ASHRAE Journal - March 2010 - 53

to prevent cold dry air being introduced, which subsequently needs to be humidified. Without this low dew-point switch, the economizer can increase the facility energy use. With ultrasonic and direct evaporative humidifiers, this is not an energy issue as the water evaporation provides beneficial cooling. With all of the technologies, the economizer increases the amount of water used for humidification. Dehumidification, if actively controlled, is only an issue if reheat coils are used to maintain warmer supply air temperatures. These were commonly used on older CRAC and CRAH units in low density data centers. Reheat coils add to the first cost of the units (both for the equipment and the increased electrical service) and increase energy use. They are almost never needed since the data center loads provide all the reheat necessary to maintain upper humidity limits. Accuracy 3,000 41.9°F 2,500 2,000 1,500 1,000 500 0 San Francisco Bay Area Albuquerque, N.M. Seattle 59°F Hours/Year Figure 2: Bin data for dew-point temperature for case studies. 7.5°F 12.5°F Dew-Point Temperature Bins (Reported at Mid-Point) 17.5°F 22.5°F 27.5°F 32.5°F 37.5°F 42.5°F 47.5°F 52.5°F 57.5°F 62.5°F 67.5°F 72.5°F The National Building Controls Information Program (NBCIP) performed laboratory and field tests on the most common commercially available humidity transmitters. In this study10 they found that most of the transmitters could not meet the manufacturer’s printed statements of accuracy. Figure 1 is an example graph from this report. It shows two different transmitters from one manufacturer, comparing the deviation from the laboratory grade reference humidity transmitter in field tests over a wide range of humidity conditions. The box in the center of the graph is the manufacturer’s published statement of accuracy for this transmitter. The deviations measured were in excess of 24% error over a significant portion of the test. Similar results have been observed in the field. Table 2 is a survey of six water-cooled CRAC units in a data center. The reference probe had just been recalibrated at the factory and the CRAC unit readings were taken from the front panel of each unit. For the reference probe and CRAC unit sensors, the temperature and relative humidity readings were converted to dew-point temperatures. As shown in Table 2, the reference sensor had ~1°F (0.55°C) variation in dew-point measurements, but the CRAC unit sensors (the ones being used for the controls) varied by as much as 8°F (4.4°C). Since this was an open space the vapor pressure (and dew-point temperature) should have been relatively constant across all of the units. As can be seen in this survey, two problems exist with the controls: the control sensors are inaccurate and unable to precisely control the humidity; and the units are fighting one another due to the inaccuracy of the control sensors and the use of relative humidity for control (the controllers in these older CRACs could not be configured for absolute humidity control). During this survey, CRAC unit 006 was actively in dehumidification mode while CRAC units 008 and 011 were actively humidifying the data center. Unfortunately, this behavior with distributed humidity controls is quite common. Looking at Table 2, it is interesting to note that AC 006, the unit that was dehumidifying, ostensibly had the lowest dew-point temperature of all six CRAC units. This is because controls work on relative humidity not dew-point temperature. The 2008 update March 2010 to the environmental guidelines3 revised the previously published recommendations to use dew-point temperature, in part, to address this control issue. If humidity is to be controlled, it should be controlled using one or more high quality dew-point sensors. As can be seen from the measurements in Table 2, relative humidity varies across the data center but dew point is relatively constant. Reliability In theory, the purpose of humidifying the data center is to improve the reliability of the IT equipment operations. In practice, humidifiers are often a source of problems that can reduce the reliability of data center operations. Issues include: the introduction of water into the data center (which increases exposure to water damage from leaks); inaccurate sensors, leading to overhumidification of the data center (a potential corrosion problem if you have gaseous contaminants or salts in the air); and fouling of smoke detectors (which can cause false fire alarms). Do We Really Need Humidity Control? A number of data centers currently operate without active humidity control. On the West Coast, the authors are aware of dozens of facilities including: • A Fortune 500 bank (all facilities); • The main data centers of a major health service provider; • Several IT equipment manufacturers; • Several software development firms; • Several collocation facilities; • A major chip manufacturer; • Several major animation studios; and • Several supercomputer facilities. Recently, two major IT manufacturers published studies of IT equipment operating without any conditioning. The first11 was an experiment to run servers in a tent in Washington State from November 2007 to June 2008. Although a fence knocked into the tent during a storm, water dripped on the racks, and a leaf was sucked into a server, the authors of this study noted “zero failures, or 100% uptime.” ASHRAE Journal 53

ASHRAE Journal - March 2010

Table of Contents for the Digital Edition of ASHRAE Journal - March 2010