Blog |

Delta T – The air-side temperature difference

A server in a Data Center takes in air at a certain temperature. Once inside the server, this air warms up due to the heat produced by all the components in the server. The air that then exits the server is roughly 10 °C to 15 °C hotter.

An air conditioning unit in a Data Center also takes in air at a certain temperature. Inside the air conditioning unit, this air is cooled and the extracted heat conveyed to the outside. So the air that exits the air conditioning unit is approximately 10 °C to 15 °C cooler.

That all works out fine then, doesn't it? Unfortunately not.

The above-mentioned 10 °C to 15 °C is the so-called air-side temperature difference, or Delta T.

In a theoretical ideal scenario – a closed circulation of air between the server and the air conditioning unit – there would be a certain air-side temperature difference, and the air conditioning unit would work at its planned maximum level of efficiency.

In a real Data Center, this is sadly not the case. The cold air exits the air conditioning unit, flows through the raised floor, enters the cold aisles through the perforations in the raised floor grilles, is sucked in by the server, heated, blown out into the hot aisle, and then begins its journey back to the air conditioning unit. However, air is stupid and lazy. It doesn't know that this is the route it has to take, and showing it the way with blue and red arrows is no help at all.

Some of the air finds openings in the raised floor, e.g. cable cut-outs in the hot aisle that have not been sealed, gaps between the raised floor grilles or even grilles missing altogether below the racks. It then takes one of these shortcuts back to the hot aisle and straight back to the air conditioning unit, without ever having seen a server from the inside or having taken any of its heat away with it. Other bits of air take the planned route into the cold aisle, but then sneak between the servers through unused rack surfaces, or to either side of the servers, hot-footing it straight to the hot aisle and back to the air conditioning unit. This air does absorb a little heat, which the servers radiate to the outside.

Air that takes in only very little heat on its trip through the Data Center lowers the air-side temperature difference and therefore the efficiency of the entire air conditioning system.

Here's an example:

With an airflow of 45,000 m³/h and a Delta T of 15 °C (return air 35 °C, supply air 20 °C), an ASD 2010 CWU air conditioner from STULZ manages a capacity of 228 kW for a power consumption of 6.2 kW. The result is an energy efficiency ratio (EER) of 36.8.

Now, if the actual Delta T is only 10 °C (i.e. return air is only 30 °C) with the same airflow, power consumption and water temperature, capacity drops to 155 kW and the EER is cut to 25.0. As a result, the air conditioning unit has an efficiency 32 % below its possible or planned level.

The aim of an efficient Data Center business must therefore be to design air conduction through the Data Center in such a way that the planned air-side temperature difference is always achieved. Likewise, the airflow circulating in the Data Center must be constantly adapted in line with the servers' need for cold air.

About the author

About the author

Benjamin Petschke was born in 1969 in Germany. After studying physics he joined STULZ in 1996 and worked since then in the R&D, Export and Marketing department on different positions. With 19+ years' experience in the Data Center cooling industry, Mr Petschke is specialised in Data Center cooling design, energy saving and acoustic issues.

He works closely with the Joint Research Centre of the European Commission for the Code of Conduct on Data Centres on the Best Practice section and recently with the German DKE in development of the DIN EN 50600, Information technology – Data Centre facilities and infrastructures.

Mr Petschke authored White Papers on subjects like Best Practice for Data Centre Cooling and Indirect Free Cooling with Dynamic Control Logic.

Comments (18)

  • Joe HU

    at 24.11.2015

    Good Material. That's true. I suggested my client that close the air inlet when air conditioning unit shut off. But I have some questions, how could we measure the cooling capacity of one air conditioning unit accurately? And which metric would be better such as kW
    or kWh?

    • Benjamin Petschke

      at 25.11.2015

      Thanks for your comment.
      The actual cooling capacity of one air conditioning unit can be determined by measuring the return air temperature and humidity, the supply air temperature and humidity and the airflow. With these values and the corresponding formulas you can calculate the cooling capacity. The accuracy is an issue as the on-site airflow measurement is challenging. Contact your local Stulz representative and they will be able to help, if needed.

    • Travis

      at 26.11.2015

      We check the delta t, on the acqc program and in some cases energy upgrade advanced, we also had to test the return flow in order to balance system and size correctly, i have never had to test to determing the capacity or eer, good to know thanks.

  • Anthony

    at 27.11.2015

    Good article. In my experience I never seem to see the sever Delta T ever reaching the 10-15 degree Celsius mark. Where did you get your data from to make this assumption? Many people in the HVAC industry mention these figures as average but are they realistic?

    • Benjamin Petschke

      at 30.11.2015

      Thanks for your comment. In practice we also see far lower Delta T's. During the design phase of a data center in many cases the Delta T is in the mentioned range. There is typically a big difference between design values and those that happen later in reality.

  • Ronald Pace

    at 29.11.2015

    I like your article it show how the data rooms where done. The new design is with one cold aile and two hot ailes. The cold aile is seal off from the hot aile. There are some data rooms still using the old set-up.

  • Peter J. M. Hardie. MInstR.

    at 02.12.2015

    Would it not be better to actually sit the servers directly over the chilled incoming air grilles to maximise the cooling effect and if not should the ceiling over the servers not be sealed to force the air through the units then onto the CRAC, for we see a lot of short circuiting of air down the isles so never hits the revers and also consequent negative pressure in the room.

    • Benjamin Petschke

      at 02.12.2015

      Thanks for your comment.
      The target should be a clear separation between cold supply air / server inlet air and warm return air / server exit air. There are various different possibilities available in the market to realise this separation. Some examples: Cold air supplied directly from the raised floor into the front area of the rack, duct connections between the top of the rack for the exhaust air to be directed through the suspended ceiling back to the CRAC, containment for the cold aisle, curtains above the racks for separation between cold and hot aisle...Important: Guide the air through the data center.

  • Kevin Farnsworth

    at 06.12.2015

    In an air cooled data center system it is very important to separate the supply air from the return air in any way possible. In non-raised floor designs it is better to contain the hot aisle or provide chimney racks. In raised floor plenum designs we find that cold isle containment works well. In any case it is essential to make sure the smoke detection and fire protection is designed to work with any containment option. Each isolated (segregated) space must have proper detection and suppression when using very early smoke detection (VESDA), standard point detection, pre-action sprinklers and/or clean agent. Too often data centers are modified to create hot/cold air isolation, yet the smoke detection and fire suppression is not adjusted to function properly under that new configuration. By just putting melt away clips on air curtains or melt away ceilings above an aisle does not meet NFPA requirements. Safety is always first before reliability or efficiency.

    • Benjamin Petschke

      at 07.12.2015

      Thanks for your comment. I fully agree with your statement, safety first before reliability and efficiency. For sure the smoke detection and fire protection must work properly with the selected containment option.

  • ali atif

    at 22.12.2016

    Dear Sir

    Subject estimated solution is required on the basis of below mentioned data

    Client: Defense

    Site Nature: Data Center

    Number of Racks: 02 Server Racks & 02 Network Racks

    Number of UPS: 02 VT 40 KVA By Schneider Electric

    Area of Data Center: H: 20 W: 15 H: 10 Feet

    Note : Solution is required for PC – I Level


    Ali Atif Bhatti

    • dibyendu pandit

      at 20.04.2020

      Please ALI ATIF you 1st mention your 4 racks power utilization.then we calculate colling and total power supply.

  • Daniel Bodenski

    at 02.05.2018

    Nice article and very interesting and important, with the increase in server power densities and the continuing need to manage airflow in data centers. Question, is there a method to measure Delta-T across a rack, row, or data center and use the data to control supply air flow [cfm]? Call it an integrated solution?

  • Sudhakar

    at 13.02.2019

    The phenomenon is very similar to what happens with chiller in an HVAC system (it is widely called as low delta T syndrome). My understanding on this phenomenon is that, this is due to reduced heat transfer across hot side (air in CRAC) and cold side (refrigerant in CRAC) mainly due to reduced temp difference. So for the same electrical energy supplied, due to reduced heat transfer it produces less cooling power and hence the efficiency.

  • sonam

    at 29.07.2019

    Each and every topic are covered in this blog about cooling.
    My all queries are solve by reading this blog.
    keep sharing

  • Eep Saepudin Hambali

    at 16.06.2020

    I want to ask, why does the temperature delta parameter have to use (K), not Celsius ...

    • Benjamin

      at 16.06.2020

      Typically Kelvin (K) is used for temperature difference. As °C and K have the same scale it doesn't matter which one to use. Farenheit btw. has a different scale.

    • Eep Saepudin Hambali

      at 18.06.2020

      Thank's a lot Benjamin.....

Write new comment

Press Contact

Please choose your contact

  • Press Contact

    For press information please contact:

    Riba:BusinessTalk GmbH
    Klostergut Besselich
    56182 Urbar / Koblenz
    Tel.:+49 (0)261-963 757-25
    Fax: +49 (0)261-963 757-11

    STULZ GmbH
    Jan Pohlgeers
    Holsteiner Chaussee 283
    22457 Hamburg
    Tel: +49 (0)40 55 85 - 0

More contact options

Contact us

Holsteiner Chaussee 283
22457 Hamburg 

Central Office: +49 40-5585-0
Fax: +49 40-5585-352

Contact us
To top