Testing methodology question

[enlightenedby42]

I was reading the Corsair face-off on the main page, and had sort of a general "just out of curiosity" question about PSU testing:

How do people attain the temperatures necessary to test efficiency claims? 45 C works out to roughly 113 degrees F, and I doubt open air testing in Death Valley or heating a room up that high is an option. How are those environmental conditions simulated on the test bed?

Thanks.

 

[Bbq]

In an incubator.

 

[enlightenedby42]

In an incubator.

Well I suppose that's a pretty simple and obvious answer now that it's in front of me...

For a "blow shit up" follow up....at what sort of temperatures do massive failures start happening? Does efficiency drop off really quickly after the rated temp is reached, or do they tolerate unreasonable heat loads fairly well?

 

[[Spectre]]

Well I suppose that's a pretty simple and obvious answer now that it's in front of me...

For a "blow shit up" follow up....at what sort of temperatures do massive failures start happening? Does efficiency drop off really quickly after the rated temp is reached, or do they tolerate unreasonable heat loads fairly well?

It depends. Usually most any power supply will start and run at 45c. It's just the temperature causes the units to lose capacity (de-rate) for evey degree over the temperature that its capacity was determined at. For intstance the FSP Epsilon based units have a rather nasty derating curve of 10w/c over 25c. So a 1000w unit at 45c actually is only capable of ~800w. Other power supplies de-rate at different rates some better some worse. Most of the time a unit when it overheats or is pushed beyond the derating curve will just shut down. Then after cooling off will come back on later and run the lower loads so it never really has a massive failure. Other units will die when this occurs but in a controlled manner. Other less well made units would for all intents and purposes be described as "explode" or have a massive failure. In general I can spot these coming as the exhuast temperature will sky rocket, or the voltages plunge, or the oscope trace goes south. So i guess the moral is good units tolerate the heat well and if they die they do so in a controlled manner, bad units melt, explode, catch on fire, or let out the magic smoke. As for efficiency it again depends on what the unit was rated at but in general I see a 1-2% drop over room temperature for good units.

 

[enlightenedby42]

Good info, thanks.

I'm not really sure why I was suddenly interested in the performance of PSU's under extreme conditions that are never encountered in real life, but thanks to everyone for indulging the random query.

 

[jonnyGURU]

It depends. Usually most any power supply will start and run at 45c. It's just the temperature causes the units to lose capacity (de-rate) for evey degree over the temperature that its capacity was determined at. For intstance the FSP Epsilon based units have a rather nasty derating curve of 10w/c over 25c. So a 1000w unit at 45c actually is only capable of ~800w. Other power supplies de-rate at different rates some better some worse. Most of the time a unit when it overheats or is pushed beyond the derating curve will just shut down. Then after cooling off will come back on later and run the lower loads so it never really has a massive failure. Other units will die when this occurs but in a controlled manner. Other less well made units would for all intents and purposes be described as "explode" or have a massive failure. In general I can spot these coming as the exhuast temperature will sky rocket, or the voltages plunge, or the oscope trace goes south. So i guess the moral is good units tolerate the heat well and if they die they do so in a controlled manner, bad units melt, explode, catch on fire, or let out the magic smoke. As for efficiency it again depends on what the unit was rated at but in general I see a 1-2% drop over room temperature for good units.

Apparently sometimes there is no de-rate. For example, the 800W I'm currently working with is rated at 40°C. I asked the OEM for de-rate data between 40°C and 50°C and apparently there is none. The explaination to me is that at the higher temperatures the MTBF drastically changes. The unit is rated at 100,000 hours at 25°C (one side of the spectrum) and at 20,000 hours at 40°C (the other side of the spectrum) and supposedly even less at > 40°C. Yet the "operating temperature" of the unit is up to 50°C.

 

[[Spectre]]

Apparently sometimes there is no de-rate. For example, the 800W I'm currently working with is rated at 40°C. I asked the OEM for de-rate data between 40°C and 50°C and apparently there is none. The explaination to me is that at the higher temperatures the MTBF drastically changes. The unit is rated at 100,000 hours at 25°C (one side of the spectrum) and at 20,000 hours at 40°C (the other side of the spectrum) and supposedly even less at > 40°C. Yet the "operating temperature" of the unit is up to 50°C.

Well that would seem to be more that the unit is actually rated at a higher temperature then and listed with a lower rating to meet the MTBF.....as in general units derate above their rated temperature. But hey ship me one and I'll see if it derates or not. I can crank the temp up to ~65c with the current setup

 

[jonnyGURU]

Well that would seem to be more that the unit is actually rated at a higher temperature then and listed with a lower rating to meet the MTBF.....

It may very well be, but in this world of marketing and number crunching mumbo jumbo it's hard to know what numbers to work with.

For example; the OEM markets the unit with the MTBF at 40°C. But most companies, even if they rate the wattage at 40 or 50°C, rate MTBF at room temperature because you get more hours. So the unit could "stack up" against it's peers in the race for bullshit stats, I had to get the (otherwise non-published) 25°C MTBF numbers so I could put those on the BFG box. Although my conscience is telling me I should actually put BOTH numbers on the box.

You know I'll get you one... I just have to wait for production to ramp up. My only two had to go to Nvidia.