Help hard drives keep clicking and dying

[Arno Wagner]

Arno looked up acceptable ripple voltage having finally grasped what
was posted so many times previously.

Wrong. As anybody with reasonable experience with electronics knows,
ripple and DC average have to be considerd separately. I just did
not make your beginners mistake...

Numbers he should have know when this discussion first started.

See above.

Ripple voltage should not exceed 120
mv. Therefore when power supply is failing (even overloaded), then
ripple voltage increases.

Hahaha! This is completely dependent on the failure mode. There
is no need at all for the ripple to increase when a PSU fails.
You are reading the implication wrong.

Correct: Ripple > 120mVpp ==> PSU broken
Your wrong statement: PSU broken ==> Ripple > 120mVpp

Arno tells us that ripple voltage was not
excessive when meter reading suggests otherwise.

Huh? I just said the ripple was completely unknown given the data
at hand. And I maintain that this is the case.

Well that 'exampled'
power supply now with 300 mv of ripple measures 4.8 volts. Why 300 mv?

With 300mVpp, the voltage oscillates between 4.95 and 5.25V with a DC
average of 5.1V. That assumes symmetric ripple, of course, which
you usually do not have with switching mode PSUs.

It also seems that you still do not understand what peak-to-peak
means...

Because power supply is failing. According to Arno, that reading is
completely normal when he cites 4.75 volts as acceptable. Arno ignores
how meters operate and why ripple voltage increases - becomes too
great.

Why would the ripple increase? It is not clear that the PSU is defect.
Even if it were, that would not necessary imply increased ripple.

Meanwhile, a voltage below 4.87 suggested a defective power
supply.

No. Not at all.

In the OP's case, 11.655 also suggests a defective supply for same
reasons. A conclusion that is confirmed by a voltage measurement when
disk drives are removed. Also suggesting a failure is excessively low
12 volts while 5 volts is unusually high - 5.12.

Arno now says the OP must use an oscilloscope. Bull.

And wrong. I did not say that. I just said that measuning the
ripple requires one. My advice to the OP is to try with a
different PSU. An oscilloscope is not only expensive, but
requires experience and understanding to interpret the
measurements taken with it.

Arno again
displays lack of experience. Yes, we often used an oscilloscope when
it was available. Therefore we learned how meters work. Many here
clearly don't have that experience and therefore deny acceptable
numbers. An oscilloscope does nothing for the OP. Numbers such as
3.23, 4.87, and 11.7 take into account both excessive ripple voltage
AND how meters work.

They don't, except one follows your completely arbitrary assumption
of 600mVpp ripple. What if the output filter capacitor is broken
and the ripple is higher, e.g. 5Vpp? Then your "obvious" approch
goes right out the window.

The OP does not need a scope. Taking the same
measurements of all voltages (purple, red, yellow, and orange wire)
with disk drives removed will confirm the problem. OP was informed by
one who did this stuff for a few decades - at the design level.

Well, now I understand why we have all these problems with PSUs.

Number
of contrary replies demonstrates how many just know - did not first
obtain experience - or even use an Oscope.

This is more voodoo than elecrical enginnering. Of course you can specify
extreme health conditions and throw anything away that does not match
them. Sometimes this will work. It will also be pretty good if you
sell replacement PSUs to non-expert customers. But in th latter
case it is higly inethical and borderline criminal. Unfortunately
this type of business practice is still widespread in the PC
industry.

Arno confused the reader by claiming the power supply is working
normally -

Wrong. I said it is not possible to tell with the available data...

assumes that ripple voltages are normal in a failing supply.
Meter suggests otherwise. Arno makes ripple voltage assumptions so as
to not admit to 'measured limits' - 3.23, 4.87, and 11.7 volts. Meter
is suggesting a power supply so overloaded as to be defective - and to
output excessive ripple voltage.

Well, seems to me you do not have much experience with PC PSUs.
The transformers are frequently not wound precisely.

Arno agrees the power supply is suspect. One voltage apparently
cannot provide sufficient power and/or has excessive ripple voltage.
Either way, that would explain disk drive problems. Meter in but a
minute can quickly identify failures if correct spec numbers are used -
as I have provided. Just because a meter reports the Intel spec number
does not say a power supply is functioning properly. And that
unusually high 5 volts in combination with excessively low 12 volts
further suggest a problem.

Or too few windings on the 12V path. Have seen that several times in
PC PSUs that worked fine.

Two factors define why a disk drive is failing intermittently (and
therefore now must be tested with a disk drive hardware diagnostic).
One: 11.655 volts is too low. Two: 5 volts is unusually high when the
12 volts is so low. That is what my original post stated when but a
few others also warned the majority of their technical errors. Notice
we had a classic example of the large majority being so wrong. In
computer newsgroups, a clear majority don't even have basic electrical
knowledge. 11.655 volts suggests the power supply is failing - in
direct contradiction to what a majority without sufficient experience
so voraciously denied.

Ok, I have enough of this.

I have sufficient experience. I say you have no real clue and are using
some black magic rules-of thumb you do not really understand yourself.

The claim that you can determine PSU health with a meter alone is
completely ridiculous. You continue to ignore that the +12V line
has 10% acceptable tolerance, most likely because it blows your
argument away. You have not understood what Vpp (peak-to-peak
Voltage) means, since you continue to use it wrongly in examples.
Then you claim that a failing PSU will have increased ripple, while
this is only one potential failure mode and one that cannot be
detected with a DC meter.

I don't know why you insist on your wrong calculationw and statements.
Possibly because you would need to find out that you did it wrong all
those "decades of experience". I also do not really care. Don't get me
wrong, your approach has some chance of working. But your reasoning is
wrong and you will miss some serious problems with it. Also you will
jump to conclusions that the facts do not support, as you have done
in this "discussion" here.

I don't really need to be a "better engineer" than you, but I
resent when people post wrong information, prying on those that
do not have the expertise that allows them to see what is
valid and what is not. Stop doing so.

Arno

 

[Paul Rubin]

Or too few windings on the 12V path. Have seen that several times in
PC PSUs that worked fine.

These PSU's aren't regulated using feedback against an internal
voltage reference?

 

[w_tom]

. My advice to the OP is to try with a different PSU.
An oscilloscope is not only expensive, but requires
experience and understanding to interpret the
measurements taken with it.

In that long post, you insist power supply voltages are OK. Why then
do you recommend he replace the PSU? Why should he replace a PSU when
you claim his power supply voltages are just fine? Do we call that a
contradiction - or a concession that 11.655 is too low?

Nobody said he should use an oscilloscope. An oscilloscope requires
so much "training" that I was using one in the days of John Kennedy.
How did I use something so complex - especially since I have no idea
about 11.655 volts. And clearly the OP should replace his PSU when
Arno knows all voltages were OK. What a contorted world we live in
when facts need not comply with conclusions.

 

[Arno Wagner]

These PSU's aren't regulated using feedback against an internal
voltage reference?

They are, but +3.3V, +5V and +12V are regulated together, i.e.
a mix is regulated.

Arno

 

[Arno Wagner]

In that long post, you insist power supply voltages are OK.

I do not. Quite obviously and directly, actually. You seem either not to
have read or understood what I wrote.

Arno

 

[Arno Wagner]

or a concession that 11.655 is too low?

BTW, why do you keep posting "11.655V"? That would require a 4+1/2 digit
DMM. Quite a professional instrument and expensive. The OP has written
11.65V.

Arno

 

[Paul Rubin]

They are, but +3.3V, +5V and +12V are regulated together, i.e.
a mix is regulated.

I'm having a hard time picturing how this is done, given that the mix
of loads for the different voltages can vary tremendously.

 

[Arno Wagner]

I'm having a hard time picturing how this is done, given that the mix
of loads for the different voltages can vary tremendously.

And there you have identified the problem with this design...

It is not quite as bad as it sounds, since the different voltages
are all derived from one transformer. If the windings are precisely
matched, you get the right AC output voltages, evem if only one
output is regulated. The reason is that the load on that output
will weaken the magnetic field and decrease the voltage on all
outputs in the same fashion.

So far the theory. In practice there are differences that are not
coupled trhough to the other windings by the magnetic field in the
thransformer core. One is winding resistance ("copper loss"), which is
a normal ohmic loss. More current - less voltage. No effect on the
other voltages. An other is the output rectifier. After all we cannot
run the AC into the mainboard, have to make that DC first. The loss
voltage is logarithmic in the current. On first glance this sounds
good. But on second it is really a problem, since it makes the whole
system non-linear and differently so for each ouput voltage. Then
there is the copper loss in the output indictor used to in this
type of switching regulator. This loss is also independent
for each output.

Now the critical thing is, that you can minimize these losses by
spending more money, primarily on the transformer and inductors. For
obvious reasons (cheap, cheap design) this is only done so far that it
just about works. The allowed tolerances (and for +12V it is 10%, you
can check in the standard yourself) are usually needed. The PSUs are
usually designed for 50% load, so at extreme conditions (high or low
loads) the voltages will be less precisely met.

Basically, I think this design sucks badly. But it can be built
cheaply and I have seen better designs so far only in high-load
quality PSUs.

Arno

 

[Rod Speed]

Arno Wagner wrote

In that long post, you insist power supply voltages are OK.

No he didnt, he JUST said that your stupid pig ignorant
claim about what the rails must read with a 3.5 digit
multimeter to be within spec is just plain wrong and
that that flaunts your complete pig ignorance in spades.

Why then do you recommend he replace the PSU?

Because the OP got two drives clicking and that while
that is possible with a coincidence, its an easy test for
the possibility that the PSU is out of spec, even tho the
two rail voltages measured later are within spec.

Why should he replace a PSU when you
claim his power supply voltages are just fine?

See above.

Do we call that a contradiction - or a concession that 11.655 is too low?

Neither, just a sensible precaution with two drives clicking.

Nobody said he should use an oscilloscope.

The ATX spec says just that if you want to see the ripple level.

An oscilloscope requires so much "training"

Only for the terminal boneheads like you. It aint that hard,
particularly for measuring ripple on a power supply.

that I was using one in the days of John Kennedy.

I was using one well before that.

How did I use something so complex - especially
since I have no idea about 11.655 volts.

Even a terminal bonehead can be 'trained' to use something like
that when he cant manage the basics like what the multimeter
will read when the ripple is out of spec with an ATX power supply.

You're so stupid that you cant even manage to grasp
that a multimeter measures the AVERAGE voltage
on the rails, not the DC level plus ripple peak.

And clearly the OP should replace his PSU
when Arno knows all voltages were OK.

He never ever said that the voltage are
always ok when not being measured, cretin.

What a contorted world we live in when facts need not comply with conclusions.

You in spades with your pig ignorant drivel about what the
multimeter will read when the ripple level is out of spec.

Not a ****ing clue about even something as basic as that.

 

[chrisv]

Again, spec numbers of 3.1, 4.75, and 11.4 means the meter must
report numbers above 3.23, 4.87, and 11.7 volts.

You are cracked.

No mystery here.

Actually, your assertions are quite mysterious.

 

[Mike Tomlinson]

You have just identified the problem. 12 volts is too low. Also the
5 volts being on the high side implies the power supply has hit its
limit.

You're talking complete crap again w_tom. 11.65 is not too low for 12v
and 5.12 is not too high for 5v. Go and read the ATX specs again.
Wanker.

 

[Mike Tomlinson]

And wrong. I did not say that.

w_tom's standard technique, when cornered, is to start lying and
twisting the meaning of previous posts to support his argument.

 

[Mike Tomlinson]

Also the unusually high five
volts was another fact that suggested defective supply.

It's only "unusually high" according to you. Not to the ATX spec or to
people who have far more real-world experience of these things than you.

Again, a case of you making assertions without the factual evidence to
back them up - something you are very quick to accuse others of.

And your use of top-posting to avoid responding to specific points
raised by others - while you are happy to use bottom posting when it
suits you - only serves to demonstrate your profound dishonesty.

Don't you ever wonder why every time you post you're shot down in flames
by those who can see straight through your overt disingenuousness and
dishonesty? Doing so repeatedly - you've done it for many years -
suggests some form of mental disorder to me. May I suggest you seek
help from a qualified professional, preferably someone who will be able
to resist the temptation to beat the shit out of you once they've
suffered more than ten minutes in your company?

 

[Arno Wagner]

w_tom's standard technique, when cornered, is to start lying and
twisting the meaning of previous posts to support his argument.

Yes, and ignoring anything that exposes his errors, no matter
how clear and concrete. I had forgotten that we have seen this
guy here before. I think I even got into an argument with him
last time...

By now I think that "I have been involved at the design level"
is either a complete lie or he was the lab assistand doing the
measurements, but none of the interpretation or decisions.

Thanks for the remainder.

Arno

 

[w_tom]

You're talking complete crap again w_tom. 11.65 is not too low for 12v
and 5.12 is not too high for 5v. Go and read the ATX specs again.
Wanker.

But again, same facts reposted for the many that did not grasp the
details. A meter measuring 11.65 is probably measuring a voltage that
repeatedly drops down below 11.4 volts. They are not RMS meters. If
the meter says less than 11.7 for a loaded 12 volts, and especially
when it reads that number when 5 volts is above 5.12; then power supply
probably is probably defective.

Also what Mike did not read is what follows as a result of that meter
reading. Mike - why do you only read one paragraph and then respond?

Many magically assume that a meter reading 11.65 means voltage is
only 11.65. Nay-sayers should have first learn how meters measure
before only reading numbers from Intel specs sheets.

Intel spec sheets says 11.4 minimum for 12 volts. Therefore my post
was quite accurate. A power supply under full load that is failing -
that is repeatedly serving up less than 11.4 volts will also measure
*above* 11.4 volts on that meter. A reading of 11.65 volts - and again
- especially when the 5 volts reads 5.12 - strongly suggests a
defective power supply. Mike - learn why RMS meters cost many times
more money. Learn the technology before attacking.

Meanwhile, voltages on that meter reading below 3.23, 4.87, and 11.7
suggest power supply failure. What does one do next? That is posted
eariler in paragraphs that so many nay-sayers did not read or
comprehend.

Mike - I understand why Arno spouts attacks. He repeatedly
demonstrated a naive grasp of electrical principles. But I thought you
had better experience. You actually believe a meter reading 11.65
means voltage is constanly above 11.4? Simple use of an oscillscope,
some experience, or why RMS meters cost so many times more money -
each would make it obvious why 11.65 volts and 5.12 indicate a failing
supply.

 

[w_tom]

w_tom's standard technique, when cornered, is to start lying and
twisting the meaning of previous posts to support his argument.

But again, same facts are reposted for the many that did not grasp
details. Mike - did you bother to read the entire quote before falling
for sound byte reasoning posted by Arno? Did you ever do power supply
designing before claiming superior experience? Let's teach you what
you should have known before posting. A meter measuring 11.65 is
probably measuring a voltage that repeatedly drops down below 11.4
volts. Therefore computer suffers intemittent or repeated failures. If
the meter says less than 11.7 for a loaded 12 volts, and especially
when meter reads that low 12 volts when 5 volts is above 5.12; then
power supply probably is probably defective.

Mike also did not understand what following paragraphs were
describing; actions as a result of that meter reading. Mike - why do
you only read one paragraph and then respond in expert mode? Why do
you assume a meter reading 11.65 volts means voltage is always 11.65?
Did you read how meters work - or just attack?

Many magically assume that a meter reading 11.65 means voltage is
only 11.65. Many computer experts don't even use meters - because they
don't know how to read them - what the numbers are really saying.
Those nay-sayers should first learn how meters work before just
assuming numbers from Intel specs sheets. Lurkers should appreciate
how many here just know - did not first learn the technology. Many who
are accusing also recommend shotgunning - because they never learned
how to use the meter. But somehow they magically know 11.65 on that
meter is OK.

Intel spec sheet says 11.4 minimum for 12 volts. A power supply
under full load that is failing - that is repeatedly serving up less
than 11.4 volts - will also measure *above* 11.4 volts on that meter.
11.65 volts - and again - especially when the 5 volts reads 5.12 -
suggests a defective power supply. Mike - learn why RMS meters cost
many times more money. Don't post as ignorantly as Arno. Instead
learn the technology before attacking.

 

[Rod Speed]

Mike Tomlinson wrote

But again, same facts are reposted for the many that did not grasp details.

And there you go again, doing precisely what he said you do.

Mike - did you bother to read the entire quote before
falling for sound byte reasoning posted by Arno?

Not necessary, even just the mindless pig ignorant drivel about
what the rail voltages should read when read with a multimeter
proves to anyone with a clue that you dont actually have a ****ing
clue about what the measurement of power supplys is about.

Did you ever do power supply designing before claiming superior experience?

Dont need any of that to see that your claims are pure drivel.

Let's teach you what you should have known before posting.

You can repeat this drivel till the cows come home, changes nothing.

A meter measuring 11.65 is probably measuring a
voltage that repeatedly drops down below 11.4 volts.

You dont know that. ALL you know is that the average reading is 11.65.

AND you plucked those numbers you claimed should be read out
of your arse and have mindlessly pig ignorantly added the allowed
ripple to the minimum rail voltages the ATX standard allows, with
the exception of the 12V rail where you didnt even notice that its
allowed a margin of 10% as maximum load on that rail.

All the rest of your stupid pig ignorant repetetive drivel flushed where it belongs.

 

[Arno Wagner]

Just a note for all others here: A complete and detailed explanation
why w_tom does not have a clue (despite sounding to some like he has)
is contained in the other postings of this thread. Why w_tom insists
on his peculiar understanding of PSU mechanics and electrical
measurements is beyond me.

Site note: Below he demonstrates that he does not understand the
term "RMS".

Arno

But again, same facts reposted for the many that did not grasp the
details. A meter measuring 11.65 is probably measuring a voltage that
repeatedly drops down below 11.4 volts. They are not RMS meters. If
the meter says less than 11.7 for a loaded 12 volts, and especially
when it reads that number when 5 volts is above 5.12; then power supply
probably is probably defective.

Also what Mike did not read is what follows as a result of that meter
reading. Mike - why do you only read one paragraph and then respond?

Many magically assume that a meter reading 11.65 means voltage is
only 11.65. Nay-sayers should have first learn how meters measure
before only reading numbers from Intel specs sheets.

Intel spec sheets says 11.4 minimum for 12 volts. Therefore my post
was quite accurate. A power supply under full load that is failing -
that is repeatedly serving up less than 11.4 volts will also measure
*above* 11.4 volts on that meter. A reading of 11.65 volts - and again
- especially when the 5 volts reads 5.12 - strongly suggests a
defective power supply. Mike - learn why RMS meters cost many times
more money. Learn the technology before attacking.

 

[Arno Wagner]

How did you deduce potential 'excessive ripple' from a single DC
voltmeter reading?

That is what everybody else is wondering too.

Arno

 

[Arno Wagner]

Mike - learn why RMS meters cost many times more money.

In a sense, this is a gem. w_tom took the word RMS from my posting
without the slightest understanding what it means. An RMS meter
will not help in the least to determine ripple voltage. The only
true thing here is that they are expensive...

Arno