Server Room AC

[ChrisConley]

How much heat should I expect to be rejected from a server rack rated at 7.2 kVA?

I converted kVA to kW for my initial calculations, but is it reasonble to expect 100% of the load will be rejected as heat into the space?

 

[willard3]

It's probably conservative to consider the diversity to be 60% or ~4.5 kva or 15.3 mbh. 1-1/2 ton unit is adequate.

 

[tombmech]

A couple of things:

1. I have not found diversity to be a dependable assumption regarding most computer equipment. I doubt that the electrical engineer will take that much diversity on the servers' electrical load.

2. You cannot simply size an A/C unit or coil with a one-to-one relationship of total Btuh. Except for specifically designed computer room A/C's or AHU's, every A/C unit and coil on the market will have a Sensible Heat Ratio of approx. 0.65, give or take a few points. That means to handle the 100% Sensible load of the servers, you must size run-of-the-mill A/C units at 150% of the total Btuh.

In other words, a 1.5 ton unit will only provide approx. 0.975 tons of Sensible capacity.

 

[designbuild]

Tombmech you make a good point on the diversity, especially regarding computer rooms. Generally the people setting up servers want to squeeze as much performance out of them as possible. This means that the equipment will run a lot closer to the rated output.

However I don't agree with your statement about sensible heat ratio. It is actually the reverse of what you are saying. That is a coil's Latent capacity is limited by the sensible heat ratio. It is prefectly reasonable given the appropriate entering air conditions that you can have any tonnage of total cooing be all sensible. However any specific coil is limited in it's latent capacity. If you look on your psych chart you can see that at 75 F/ 45% RH you can very easily leave the coil at as low as ~50 F without doing any latent cooling. Hence your total btu would be all sensible. However if you're entering at 75 F/ 70% RH you can see that a large portion of the cooling will be latent to get to a leaving coil condition of 55 F. This is were the SHR comes into effect because it is an easy one number check to see if a particular coil selection can handle the latent load.

 

[walkes]

Chris,

I have been doing some research on the server racks lately. Some of our clients measure the amp draw on their existing equipment. The power associated with the amp draw is equivalent to your sensible cooling load. The amp draw rate can be as little as 40% of the nameplate rating. However, newer technology such as blade servers can reject heat at 100% of the nameplate rating. (A fully loaded blade rack can reject 16kW of heat)
If you have a local Liebert or APC rep they should have current information as to the heat rejection from server equipment. I know Liebert has a test facility where they test and monitor various equipment loads.

 

[tombmech]

designbuild,

No offense, but you miss the point entirely. That is not how A/C units' (not computer-room type) capacities are cataloged. If you select an off-the-shelf A/C unit (or a coil) for a total Btuh, you will not be buying all of that capacity in Sensible load. Your unit selection will be busy drying and dehumidifying the last bit of air while your computer components overheat, instead.

Further, even if you size a custom chilled water coil so that you "control it" within a sensible range - unless you are purchasing a specifically custom-made, high SHR coil - you are purchasing it for a reduced capacity. That doesn't even count the fact that you are implying a super-sophisticated set of controls to measure entering air conditions and compensate by adjusting the delta T accordingly - just to ensure that it never enters into a dehumidifying regime.

At 1 1/2 tons, it is most likely that a DX refrigerant coil would be used instead - and there's no way of making it work that way.

Finally, why would companies such as Liebert go to great lengths to specially design A-Frame coils, custom units, custom controls, etc. and form a special "niche" market if it was really that simple?

 

[ChrisConley]

One additonal point on the sensible heat issue.

I'm considering the XD (eXtreme Density) solution for the project. R-134a is pumped around to rack mounted fan coils which draw hot air from the servers and cool it down to ~68F. Due to the temeperatures of the refrigerant no condensing occurs on the coils and the coils are rated at 100% sensible.

Having said that I will also be using an air handling system to heat, humidify, cool and dehumidify the room.

 

[tombmech]

Now, that sounds cool!

 

[designbuild]

Tombtech
If you select an off-the-shelf A/C unit (or a coil) for a total Btuh, you will not be buying all of that capacity in Sensible load. Your unit selection will be busy drying and dehumidifying the last bit of air while your computer components overheat, instead.
My whole point was that it depends on your specific conditions. Depending on those you very well could have a 1 1/2 ton selection (120 Ton for that matter) that operates as a dry coil and has no latent capcity. Look at any manufacturers catalog and check for the astricks denoting a dry coil condition. A particular tonnage, and assoicated coil has an absolue SHR but it works the opposite of what you stated. Meaning you can't go beyond that point and get additional latent capacity however you can get additional sensible capacity. Basically your total capacity stays fixed. Your sensible capacity can vary anywhere from total capacity to a number such that your sensible and latent equal the total load and the ratio is the coil's SHR. It is all dependent on the entering conditions and airflow. (in real life the total will vary some but it is a relatively small amount)

Further, even if you size a custom chilled water coil so that you "control it" within a sensible range - unless you are purchasing a specifically custom-made, high SHR coil - you are purchasing it for a reduced capacity. That doesn't even count the fact that you are implying a super-sophisticated set of controls to measure entering air conditions and compensate by adjusting the delta T accordingly - just to ensure that it never enters into a dehumidifying regime.
Once again there is no special control or coil required if you have a 18,000 BTU sensible load with no latent component. Very likely in a computer room see my example in my previous post.

At 1 1/2 tons, it is most likely that a DX refrigerant coil would be used instead - and there's no way of making it work that way.
See above, there is nothing special about making it work that way. If your conditions are such that you have no latent load it does just work that way. No special controls required.

Finally, why would companies such as Liebert go to great lengths to specially design A-Frame coils, custom units, custom controls, etc. and form a special "niche" market if it was really that simple?
They specially design all of those things because computer rooms tend to have extremely tight tolerances. I have designed many a computer room that used a standard split system dedicated to it, and there was no special coil selection. I could do this because I had the space and the owner did not have extrememly tight requirements for the space. However, the need for a leibert or similar arises in all the extra things that a computer room may require that a standard unit can't provide. Extremely tight temperature control, within 1/2 a degree. Extremely precise humidity control, +/- 1%. 99.9% reliability with auto failsafes, alarms, and redunancy control all without needing an extertnal control system. Specialized cabinets are required to fit computer room floor plans, ceiilng spaces, etc. Specialized distribution plenums (under floor, etc). All requiring special coils, not due to loads but due to cabinet sizing.

 

[willard3]

You know, the days of extremely close wb and db control are really over for computer rooms, either that or the original poster neglected to include that information.

I have designed many computer rooms with Liebert units, but new stuff is much more robust.

I am probably wrong on the 60% diversity and I once challenged a client to demonstrate that much diversity with computer loads.

The tonnage on my earlier post assumed a high shr which is why it's oversized. You're going to get a 1 or 2 row coil anyway which has inherently higher sensible cooling.

 

[ChrisConley]

Another point to add is that I've actually got 44 of these server racks in a very small space - which is why I'm considering the XD solution. The client has decided that they don't want to use a raised plenum so the only solution I could find to provide ~75 tons of cooling in 800ft2 of space was the server mounted fan coils system.

It was high load density, not tight control, that pushed me down the Liebert path, although I have also used Liebert's in tight temperature control areas.

 

[tombmech]

designbuild,

We disagree. What you say is true as far as it goes - "It is all dependent on the entering conditions and airflow." However, the only thing fixed about the entering conditions and airflow are your peak load calculations - that's it.

Under all conditions, the coil will drive toward saturation with Sensible cooling. Depending on the physical configuration of the coil, the latent turn down will occur as the cooling nears saturation. Those specific points may vary, but not the direction: Latent cooling.

Trying to limit that effect under all combinations of entering conditions is exactly the point. However, since the typical computer room has -0- latent load, you are starting off with a huge disadvantage - more accurately, an impossible mis-match.

Put simply, in small applications of 1-3 tons or a little more, you can get away with a somewhat oversized "normal coil" unit - so that the coil never gets very close to saturation. You will probably find that the differences are not enough to cause problems. That is essentially the same as recommending to size a unit by its sensible capacity. Too bad I tried to offer an explanation as to why one would do that.

As far as your argument about Liebert units, a high sensible coil is the first requirement for the specialized application. The other features are options, not product definitions. Even so, if you claim "precise control", one might ask, "How is that precision obtained?" It is fundamental, because the coil is designed to match the process!

Finally, no offense - but if you have used standard split systems for computer rooms, they can't have been too big. As with the rest of this discussion, for smaller applications of 1-3 tons, that is entirely possible. Slightly oversizing for a safety factor has the almost the same effect as sizing it for a sensible range, only. I have done it on many occasions myself.

That doesn't refute the underlying principles, or mean that you wouldn't have problems as you went to 10, 20, 30 tons and more.

Good luck on your future projects.

 

[UtilityLouie]

I think that everyone's talking about the same thing here.

You can't just pick a coil based on total capacity.

You need to do your load calculation.

You need to determine from a psyc chart what your true SHR and sensible and latent loads are (in my industry - we need pressurization air in these rooms - so I have to add in scrubbed OA to my calcs).

You then specify a unit to meet these conditions and double check equipment selections from a vendor - Liebert, York, Carrier, McQuay, etc.

Then, you install the unit and hope that no one was lying to you when the told you the loads.

These low latent rooms are a little more persnickity than standard HVAC (especially with high air flows in small areas) - but standard rules of selection apply.

I always count on my reps to give me good selections and if I give them enough of the right information - they do.

 

[CarolinaPE]

Does anyone know of good information as to what to estimate for cooling load for transformers? For instance a 10 kva transformer?

 

[walkes]

An electrical engineer once gave me a handwritten chart with the percent heat rejection for transformers. For a 10kVA unit it varies from 2-4.5% depending on the type of transformer.