PSU temp measurements??

[MikeC]

what temp for a PSU is consider danger ??

This a is good question that I have been grappling with.

There are parts in PSUs that hit 80C+ and other that never get close to that. Which temp is important? My current technique is to insert the thermal probe into the HS fins closes to the large coils, which do get really hot. But it is not really scientific as it does not account for variances between different PSU. Another might get hottest in a very different spot, for example, and not so hot in that one.

One thing I am thinking about implementing is to set up a standard heat source -- like a 60-100W light bulb -- in an empty case to simulate CPU and other component heat, mount each PSU in the normal place, then measure the temp of the PSU exhaust air. Even this airflow is not free of hot/cold spots, I've determined through experimentation. To get the exhaust air to mix and even out in temp, I am thinking of a tube ~80mm diameter & maybe 15-25cm (6-10") long that would go over the PSU exhaust. The thermistor would be positioned at the center of the end of the tube.

The result would not tell us how hot the PSU internals are, but it WOULD tell us whether one PSU runs hotter than another. This info would only be useful after a few models were tested this way. As the database grows, it *could* become a reasonable predictor of PSU reliability. Temp is only one factor though -- electronic components (that go into PCBs) rated for higher power can sustain higher temps that those rated for lower power...

Comments on this proposed addition to the PSU reviews?[/b]

[quokked]

MikeC, this is a really good idea, build a entire Case simulator setup, with a Simulated CPU with something like a precision resistor, a simulated HDD like a resistor in a HDD case and a simulated PSU this could be the ultimate simulated case for testing everything

[crisspy]

Hey Mike, some thoughts:

-suggested principal is to isolate any extra possible variables out of the test system. Just test the PSU.
-use a small well insulated box to eliminate ambient influences. It's not the size that counts, but the heat applied
-load the PSU with resistors inside the box, total load similar to an average system. This tests PSU under realistic but standardized conditions: at load, and sucking back it's own heat, or maybe with one standard exhaust fan too?.
-full agreement on exit duct/s, but measure at 1/2 radius (?) in from edge to avoid any possible center lulls caused by spiralling airflow.
-line the box and outlet duct/s with major foam/sound dampening. Sort of anechoic like? Run nearfeilds during test, both inside box and out back of duct/s. Idea is to gain realative but accurate insight into comparative noise of PSU alone as a solo component.

But on the other hand, I think that the whole test could be done just as accurately by electrical testing; maybe less the acoustic aspects:

- use stable load resistors pulling a typical-system power draw. Measure current & voltage into PSU for total power dissapated: no way to quibble about this, it must equal total heat generated.
- measure current & voltage on all loads for total delivered power output.
- difference between above two is wasted power, waste heat.
- simple calcs will tell precise temp rise in X cfm. of air for whatever wattage.

Now here's where it might get more interesting and desireable to do open tests:

-use a non-contact infra-red thermometer to measure around inside the PSU with the lid off. These are pretty accurate, and would allow you to sweep around for hot spots that might become potential problems.
-Big Industry uses thermal cameras, but think of this as the hand-held scanner mini equivalent. Some even have a laser pointer. Kewl.
-look for fast response over accuracy in chosen meter, to allow easy sweeping for hotspots.
-would tell a lot about potential moddability. Can problem hot-spot-parts be moved to better cooling?
-I bet that lack of bad hot spots would be the single best indicator of probable long term reliability. Cheap = minimum-specs = maxed-out = hot = soon-dead.
-big electrolytic caps are a top failure source. Are any of them running hot? If so, PSU has skimped on cap ratings, and is probably a looser. They are wetware, and probably don't like heat much.

[MikeC]

use stable load resistors pulling a typical-system power draw

Maybe you haven't looked at the Vantec PSU review? Check out our lab's spiffy PSU load test instrument shown on page 2: http://www.silentpcreview.com/goto.php?t=s&id=43&a=2

Efficiency is easily calculated on the basis of AC power draw (with Kill-a-Watt) and total DC load on the load tester. I know that tells me how much heat is generated -- 70% efficiency while delivering 100W means 42.8W is lost as heat, mostly inside the PSU but also in the cables & connectors. But it doesn't tell me how well that heat is handled by the PSU.

Which is why I am interested in the exhaust temp -- forgot to mention earlier that I would also measure the intake temp. In a perfect universe, the in/out temp difference would correlate to efficiency solely, but I'll bet it doesn't in ours. Worth experimenting with this a bit; will see if differences between PSUs show up. If so, may include this in the PSU test procedure. One set of in/out air temps at min power and another at 90W without external heat.

Why 90W? All PSU tests till now were limited to the 90W max power the test platform could pull. It seems a decent typical mid-range system.

I do think the additional heat is an important real-life simulation. A CPU that generates 50W+ is usually within an inch or two under (or beside) the PSU, and its heat is definitely a factor. From what I have seen, thermal behavior is NOT linear, we can't just add some calculated C/W factor for the CPU heat, and empirical testing can actually be done fairly easily, so why not. I already have the rig set up: an empty cheap 18" tower case and a simple light bulb fixture on a board that can be positioned about where the CPU would be.

more interesting and desireable to do open tests

Time & money are big issues. Also the moddability issue is cool to some people, but I am starting to feel a bit anxious about encouraging people to open up and muck with PSUs. Too many newbie questions in these forums that suggest someone might get into serious trouble one day.

ACOUSTICS -- As I get more exposure to the UBC sound lab, I realize more & more how inadequate my old Heath SLM is. I need a calibrated SLM that's good down to 20 dBA or better. UBC is not that easy for me to access, so sooner or later, this will require a purchase -- or donation from a generous company? These things start at $600 and keep going up. I was working with a $7000 SLM in the chamber today -- limited by the high ambient (~20 dBA) even in there, with big machinery running in the bldg where the chamber is. We did measure a production PC at 23 dBA @ 1 meter. Pretty impresive, but my own best one is much quieter still, I think it must be in the 12 dBA range. But 23 dBA for a XP2000+ machine is pretty damn good. More on that in a review soon...

[crisspy]

Oops, forgot you had a load tester. I do assume you will test PSU exhaust temps with the PSU loaded to the same 90w as your internal heat source? That was why I had thought of load resistors inside the case, powered by the PSU under test. And point well taken that the number you really want is the real world temp, so might as well just measure them.

Time & money are big issues. Also the moddability issue is cool to some people, but I am starting to feel a bit anxious about encouraging people to open up and muck with PSUs. Too many newbie questions in these forums that suggest someone might get into serious trouble one day.

People advocate anything from sleeping (safe) to suicide (fatal) on the internet, and almost no-one gets the blame. Ignorance is all too common, often not bliss, and accidents will happen with or without our help. I wouldn't worry about causing trouble with your contributions here Mike. By the very high quality of your work so far, I think you would do much more good than harm, to most any topic you address. The rest is caveat emptor.

I think the more important issue is with larger scale direction of where quiet PC technology is headed, especially with respect to the marketplace. It is lamentable that most of our available quieting tech is still work-around or obscure, instead of simple mainstream standard practice. The industry moves, but slowly, and only when sniffing on the trail of $$$. If you concentrate your efforts on especially esoteric work-arounds like PSU modding, you may show a few manufacturers a thing or two. If you show that given accessible products and technology that actually works, there is a huge potential marketplace worth a very large slice of the future profit in the industry: then you will lead them in droves by the one fabric that truly binds them.

So I think you're on the mark to stay on the conservative side of the street with SPCR, and avoid looking quite as esoteric and kooky as some of the extreme modding crowd. Emphasize readily acheivable quiet configurations of normal hardware, that beg to be sold that way in the first place. Feed the industry with the product ideas we all want to buy in the first place. You will get more of that much needed respect from the industry, maybe even a few engineers They might pick up the scent and get more dilligent about high quality, truly quiet PC technology, mainstreamed at affordable prices. We would all benefiet from that.

[siliconacoustics]

Mike,
There are a couple difficulties with this methodology. To illustrate the let me present two thought experiments using air temp as the test indicator.

For the first experiment assume two otherwise identical PSUs, one with a high CFM fan and one low CFM fan. The PSU with the higher CFM will have a lower air output temperature. What you will be measuring is how much air the fan moves and not how much heat the PSU generates.

For the second experiment assume again two electrically identical PSUs (identical fan), one with large heatsinks and one with small heatsinks. Naturally the PSU with larger heatsinks is more efficient at moving heat into the air. It is said to have greater thermal efficiency. The temperature of the air exhausted from the high thermal efficiency PSU will be warmer then the PSU with lower thermal efficiency, because more heat is transferred out though the exhaust fan.

The air temperature difference will be expressed to an even greater extent if you consider that the smaller heatsinks are hotter, and a hotter heatsink will cause a thermister-controlled fan (attached to said heatsink) to spin even faster (higher CFM), which will introduce the first mentioned experimental effect to an even greater extent.

The results of these experiments will lead you to conclude that the more thermally efficient PSU "runs hotter" then a less efficient PSU, which is of course the exact opposite of the truth of the matter.

So the problem is not only that there are too many variables affecting air temp, but also that some are opposite in there correlation.

There is a negative correlation between electrical efficiency and air temp.
There is a positive correlation between thermal efficiency and air temp.
There is a negative correlation between CFM and air temp.

In a thought experiment you can isolate for these variables but in practice you cannot. It is simple to use electrical efficiency as a means of determining which PSU “generates more heat”, but trying to measure which PSU “runs hotter” will be much more difficult. I’m not convinced it’s a valuable tidbit of information anyway since there’s no clear way to interpret what it even means while comparing PSUs of differing designs and quality. Seems like something that is easier to misinterpret then interpret correctly.

Jonathan
http://siliconacoustics.com

[MikeC]

trying to measure which PSU “runs hotter” will be much more difficult. I’m not convinced it’s a valuable tidbit of information anyway since there’s no clear way to interpret what it even means while comparing PSUs of differing designs and quality. Seems like something that is easier to misinterpret then interpret correctly.

Jon, You're right. These thoughts had occurred to me in the past but were savagely suppressed due to some insane desire to do more work. Temp measurements are out, efficiency calculations are in. Thanks for saving me the headaches, Jon.

[crisspy]

Egads! Yet another longish post from Crisspy for Mike... sorry Mike...

For the second experiment assume again two electrically identical PSUs (identical fan), one with large heatsinks and one with small heatsinks. Naturally the PSU with larger heatsinks is more efficient at moving heat into the air. It is said to have greater thermal efficiency. The temperature of the air exhausted from the high thermal efficiency PSU will be warmer then the PSU with lower thermal efficiency, because more heat is transferred out though the exhaust fan.

Many other points in the post are right-on, but this one skipped a beat. Total heat in = Total heat out. The smaller heatsink will get hotter because it's shedding the same watts in less area. The hotter heatsink might fiddle the electrical performance a wee bit, albeit not much. It might sink a bit more into a metal case too, again very minor. Still, after warm up, the air temp out will be the same for both PSU's because there is the same power output from both, going into the same amount of air.

The differences in fan volume between PSU's complicates these tests badly though.

A valid solution is to measure and report airflow. How to do it accurately? Would it work well? I don't know, haven't done my homework. Check this out. $120usd, might work. In the spirit of what a real-world-ish empirical test can expose, you would see straight-off what the strengths and weaknesses of different PSU's are. And that is very valuable information.

Otherwise you could go nuts and swap in standard panaflo's. Then any differences would be due to airflow impediments that you could rightfully fault the PSU for, and even suggest remedies for. You would still want to measure airflow for a well rounded picture. This isn't too far out, it's the most expected and accepted repair/mod for PSU's. And some PSU's might just be a-beggin' for a quiet fan. Their makers might oughtta know in time for the next model release. But then there's so many thermal control reliant ones

Testing in a real case would likewise be smart since sinking heat to the case is a potentially valid stategy, even a primary method for some small form factors that are definitely within the scope of our interests.

What we want is quiet computers. We want to know how a real PSU performs as part of a real system. What it's weaknesses are, and why. A PSU that is quiet with high airflow, but a little inefficient, might waste a few more watts, but still help cool a whole system better, and therefore be a really good buy. It might even avoid a second exhaust fan. Maybe it would allow a quieter system than one with a quieter PSU + second fan. It's a big mixed bag of conflicting priorities, and the real trick is to find out what works well, for real, and why. And if not, why not. Empirical testing is the only thing that might sort it all out.

So I suggest (revisions revisions):

basic hardware:
-realistic testing case.
-well controlled sealed air flow.
-fixed area passive air intake at bottom front, well larger than both fans.
-mixing ducts out back.
-case exhaust fan/s (std. panaflo) at upper-middle back (typical)
-PSU inside powering reference load resistors for a typical system load.

testing/measuring:
-airflow from each fan. ( case fan/s won't vary significantly for a given voltage), and at intake. F1+F2=F3?
-load voltage. +12+5+3.3, actual load wattage use V*V/R (measured more to check for working regulation)
-mains voltage and amperage for total input wattage.
-temps: each exhaust stream, inlet stream, some spot near where CPU would get air?
-sound levels out back, front, and side

trials / things to fiddle and combine:
-case panaflo/s @ 12v, 7v, 0V/sealed off
-PSU at idle, 100w load, and 200w load
-stabilized PSU fan speeds before the measurements are taken, or attempt to record highs/lows.

what it all tells us:
-efficiency of PSU in a realistic setting.
-performance of PSU in a realistic setting.
-PSU's thermal contribution to a realistic system.
-temperature performance of a realistic system.
-thermal viability of simple & realistic configurations.
-might we need a second case fan? if so, easy to approximate how much.
-how badly will our ears bleed at the end of the day?
-we might even get a clear idea as to what PSU's are worth buying and why.

Is it all worth it? That's obviously up to you Mike. I don't think your compulsion to work too much was completely wrong on this idea.

As for realiablility / hot spots: off the lid; off the fan; standard medium reference breeze from a fan; plug into load tester; power up; look for hot spots with IR thermometer. That's basically what they do during good design, but with an IR camera. Would certainly be interesting, and could at least indicate an all's-cool / warm-spots / hot-spots / run-she's-gonna-blow type rating. Note too that most component types have fairly generic temp ratings, so a generic above-sane-hot-spot warning could be indicitive of likely trouble. A thermal under-the-hood with a "Don't do this at home kids" lable. But still, do we really need to know?

I am reminded of quokked's sig. And I promise I'll run out of time soon, or get a life, or something... At least I'm enthusiastic.

[siliconacoustics]

Many other points in the post are right-on, but this one skipped a beat. Total heat in = Total heat out. The smaller heatsink will get hotter because it's shedding the same watts in less area. The hotter heatsink might fiddle the electrical performance a wee bit, albeit not much. It might sink a bit more into a metal case too, again very minor. Still, after warm up, the air temp out will be the same for both PSU's because there is the same power output from both, going into the same amount of air.

The same amount of heat will ultimately leave both PSUs one way or another but it won’t all leave through the exhaust air. With a less efficient heatsink an increased portion of heat will escape through secondary heat paths and radiation. You can feel this when you touch the warm case of a low airflow PSU under load. Do you think this effect won't be measured?

I would agree this is a MUCH smaller effect then CFM but it does exist. And since the temp of the heatsink has a direct effect on CFM in any thermally controlled PSU it can't be practically separated without bypassing thermal control. Ultimately the efficiency/size of the heatsink has strong correlation to air temperature that is very counter intuitive and would result in misleading test results.

Jonathan

[Melchior]

[quote="siliconacoustics"][quote]Many other points in the post are right-on, but this one skipped a beat. [b]Total heat in = Total heat out.[/b] The smaller [i]heatsink[/i] will get hotter because it's shedding the same watts in less area. The hotter heatsink might fiddle the electrical performance a wee bit, albeit not much. It might sink a bit more into a metal case too, again very minor. Still, after warm up, the air temp out will be the same for both [i]PSU's[/i] because there is the same power output from both, going into the same amount of air.[/quote]

The same amount of heat will ultimately leave both PSUs one way or another but it won?t all leave through the exhaust air. With a less efficient heatsink an increased portion of heat will escape through secondary heat paths and radiation. You can feel this when you touch the warm case of a low airflow PSU under load. Do you think this effect won't be measured?
[/quote]

I doubt that radiative heat transfer will make any sort of a difference in the measured temperature of the PSU exhaust air. Lemme pull out my old heat transfer textbook and I'll do some calculations because I'm curious:

q=AEFZ(T1^4-T2^4)

Where:
q=heat transfer in Watts
A=radiating surface area (in square meters)
E=emissivity of the radiating surface (between 0 and 1)
F=face factor (1 for this situation)
Z=Stephan-Boltzmann constant (5.67x10^-8 W/m/m/K). That's ten to the power of negative eight, which is what truly kills the radiative heating aspect.
T=absolute temp in degrees Kelvin (*C + 273).

I'm going to guesstimate some numbers here:
T1 = heatsink temp = 75 *C = 348 K
T2 = temp of the walls of the PSU = 30 *C = 303 K
A = heatsink radiating area, say 5cm square per side of the heatsink = 0.005 m*m
E = assume perfect black-body = 1
F = 1 for the situation of a heatsink in a PSU case

Thus, q = 1.7 W

That 1.7 W number is higher than I was expecting to come out with-- and indeed, it is higher than it would be in reality, since a 75*C heatsink is damn-freaking-hot, the emissivity will probably be closer to 0.3-0.4, and the heatsink probably won't have that much area to radiate from (radiating surface is quite different from a convection surface, so having lots of fins doesn't go very far).

Next, that 1.7 W of heat has to get from the PSU casing to the outside world-- it can travel through bolts to the tower case, or it can radiate (there'd be a very small difference in temps between the tower case and the PSU case), or it'd be transferred to the air via convection, and then it might end up being sucked out through the PSU fan anyways.

One factor these calculations don't take into account, however, is that as the individual components in the PSU get hotter, the PSU's overall efficiency may decrease, meaning that there's even more waste heat to get rid of, further compounding the problem....

All I'm thinking of suggesting here is opening up the PSU and using a heat-gun to get an idea of what the temps of individual components are within the PSU, as others have suggested above.

Dinner time, I must eat...

[MikeC]

A basic problem with opening up the PSU to measure temps: the temps change as soon as you remove the cover, as the normal airflow through the case is totally disrupted. I have already measured this effect.

[crisspy]

Yeah Mike, I agree. You would probably catch flack from manufacturers for unfair testing. Not good. Results might still be interesting, but wouldn't be very representative of real life application. Not worth the hassle.

[Melchior]

[quote="MikeC"]A basic problem with opening up the PSU to measure temps: the temps change as soon as you remove the cover, as the normal airflow through the case is totally disrupted. I have already measured this effect.[/quote]

True, it probably wouldn't be very accurate, though the results could still be interesting-- can you get the PSU open quick enough to find out if the hot spots are still hot?

[crisspy]

Hey Mike, I found this cool tool:
seller -> 405 Velocity Stick Pocket Thermal Anemometer
maker -> Testo Inc. Look up under velocity.

$119usd at Davis Intruments, a very interesting on-line shop for us tekkies. The velocity stick has .1fpm resolution, 5% accuracy. Seems like it's perfect for accurate fan and case airflow measurements, and realatively cheap. Seems to be the only thing in it's class, and thermal is the only way to do ultra low-flow. Simple math says:

Round ducts: 80mm / 7.8sq.in. > 18.5fpm=1cfm - - - 92mm / 10.3sq.in. > 14fpm=1cfm - - - 120mm / 17.5sq.in > 8.3fpm=1cfm

Square ducts: 3"*3" / 9sq.in. > 16fpm = 1cfm - - - 3"*4" / 12sq.in. > 12fpm=1cfm - - - 4"*4" / 16sq.in. > 9fpm = 1cfm

Would certainly tell a tail or two about numerous PSU's, fans, cases, and whole systems. Also imagine a tall 80mm intake chimney on top of the panaflo for HS testing. Airflow at different fan V's would explain some of the results methinks.

[mpteach]

Is there any way to tell what psu intake temperatures are safe for given number of cfm's?

I'm considering on panamodding a psu and i need to know if i have to build a cool air psu duct. I dont have any front bays open so the duct would have to go thru the top of the case and would be a pain.

[MikeC]

Is there any way to tell what psu intake temperatures are safe for given number of cfm's?

Not really... way too many other variables. Try reading through the discussion above fully, especially the points raised by Jon of Silicon Acoustics.

[Bluefront]

Most of the posts in this thread are almost two years old. I wonder if there is any new thought on this subject.

I am using the PSU as the only exhaust point for my latest project. I have measured PSU tempersture at several internal points (using wired probes) and at the final exhaust point at the rear of the case. About 36C is as high as I can see.....even at 100% cpu load.

I wonder if I can find an accurate temperature point at which I should be concerned?

[MikeC]

Bluefront --

The thread hasn't become any less valid with the passage of time. I do provide PSU intake / exhaust temps in the reviews now, but more for interest than anything else, really. It's hard to interpret. And no, I don't think you should be concerned.... although 36C seems amazingly low.... Surely there would be some points in the PSU that are always much hotter than that.

[Bluefront]

You're right....most of the PSUs I test at the output are higher than that. But this one is modded for better airflow. And I suspect that is what is lowering the output temp readings.

I did place two probes inside the PSU....one on the heatsink facing the bottom fan, and another on the top of the heatsink out of the airflow. There is another probe at the exhaust point on the rear of the case. All three probes read the same temp at all loadings. So I guess it's plenty cool.

Considering this PSU is also exhausting all the CPU heat (P4 2.66), I am really surprised also.

[anli]

I dont't remember another such case in my life - I mean reincarnation of so old thread Probably, there are "new generation" threads wrt PSU temperature. At such case - please excuse me, and point to those new threads. I have found this one.

The question is very simple and very common. Are there (accumulated during this long period) some simple empirical indicators telling "don't be crazy, don't reduce fan rpm further!" ?

Of course, I don't mean something absolutely safe, rather something like "I don't know bad PSU strories when output air flow (measured with multimeter thermopair) is below 45C".

Has anybody such empirical observations?

P.S. At my case PSU is located in dedicated chamber in Antec Mini P180.

[MikeC]

anli -- The age of the thread does not matter; just its relevance.

Take a close look at the temperature data provided in every PSU review we do. We have an intake air temp probe and an exhaust probe, and both are cited through about a dozen power points for every PSU.

You'll find this general consistency: If you factor in the variances in ambient temperature, the intake temperature is closely tied to the test load. For example, at 150W load, the intake is usually 25~27C (with fan-cooled PSU and typical 20-22C ambient). At 300W, it's usually 30~33C.

The exhaust temps vary much more, as it depends on efficiency and airflow in each PSU, as well as the maximum power we tested the thing at. Usually, the higher the power level, the hotter the output air -- although with PSUs that have fans that ramp up aggressively, this is not true in a linear way.

In general, you need to pay attention to what the manufacturer's maximum rated temperature is. Some cite max power at 25C. Others at 40C, still others at 50C. If you are unfortunate to have a 25C rated PSU, then you should assume the full power rating is a joke and never put a load more than say 60% the rating at maybe 40C intake temp.

At this point in time, assuming you are not modifying the PSU's own fan, it's not the exhaust I would measure but the intake. As long as the intake is below say 45C, you're probably safe with most PSUs. But to be safe, try to keep it at ~35C.

If you are modifying the PSU fan, then you need to pay attention to both intake and exhaust temps -- both are affected at least a little by the PSU fan. (Note: The best quiet PSUs manage to stay pretty quiet yet keep less than ~15C temp rise between in/out even at >400W.)

Hopefully, this info helps.

[anli]

Mike,

Thanks for these concrete estimating points (I mean 45C and 35C)!

Intuition is a help (I mean own audio DIYing), but it is much more helpful to get infor from first - your - hands familiar with plenty of different PC hardware units