temperature controller

Racking fun

Saturday, August 11, 2007 

fans

IMG00065.jpg

 

The rack system we use to house our servers….

 

The house is not air conditioned and the rack is effectively in the garage, so it needs both dust filtering and self-adjusting cooling to compensate for summer daytimes without being insanely noisy all the time. The fans have taken a little trial and error to get configured right. They are connected to this industrial controller that turns on fans sequentially as the temperature rises.

 

I just got the door fans repaired after a minor accident that sent one blade flying, and came to two observations:

 

  1. The fans are loud.
  2. I’m not optimally supporting the designed cooling system for the servers.

I visited a massive colo facility a while back and they went on about using CFD to calculate the flow rates through their cages to keep 20kw racks from melting down (about 15 space heaters inside a box) and one thing they mentioned was sealing all the gaps between the front and back of the rack. Clearly I have not done this. The bit I didn’t really think of is that the rack isn’t so much a chimney filled with rising hot air as a jet with the intake at the front and the exhaust at the back. The fan system needs to enhance that jet flow, not some chimney effect with hot air exiting out the top.

 

That said, normally racks are put in environments where dust isn’t a problem. So I need positive pressure fans to pressurize the front of the rack through filters and compensate for the pressure drop through the media. That should provide plenty of clean, ambient temp air for the server fans to blow through the rack (and provide enough air that the server fans don’t create negative pressure in the front and draw dirty air in though other openings in the front half of the rack).

 

If that positive pressure turns out to be inadequate, then the temperature controller will turn on negative pressure fans at the back of the rack to exhaust the heated air, which will necessarily increase the flow through the servers and presumably drop the pressure in the front of the rack below ambient and thus, worst case, draw in a mix of dirty and clean air. The compromise is that a bit of dust is better than overheating on hot days.

 

The temp controller is currently set to turn on supplemental cooling at 85F (inside the rack temp). Since ambient hits 85 on the worst days, that seems like the lowest reasonable temperature to set the controller to.

 

In the end, if that’s not enough, then a ton of chilled water should do the trick, but I’m hoping not to have to go there…

Posted at 14:15:39 GMT-0700

Category: phototechnology

Inverting SSR

Sunday, July 22, 2007 

IMG00035.jpg

This is part of a cooling system for the cabinet. This relatively inexpensive (on ebay) Watlow series 900 temperature controller has a single output and two alarm relays. The output and one of the alarm relays have snubbers inside and can trip fans on their own. The second output was NC and needed a relay to invert it. I used a solid state relay (SSR) to do it and it is currently controlling the halogen light. It turns on if the temperature goes over 81. Woo! Three more simple wiring harnesses to fabricate and then the whole mess to install in the cabinet. The concept is that a set of filtered fans will run full-time to drive fresh air into the cabinet (not quite working yet) to achieve ambient temp, say below 85 or 90 degrees. If the temperature goes above that, a single extraction fan will turn on. If it goes up another 5 degrees a second extraction fan will turn on, and 5 degrees above that a third, honking big fan will turn on. The extraction fans will probably suck in some dust, but better that than melting down.

Posted at 17:10:15 GMT-0700

Category: phototechnology