Country missing? Please select your nearest region...
Published on 15th February 2012 by Ian Bitterlin
A few months ago I posted a blog that challenged the relevance of the 12-year-old ITI (CBEMA) curve. I was surprised that I got very little return traffic via LinkedIn – and certainly almost no comments compared to similar discussions about fresh-air, humidity, corrosion and zinc-whiskers usually generate!
My point was that the 2000 version of the voltage immunity curve for IT hardware must (surely?) have been overtaken by progress. It said (and still does) that ICT hardware can withstand 20ms of zero-volts without dropping off-line. OK, we all know that it’s much more complicated than that and IEEE Std 1100-1999 does a great job of explaining the rest – but the ‘zero-volts’ threshold is pretty key to the design of UPS etc.
The history of the CBEMA (Computer Business Equipment Manufacturers Association) curve goes well back into the mainframe days but it was accepted as applying to servers by the time it appeared in the 1992 edition of IEEE Std 1100 (and the excellent ‘emerald’ book). As we have no indigenous ICT hardware OEMs in Europe we never seemed to have had an equivalent reference standard but the bottom line was a ‘zero-volts’ immunity of around 10ms. I say ‘around’ because the curve was just that, a curve, but plotted on a logarithmic scale so it was hard (and certainly debateable) to say with certainty where the curve touched the x-axis. It’s probably worth mentioning that one problem we have had in Europe has been the fact that the curve has always been related to 120V-60Hz equipment so we have to take a leap of faith in our 230V-50Hz environment. Anyway it got updated in 1996, renamed ‘New ITIC (CBEMA) curve’, and it appeared set in stone in the 1999 revision of IEEE Std 1100. By that time it was not a curve but a well labelled step diagram that had an ‘improved’ zero-volts tolerance of 20ms. It’s worth contemplating that one cycle in 50Hz is 20ms whilst at 60Hz it is 16ms so ‘maybe’ the immunity was really longer for us? For the record it was updated again in 2000 and renamed ‘ITI (CBEMA) Curve’ published by the Information Technology Industry Council (ITI) and still found on http://www.itic.org with some interesting notes. However I can’t see any difference at all between the 2000 version and the 1996 version – still 20ms zero-volts.
So, I hear you ask, what am I scribbling about this time? Well, with the dramatic rise in energy costs in Europe (compared to the low rates available in the USA) European UPS vendors are all offering ‘eco-mode’ UPS that rely on taking advantage of our relatively stable grids and running in ‘reverse’ – in short, switching to UPS when the mains fails instead of switching to grid when the UPS fails. Of course the time taken to switch is a key factor! The average time is currently 4-6ms with the ‘fastest’ OEM achieving a staggering 2ms switch-over time. Now comparing this to 20ms (as the ‘worst case’ survival) shows us what risk we may take. The ‘reward’ is a huge hike in power efficiency to 99% - from c92% in the US or c95% in Europe. So it all looks promising for those organisations that are willing to accept the reward and accept the risk. Actually we could discuss the ‘risk’ at length here but I won’t as it is very subjective and open to opinion – although it does depend upon the stability of your local grid and your connection to it...
But in the course of discussing ‘eco-mode’ at high industry levels in the USA DELL have burst from the undergrowth with the surprising statement that their current hardware can only match the original pre-1992 CBEMA 10ms immunity level. Now I for one don’t actually believe them since I can’t believe that they purchase ‘inferior’ or ‘cheaper’ switched-mode power supplies and I think it’s a play-safe game to avoid warranty issues. I hope so anyway, because the introduction of running UPS in eco-mode UPS will save huge amounts of energy and, to my mind, reflects exactly the same trend in data-centre cooling; wider tolerances enabling energy savings – just by CBEMA instead of ASHRAE.
What do you think? DELL can’t have regressed 20 years, can they?