There is an expectation for data center servers, of 24x7x365 uptime. We expect the servers to run all the time.
One of the challenges to this availability goal is the fact that utility power is inherently dirty. Utility power is not a clean, smooth, sinusoidal wave as it’s commonly drawn. Utility power has noise, sags, swells, spikes, frequency changes, and all sorts of flaws. We expect the ICT equipment to continue to operate through all of this, but the tolerance of ICT equipment to degradations in input power quality is limited.
Susceptibility curves are specifications for how IT equipment perform as the quality of input power degrades. The figure shown here is an example of a susceptibility curve. Let me explain how to read this.
The vertical axis is the percentage of input power supply voltage, expressed as a percentage of nominal. The 100% line is the point at which input voltage to the ICT equipment is nominal, that is, it’s at the level it should be. At this point, the equipment operates reliably.
If we go higher than the 100% line on the graph, we are describing over-voltage conditions. The curve shows, for example 200%, 300%, 400% and 500% over-voltage scenarios.
The horizontal axis on the graph is duration of time, and this is a logarithmic scale. These are very small durations of time as you can see with times of a microsecond, 1 millisecond, 20 milliseconds, and so on.
What we are describing then, with a graph like this, is some magnitude of input power voltage of a short duration. A power spike might be 3x the nominal input voltage for a microsecond, for instance. A short drop in power might be 10% of the nominal input power for 1 milisecond. The susceptibility curve describes how tolerant is the ICT equipment to these deviations from ideal input power quality.
There are two critical regions on this curve. The upper green area shows the cases of over-voltage duration which will cause the ICT equipment to stop operating because of damage to the electronics. A 200% over-voltage condition that lasts for 20 miliseconds will damage the equipment according to this curve.
The lower green area describes under-voltage conditions. These conditions will cause the equipment to simply stop running because the electronics become de-energized. According to Figure-1, if the input power falls to less than 70% for more than 20 miliseconds the equipment will shut down.
The most common susceptibility curve in our industry is the ITIC/CBEMA curve (commonly referred to simply as “the CBEMA curve” pronounced “Suh-beem-ah). It has been used for the past 30 years to describe input power quality susceptibility for all electronic data processing equipment.
By Bob Landstrom
(This is Part-2 of a 9-part series on a modern assessment of the ITIC/CBEMA curves)