Why does PSU efficiency drop at low loads?

[TELVM]

Take for instance the Seasonic SS-250SU:

.....

http://www.xbitlabs.com/articles/cas...undup-2_4.html

Below 50W draw efficiency drops under 80%.

Please enlighten me: exactly why does this happen?

[Wester547]

Re: Why does PSU efficiency drop at low loads?

There are a number of reasons, but the ones that are most prominent to me are:

1) Output rectifiers (diodes), like bipolar transistors and electrolytic capacitors, have a negative temperature coefficient. This means that as the temperature increases, the voltage drop (resistance) decreases, so less power is squandered in turn and efficiency improves (though as a byproduct of that, less resistance = more current flowing through the device).
2) The duty cycle. Lower loads tend to result in lower duty cycles and higher loads higher duty cycles which result in a reduction of switching losses which means higher efficiency.
3) The fan speed increases with the load and temperature.
4) Crossloading. Unusually small loads on any one rail, in a less than optimal design, can lead to crossloading which usually means lower efficiency.
5) Forward topology, because of its use of a half-wave rectified output, does not fully take advantage of the dual diode package as one of the two must be used for freewheeling during "off" time so as to discharge the coils and complete the current path, and that also means less efficiency (though I don't think this applies as much to synchronous rectification).

[eccerr0r]

Re: Why does PSU efficiency drop at low loads?

The simple way to think of it is that as the power supply itself uses energy (the control logic, etc.). As the load goes down the ratio between the real load and the PSU usage goes down, the efficiency goes down. A PSU plugged into the wall with no load and turned on will eat some power and is hence 0% (versus 0/0 which is indefinite) efficient by definition.

If you only have a light load, a smaller supply will have smaller control logic and may help you out at lower loads, but if your load greatly varies you will have to pay the penalty across at low usage.

Another real world example for high efficiency PSUs is that relay that shorts out the inrush NTC thermistor. Power that's used for that relay is wasted power that's fixed regardless of the load (versus the variable, somewhat load dependent power wasted by the thermistor).

[TELVM]

Re: Why does PSU efficiency drop at low loads?

... Output rectifiers (diodes), like bipolar transistors and electrolytic capacitors, have a negative temperature coefficient. This means that as the temperature increases, the voltage drop (resistance) decreases, so less power is squandered in turn and efficiency improves ...

I've just perpetrated a little experiment on the subject.

Took a sacrificial hopeless piece of crap PSU I've got around and loaded it with two 21W car bulbs, one on +12V and the other on +5V (~25W total DC draw).

- With the 80mm fan running, the kill-a-watt showed 40.3W AC draw from the wall ( = ~64% efficiency).

- Without the 80mm fan (passive operation ) the numbers in the kill-a-watt dropped steadily as components got warmer than before, until draw from the wall stabilized @ 36.4W ( = ~71% efficiency) . That's a ~7% improvement in efficiency. The piece of crap didn't seem to heat up much and survived the ordeal.


Do you think a modern decent ~300W 80Plus bronze PSU could work without for long periods on passive cooling (no fan, just natural convection) IF comp draw is as low as ~50W tops, 25W typical?


Every bit would help, for most low-level PSUs are gloriously inefficient at very low loads :

http://uk.hardware.info/reviews/4683...-225-watt-test

[Wester547]

Re: Why does PSU efficiency drop at low loads?

Do you think a modern decent ~300W 80Plus bronze PSU could work without for long periods on passive cooling (no fan, just natural convection) IF comp draw is as low as ~50W tops, 25W typical?

Definitely, yes. After all, people manage to run power hungry Prescott CPUs and nVidia graphics cards on "500W" labeled-in-reality-150W cheapos with two 3A diodes on brackets for the +12V rectification (because the derating curve for diodes is faster than the temperature curve) so an overbuilt 300W 80+ unit should easily handle a modern system even passively cooled, bearing all but the most power hungry setups.

[TELVM]

Re: Why does PSU efficiency drop at low loads?

^ Thanks Webster.

Thinking about the rectifier bridge ... would unsinking it increase overall efficiency at these very low 25~50W DC loads?

[Wester547]

Re: Why does PSU efficiency drop at low loads?

I'd say the bridge rectifier doesn't really have an appreciable effect on efficiency considering that the main switchers and secondary rectifiers tend to always fail first short of the main transformer saturating (not to say the bridge rectifiers never short, just that it seems to happen for reasons other than being overloaded). I think they add a heatsink to increase the current handling ability of the bridge rectifier. Even a 4A bridge without a heatsink should be more than enough for most systems. Need further proof? Older CRT monitors that consumed as much as 120W regularly either used 4 diodes or a 4A bridge with no heatsink and no airflow and they did fine for a very long time and these CRTs obviously didn't have the most efficient PSUs.

[mariushm]

Re: Why does PSU efficiency drop at low loads?

Do the math... for 50w, with 230v mains you're looking at about 0.2A of current.. a 10-20A bridge rectifier will have about 0.7v drop per diode at such low current so the power dissipated would be what... probably under 1w

That's unlikely to make a difference to the actual package temperature, with or without heatsink. And even if it will heat up a bit, the forward voltage will change by what, 0.1v? And the power wasted will change by what, 0.1w?

In the future, it may make sense for manufacturers to implement some kind of bridge rectifier using mosfets for example, to squeeze that Titanium efficiency or something like that.

It's already done, there were papers about it since years ago (this is one of them: https://cdn.badcaps-static.com/pdfs/...ea48825147.pdf ) and there's some controller chips that actually do it but afaik they're only rated for maximum 72v or something like that.. yeah, here's one: http://www.linear.com/product/LT4320

[cheapie]

Re: Why does PSU efficiency drop at low loads?

In the future, it may make sense for manufacturers to implement some kind of bridge rectifier using mosfets for example, to squeeze that Titanium efficiency or something like that.

It's already done, there were papers about it since years ago (this is one of them: https://cdn.badcaps-static.com/pdfs/...ea48825147.pdf ) and there's some controller chips that actually do it but afaik they're only rated for maximum 72v or something like that.. yeah, here's one: http://www.linear.com/product/LT4320

From what I've heard, some of the more efficient power supplies do something like that on the output. Also, some power supplies do have an option to allow the fan to stop at low loads.

[TELVM]

Re: Why does PSU efficiency drop at low loads?

... Also, some power supplies do have an option to allow the fan to stop at low loads.

Yep, for instance this Seasonic 1200W Platinum only activates fan above ~500W DC draw (~40% of max output):

http://www.jonnyguru.com/modules.php...tory3&reid=385

So I hope a decent 300W 80+ Bronze should live long and prosper while fanless @ 50W DC tops (~17% of max output).