Computer room air conditioning is an issue that is preoccupying people at all levels of datacentre management as they grapple with the dual tasks of energy reduction and carbon saving. Julian Jowitt, business development manager of independent HVAC energy saving engineering company Efficient Air has been in the thick of recent high profile projects, and offers some simple guidance that could guide readers through the jungle of possible solutions.

Not all organisations are in a position to build a brand new, state of the art, free-cooling data center with integrated energy and carbon saving systems. Even if they have future long terms plans to so do, that leaves many organisations with an urgent need to consider ways to optimize energy usage. They will have read about the myriad devices, theories and equipment that purport to provide instant solutions and quick wins.

Whilst choice is always a good thing, what may suit one application can be wholly inappropriate for another, and we would caution premises or energy managers against an all too frequent tendency to rush into the “headline fixes” that so many familiar energy reduction methods offer through lighting, BMS or BEMS or inverter systems. There is no single panacea, but a good dose of common sense helps.

For example, you are probably familiar with and have experience of using inverters. In theory they are an excellent concept as they provide control - why use what is not required? For instance, only using pumps and fans to do what they are supposed to do and no more. Inverters are usually thought of as producing only one result : optimizing speed = optimizing energy consumption.

Heart
Or let’s look at it another way. In this case the fan is seen as the heart that pumps the blood round the body. Make the heart more efficient by all means, but don’t reduce the flow unless the body is asleep or unconscious. This may not be a great analogy, but it shows that the heart/fan is actually only part of the bigger picture. The fan, with its associated air movement volume, velocity and temperature is only part of the system which supports the distribution of the coolth and humidity to the servers.

It has to overcome resistance to do this effectively, it supports the refrigeration circuit by giving up heat to the heat exchange and indirectly supports the dry air coolers or condenser heat rejection equipment. This is all indicative of the system’s coefficient of performance. It also interacts with the other CRAC units, all operating the same way, blowing into the same plenum, which creates a conundrum of pressures, temperatures and air flow patterns.

Other aspects, which need to be addressed when looking at fan savings, fall into four broad categories: savings; environment, control and equipment.

With regard to savings, questions arise over whether the savings are direct or operational, whether they are guaranteed or perceived. Environment refers to building usage, not the wider environmental planet, so questions to be considered are whether the environment will allow the equipment to be slowed down and how practical and measurable is this action. The control question would refer to what would control the reduction in speed, pressure, temperature, humidity, CO2 or the time schedule. Equipment would consider the effect of slowing down the fans and how the stability, noise, efficiency  and maintenance would change. In data center terms this comes down to the resilience of the equipment.

Pros and cons
“RISK” is the one word in datacentre vocabulary which encompasses all of the above questions and final decisions. Below I set out the pros and cons of fitting inverters as the first means of reducing datacentre energy, rather than retrofitting high efficiency fans.

The first affected area when fitting an inverter and reducing the frequency is blade tip speed and the slowing down of the impeller.  Most existing CRAC units, with the exception of new high end, top of the range CRAC ones, have forward curved blades on the impeller. These, unlike backward curved blades, have an unstable air flow characteristic. In the diagram on the previous page (diagram 1) one can see that there are two volumes where the fan curve can sit at any one given speed.
These forward bladed fans are selected according to the volume required by the cooling system, the estimated pressure that is has to work against, and the best efficiency and size of fan that will fit into the limited space within the CRAC unit.

The best efficiency is rarely on the tip of the fan curve and will normally fall down the left hand bank of the curve. Unfortunately on forward curved fans, this point on curve is mirrored on the right hand side also, the pressure and the fan speed is the same. But what stops it hunting from the left to the right? The answer is, nothing.

Speed
What happens when the speed is reduced?  It is impossible to predict changes in pressure, temperatures or air flow patterns while the fan is hunting, but they will then undoubtedly cause this CRAC or indeed opposing CRACs to compensate. There is a natural delay in the rectification of the temperature, pressure etc and this start to lead to over compensation. Temperatures will fluctuate, causing unnecessary overshooting of cooling and air flow, and in the worst case motor trip out, or not enough air flow across a DX coil which could cause the freezing of the coil, which in turn will block the air flow and cause the refrigeration circuit to trip out on low pressure.

By contrast, you can see from the following backward curved diagram [diagram 2], that the curve is one sided, with nowhere for it to hunt and therefore ideal for speed control. No hunting means no over shooting, no over compensation and no risk!
Further points of risk to consider are that fitting an inverter is effectively building in a single point of failure. This is not an elevated risk on single fan units, but as most CRAC units have 2, 3 or 4 fans, if retrofitting a single inverter, as is common practice, to feed all fans, one then introduces the single point of failure. If the inverter falls over, the CRAC unit will stop functioning. Fitting high efficiency fans will actually build in a fail-safe contingency. If one fan fails, the others can be ramped up with a simple 0-10 volt signal, to compensate. Another risk factor is the harmonics emitted by the inverters; they can be suppressed but generally they are passive and do not change with fluctuating loads

Savings
Energy savings are the major reason for introducing these changes. It is important to emphasise the distinction between operational and direct savings, and although both are valid, it is no good achieving one when both could be realised. The ideal is to start with direct, which means energy saving due to equipment efficiency. The second step is then to optimise the newly efficient equipment to get operational savings, such as slowing it down where appropriate.

It is not unusual to expect 30 to 45 per cent direct savings from the replacement of correctly selected, high efficiency fans, with paybacks in the region of 2 years. The replacement fans are also EC (electronically commutated), brushless motors and therefore produce less heat than the traditional fans, so there is less to cool.

Further ‘value added’ savings are also on offer from the removal of belts. Traditional fans are belt-driven which can absorb 5 to 15 per cent energy even when they are correctly installed, plus the cost of replacement and regular servicing. No belts also means no belt dust, thus removing a major concern in maintaining the sterile environment of datacentres.

A further advantage of using these alternative fans, is the vastly improved airflow across the cooling coil. Most CRAC unit fans are situated just under the cooling coil, which causes three unnecessary issues:  firstly they block part of the coil, thus reducing the surface area and thus its efficiency. The second is that the airflow has to split on either side of the fan, and then turn 90° before it can be distributed. The third is that that air is distributed into the floor slab, causing turbulence and resistance, which uses up energy. There are often guide vanes, but as the floor voids are usually around 450mm deep, it doesn’t leave a lot of time for the air to turn. 

Fan
The typical sort of fan that would be retrofitted has a totally different design: firstly it can be located much lower down, on or in the floor of the CRAC unit, thus removing the dead spots over the coil’s surface. Air flows directly to it, so there are no unnecessary 90 degree air paths The third is that it pressurises the floor plenum, so doesn’t blow at the floor slab. All these advantages, together with an advanced engineered backward curved blade design, can result in a far superior solution.
After savings have been made through the retrofitting of high efficiency equipment, there may then be the opportunity to reduce the speed of the fan to compensate and so achieve savings due to over capacity, the servers being moved, or simply passing seasons and their effect on the building fabric. The new fans are ideal for this. The replacements are EC and although supplied with a 415v signal, are internally controlled by a single DC (direct current) supply. This is good for speed control and can again be ramped up and down on a 0-10v or 4-20ma signal from any sensor such as a temperature or pressure type. The cost of this is minimal, and negates the need for expensive control systems.

So to conclude, if both direct and operational savings are desired, then the retrofitting of high efficiency, backward curved, electronically commutated fans is the clear solution. However, selection of the correct fan is paramount, and the installation critical. Use a recommended, qualified installer with a proven track record.

Follow this advice and the advantages will be many, the risks zero, the savings measurable, and you will end up with CRAC units operating at their maximum efficiency for years to come. And I would hope that that is the objective that you set in the first instance.