ASHRAE Journal - May 2010 - 8

Humidity Controls For Data Centers
I was intrigued by March’s “Humidity Controls for Data Centers” by Mark Hydeman, P.E., and David E. Swenson. One interesting bit of information included in the article is a field test on Humidity Controls the accuracy of two For Data Centers commercial humid- Are They Necessary? ity sensors used for A humidity control in a data center. The two sensors are factory rated at ±3% RH between 20% and 80% RH. The results indicate that the sensors rarely provided readings within the factory claimed accuracy limits. The standard deviation of both sensors was about 10%, and a significant number of measurements for each exceeded 24% RH errors. The other interesting part of this is that readings for both sensors varied from more than 24% above actual to more than 24% below actual, but for both the average of all readings taken was within 2% of actual. The sensors tended to read low at higher humidity levels and high at lower humidity levels. The lack of accuracy and precision of these sensors is quite concerning. In practical terms what this means is that these sensors cannot and should not be calibrated with offsets to their readings because they may read 20% high one day and 20% low the next. If you calibrate them when they are 20% high, they will be 40% off the next day, and the average reading would now be about 20% low. This also leads to a deeper discussion on accuracy and precision. Accuracy refers to the deviation of the readings from the actual value. A temperature sensor with an accuracy of ±5°F should always read within 5°F of actual temperature. That same device could have a precision of ±2°F, and all readings would then be within 2°F of the midpoint of the readings. That is, all readings should fall within a 4°F span. By field calibrating that device you could then theoretically achieve ±2°F readings. Whenever we calibrate field devices, we are assuming that the precision is better than the accuracy of the device. If the precision is not better than the
By Mark Hydeman, P Fellow ASHRAE; and David E. Swenson .E.,
The Current State of Practice

SHRAE’s Special Publication Thermal Guidelines for Data Processing Environments1 recommends humidity control in data centers in all climates.

The same is true for ANSI/TIA-942-1 2005, Telecommunications Infrastructure Standard for Data Centers2 developed by the Telecommunication Industry Association (TIA). In contrast, telecommunications facilities covered by Telcordia’s Network Equipment Building Systems (NEBS) requirements have no lower humidity

Table 1 shows the current published recommendations for humidity control in ASHRAE, NEBS and TIA. As shown in this table, only ASHRAE and TIA have lower humidity limits. NEBS has no lower humidity limit but has recommended personnel grounding practices to prevent electrostatic discharge events. All three documents have upper humidity limits. In About the Authors

limit.1 This article examines the background of these guidelines, the issues they

try to address, and the practical impacts of controlling humidity in data centers.

The background discussed in this article was discovered in the development of TC 9.9’s Research Work Statement 1499, The Effect of Humidity on the Reliability of ICT Equipment in Data Centers. The research
48 ASHRAE Journal

proposal investigates the relationship between humidity and the severity of electrostatic discharge (ESD) events in datacom environments over a range of thermal conditions and grounding of the floor and personnel.
ashrae.org

Mark Hydeman, P is a principal and founding .E., partner at Taylor Engineering in Alameda, Calif. He is a member of ASHRAE TC 9.9 and vice-chair of SSPC 90.1. He was lead developer of TC 9.9’s Research Work Statement 1499 on humidity control. David E. Swenson founded Affinity Static Control Consulting in Round Rock, Texas. He has been a member of the ESD Association since 1984 and serves on the standards committee and the ANSI/ ESD S20.20 Standard Task Team. He is a member of the Electrostatic Society of America and is a U.S. Technical Expert to the International Electrotechnical Commission (IEC) Technical Committee, TC101—Electrostatics.

accuracy, there is truly no point in trying to calibrate the device, and in fact that calibration is detrimental. If the theoretical sensor described above is calibrated with a single reading, the result would be a reliable accuracy of ±4°F with a precision of ±2°F. To improve upon that you should take a statistically meaningful sample of readings to determine the average systematic error of the sensor and adjust/calibrate by that amount. Clearly, this is not economically feasible on a real world commercial project. Loek (Laurens) Vaneveld P.E., Member ASHRAE, Menlo Park, Calif.

March 2010

The Author Responds
Thank you for your comments and observations. I agree that the graph from the Iowa Energy Center report is disturbing. In addition to the research they performed on duct-mounted humidity sensors, they have also begun the testing of economizer high limit switches. Our conclusion on review

of their research is that we should avoid using humidity or enthalpy sensors except where it is absolutely necessary. In those cases where it is necessary (e.g., control of a museum or archive), we use only the sensors that tested well in their research. At the ASHRAE conference in January, there was an excellent session on this topic: Seminar 41, Measuring Humidity: Does Accuracy Matter? Steve Taylor’s presentation, “Impact of Humidity and Enthalpy Sensor Error on Economizer Performance,” concluded that we ought not to be using humidity or enthalpy sensors for economizer high limit switches in any climate given the lack of precision and accuracy in the typical sensor. At the same meeting, in an interesting twist of fate, the SSPC 90.1 issued a new addendum cy on economizers that prohibits dry-bulb high limit switches in Climates 1a, 2a, 3a and 4a (humid climates)! Mark Hydeman, P.E., Fellow ASHRAE, Alameda, Calif.

Demand-Based Control of Lab Air Change Rates
I read with interest Gordon Sharp’s February article, “Demand-Based Control of Lab Air Change Rates.” Given the life-safety nature of laboratories, I think the readers should be aware of additional factors that must be evaluated when considering the use of demand-based control as advocated. As acknowledged by Mr. Sharp, lab airflow rates, and therefore air changes, are driven by exhaust devices (i.e., hoods), heat loads, or air changes. This implies that the best candidates for demand-based ventilation control are labs with few or no hoods and low heat loads, which leaves us with the question of what are appropriate air changes and heat loads. Mr. Sharp suggests that air change rates can be reduced to between four and two per hour. With a 55°F supply air temperature, the conversion of supply air to cooling capacity in a typical lab relates to approximately 0.75 W/ft2 per air change. Generally, two air changes are required to carry just the heat from the lights (labs do not use ceiling returns). And, given the intensity of analytical equipment, refrigerators, and PCs in most labs today, it is not uncommon to need six to eight just to handle the heat loads from equipment and lights. This does not disqualify the
ashrae.org

concept, but it severely reduces the hours where savings might be possible. More important is the issue of establishing the ventilation “demand” by sensing the presence of chemicals in the room environment. There is a life-safety component in labs, which transfers the responsibility for approving solutions like this out of the domain of ventilation engineers and into the environmental health and safety (EH&S) department. Hygienists do not talk in terms of TVOCs but rather TLVs (threshold limit values) and PELs (permissible exposure limits). Both are used to quantify the levels at which chemicals can harm humans. From a layman’s standpoint, the problem is that different chemicals in the TVOC class may have toxicity levels that are many orders of magnitude different (i.e., 2 ppm versus 200 ppm), yet the type of sensor cannot distinguish between the specific chemicals in the class. This means the “demand response” trigger levels need to be set low enough to provide a margin of safety for the chemical with the TLV in the lowest concentration. There also may be some classes of toxic chemicals that a TVOC specific sensor does not measure.
May 2010

ASHRAE Journal

ASHRAE Journal - May 2010

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