Increase PSU Efficiency?

[ishelly404]

I have a HP/Compaq DC5000 SFF pc, that uses a proprietary PSU that runs HOT!! It runs so hot that I have had to recap it twice. It even managed to bloat some Panasonic FC capacitors after about 4 years of 24/7 use. I tried connecting the fan directly to 12v, but found that it was very noisy and not very effective. I have also removed the brown glue that goes conductive over time.

Anyway, I was wondering if it would be possible to increase the efficiency of the supply, and reduce the heat output, by replacing the primary side mosfet, which is a single Fairchild 9N90C 900v 9A part with a RDS of 1.4 ohms with a part with a much lower RDS. I was thinking something along the lines of a Fairchild FCPF22N60N, which has a RDS of 0.165 ohms and a better (22A) current rating. However, the voltage rating is lower, only 600v, and I was wondering if that would cause a problem.

I was also thinking of replacing the secondary side rectifiers with parts with a lower forward voltage drop. I assume as long as the voltage and current ratings are the same or better, that I should have no problem there.

Fairchild 9N90C Datasheet http://pdf1.alldatasheet.com/datashe...A9N90C_06.html

Fairchild FCPF22N60N Datasheet http://www.fairchildsemi.com/ds/FC/FCPF22N60NT.pdf

[kaboom]

Re: Increase PSU Efficiency?

I tried connecting the fan directly to 12v, but found that it was very noisy and not very effective.

Interesting.

Does the case have adequate free area for make-up air?

If you can, test the power supply outside the computer case with a reasonable load on 3.3, 5, and 12. You'll be better able to judge both the heat put off by the power supply alone, as well as the performance of its fan.


As you've found, higher voltage MOSFETS with low(er) RDs are not easy to find. I'd suggest not going lower in voltage. The best thing to do would be to scope the drain and give yourself some margin. Then go with the next-highest-voltage device.

There's probably a diode, ceramic cap, and one or more resistors forming a snubber on the drain of that MOSFET. Lose the snubber and the switcher will avalanche. Make sure none of those small components are loose/damaged, etc.

As far as secondary side rectifiers, are you suggesting schottkey diodes? Also, see if there are any load resistors against any caps near the output leads.

[ishelly404]

Re: Increase PSU Efficiency?

Ok, I will test the PSU outside of the case and see how much heat it puts off.

The problem with the fan, and cooling the psu in general is that it is so packed full of components and wiring that even with a powerful fan it is almost impossible to get adequate airflow. It also means that many of the caps are touching or very close to the heatsinks. Last time, I tried insulating the capacitors from the heatsink with heatshrink tubing but the psu is so hot inside that it didn't seem to make much of a difference.

I was thinking of replacing the ultra fast recovery diode rectifier on the 12v rail (for some reason all of the other power rails have shottkey rectifiers) with a shottkey rectifier that has the lowest voltage drop I can find.

[Behemot]

Re: Increase PSU Efficiency?

Yes, it is generally good idea to swap ultrafasts for schottkies. As for the MOSFETs, I think 600 V would be enough even for 230 V AC (with voltage doubler). If you can guarantee high voltage will not ever appear on input capacitors (e.g. by adding varistors if they are not present), you won't get anything over like 450 V (varistors start to conduct on some voltage above their normal operating value).

Also interesting question, I am both lazy to search for datasheets and I am also on EDGE now, does the computer have some additional output fan, or does it also pull all the wasted power from other components through the PSU? This may be the greatest problem.

Worst case scenario would be to swap it for another system if you need to run it 24/7, preferably some older server maybe? Because even that it is general rule, of course not 100 % of business class computers are so flawlessly design you could put so much on them…it may work for 5-6 years running usuall time (8-12 hours a day), but you see, there may be bigger room between "able to run many years with typical usage" (desktop) and "able to run many years with 24/7 load" (workstation/server). Some desktops can handle both, well, some not.

[Th3_uN1Qu3]

Re: Increase PSU Efficiency?

Yes, it is generally good idea to swap ultrafasts for schottkies.

Yes, that is a good idea. The output diodes have the most impact on efficiency. Because if the diodes drop a lot of voltage, then the primary section has to work harder too to keep the supply in regulation.

As for the MOSFETs, I think 600 V would be enough even for 230 V AC (with voltage doubler).

No, it won't be. Real-world inductors and transformers aren't perfect, and in a single forward supply the maximum voltage presented to the switch is twice the input voltage (~320VDC*2 = 640V) plus a leading edge spike caused by the leakage of the transformer. This can easily be a couple hundred volts - bringing the total pretty close to the 900V rating of the transistor.

[Behemot]

Re: Increase PSU Efficiency?

Ahh, true, would be best to know the topology of that thing before making judgements, good point…

[ishelly404]

Re: Increase PSU Efficiency?

Ok, yes I will replace the ultrafast rectifier, I have found another mosfet (ST STP21N90K5) with a slightly higher RDS (0.25 ohms), but a 900v 18.5A rating. That should work right?

STP21N90K5 Datasheet http://www.st.com/internet/com/TECHN...CD00255284.pdf

[Th3_uN1Qu3]

Re: Increase PSU Efficiency?

Another parameter to watch for when swapping MOSFETs is gate charge and gate capacitance (same thing, different units). If the gate capacitance of the replacement is significantly larger, the driver may not be able to offer good switching times, in which case the advantage given by the lower RDS will be negated by higher switching losses. +20% is okay but not more.

In this case, the ST mosfet you linked to is a perfect match - its gate charge is actually slightly lower than that of the original Fairchild part. Go ahead and buy it.

[ben7]

Re: Increase PSU Efficiency?

Another parameter to watch for when swapping MOSFETs is gate charge and gate capacitance (same thing, different units). If the gate capacitance of the replacement is significantly larger, the driver may not be able to offer good switching times, in which case the advantage given by the lower RDS will be negated by higher switching losses. +20% is okay but not more.

In this case, the ST mosfet you linked to is a perfect match - its gate charge is actually slightly lower than that of the original Fairchild part. Go ahead and buy it.

Would a lower gate charge MOSFET dissipate less power, since the gate driver would switch faster?

Could we see some pics of this PSU?

Perhaps it might help to have bigger heatsinks, or connect the output wires differently to have more room inside for airflow.

[Behemot]

Re: Increase PSU Efficiency?

or connect the output wires differently to have more room inside for airflow.

Or even remove some if you know you won't need them.

[tom66]

Re: Increase PSU Efficiency?

Don't quote me on this but... gate charge isn't the same as gate capacitance... gate charge (Qg) accounts for the overall charge of the device - this includes drain-gate and source-gate capacitance, as well as the "dead band" in the middle of the charge curve - it also accounts for the MOSFET's threshold voltage(s). (One where the transistor begins to conduct, and one where it is considered saturated, I think.)

[ishelly404]

Re: Increase PSU Efficiency?

I have ordered the STP21N90K5 along with some other parts, from mouser and I will report back on how it goes once the order arrives. Hopefully, it won't .

Here are some pictures of this psu.

[Behemot]

Re: Increase PSU Efficiency?

Joke-type fan? You can try to tie the wires together inside, may get a small bit of airflow.

[momaka]

Re: Increase PSU Efficiency?

Joke-type fan?

Yes, indeed (even though they are actually high quality NBM, Nidec, or Delta fans).
HP did a very good job with these DC5000 SFF computers in terms of planed obsolescence. We have about 50 of them at work, 30 stacked and waiting for a motherboard recap. Guess who's in charge of that task? .... yup . I have done 7 so far. YAY!
As far as the power supplies go in these: at least the ones I opened from the 7 machines I recapped, only one had bad caps in the power supply and that's because that one had the fan not working. The other PSUs seem fine (these PSUs usually use Ost or Samxon, and occasionally United Chemicon - depending if Lite-On, Delta, or Chicony/HiPro made them).

Funny thing is, there's actually about 20x3x5 cm of extra space in the case above the power supply. So HP could have actually designed those PSUs to be a little larger and use a larger fan. But I guess they didn't want it to last. Also, the D530 SFF shares the same case as the DC5000, but the D530 has an AGP slot and 2x extra full-height PCI slots via an add-on PCI card that positions these full PCI slots perpendicular to the motherboard and right above the PSU free area - so that's most likely another (dumb IMO) reason why HP made the PSUs so small.

@ ishelly404: have you recapped the motherboard on yours yet? If not, you should. Most have Teapo caps scattered on the board that eventually fail due to the heat in the case. The Rubycon MCZ by the CPU also like to fail quite a bit. I have successfully done a full poly mod on 1 and it's 100% stable so if you need info on what caps to get, let me know.

[ishelly404]

Re: Increase PSU Efficiency?

momaka: No, I have not recapped the motherboard yet. If you could give me the list of caps for the poly mod, that would be appreciated. I just noticed on my other dc5000 has two bulging caps by the memory slots. I figure I should recap both boards.

[ishelly404]

Re: Increase PSU Efficiency?

I just got around to recapping my DC5000 and replacing the mosfet + rectlifier and now it is running much cooler. Instead of scorching hot, it is now only barely warn to the touch. I would call this experiment a success!

[Behemot]

Re: Increase PSU Efficiency?

Nice

My experiment didn't work out as good. Everything was fine till the moment I started playing with halogen lamps, forgotting they have 10 times bigger startup current spike than their nominal value Had to swap +12 V rectifier Now there is 30 A/50 V instead of 20 A/45 V. Rated current 15 A so hopefully it can do some overloading before core saturates. Only thing limiting it is the transformer.

I am going to shop today for some cuprexit, finally soldering all the load equipment in. I got NTCs for the halogens lamps so it should not do ANY spike at all when it is finished. Gonna be at least 400 W passive load only on +12 V for the beginning with active load added in future.

[momaka]

Re: Increase PSU Efficiency?

I just got around to recapping my DC5000 and replacing the mosfet + rectlifier and now it is running much cooler. Instead of scorching hot, it is now only barely warn to the touch. I would call this experiment a success!

Cool (pun intended).

I guess I still haven't gotten "lucky" yet with one of those PSUs blowing up at work .