warmest ) setting.

True. Another concept that should be looked at is a separate
temperature control and monitoring system.

For that purpose I would highly recommend looking at a CFA-633
from Crystalfontz. It has 4 fan controls and up to 32
temperature probes. Plus it has good fail safe features built
in that would allow rebooting the control system or shutting
down based on a watchdog.

The downside is that it would require custom software. I'm not
sure what is available for various platforms, but for example
I've written software that runs on Linux (or any Unix). It's
only about half finished, but the hard part is done (interfacing
to the CFA-633). Currently I'm not controlling anything with
it, but just monitoring several temperature probes. Adding
controls is easy. For a more complex setup than I am looking at
(just a water cooled dual cpu box) a separate I/O module would
be needed if multiple control outputs are required. But as is
the unit can control 4 relays with the fan controls, or if one
wanted to... those 4 controls can easily be multiplexed (with a
handful of diodes and resistors) to control up to 16 relays.

The unit comes with either a RS-232 or a USB interface.

  http://www.crystalfontz.com/products/633usb/index.html

The latest data *** is available at

  http://www.crystalfontz.com/products/633/CFA_633_1_9.pdf

Obviously a very complex control system could be developed,
and various fans and the refrigeration system itself can
all be controlled to provide the environment required. 

-- 
Floyd L. Davidson           <http://web.newsguy.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska)                         floyd@barrow.com