Fan Performance vs. Thermal Reliability for 1U Custom Chassis
Due to their restrictive height, 1U chassis are fan challenged. When we look at today’s 1U servers, RAID or 40GbE+ switch enclosure, etc., they are jam packed. There is little space left for air and still they need to be cooled. These dense systems are also pushing limits on power consumption. Many of these are considered mission critical and failure is not an option. Therefore, adequate sustained cooling is crucial. But when a fan fails, what do you do? Are you dead in the water, can you schedule to take the system out of service, remove it from the rack or what? Well, what if the fans were field replaceable and low tech to replace in the field without a service shutdown?
Fan Strategies for Thermal Reliability
- N+1 Fans
- Counter Rotating Dual Rotor Fans
- Reducing Speed to Increase Fan’s MTBF
- The Dirty Truth
- Hot-Swap Field Replaceable Plug In Fan Modules
Or . . . Schedule and switch systems provided it has adequate airflow capacity to sustain itself. The amount of time before system swap could be long or short depending upon its deployment.
This should always be the case in any mission critical chassis. Knowing how close to the edge of adequate airflow is also important and needs to be taken into account for all deployments. The greatest factors in deployment are maximum ambient temperature and elevation – both affecting air density and pressure. (Relative humidity also plays a part.) The higher the elevation the less efficient your cooling system will be. For example: at 70°F/21°C at sea level, air is 14.7psi and at 10,000 feet it drops to 10.1psi. So airflow volume needs to compensate.
When a fan dies, say locked rotor, you loose more than the output of a single fan. If the fans are exhausting hot air then the dead fan becomes another intake because the chassis has a negative interior pressure. This reduces the efficiency of the remaining fans. When using single rotor fans, with one in lock rotor condition, will not maintain the system temperature then this is where dual rotor fans may be considered.
Counter Rotating Dual Rotor Fans
These are a fairly new breed of axial fan. A typical 40mm square fan may be 56mm long then a more typical size of 28mm or less. Each rotor is independently powered and controlled. If one rotor fails the remaining rotor continues to power through the dual rotor venturi. This eliminates the airflow reversal that occurs when a single rotor fan locks. Doing a quick search on DigiKey I found over 200 40mm single rotor axial listings (in stock and RoHS compliant). In contrast, there were only 9 listing of dual rotor fans. Choosing wisely, one could still second source dual rotor fans.
Reducing Speed to Increase Fan’s MTBF
Another strategy is to employ that under general conditions may be throttled down to a less the maximum speed. This will increase the fan’s MTBF. This also provides the option that if conditions radically change, i.e. the air conditioning fails in the building, the fans can be sped up to better support cooling under the higher then planned ambient conditions.
The Dirty Truth
Most systems are installed in a rack buried in a closet, NOC, etc., never to be seen again until something goes wrong. What about the fans and is there something here that should be taken into account when initially designing the system beyond maximum ambient and elevation (aka worst case air density)?
Fans will over time load up on particulate. If there is a dirt/dust load factor of 10% then airflow will drop 10% . . . 15% = 15% loss and so forth. The knee jerk reaction would be to say “lets add filters.” If no one is their to remove the dirt/dust from the fans then who’s going to be there to replace the filters (where are they installed, how does one replace them)? Filters, depending upon their efficiency will load up much more quickly the letting the vast majority of dirt and dust just pass through the system and fans.
This begs the question, how much additional drop in fan efficiency do you design for do to fan dirt/dust fan loading to determine a useful predictive field life between installation and maintenance?
One more thing, dirt/dust loading on heat sink fins – how or do you account for that? Frequently BGAs, etc., have the ability to report their temperature . . . is anyone listening?
Hot-Swap Field Replaceable Plug In Fan Modules
We routinely suggest using one of our field replaceable (removable) fan modules that allows for quick, simple routine scheduled maintenance. Depending upon the system (normal operating temperature and it’s max temp) and the speed at which your tech works, the system may remain live in many cases during the fan removal, clean and replace procedure. We have the same capability in 1U Chassis as in 2U, 3U or more.