cooling a small server room

[serverroomcooling]

Input Voltage: 209.6 VAC
Input Frequency: 60.00 Hz
Maximum Input Voltage: 209.6 VAC
Minimum Input Voltage: 209.6 VAC

Output Power Status



Output Voltage: 210.6 VAC
Output Frequency: 60.00 Hz
Output Current: 21.3 Amps
Output Watts at n+0: 055 % Watts
Output VA at n+0: 037 % VA

[serverroomcooling]

the temp probe which is connected to my UPS has been constant between 81.50 and 81.95 for the past 4 hours. I get a reading every minnute.....this tells me that the temp is pretty stable in there....I wish I remeber exactly where the probe is....I know it'd somewhere in the rack in the back of the room

[serverroomcooling]

ok here it is....with the outside temp of 77 and relative humidity outside of 84% here are my on/off cycle information from this morning with the tstat set to 68

10:05 cycle off temp reading 68
10:10 cycle on temp reading 72 (actually hit that about 10 seconds after the handler turned on

10:18 cycle off temp back down to 68
10:21 cycle on temp read 71
10:33 cycle of temp read 68
10:36 cycle on temp read 71
10:48 cycle off temp read 68
10:51 cycle on temp read 72

here is the temp probe information from my probe that is in the rack at the back of my room (the one which contains my UPS and a bunch of server:

looks to be pretty stable

Time TIambF
10:57:56 81.50
10:56:56 81.50
10:55:55 81.95
10:54:55 81.95
10:53:55 81.95
10:52:55 81.95
10:51:55 81.95
10:50:55 81.95
10:49:55 81.50
10:48:55 81.50
10:47:55 81.50
10:46:55 81.50
10:45:55 81.50
10:44:55 81.50
10:43:55 81.95
10:42:55 81.95
10:41:55 81.95
10:40:55 81.95
10:39:55 81.95
10:38:55 81.95
10:37:55 82.40
10:36:55 81.95
10:35:55 81.95
10:34:55 81.50
10:33:55 81.50
10:32:55 81.50
10:31:55 81.50
10:30:55 81.50
10:29:55 81.50
10:28:55 81.95
10:27:55 81.95
10:26:55 81.95
10:25:55 81.95
10:24:55 81.95
10:23:55 81.95
10:22:55 81.95
10:21:55 81.95
10:20:55 81.50
10:19:55 81.50
10:18:55 81.05
10:17:55 81.05
10:16:55 81.05
10:15:55 81.50
10:14:55 81.50
10:13:55 81.50
10:12:55 81.50
10:11:55 81.50
10:10:55 81.05
10:09:55 80.60
10:08:55 80.60
10:07:55 80.60
10:06:55 80.60
10:05:55 80.15
10:04:55 80.15

[serverroomcooling]

the internal temp on my UPS unit, which is an APC Symetra 1200, in case you were interested used to get as high as 117....now it is stable at 101

I would call that progress

[serverroomcooling]

with Christmas right around the corner, I don't think I'm going to bug him all that much to come by. He reassures me everything will be fine and that a 12 minute on, 3 minute off cycle will be fine and my unit can certainly handle that, and I'll take his word for it.

In the meantime, I still want to gather those readings for the thread....what equipment do I need and how do I gather the readings?

maybe I'll bring in a third party to take those reading for me

[john dalton]

Dear Serverroomcooling,

First of all, thanks for continuing to update this forum with your posts.

I’ll try to answer your questions:

The reading of 26 amps from both your power lines upstream of your UPS is the amps of the system……26(one not both readings). The electron flow (electricity) is actually moving through both lines in an alternating cycle, flowing on way and then cycling 180 degrees and flowing the other way, this happens at a rate of 60 times a second, hence 60 cycles.

From your data it would appear your current computer equipment load would be as follows:

210.6 VAC x 21.3 Amps = 4,486 watts

In addition the UPS is consuming the following:

(209.6 VAC x 26 Amps) - (210.6 VAC x 21.3 Amps) = 964 watts

Therefore the total equipment load in the room is as follows:

4,486 watts + 964 watts = 5,450 watts

If we wanted to convert that to a BTU, then we would do the following:

5,450 watts x 3.412 = 18,594 BTU/Hr of sensible heat

If we wanted to add the latent heat factor, we would do the following:

18,594 x 1.4 = 26,032 BTU/Hr

If we wanted to find the “approximate” structure and light load at 77 degrees F / 86% RH(Current outside conditions), we would do the following:

36,000 BTU/Hr(system cooling capacity) x 80%(Percent of on time for the compressor)

= 28,800 BTU/Hr - 26,032 BTU/Hr(computer and UPS equipment load) =

2,768 BTU/Hr.

Therefore the entire heat load of your server room under the existing conditions is 29,800 BTU/Hr = 2.4 tons of cooling.

Based upon the lack of temperature fluctuation in your UPS probe I’m assuming this probe is monitoring a function of the UPS electrical circuitry rather than an actual air temperature and should therefore be used for comparison ratings only.

You should obtain an “AIR” temperature probe that could monitored over the internet along with data logging capability(if your budget allows).The placement of the probe could be in the return air grill of the cooling unit, thereby sensing an over average of the server room temperature.

Based upon your cycle on and cycle off times the overall cycle rate of your system is 4.0 cycles/Hr. Industry standards are 4 to 6 cycles per hour, therefore this cycle rate is satisfactory at this time….it will change though based on internal and external forces. I am concern on the off time for the system.

Respectfully Submitted,
John J. Dalton

[john dalton]

Dear Serverroomcooling,

The off time for the compressor does raise my eyebrow at this time though. Currently the off time of three minutes boarders on the lower side of the industry rate of 3 to 5 minutes. The off time is important for the system because the suction and discharge pressures need to equalize as close as possible and the start windings of the compressor need to cool down satisfactory for continued reliable starting cycles of the compressor through the course of the day(and night in this project). As the outside temperature increase, especially during times that the building’s HVAC system is not running, and as you operate more computer equipment, then the off time is going to decrease accordingly. I think this three minutes off time has the potential of decreasing too low.

At this time an anti-short cycling timer would be a great component to think about. This would be a delay on break timer that would prevent the compressor from cycling too quickly. This feature may already exist in your system though. Some higher end thermostats have this anti-short cycling circuit built into them. If you give me the make and model of your thermostat, I’ll check in our data base to see if this feature is included. Then all you would have to do is make sure your contractor configured the thermostat with this feature on.

The best person to obtain the commissioning readings is your HVAC contractor. Almost all of these reading and measurements are needed by any contractor to ensure the correct operation of your split system. Some of these reading, such as the refrigerant pressures, can only be obtained by someone with a type II or type UNIVERSAL EPA 608 certification card.

Hope to hear from you soon.

Respectfully Submitted,
John J. Dalton

[serverroomcooling]

the short off cycle bothers me as well, but he does not seem bothered, he says this system can be on for years with no problems.

If I am uderstanding you correctly, I should be making sure that the system is off for close to 5 minutes before turning on. From looking at my thermostat, I don't think it has the feature built in, it looks kind of cheap.

If I cool the room down to 65 degrees, and tell the thermostat to kick on at 70 degrees, I think that will yield the 5 minutes off cycle. But cooling back down to 65 will probably take upwards of 20 minutes, which would take us out of the 4-6 cycles per hour, it will be between 2 and 3 cycles per hour, is that satisfactory??

The probe I have I believe is an air probe. It has a serial connector and is a 1x1x1 box with slits in it, it also reads humidity....I do not believe it has anything to do with the electric current.


Here's the next wrinkle, now that the AC is working, I will be putting my racks back together. I will be putting the walls and doors back on. What this does is seals the racks, and this allows the fans to suck out all the heat through the side panels of the rack, and up toward the ceiling. Will this change anything?? Don't all shoot me at once

[serverroomcooling]

I put the doors back on the racks, and now the cycle appears to be 10 minutes on, 4 minutes off....seems a bit better....right??

the temp probe seems to be going down as well....looks like it is hovering between 79 and 81 now....I'll let you know after a few hours of monitoring

[twcpipes]

With very little ducting, the blower may have very little static pressure to work against. Consequently the system may be doing more than, say, 400cfm/ton. If this is the case, slowing the blower and reaching a new balance point of recirculated air temperature and load may well extend the run time. But it is not acceptable to be under 350cfm/ton.

If a variable speed condenser fan control was installed, the setting of the controller can be reduced to a lower pressure decreasing the expansion valve tonnage which will increase run time. The expansion valve superheat can also be adjusted to a higher value.

Off time can be extended if the thermostat can be set with a wider on/off range such as 67off/75on, as an example. The maximum "on" can be checked by measuring the heat from the equipment in the room and derived at by choosing a maximum value the equipment can safely operate at.

If "off" time cannot be extended, insure there is a start capacitor for the compressor. This will extend the compressor life as it is diminishing with so many starts per hour.

It is all a matter of proper set up of the system to insure you have the proper cooling but with the minimum "wear and tear" on the compressor so you don't experience a burnout down the road.

Sidebar: Comments were made about the difference between Hollywood, Florida, Hollywood, California and San Diego. This is quite incidental as a project is always designed for the area the project is located. Also, being a low rise 7-story building I would hope the relative humidity is under control with the limits of the envelope or the building has a considerable problem being wet. Thus the relative humidity is less of a consideration given said RH is in fact controlled by the main unit(s).

As the tonnage is always an ongoing consideration, using a condenser blower controller to keep the head pressure up allows the expansion valve to always see its design PD as related to its design tonnage. It is always a critical factor to use an external equalized, adjustable valve to fine tune that tonnage.

For what it is worth, I spent eight years in the Dallas, Texas area designing projects so I am very familiar with relative humidity designing Day Surgery, MRI/Cat Scan, server rooms and hospital facilities.

With all that said, thank you for entertaining the above post if you got this far down.

[john dalton]

Dear Serverroomcooling,

It’s better for the compressor for the off cycle to be five minutes, but I agree with your HVAC contractor that if, and I mean if, the off cycles stays at three minutes, then the compressor should run satisfactory. It’s just that, as I pointed out in my previous post, if the cycle rate falls below three minutes then you’ll have a problem.

If in fact to do have an “economical” type thermostat, then you wont have the option to program the “operational span” of the thermostat(the difference between the on and off events of the thermostat).

The 4 to 6 cycle rate is the industry standard for an office HVAC system and is only an indicator of where “normal” is. If you had the ability to increase the operational span of the thermostat to 4 or 5 degrees, then a 2 to 3 cycle rate would be totally acceptable.

In regards to the following statement:

“You should obtain an “AIR” temperature probe that could monitored over the internet along with data logging capability(if your budget allows).The placement of the probe could be in the return air grill of the cooling unit, thereby sensing an over average of the server room temperature.”

I think the temperature you’re reading in the UPS is an operational temperature of some portion of the machine rather than of the surrounding ambient air. I base this on the fact that according to your readings the ambient air temperature in the room travels from 68 to 72 degrees during your cycling monitoring, giving you a 4 degrees delta T. The travel rate of the UPS was from 81.95 to 81.50 for the majority of the data logging, a delta T of only 0.45 degrees F.

Therefore, although the ambient air does affect this “air probe”, it affects it disproportionably. That’s why I would suggest another air probe temperature sensor be place somewhere else in the room that would react more actively or proportionately with the rise and fall of the ambient air in the server room, and therefore would not be buffeted by equipment surrounding it as apparently is the case of the UPS sensor.

We won’t shoot…..lol. Yes the cycle rate seems to have improved as a result of the rack panels being replaced as they should be.

Does the systems thermostat have a digital display? Do you have access to any digital display thermometer? If so, try putting it temporarily in the return air grill at the air handler (the vent that sucks air from the room into the air handler). Although you wont be able to monitor it online, you will be able to get a better overall picture of how the room air is reacting to the cooling system.

Hope to hear from you soon.

Respectfully Submitted,
John J. Dalton

[john dalton]

Dear Twcpipes,

I’ve read some of your posts here at the forum and know that you’re based in Huntington Beach, CA. You specialize in clean rooms, surgery centers, operating rooms, hospitals, and the such, and you’ve been doing this for 40 years. Who are you?????????

I work out of the Downey area, work in all the same areas you do, and have been doing this for only 30 years(lol). Do we know each other? If not I bet we know a lot of the same people.

My name id John Dalton and I work in the Engineering Department of Universal Air Conditioning Company, Inc.

??????????

Respectfully Submitted,
John J. Dalton

PS: Check my previous posts on this thread. We believers in condensing control have been shot down in flames. At least for now, we’ll see later though……lol.

[serverroomcooling]

I'm leaving that probe in there....it's a good baseline for me anyway, and I want to know what the rack temp is anyway

I have a digi thermometer I will place near the handler and monitor on my next cycle monitors, once at night, and another time during the day on the weekend when the building AC has been off for a while

I think I am going to get a digi tstat that can do all these features....If I set it to 65 and tell it to cycle on again at 70, that will give me probably 5 minutes off off time, and the temp won't get so high...does that make sense??

[seaboard]

Most moderately priced digital thermostats will have a built in 5 minute time delay. I'm not sure about an adjustable diffential? You'll have to look harder to find that. If all your computer equipment is on in the room now it seems that the tonnage is adequate for the computers. Now all you have to worry about is outside influences. There's no way for us to tell how insulated the room is to know how it's going to do in the summer. As Mr. Dalton said, if the main building BAS system cuts off the building system on Labor and Memorial days that may be the toughest test the unit will have. Us observers of this thread appreciate you answering questions posed and prompt responses.

[john dalton]

Dear Serverroomcooling,

You misunderstood me. I did not suggest you take that probe out, I suggested you obtain “another” one to use.

If you place the thermometer “near” the air handler that will not be nearly as good as putting the thermometer in the return air grill because of the following:

With the probe in the return air grill the reading you will be obtain will be an “average temperature” of the entire room. “Another” great baseline temperature to use.

With the probe in the return air grill the reading you will obtain will be that of the evaporative coil air entering temperature. This reading is especially important in diagnostic terms if you start to ice up the evaporative coil, which is common when you have either a humid environment, a system that’s operated below 72 degrees F, or if you have a system that operates at night with a high internal (server room) load, which of course you have all three of these things. Therefore, a great place for a temperature probe.

As far as a better thermostat, probably a great idea. But, don’t get a thermostat with the capability to adjust its operating span, get a thermostat that has the ability to set its delay on make of the cooling cycle. The off/on events of the thermostat is a function of the off time of the compressor, whereas a delay on make of cooling cycle is a direct link to the problem being addressed. I’ll suggest a couple of thermostats that would work with this project, you might ask your HVAC contractor for his option as well.

Look forward to the next report.

Respectfully Submitted,
John J. Dalton

PS: When you obtained the amp reading of 21.3 Amps from the output of your UPS system, was all your computers and servers in operation?

PSS: When and how much wattage will you be adding to this Server Room next?

[serverroomcooling]

I will place a probe inside the return grill

I will ask for a better tstat...any model suggestions would be greatly appreciated

tha AMPs was with everything on and ruinning, not sure of peak usage, but it's a good average

I plan on adding 2 more servers, max watts of 800 in the near future, I would like to add a Storage Area Network which would add....I'd say maybe another 2000 of max watts, by max I mean what it says on the back of server, not actual wattage....anything more will simply replace aging servers

[acjourneyman]

Do you guys work for the goverment, I can't believe how many posts it takes to install a split system.

[serverroomcooling]

I don't work for the government, I am in long term healthcare....we had a 1.5 ton unit put in which died on the first weekend of use. I wanted to put in a system specifically designed to handle server rooms, but the price tag was too high and was denied. HVAC contractors came in and told me that it can be done for much cheaper, and I was skeptical. I did tons of research before starting this thread, and appreciate all the work you guys have done to help me. I actually am part of an IT community which works like this site. I help as many people as I can when I have the time......

Anyway.....

After moving my Acurite thermometer to the air handler and placing its temp probe directly under the filer for the handler's return, I decided to watch and document the cycles, this time with the flourecent lights on

I walked in, at 11:19 and the unit was already on...

11:19 acurite temp reading 65.5 ath the return grill

11:20 65.1
11:21 65.1
11:22 64.6
11:23 64.9
11:24 65.1
11:25 64.9
11:26 64.6
11:27 64.2
11:27 cycle off 64.2
11:28 64.2
11:29 64.4
11:30 65.5
11:31 66.0
11:32 67.3
11:32:15 cycle on 69.6
11:33 70.9
11:34 69.1
11:35 67.3
11:36 66.4
11:37 66.2
11:38 66.0
11:39 65.5
11:40 65.1
11:41 64.6
11:42 65.1
11:43 64.6
11:44 64.6
11:45 64.6
11:46 64.4
11:47 64.4
11:48 64.2
11:49 64.2
11:50 64.0 cycle off
11:51 64.2
11:52 64.9
11:53 65.8
11:54 66.9
11:55 goes up to 70.9 and cycles on

I also have 2 springfield thermometers and humidity readers which read 75 behind the servers (no direct contact with vent air), and 72 in front of the servers (also with no direct contact)

this seems to be 18 minutes on, with 5 minutes off, and the air temp at the return grill ranges from 64.2 to 70.9 from cycle off to cycle on

Merry Christmas everyone....and I am not expecting responses. I am going home soon, but will be back on Sunday for more testing and monitoring

[serverroomcooling]

ok, here is an update.....the first real test was saturday, after the building AC had been off since 7 PM friday night, the system did not survive the day without freezing over. At about 9:00 the temp in the room began to rise as the system began to shut down. I was able to get to the office in time to shut the systems down, thaw out the handler and reset the tstat to 72, and all day sunday it was fine. So 72 is a good safe temp, at least this time of year. I believe that magic number was mentioned above. 72 is a bit high for me since the equipment now reads at about 83-84 during main building AC off times, and I would like that to be lower.

We mentioned that a more expensive tstat would allow me to make sure that the system does not turn on again until after 5 minutes has passed....will this prevent another freeze, or will the handler running on an on cycle for too long also cause a freeze??

Am I stuck at setting the tstat at 72 or 73??

[twcpipes]

>>>>the system did not survive the day without freezing over<<<

Isn't that amazing how the refrigerant cycle just never seems to change. 80 years and it is still the same.