+12V and +5V regulation
Franc Zabkar
significantly under different load conditions. For example, at various
times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
good, this seemingly poor regulation is to be expected from the
design. Most PC PSUs regulate using a weighted average of the +5V and
+12V rails, so that when the +12V rail rises, the +5V rail falls to
maintain a constant average.
I'm considering modifying this PSU so that it regulates the +5V output
and allows the +12V rail to find its own level. This is easily done by
changing a few resistors.
For example, in the following PSU one could remove R25, and replace
R20 and R21 with a single 5K6 resistor:
http://www.pavouk.comp.cz/hw/en_atxps.html
This would guarantee a regulated +5V rail that was unaffected by
changes in the +12V load. Unfortunately the penalty would be that the
+12V rail may at times rise as high as 13V.
I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
+13V supply be a serious problem?
- Franc Zabkar
--
Please remove one 's' from my address when replying by email.
kony
<fza...@optussnet.com.au> wrote:
>I have a generic 400W ATX PSU whose +12V and +5V outputs vary
>significantly under different load conditions. For example, at various
>times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
>good, this seemingly poor regulation is to be expected from the
>design. Most PC PSUs regulate using a weighted average of the +5V and
>+12V rails, so that when the +12V rail rises, the +5V rail falls to
>maintain a constant average.
>
>I'm considering modifying this PSU so that it regulates the +5V output
>and allows the +12V rail to find its own level. This is easily done by
>changing a few resistors.
>
>For example, in the following PSU one could remove R25, and replace
>R20 and R21 with a single 5K6 resistor:
> http://www.pavouk.comp.cz/hw/en_atxps.html
>
>This would guarantee a regulated +5V rail that was unaffected by
>changes in the +12V load. Unfortunately the penalty would be that the
>+12V rail may at times rise as high as 13V.
... or with modern systems using so much 12V power, a 12V
rail that dips quite a bit. You don't mention the
application of this modified PSU. If it's to be powering a
*typical* PC then perhaps starting out with
better-than-generic-400W would be desirable? Just wondering
if the result would be worth the time spent, unless you only
needed a switching 5V supply for other project(s).
>
>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
>+13V supply be a serious problem?
In a recent thread someone had mentioned that one of the
(suspected) failure modes of Maxtor HDDs were a couple of
small 8 pin (SO-8) chips that were overheating. I meant to
take a look at some Maxtor drives to determine what those
might've been but suspect power regulators, maybe linear.
If so, raising the 12V level might raise the thermals and
tend to be more problematic for such drive designs... which
could easily be more common than only (certain) Maxtor
models.
As for other 12V, it seems most CPU VRM circuits would work
as well at a fairly high input variance, though raising 12V
would tend to make other non-regulated devices (fans, some
LED & misc case circuits) run a little
hotter/brighter/louder/etc... which probably won't matter
given the low % change.
Franc Zabkar
keyboard and composed:
>On Wed, 09 Mar 2005 20:15:28 +1100, Franc Zabkar
><fza...@optussnet.com.au> wrote:
>
>>I have a generic 400W ATX PSU whose +12V and +5V outputs vary
>>significantly under different load conditions. For example, at various
>>times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
>>good, this seemingly poor regulation is to be expected from the
>>design. Most PC PSUs regulate using a weighted average of the +5V and
>>+12V rails, so that when the +12V rail rises, the +5V rail falls to
>>maintain a constant average.
>>
>>I'm considering modifying this PSU so that it regulates the +5V output
>>and allows the +12V rail to find its own level. This is easily done by
>>changing a few resistors.
>>
>>For example, in the following PSU one could remove R25, and replace
>>R20 and R21 with a single 5K6 resistor:
>> http://www.pavouk.comp.cz/hw/en_atxps.html
>>
>>This would guarantee a regulated +5V rail that was unaffected by
>>changes in the +12V load. Unfortunately the penalty would be that the
>>+12V rail may at times rise as high as 13V.
>
>... or with modern systems using so much 12V power, a 12V
>rail that dips quite a bit.
... which would result in a higher +5V rail in an unmodified PSU.
>You don't mention the
>application of this modified PSU.
Athlon XP 2500+, 512MB DDR SDRAM, Xabre graphics card, 120MB HD, DVD
ROM, DVD burner. Measured power consumption on AC side is ~160W.
>If it's to be powering a
>*typical* PC then perhaps starting out with
>better-than-generic-400W would be desirable? Just wondering
>if the result would be worth the time spent, unless you only
>needed a switching 5V supply for other project(s).
During my minicomputer days, the +5V rail was deemed of utmost
importance as it powered all the logic. In contrast, the PC platform
seems rather lax in this respect. In fact the only PC PSU I've
encountered that has independent regulation for each of the major
rails is Antec's Truepower. If a "better-than-generic-400W" PSU
regulates the same way as a generic, then I can't see how it would be
any more accurate. I'd be interested to know how other PSUs perform
under varying load conditions. For example, what are the +5V and +12V
readings when the HDs and DVD ROMs have spun down, as opposed to when
they are all spinning and seeking? How are the voltage readings
affected by CPU load, by multimedia apps, graphics card, etc?
Do you have any numbers for your own PSU? You may be surprised.
>>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
>>+13V supply be a serious problem?
>
>In a recent thread someone had mentioned that one of the
>(suspected) failure modes of Maxtor HDDs were a couple of
>small 8 pin (SO-8) chips that were overheating. I meant to
>take a look at some Maxtor drives to determine what those
>might've been but suspect power regulators, maybe linear.
>If so, raising the 12V level might raise the thermals and
>tend to be more problematic for such drive designs... which
>could easily be more common than only (certain) Maxtor
>models.
Yes, that's my concern also, ie whether HDs and DVDs still use linear
power techniques.
>As for other 12V, it seems most CPU VRM circuits would work
>as well at a fairly high input variance, though raising 12V
>would tend to make other non-regulated devices (fans, some
>LED & misc case circuits) run a little
>hotter/brighter/louder/etc... which probably won't matter
>given the low % change.
PC
"Franc Zabkar" <fza...@optussnet.com.au> wrote in message
news:7lat2119n98886049...@4ax.com...
Franc
3 thoughts come to mind
If you realy do want to waste time on an apparent poorly regulated Power
Supply then I strongly suggest you build a suitable (variable?) dummy load
and connect your modified power supply to this during your 'experiments'.
You don't want to risk blowing the rest of your PC if the power supply runs
away and 'overvolts'
For all the time and effort involved in redesigning someone else's work it
will be better to just go buy a tighter regulated Power Supply.
Is the circuit you give exactly the same make/model as yours? If not I would
suggest yours will be vastly different.
Cheers
Paul.
PC
Franc
Have posted other thoughts to your original post. In it I suggested a
variable dummy load for experimenting.
This may be the better way to 'test' the supplies regulation.
For my part while having built supplies (linear not switch mode) with
regulation tighter than a fishes a... I have found the biggest problem with
PC supplies at the moment is the Chinese Capacitor problem.
i.e. on a meter they will show good regulation well within 5%, but when you
look at the rails with a scope the noise can exceed 2 volts, This 'noise' is
switching transients from inside the supply that are not removed by the Lo
ESR caps in the supply. The caps of course invariably turn out to be some of
the 'Chinese' one's suffering from the pirate electrolyte problem.
Just another 2c worth
Cheers
Paul.
Dick Cheney
Franc Zabkar wrote:
> I have a generic 400W ATX PSU whose +12V and +5V
> outputs vary significantly under different load conditions.
> Most PC PSUs regulate using a weighted average of the
> +5V and +12V rails, so that when the +12V rail rises,
> the +5V rail falls to maintain a constant average.
>
> I'm considering modifying this PSU so that it regulates
> the +5V output and allows the +12V rail to find its own level.
> This would guarantee a regulated +5V rail that was
> unaffected by changes in the +12V load. Unfortunately
> the penalty would be that the +12V rail may at times
> rise as high as 13V.
13V won't hurt anything because HDs don't use linear regulators except
possibly for low power logic circuitry, and I'd expect them to run off
the 5V anyway. The small chips someone referred to are most likely
power drivers for the motor and are frequency and duty cycle modulated.
Cheap power supplies often have badly designed output chokes that don't
help the cross regulation as much as they should.
Franc Zabkar
to keyboard and composed:
>For my part while having built supplies (linear not switch mode) with
>regulation tighter than a fishes a... I have found the biggest problem with
>PC supplies at the moment is the Chinese Capacitor problem.
I suspect these caps may have found their way into some high-end
equipment as well.
>i.e. on a meter they will show good regulation well within 5%, but when you
>look at the rails with a scope the noise can exceed 2 volts, This 'noise' is
>switching transients from inside the supply that are not removed by the Lo
>ESR caps in the supply. The caps of course invariably turn out to be some of
>the 'Chinese' one's suffering from the pirate electrolyte problem.
My ATX PSU has had very little usage, but I'll check the caps anyway.
I have an ESR meter (Bob Parker's design, sold in kit form) - it's the
most useful item in my toolkit. It comes highly recommended by the
folks at sci.electronics.repair. It can test caps in circuit.
>Just another 2c worth
>
>Cheers
>Paul.
The "regulation" formula for the PSU at
http://www.pavouk.comp.cz/hw/en_atxps.html is:
(0.327 * V5) + (0.0678 * V12) = Vref / 2 = 2.50V (nominal)
The corresponding relationship for my own PSU (based on external
measurements) appears to be:
(0.2654 * V5) + (0.09436 * V12) = 2.50V
IMO, this intentional interdependence between the two rails is
bizarre. AFAICS, the same design philosophy is used in branded PSUs.
Franc Zabkar
to keyboard and composed:
>
Apart from Antec's TruePower PSU, how can I be sure that any branded
PSU will regulate any better than my generic?
Just out of curiosity, would you please indulge me by testing your own
PSU? You should be able to use MBM5 for this purpose, assuming your
motherboard supports hardware monitoring. In particular I'd like to
know what happens if the +5V rail is heavily loaded while the +12V
rail is lightly loaded.
>Is the circuit you give exactly the same make/model as yours? If not I would
>suggest yours will be vastly different.
I'd be surprised if that were the case. I haven't worked on many ATX
PSUs, but I have repaired and modified several AT PSUs (eg 12V ->
13.8V, 5V -> 6V). They all seemed to follow similar design principles,
at least in the area of interest. As for PSUs in general, I've built a
few linears, and I've repaired probably one thousand PSUs of all
kinds, ranging from 1000W switchers to sub-1W linears, in all kinds of
equipment, from minicomputers to toasters.
>Cheers
>Paul.
Franc Zabkar
put finger to keyboard and composed:
>
>Franc Zabkar wrote:
>> I have a generic 400W ATX PSU whose +12V and +5V
>> outputs vary significantly under different load conditions.
>
>> Most PC PSUs regulate using a weighted average of the
>> +5V and +12V rails, so that when the +12V rail rises,
>> the +5V rail falls to maintain a constant average.
>>
>> I'm considering modifying this PSU so that it regulates
>> the +5V output and allows the +12V rail to find its own level.
>
>> This would guarantee a regulated +5V rail that was
>> unaffected by changes in the +12V load. Unfortunately
>> the penalty would be that the +12V rail may at times
>> rise as high as 13V.
>
>13V won't hurt anything because HDs don't use linear regulators except
>possibly for low power logic circuitry, and I'd expect them to run off
>the 5V anyway. The small chips someone referred to are most likely
>power drivers for the motor and are frequency and duty cycle modulated.
That's what I wanted to hear. I would think the voice coil amp would
still be linear, though.
>Cheap power supplies often have badly designed output chokes that don't
>help the cross regulation as much as they should.
larry moe 'n curly
Franc Zabkar wrote:
> Just out of curiosity, would you please indulge me by testing your
own
> PSU? You should be able to use MBM5 for this purpose, assuming your
> motherboard supports hardware monitoring. In particular I'd like to
> know what happens if the +5V rail is heavily loaded while the +12V
> rail is lightly loaded.
I tested either a 1.6 GHz Duron, Athlon XP1800+, or 1.7 GHz Celeron in
an ECS mobo that ran the CPU from the +5V rail, and the +5V current
ranged from about 4.5-9A. A 7200 RPM HD, CPU fan, PSU fan, and CD-RW
burner were the only loads on the +12V, and the +12V rail changed by no
more than 0.1V whether the HD and CD-RW were at idle (probably only
0.5A) or when a CD-RW disk was burned at either 4x or 8x speed (about
1.6A). The PSU was a 300W Powmax but one of the older ones that was
built decently, like a better Raidmax.
larry moe 'n curly
Franc Zabkar wrote:
> On Thu, 10 Mar 2005 11:28:27 +1300, "PC" <ab...@local.host> put
finger
> to keyboard and composed:
> > I have found the biggest problem with PC supplies at the
> > moment is the Chinese Capacitor problem.
> My ATX PSU has had very little usage, but I'll check the caps anyway.
My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
from the +5V and +12V rails, but a couple of electrolytic capacitors on
the +12V rail bulged anyway, despite powering only one HD and one
CD-RW. At first, only one of these caps tested bad with the Bob Parker
ESR meter (out of circuit because the caps were in parallel), but a
week later a second one bulged and showed high ESR. Both were the
Fuhjyyu brand, one purple, one black.
Conor
says...
> I have a generic 400W ATX PSU whose +12V and +5V outputs vary
> significantly under different load conditions. For example, at various
> times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
> good, this seemingly poor regulation is to be expected from the
> design.
Are you using a multimeter or the info the motherboard reports?
--
Conor
An imperfect plan executed violently is far superior to a perfect plan.
-- George Patton
Conor
says...
> Just out of curiosity, would you please indulge me by testing your own
> PSU? You should be able to use MBM5 for this purpose, assuming your
> motherboard supports hardware monitoring. In particular I'd like to
> know what happens if the +5V rail is heavily loaded while the +12V
> rail is lightly loaded.
>
ROFLMFAO...
Let me get this right....
You're using the readings the motherboard reports? Jesus...
Franc Zabkar
finger to keyboard and composed:
>In article <7lat2119n98886049...@4ax.com>, Franc Zabkar
>says...
>> I have a generic 400W ATX PSU whose +12V and +5V outputs vary
>> significantly under different load conditions. For example, at various
>> times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
>> good, this seemingly poor regulation is to be expected from the
>> design.
>
>Are you using a multimeter or the info the motherboard reports?
Initially I used my DMM, but I'm now relying on MBM (and BIOS). The
motherboard reports readings which are consistently lower than the
metered readings at the ATX connector by ~50mV. The resolution of the
hardware monitor is 16mV per count, which gives resolutions of 27mV
and 64mV for the +5V and +12V rails, respectively. Therefore a change
from 12.74 to 12.29 (450mV) is significant. BTW, I've calibrated my
DMM against a 5.000V +/- 0.001 reference based on Maxim's MAX6350.
See the second half of this very simple circuit consisting of just one
IC plus caps:
http://www.siliconchip.com.au/cms/A_30853/article.html
Franc Zabkar
finger to keyboard and composed:
>In article <2qqv21lo7sosa0j4b...@4ax.com>, Franc Zabkar
>says...
>
>> Just out of curiosity, would you please indulge me by testing your own
>> PSU? You should be able to use MBM5 for this purpose, assuming your
>> motherboard supports hardware monitoring. In particular I'd like to
>> know what happens if the +5V rail is heavily loaded while the +12V
>> rail is lightly loaded.
>>
>ROFLMFAO...
>
>Let me get this right....
>
>You're using the readings the motherboard reports? Jesus...
Once again you've ejaculated prematurely. See my other reply ...
Franc Zabkar
<larrymo...@my-deja.com> put finger to keyboard and composed:
So what does your experience with dud caps in a branded product tell
you? ;-)
>Both were the Fuhjyyu brand, one purple, one black.
I thought you made that up until I found their website. :-)
kony
<fza...@optussnet.com.au> wrote:
>>My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
>>from the +5V and +12V rails, but a couple of electrolytic capacitors on
>>the +12V rail bulged anyway, despite powering only one HD and one
>>CD-RW. At first, only one of these caps tested bad with the Bob Parker
>>ESR meter (out of circuit because the caps were in parallel), but a
>>week later a second one bulged and showed high ESR.
>
>So what does your experience with dud caps in a branded product tell
>you? ;-)
>
>>Both were the Fuhjyyu brand, one purple, one black.
>
>I thought you made that up until I found their website. :-)
LOL.
I wondered about the spelling but I have several of those
sitting in a scap bin, though I don't recall ever seeing the
purple ones. Offhand I vaguely recall they were used in
some other major PSU brands, maybe Sparkle for awhile.
Don't remember too many of them failing though and find it
odd/rare that there were two paralleled in the circuit.
More often it seems they'll stuff one before an inductor and
one after.
kony
<fza...@optussnet.com.au> wrote:
>>If it's to be powering a
>>*typical* PC then perhaps starting out with
>>better-than-generic-400W would be desirable? Just wondering
>>if the result would be worth the time spent, unless you only
>>needed a switching 5V supply for other project(s).
>
>During my minicomputer days, the +5V rail was deemed of utmost
>importance as it powered all the logic. In contrast, the PC platform
>seems rather lax in this respect.
The primary chipsets have that 5V stepped down now, though I
don't know what else still relys on 5V as it could certainly
vary per board.
>In fact the only PC PSU I've
>encountered that has independent regulation for each of the major
>rails is Antec's Truepower.
Is it always a positive thing though? It would seem to
inherantly limit the current on any rail. AFAIK most of the
major components now use one form or another of step down so
to a certain extent it may not matter if the voltage is
exactly right as much as the noise.
>If a "better-than-generic-400W" PSU
>regulates the same way as a generic, then I can't see how it would be
>any more accurate. I'd be interested to know how other PSUs perform
>under varying load conditions. For example, what are the +5V and +12V
>readings when the HDs and DVD ROMs have spun down, as opposed to when
>they are all spinning and seeking? How are the voltage readings
>affected by CPU load, by multimedia apps, graphics card, etc?
If no weight was placed on 12V, how would there be any way
to determine under/overvoltage in such a budget-optimized
design? Even if it's not technically as necessary to keep
12V in tight tolernaces, it's still in the ATX spec, IIRC.
>
>Do you have any numbers for your own PSU? You may be surprised.
Though I don't always trust motherboard monitoring software,
I do often have it running on various systems, pariticularly
those performing aux. functions that wouldn't be (otherwise)
monitored as closely. Some of them do spin-down drives and
some are mostly idle but others often under extended full
load... voltages don't get too far out of line. Lowest of
any system here (AFAIK) is a reading of 11.5 on 12V, though
that's ~ 0.25V lower than it reads at the connector with a
DMM.
>
>>>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
>>>+13V supply be a serious problem?
>>
>>In a recent thread someone had mentioned that one of the
>>(suspected) failure modes of Maxtor HDDs were a couple of
>>small 8 pin (SO-8) chips that were overheating. I meant to
>>take a look at some Maxtor drives to determine what those
>>might've been but suspect power regulators, maybe linear.
>>If so, raising the 12V level might raise the thermals and
>>tend to be more problematic for such drive designs... which
>>could easily be more common than only (certain) Maxtor
>>models.
>
>Yes, that's my concern also, ie whether HDs and DVDs still use linear
>power techniques.
Having just had someone's CDRW drive open today, I counted:
1 - 78L05
2 - LM317
1 - 7809
(maybe another, I wasn't really trying to enumerate them at
the time) So certainly there's the potential for linear
regulators on at least CDRW drives. This particular drive
was a MSI/Acer/et al. 40X. However I don't claim to have
tested the linear regulators on *dead* optical drives but
suspect they aren't one of the primary failure points, and
none of these low amperage SO-8 like what was (potentially)
mentioned on some HDDs except the 78L05, which was only a
supply for the front headphone opamp, IIRC.
Conor
says...
> Once again you've ejaculated prematurely. See my other reply ...
>
Actually I didn't. Get a newsreader that threads conversations
properly.
Conor
says...
> >Are you using a multimeter or the info the motherboard reports?
>
> Initially I used my DMM, but I'm now relying on MBM (and BIOS). The
> motherboard reports readings which are consistently lower than the
> metered readings at the ATX connector by ~50mV. The resolution of the
> hardware monitor is 16mV per count, which gives resolutions of 27mV
> and 64mV for the +5V and +12V rails, respectively. Therefore a change
> from 12.74 to 12.29 (450mV) is significant. BTW, I've calibrated my
> DMM against a 5.000V +/- 0.001 reference based on Maxim's MAX6350.
>
ROFLMAO...
You may want to read this article...
http://www.neoseeker.com/Articles/Hardware/Reports/motherboardvoltages/
3.html
I particularly like the following bit which describes your problem you
"think" you have:
"Another interesting trend I noticed was that percent-variance on 13
out of 14 individual rails tested increased when the board underwent
load."
"In fact, I wouldn't hesitate to bet that some people have been scared
into buying a new power supply just because of some low voltage
readings that were the fault of the motherboard rather than the
original power supply."
kony
<co...@conorturton.com> wrote:
>I particularly like the following bit which describes your problem you
>"think" you have:
>
>"Another interesting trend I noticed was that percent-variance on 13
>out of 14 individual rails tested increased when the board underwent
>load."
>
>"In fact, I wouldn't hesitate to bet that some people have been scared
>into buying a new power supply just because of some low voltage
>readings that were the fault of the motherboard rather than the
>original power supply."
That is true when someone relies ONLY on the motherboard
software for a reading, but he appears to have already
compared this to a meter reading so for this particular
isolated system it is known what the variance is.
David Maynard
True, but I'll be interested to see his reply because that article suggests
the offset is not constant over load.
Conor
> That is true when someone relies ONLY on the motherboard
> software for a reading, but he appears to have already
> compared this to a meter reading so for this particular
> isolated system it is known what the variance is.
>
Oh dear oh dear oh dear. He measured it with very little load on. The
motherboard then reported a drop in voltage under a high load. The
article points out the fact that the margin of error of the
motherboards INCREASES as the load goes up.
kony
<co...@conorturton.com> wrote:
>In article <o683311hn8vebf4m1...@4ax.com>, kony says...
>
>> That is true when someone relies ONLY on the motherboard
>> software for a reading, but he appears to have already
>> compared this to a meter reading so for this particular
>> isolated system it is known what the variance is.
>>
>Oh dear oh dear oh dear. He measured it with very little load on. The
>motherboard then reported a drop in voltage under a high load. The
>article points out the fact that the margin of error of the
>motherboards INCREASES as the load goes up.
One issue not mentioned is whether the CPU uses 12V or 5V
derived power. If 5V, the 12V load won't be changing very
much at all (until drives spin down/up), there would be no
"high load" relative to the constant load.
Franc Zabkar
keyboard and composed:
>On Fri, 11 Mar 2005 08:21:30 +1100, Franc Zabkar
><fza...@optussnet.com.au> wrote:
>
>
>>>My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
>>>from the +5V and +12V rails, but a couple of electrolytic capacitors on
>>>the +12V rail bulged anyway, despite powering only one HD and one
>>>CD-RW. At first, only one of these caps tested bad with the Bob Parker
>>>ESR meter (out of circuit because the caps were in parallel), but a
>>>week later a second one bulged and showed high ESR.
>>
>>So what does your experience with dud caps in a branded product tell
>>you? ;-)
>>
>>>Both were the Fuhjyyu brand, one purple, one black.
>>
>>I thought you made that up until I found their website. :-)
>
>
>LOL.
>
>I wondered about the spelling but I have several of those
>sitting in a scap bin, though I don't recall ever seeing the
>purple ones.
I don't know whether their current product range
(http://www.fuhjyyu.com.tw/pro.htm) is representative of their past
efforts, but the only purple/violet caps I see are the 85 degC general
purpose SK series:
http://www.fuhjyyu.com.tw/pro_sk-1.htm
>Offhand I vaguely recall they were used in
>some other major PSU brands, maybe Sparkle for awhile.
>Don't remember too many of them failing though and find it
>odd/rare that there were two paralleled in the circuit.
>More often it seems they'll stuff one before an inductor and
>one after.
Yeah, I wondered about that. A continuity test with a DMM would
suggest they were in parallel, whereas an ESR test (with Bob Parker's
meter) would see the inductor as a relatively high impedance.
Franc Zabkar
keyboard and composed:
>On Fri, 11 Mar 2005 21:25:10 -0000, Conor
The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
the fans and drives, and not much else.
Franc Zabkar
finger to keyboard and composed:
>In article <65b131d47bu9b2f1o...@4ax.com>, Franc Zabkar
That's a very interesting article. The bit that I particularly like is
this one:
"It seems that the general consensus amongst enthusiasts is that
motherboard sensors should never be trusted when looking for accurate
readings of voltages and temperatures. ... Failing to find any
conclusive evidence of these statements, we have to wonder if anyone
has actually ever tested this."
Anyway I tested my system and came to the conclusion that my PSU is
very poor, at least as far as +5V/+12V regulation is concerned. MBM's
reporting appears to be reasonably close to the mark after allowing
for voltage drops in the PSU harness and on the motherboard itself.
Some pertinent specs for my system are:
Athlon XP 2500+ CPU
ECS L7S7A2 motherboard w/ IT8705 sensor
512MB DDR SDRAM
Xabre AGP graphics
120MB HD
DVD-ROM
DVD burner
CPU core voltage (1.65V) derived from +5V rail
Xabre (or maybe just its RAM?) supplied from +3.3V rail
SDRAM voltage (2.5V) derived from +3.3V rail
IT8705 at rear LHS of PCB
20-pin power connector at front RHS of PCB
Vcore regulator at mid RHS of PCB
PSU has unused 6-way auxiliary connector (P6)
PSU has unused 2x2 12V connector
MBM5 had some initial configuration issues which I corrected by
editing the .ini files. Firstly, the -5V and -12V readings were
actually reporting the +5VSB and +3.3VSB standby voltages. Other
readings were being scaled up by about 1%, probably because the person
who submitted the configuration info was compensating for the
tolerances in his own setup. I also added the CMOS battery voltage to
the report.
To assess the accuracy of the hardware monitor chip, I measured the
CMOS battery voltage. My DMM said 3.07V, MBM said 3.06.
I compared my DMM readings against 5.000, 4.096, and 2.500V precision
references (0.02%). The results were 4.96, 4.07, and 2.48, suggesting
an error of about -0.8%.
The table below shows my test results. The voltages represent
unadjusted readings taken by MBM, and by DMM at the 20-pin ATX
connector and at the appropriate auxiliary connector. The AUX readings
reflect the voltage values inside the PSU.
Test A is for an idle system with CPUIDLE running and enabled.
Test B is for an idle system with CPUIDLE running but disabled.
Test C - CPUIDLE running but disabled, DVD audio disc playing,
full virus scan of HD
MBM5 ATX AUX Test
+3.3V 3.23/3.25 3.30 3.37 A
3.23/3.25 3.29 3.37 B
3.20/3.23 3.28 3.37 C
+5V 5.08 5.10 5.15 A
4.95/4.97 5.00 5.08 B
4.89/4.92 4.96 5.05 C
+12V 12.22 12.37 12.39 A
12.54/12.61 12.69 12.73 B
12.67/12.80 12.81 12.84 C
During test B the voltage difference between ATX ground and P6 ground
was 28.5mV, and between ATX +5V and P6 +5V was 56.5mV. The CPU fan
speed changed by about 50-90 RPM. Surprisingly, even though the fan
speed was 4000RPM (verified by strobe light), the relatively minor
speed change was clearly audible.
Conor
says...
> Anyway I tested my system and came to the conclusion that my PSU is
> very poor, at least as far as +5V/+12V regulation is concerned. MBM's
> reporting appears to be reasonably close to the mark after allowing
> for voltage drops in the PSU harness and on the motherboard itself.
Once again you fail to take into account that the accuracy of the
motherboard monitoring DECREASES with load.
Conor
says...
> The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
> the fans and drives, and not much else.
>
And as the voltages are regulated down to 3.3V and lower, 5V being a
little out isn't going to matter.
Franc Zabkar
finger to keyboard and composed:
>In article <1ti731lieuttct0a3...@4ax.com>, Franc Zabkar
>says...
>> Anyway I tested my system and came to the conclusion that my PSU is
>> very poor, at least as far as +5V/+12V regulation is concerned. MBM's
>> reporting appears to be reasonably close to the mark after allowing
>> for voltage drops in the PSU harness and on the motherboard itself.
>
>Once again you fail to take into account that the accuracy of the
>motherboard monitoring DECREASES with load.
If you are referring to ohmic losses in the conductors and in the
motherboard power planes, then clearly this is evident in the results.
I've even measured the voltage drops in the cable. I've also pointed
out the positions of the various circuit blocks so that the reader
could assess the potential for inaccuracies. For example, the ATX
power connector and monitoring chip are at diametrically opposite ends
of the PCB, which is potentially a worst case scenario for voltage
measurements. However, even allowing for these voltage drops, the
hardware monitor appears to be close to the mark.
If you believe there is some factor other than ohmic loss which is
causing inaccuracies, then what could it be?
Anyway, regardless of inaccuracies, the article you referred to proves
the point that it set out to test, ie that voltage measurements
reported by motherboard monitoring software do not always accurately
reflect the actual output of the PSU, nor can they. I've said as much
in several prior discussions in Usenet. It is for precisely this
reason that I've taken the time to assess the accuracy of my own
setup. If I were the author of that test, I would have taken the next
step and investigated *why* some results (eg AOpen AK79D) were way off
the mark. In my case, I downloaded the datasheet and a reference
circuit for my monitor IC and used the information to properly
configure MBM.
http://www.ite.com.tw/product_info/file/pc/IT8705F_V.0.4.1.pdf
AOpen's AK79D appears to use Winbond's W83627HF IC:
http://www.winbond.com/PDF/sheet/w83627hf.pdf
This IC directly monitors the +5V rail via on-chip scaling resistors,
so I would have thought that its accuracy would be better than mine
(the IT8705 requires external resistors). Anyway, if that were my
system, I'd be trying to find a reason for the 0.5V drop in the 5V
rail. Perhaps the motherboard has poorly designed power planes??? Is
the ATX power connector showing signs of discolouration? How does MBM
report the same parameters?
Having said all the above, note that my +12V rail, as measured by MBM,
INCREASES when the +5V rail is loaded, even though the load on the
+12V rail does not change. Clearly this points to a regulation problem
in the PSU. I have since dismantled the unit and can confirm that it
regulates by sensing a weighted average of its +5V and +12V outputs.
This explains why the two rails move in opposite directions.
Franc Zabkar
finger to keyboard and composed:
>In article <fse5311b9mf5uu3r2...@4ax.com>, Franc Zabkar
>says...
>
>> The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
>> the fans and drives, and not much else.
>>
>And as the voltages are regulated down to 3.3V and lower, 5V being a
>little out isn't going to matter.
It definitely won't matter to the CPU, but there are other devices
which depend on this rail. Even so, you may be right, it may not
matter to them either. I come from the old school where every supply
voltage could be trimmed. In fact, flaky intermittent problems could
sometimes be narrowed down by misadjusting a particular supply rail.
Overclockers also recognise that minor supply adjustments can
sometimes improve the stability of their setup.