voltage regulator and psu

hello,
baby-at (1995)
P54C (100 mhz)
stopped working....
power led, and case and psu fans work and right when it's powered on the kBoard leds flash.
1. the psu (200 Watts) has a chart with the voltages for the different wires.
if the readings i get are for ex: the power good is 4.8 volts (vs 5); the hd (yellow) is 11.8 (vs. 12) and the isa(?) is -12.2 (gound is zero) would this indicate that the psu is <A> the problem?

2. the pc has a voltage regulator module (?linear) (like a small wall next to the cpu (which has a heat sink only) with a perpendicular s.crew)
this gets warm within a minute (after several minutes it gets ~ 95 - 100 deg F) then i shut it off...
is this normal for the vrm to get warm?
could it be due to a faulty psu?
THANKS very much in advance!

 

Hi Tanya

re: 2

unfortunately this feels like something seriously overloading (?shorting out) the output of the voltage regulator causing it to get hot - rather than failure of the regulator itself.

is there a large three terminal device on that heatsink pls.? (usually three legs, sometimes 4 or 5)

try to measure voltages on the pins of this device, w.r.t. chassis will do for the moment...

also please let us know markings on this device

(probably a voltage reg IC but possibly a series-pass transistor)
(3 terminal regulators will current limit - but get exceedingly hot as they do so)

best wishes, HJ.

(will think about #1 depending upon the answers to the above, it's too much of a coincidence they are all out by 0V2)

 

I agree with HJ on this one. The VRM get very hot in operation.

 

hello Hugh Jarss and tel33,
THANK YOU for answering!
i have taken some photos of the vrm and the cpu (heatSink)

http://pws.prserv.net/cainet.tjtmd/koPCVRM.html
(VERY sorry -- size and res)

here's the board (fwiw)
http://pws.prserv.net/cainet.tjtmd/kpPCRedo.html

i don't se anything on the heat sink but next to it there is a *triPart* object which (IMVeryHO) is the vrm...

i don't see pins ON it though
or anything i could use the meter on

the markings: are EXTREMELY small however there is a clear one with
D882-Y
another part has:
9636
N[D/0/O]P-406[D/0/O]L
(nb: the [D/0/O] it is either 1 of the 3 possibilities (it is FAR TOO small))
i had found some refernces to what could be a vrm on google using the D882-Y

the board is clean -- have removed it -- no areas that are burnt (darker) no warping no cracks (magnifying glass plus shone light through it (maybe a microCrack though))

very grateful for the replies!
thanks again
sincerely
Tanya

 

Hi there

What you have in the pictures are the VRM for the CPU, the long black heatsink with the single transitor on it, that does get very hot, its normal.

It contols the voltage to the CPU core and it is a hard job to do for it that why its warm. If it was faulty the pc would not work at all believe me.

Have you looked at the power supply yet for damage on the printed circuit board or especially the capacitors bulging etc?

Regards
tel33

 

hello and thank you!
the pc doesn't work past the fans (psu and case), the kb leds flash right when it's powered on, and the power led comes on.
(there is no POST nor any beep codes with no mem)

do mean a completely dead pc -- no lights, no fans?

i have not looked in the psu of that one at all yet.
i still have to test the voltages again

i'll test the voltages and post back
THANKS AGAIN!
sincerely
Tanya

 

From that I would say it was the power supply that has gone, get a new one.

 

thank you for the advice
sincerely
Tanya

 

hi
still working on (hoping for) this one...
when i reMeasured the voltages they were all a bit high.

even though the pc speaker is plugged in there is NO POST...however the kBoard leds flash briefly.
last i tried, it had the 4 simms chips, and a new floppy drive and floppy data cable. (no video card though)

left it on awhile and the vrm got really warm (i'd say hot) and the heat sink warmed up. (neither were too hot to touch)
i tried another psu -- same results.

questions:
1. there are capacitors on the motherBoard... they have "x "s on top. HOW TO CHECK THEM?

2. can i remove the voltage regulator?

3. the heat sink getting fairly warm -- normal?
would cleaning it and reApplying it (i have the correct adhesive and remover) help?

4. someone told me i could check the moBo's integrity <spellCheck -- there isn't 1 here is there?> with an AM radio moved over the board
has anyone heard of this?

5. i tested then pinOuts of the mem banks and (although i question the accuracy of my dvm(s)) some had a charge where they were not said to have one (not where they were supposed to read ground but elseWhere)
bad news?

any and all help is GREATLY hoped for and appreciated!
sincerely
Tanya

 

Hi again

#2 yes probably not too difficult but I don't think it's what you're hunting, see #3

#3 although the voltage regulator heatsink is getting pretty hot not necessarily too hot - Tel said to expect it hot and I've found a few figures to confirm this -
using Vcore=3V3, Icore(pk) =3A7, assuming single TR coming down from +5V (series pass configuration)
P ( "that" transistor) = (5-3.3)*3.7 ~ 6W3
looking at the heatsink I'd give it 4C per watt , say 25Celsius temp rise or 45 deg F. Add on ambient temp ~ 70F, plus the 45, comes out at 115 deg F - not far adrift from what you measured.

(they seem to have given the CPU its own little personal radiator lest it get cold in winter)(shut up HJ)

(from the markings:
D882-Y probably Fairchild NPN 3A (pushing it a bit at 3A7 )
9636 is prob a date code, = 36th week of 1996)
>would cleaning it and reApplying it (i have the correct adhesive and remover) help?<
no - if the transistor had failed it would fail to short and fry the CPU - but this (probably**) hasn't happened - if the transistor were s/c it wouldn't have volts across it and thus wouldn't be getting warm. Got to have both voltage across and current through to make watts.
but if that's a nylon threaded fastener watch out - they give up and disintegrate after a while
**if there's a truly massive current flowing it would warm up a bit
but if that were the case the CPU would be frying (and toasted) and all sorts of things would be getting hot.

it would be useful to confirm the voltages on the three pins if the transistor on the heatsink though - but you say hard to get at - but if you could confirm 3V3 on the emitter you would have checked CPU Vcore.

#4 you can determine some things with portable radio but only really by comparison - expect tons of radiation all over the bands (AM and FM) when you have the case open

#5 not sure what you mean - did you remove the RAM and try with the meter on the pins of the sockets, try the modules when unplugged, or??

meter testing across RAM module pins (when not plugged in) is not a good idea, can bring them to an untimely end. Static and stored charge are your enemies here - the data and address pins are very fragile. Will sort you some notes about this.

not very constructive yet I'm afraid - sorry - interim reply! to warn you about the hazards of #5; still pondering which way to go here.

I think it might be a very good move - paticularly if you are doing several of these PCs - to put together a test jig (dummy load) so that you can prove PSU units without having them connected to a motherboard. Might make things easier in the long run? I get really paranoid about blowing up all the expensive bits with a PSU that's blasting out the wrong voltages... and it would surely get things a bit more under control.

best wishes, HJ.

(will have a go at #1 later!)

 

from the pics in your top link (koPCVRM) either topmost or bottommost picture, the emitter's on the left, it goes E,C,B

If my guess ID etc. for the 882-Y is correct you would be expecting 3V3, 5V0, 4V0 respectively

best wishes, HJ

 

hi and many thanks for answering this thread too!
just briefly:

i measured the pins (holes) of the sockets (similar to measuring the pin holes of the isa slots, the vesa's and the pci slots)
i read how to measure isa slots in a book (the others i found as "pinOuts" on the internet so i am not sure there was much point.)

i did not measure / touch the mem chips with the voltmeter whether in or out of the motherBoard

(also when doing the measurements (with the psu on) i had a wrist strap on and the board was on a rubber mat))

i still do not see anything on top of the heatSink aside from the fastener. i see the 3 tiers of the vrm next to the heat sink.

and i guess if there was (is) nylon it is on the s.crew that goes at 90 degrees to the long axis of the vrm
?

i'll try to figure out where to measure the vrm voltages....

this is all good news as there's obviously power travelling where it should across the board.

THANKS AGAIN!
sincerely
Tanya

 

trying to find a decently clear diagram/drawing... HJ

 

Tanya

haven't been able to find a good picture yet - this naff one's from the 882 datasheet. Note that the middle leg is slightly wider than the outside two - this will separate the device from the motherboard - it should stand up 2mm or so. With luck that might give enough clearance to sneak a prod for a connection, between the bottom of the device and the motherboard.

The heatsink may be at +5v - probably not but... (not a bad idea to find out if it's 0V, +5V or disconnected anyway)

it's the left hand leg you want to get at if you can (DC w.r.t 0V) - but don't risk shorting out to top tracks on the motherboard 'coz that's the CPU Vcore you are checking.

If in doubt, don't! (my guess is that it's fine, because the heatsink's getting to that temperature; it was pretty well bang on prediction)

gotta go - getting busy - will try to put some bumf about replacing the elcaps when I can get back

best wishes, HJ.

 

hello Hugh Jarss,
just wanted to thank you very much for replying and for the link!
very much appreciated!
sincerely
Tanya

 

Hi Tanya

about the motherboard elcaps - I don't think you have to worry as much about them like the Cs in the power supply. The ones in the power supply have to take far more "ripple current" which is why they often fail

In terms of DC voltage the motherboard would probably be quite happy 4.7V to 5.2V - particularly because the CPU Vcore has its own regulator, which will stabilise Vcore fine even if the +5 varies quite a bit

unfortunately the specs. which matter for PSU caps have notalot to do with the numbers printed on the can (capacitance and voltage)

- what matters are ESR (effective series resistance) and max ripple current - not usually written on the can, if you want to check what they should be good for identify the style of elcap and look it up. Look for signs of overheating and bulging cans, is your best bet.

Code:

C        V        ESR      I(max)   £    
100     10      0.03     2.65      1.30
100     10      3.0       0.1        0.07

same value, same voltage, ~same size can...

sorry I'm so ridiculously busy - how did you get on zeroing that analogue meter?

here's slightly better pix of
TO126 and TO220 case styles - to compare with the device on the heatsink


best wishes, HJ

 

realising I'd left out the bit that mattered

so if I have a PSU application which means the elcap has to take 2A ripple current...

with the first one (in post above) I'm fairly OK, 2A ocross 0.03 ohms will leave me with 0.06V ripple and the thing will warm up 0.12W - all pretty much under control.

but say I then replace it with one that looks pretty much the same but is actually the lower sort...

2A across 3 ohms means 6V ripple (ooops) and worse, it'll try to warm up 12W and go POP rather quickly. WAHEY!

This is reflected in the manufacturer's ripple current rating, I shouldn't be trying to use it for more than 0.1 Amp

moral - if you really have to find replacements try to ascertain manufacturer, style etc. - enought to look it up with or at least to make sure what you are replacing with can take the required amount of welly!

best wishes, HJ

 

which finally gets round to the thing about the meters even though it's in the other thread because the question about measuring the capacitors is in this thread!

what matters for PSU capacitors is how well they are smoothing the power rail, which is largely down to the ESR even though you won't find numbers for it on the can. Nice low ESR and you have a smooth power rail, ESR gets higher and the power rail isn't as smooth, and the capacitor gets hot.

You can suss how "unsmooth" the power rail is by measuring ACV with a meter which blocks DC on its AC ranges - about the only way short of using an oscilloscope.

high AC component - probably naff smoothing caps.

And it looks like your digital ones are OK for this (avoid autoranging) - get used to what they read on a healthy computer first though!

(Analogue meters usually don't block DC on their AC ranges.)

Phew, Hugh.

(it's lamp, not light bulb )

 

hello HJ,
thanks again for replying (sorry for delay)
i can answer ONE question:

by using a screwDriver -- there's a plastic scroo on the front... i didn't use a mirror since i won't be using it w/ a mirror

the rest of your posts (which i REALLY appreciate A LOT) i am unClear on for now... and since i don't have a supply of building material / cleaning material or paint etc... i don't want to follow in the footSteps of the twerp

i am grateful for the links -- very helpful!

and so, i'll work on the rest

thank you again -- all this info is GREATLY appreciated (and the links!!
sincerely
Tanya

 

hi Tanya

re: the pix TO126 and TO220...

...if I'm correct with the device on the heatsink it should correspond to the TO126 picture. If you reckon it's more like a TO220 - or something else altogether - it means my "best guess" is off-beam and we have to think again.

the TO126 link shows the wider sections of the "legs" quite well, these are the wider bits I was hoping would make the device stand up above the board a little, would give you a chance to get a test prod on to measure Vcore

==

re: the "mirror" thing, even cheapo analogue meters commonly have mirror scales - there's a mirror actually in the scale, where you read the numbers off. I was hoping you had one of those... if you find a small mirror where the numbers are, use it to line up your eye directly above the meter (get the pointer on top of its own reflection). But maybe yours doesn't have one.

good luck and best wishes, HJ.

mirror scale another another