new system bootup woes

[Paul]

How the hell does one rip caps of the board during assembly??


(.....and why would someone think it was still under warranty after doing so?!)

I think he's referring to the owner using a screwdriver
to push on something. The screwdriver slips, and hits
the motherboard. If a surface mount ceramic is in the
path of the screwdriver blade, it's scraped right off the board.
Sometimes, if you're lucky, you can compare a reference
photo, to the mess someone made, and spot the missing
small ceramic cap or surface mount resistor.

You can see someone here, rolling the dice, with
screwdriver poised for action.

http://www.techimo.com/forum/attach...082242512-installing-heatsink-fan-amd-hsf.jpg

Some socket families, the retention is done in a way that
screwdrivers are not required, and the risk goes down
for a scraping injury.

On some of those dangerous cases, you use a hex driver instead
of a flat blade screwdriver, as a means to get a grip on
the clamp. It all depends on the shape of the metal on the
end. This one, for example, is just begging for a slothead
screwdriver. Fortunately, the first picture is the passive end,
and the other side has the rotating lever for finishing the job.
So the user doesn't need all of their strength, to guide
the passive end into place, under the tab.

http://www.ninjalane.com/images/install939/heatsink_clip1.jpg

http://www.ninjalane.com/images/install939/heatsink_clip_close.jpg

Paul

 

[Kirk_Von_Rockstein]

I've had a screwedriver slip which hit the mobo and cut a trace. The
coating makes resoldering difficult so you have to find a cheat on how
to reconstruct the broken trace.

There are lots of defects to check for when inspecting a mobo.

On that note,
I had recieved a motherboard that had a very very fine
shallow narrow scratch on the backside that had
cut through several of the very very small close
traces which were not noticable until I studied
the board with a high magnification glass.
The area of the damage was so thin, small and whited
that I had to apply a little water to actually
see through the damaged insulation
to see clearly that the traces were infact cut
and that it was not just the insultion layer
that had been penatrated alone. Very informative thread.

 

[DecadentLinuxUserNumeroUno]

The "rice grain method" ?

No... It's the Oh COME ON USE A LITTLE COMMON SENSE, METHOD.

Can you really be that dumb?

 

[DecadentLinuxUserNumeroUno]

Before you press the button on the front of the computer,
the power supply is not able to monitor what kind of load may be present
on the supply outputs.

Think about what you said.

BEFORE you press that button, it doesn't need to.

WHEN you press that button, it most certain can tell.

If the slew rate for the current signature rises too fast, it will shut
off before it even reaches it's rated limit.

IOW, before it even turns on.

Another reason to segregate the PS from the rest of the system. If it
works, THEN you have a problem at the motherboard location.

Trying to make that determination with them paired is ludicrous.

 

[DecadentLinuxUserNumeroUno]

On that note,
I had recieved a motherboard that had a very very fine shallow narrow
scratch on the backside that had cut through several of the very very
small close traces which were not noticable until I studied the board
with a high magnification glass.

That would have to be some scratch, considering the basic cladding
thickness of a typical circuit board trace.

More like a gouge.

 

[Jonathan N. Little]

How the hell does one rip caps of the board during assembly??

Some motherboards have very poor layouts which put the tall electrolytic
caps so close to the CPU heatsink where in combination with those spring
loaded across the heatsink retainers that clip to the socket, one slip
can cause yo to slip and bend the caps over. AMD motherboard love to use
those old style heatsinks like the 486, P1 days. Personally I prefer the
bolt through the board heatsinks. Seen some boards where the caps were
actually touching the the socket's mount point such that the only way to
put on the heatsink was to bend the caps.

(.....and why would someone think it was still under warranty after doing so?!)

You'd be surprised. My daughter had to explain to a customer that
Gorilla Glue was not a suitable thermal paste and voided the warranty.
(not a joke)

 

[Jonathan N. Little]

Thanks, definitely not a mobo/case shorting problem.
See my earlier post for more details/measurements.

RMA time.

 

[Jonathan N. Little]

That would have to be some scratch, considering the basic cladding
thickness of a typical circuit board trace.

More like a gouge.

More typical issues are cold solder joints and poorly soldered
feedthroughs, (the tiny plated holes that connect traces on different
layers as circuits meander through and across the board).

 

[DecadentLinuxUserNumeroUno]

More typical issues are cold solder joints and poorly soldered
feedthroughs, (the tiny plated holes that connect traces on different
layers as circuits meander through and across the board).

Vias used to be a problem. Folks doing layouts used to have to worry
about "via count". Modern PCB mfg tech insures nearly any 4+ layer board
is going to have viable vias. Nobody even counts any more, and
diagnosticians rarely suspect them any more.

With RoHS, cold solder is the predominant failure mode.

I have a USB stick I could have sworn I broke by flexing the connector
header attachment to the internal PCB. Looking at it, my flexure
apparently broke free one or two pins from the connector side row of one
of the two memory chips. Once I reflow that, I can prove it, but without
a microscope, I can see what appears to be unsoldered feet on a couple
pins.

I used to do post reflow inspection too, so I am reasonable certain that
I am looking at non-connected pins. I need a scope though and a soldering
iron from some box in the garage. I have the $200 a roll lead free solder.
(even though repairs are allowed to be with leaded solder).

 

[Adam]

Adam wrote:

Did you spread a thin transparent layer of thermal paste on the cpu or
did you glop it on like paint?

The nice thing about AMD CPUs is they come with
thermal paste pre-applied at the factory.
No need to worry about thermal paste unless reinstalling the CPU,
which is my case now. Ick!!!

With the PSU connected to the mobo, do both the CPU and PSU fans even
jerk or twitch a bit when you short the PWR pin on the mobo's front
panel header?

No twitch from PSU fan.

If CPU fan speed is too low or zero, the BIOS will shutdown the system.
This is to prevent the CPU from burning up. Did you separately test the
CPU fan connected to a 12V output from the PSU to make sure it will spin
under power? Did you connect the CPU fan to the correct mobo header?
Are you trying to use a 3-wire CPU fan on a 4-pin mobo CPU header?
While the thermal checks in BIOS may take a few seconds to kick in, like
you forgot the thermal paste or the heatsink wasn't put onto the CPU, a
zero RPM CPU fan often causes immediate shutdown.

4-wire CPU fan is connected to 4-pin CPU_FAN header.

When you did the PSU+HDD test (no mobo) and when shorting PS_ON to
ground which got the PSU fan spinning, did you test the 5V standby
voltage from the PSU? Even when you "power off" the computer but as
long as the PSU is connected to an A/C power source, the PSU will
continue to supply 5V as standby voltage to the mobo. The old AT PSUs
used a case switch that went directly to the PSU. The ATX PSUs rely on
soft power logic on the mobo to tell it when to fully power on. That
is, the case switch goes to the mobo to have logic on the mobo decide
when the PSU will fully power up. With the computer "off", the PSU
should still output 5V (aka 5VSB). Without the 5V standby voltage from
the PSU, the mobo logic is dead.

Did you yet get a multimeter so you can check if the PSU, when connected
into an A/C power source, is putting out 5V standby voltage (both when
it is disconnected from and also when connected to the mobo)? It's the
"+5VSB" line shown in the linked photo I gave before.

Yes, the +5VSB purple SB_PWR has never been a problem as
the green LED is lit whenever the PSU is switched ON.

 

[mike]

Before you press the button on the front of the computer,
the power supply is not able to monitor what kind of load
may be present on the supply outputs.

From T=0.0ms to T=35ms, the current limiter is turned off.
This is to give the supply time to charge the output capacitance.
The supply is allowed to have around 5000uF on the load side,
and remain stable from a control loop perspective. For example, the
motherboard has 100uF on each USB header pair, for inrush,
and all of those have to be charged.

It's only after T=35ms, that the current limiter is enabled and
the output current monitored.

In that brief 35ms period, the available power (if not completely
shorted out) can cause the fan motor to start spinning. That's
where the "twitch" comes from, on a power supply which is
capable of driving a load, but is eventually overcome by a
short circuit.

Paul

Thanks for the very informative explanation.
Just for fun, I took the lid off an ATX power supply and measured
zero ohms between the + fan terminal and the yellow 12V output wire.
So, "completely shorted out" is a real possibility, and the fan shouldn't
twitch under those conditions.
I never saw an answer to my question about measuring ohms
at the motherboard power connectors.

 

[Dustin]

DanS <[email protected]>
Tue, 23 Dec
2014 12:24:15 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:

How the hell does one rip caps of the board during assembly??

They didn't rip them off. They provided power from a bad ps that was
providing well over 12 volts.. they.. blew the caps. heh.

(.....and why would someone think it was still under warranty after
doing so?!)

That, I don't know. I'm not them.

 

[Dustin]

Paul <[email protected]> Tue, 23 Dec
2014 13:58:06 GMT in alt.comp.hardware.pc-homebuilt, wrote the
following message:

I think he's referring to the owner using a screwdriver
to push on something. The screwdriver slips, and hits
the motherboard. If a surface mount ceramic is in the
path of the screwdriver blade, it's scraped right off the board.
Sometimes, if you're lucky, you can compare a reference
photo, to the mess someone made, and spot the missing
small ceramic cap or surface mount resistor.

You can see someone here, rolling the dice, with
screwdriver poised for action.

http://www.techimo.com/forum/attachments/processors-memory-overclock
ing/10602d1082242512-installing-heatsink-fan-amd-hsf.jpg

Some socket families, the retention is done in a way that
screwdrivers are not required, and the risk goes down
for a scraping injury.

On some of those dangerous cases, you use a hex driver instead
of a flat blade screwdriver, as a means to get a grip on
the clamp. It all depends on the shape of the metal on the
end. This one, for example, is just begging for a slothead
screwdriver. Fortunately, the first picture is the passive end,
and the other side has the rotating lever for finishing the job.
So the user doesn't need all of their strength, to guide
the passive end into place, under the tab.

And thats the other way of removing one. I was actually referring to
some one using a ps that was no longer holding no more than 12 volts
on the rail. This one jumped a bit; burned a few caps up in the
process.

 

[Dustin]

"Jonathan N. Little" <[email protected]>
Tue, 23 Dec 2014 14:57:51 GMT in
alt.comp.hardware.pc-homebuilt, wrote the following message:

You'd be surprised. My daughter had to explain to a customer that
Gorilla Glue was not a suitable thermal paste and voided the
warranty. (not a joke)

Ouch

 

[Paul]

Thanks for the very informative explanation.
Just for fun, I took the lid off an ATX power supply and measured
zero ohms between the + fan terminal and the yellow 12V output wire.
So, "completely shorted out" is a real possibility, and the fan shouldn't
twitch under those conditions.
I never saw an answer to my question about measuring ohms
at the motherboard power connectors.

While we characterized some products at work, with an ohms
test of their power input, I wouldn't even consider it on random
motherboards. You don't have enough units to sample and get
a statistically satisfying number. Should the number be 100 ohms,
or should it be 3 ohms ? Can you justify why that number is
appropriate ? Etc.

The board I worked on at work, it would measure 3 ohms on the 3.3V
rail. And the board had probably a thousand pullups, and little loads
on it (diff ECL section). A motherboard, by comparison, is too cost sensitive
to be that wasteful. Our product, the ohms measurement was pretty
consistent from unit to unit. So in our case, a dead short on 3.3V
to ground, would be different than the 3.0ohms you'd read when
the thing was fault-free. I didn't even attempt to justify
where that number comes from, as I had more on my plate to do than
that. Like, make it work. That test was intended as part of
incoming inspection, so you wouldn't waste time on the defective
ones, and only work on the ones that looked like good candidates.

It's also pretty hard to ohm something, with 1000uF of
capacitance across it. In any case, for us, it was simple
chance that we noticed our board (assembly) was easy to read,
and consistent from unit to unit. So we could throw it in as
a quick check. Before connecting power to it. That thing
probably drew 100-150W or so. And had externally applied
forced-air cooling.

Paul

 

[VanguardLH]

The nice thing about AMD CPUs is they come with
thermal paste pre-applied at the factory.
No need to worry about thermal paste unless reinstalling the CPU,
which is my case now. Ick!!!

That's not thermal paste. That's thermal tape. Under compression and
at heat, its elasticity changes to make it more liquid and conform to
the irregularities between CPU plate and heatsink. It was designed to
provide a thermal compound with the CPU (rather than including a tube of
paste) and eliminate neophytes that would apply too thick a layer of
paste (which can actually eliminate the much preferred metal-to-metal
contact). Paste (properly applied) is better than tape. Still, good to
know that there was some thermal compound between CPU and heatsink.

If you end up removing the CPU from the heatsink, the thermal tape is
non-reusable. You will have to get more thermal tape or clean both
surfaces (99% isopropyl alcohol should work) and use paste.

No twitch from PSU fan.

And why it could be the soft-logic on the mobo that is defective in not
telling the PSU to fully power up when you short the PWR mobo header pin
to GND (you also mentioned the +5VSB line was okay that powers the soft
logic on the mobo). The soft-logic on the mobo pulls low the PS_ON line
which tells the PSU to power up. If the PSU is supplying the required
5VSB to the soft logic and if shorting PWR pin to ground doesn't drop
PS_ON to low then there's a problem on the mobo.

As you noted, with the PSU not connected to the mobo, dropping PS_ON
turns on the PSU (fans spin, 12V and 5V outputs go live). Doesn't look
like a PSU problem (unless it cannot handle any load). With the PSU not
connected to the mobo, have you connected an HDD on the PSU to put some
load on it?

4-wire CPU fan is connected to 4-pin CPU_FAN header.

Did you separately test the CPU with a 12V supply (e.g., car battery
with car NOT running or detached PSU with PS_ON grounded) to make sure
it will spin? No spin = no RPM = BIOS immediately shuts down computer.

I have received new fans (not used, not refurbished, not lot sales) that
were defective on arrival. Many have the fan hub (center and blades)
pressed on and I've had a couple that weren't pressed fully onto the
spindle. The result was the fan hub was slightly atilt and the blades
would hit the fan's case to bind up. I've had some that you could tell
there was a friction problem when rotating the blades by hand (the
scratching could be felt). I've had where the wiring broke loose from
the solder joints on the connector. Make sure the CPU fan will spin by
itself when supplied with 12V power.

 

[DecadentLinuxUserNumeroUno]

On Tue, 23 Dec 2014 20:33:29 +0000, Dustin wrote:

snip

And thats the other way of removing one. I was actually referring to
some one using a ps that was no longer holding no more than 12 volts on
the rail. This one jumped a bit; burned a few caps up in the process.

Ceramic caps? From just higher than 12 volts. No way. Even the ELs
are rated at 16 or 20 volts in those circuits. And those would take way
higher than that to blow them.

 

[Godzilla]

On Tue, 23 Dec 2014 20:33:29 +0000, Dustin wrote:

snip


Ceramic caps? From just higher than 12 volts. No way. Even the ELs
are rated at 16 or 20 volts in those circuits. And those would take way
higher than that to blow them.

Nice newsreader, Fixadent.
Good job.
If your gonna use Linux, suffer with the rest of us.

--
_____ _______ ____ __ __ _____ _
/ ____|__ __/ __ \| \/ | __ \ | |
| (___ | | | | | | \ / | |__) | | |
\___ \ | | | | | | |\/| | ___/ | |
____) | | | | |__| | | | | | |_|
|_____/ |_| \____/|_| |_|_| (_)

 

[mike]

While we characterized some products at work, with an ohms
test of their power input, I wouldn't even consider it on random
motherboards. You don't have enough units to sample and get
a statistically satisfying number. Should the number be 100 ohms,
or should it be 3 ohms ? Can you justify why that number is
appropriate ? Etc.

The board I worked on at work, it would measure 3 ohms on the 3.3V
rail. And the board had probably a thousand pullups, and little loads
on it (diff ECL section). A motherboard, by comparison, is too cost
sensitive
to be that wasteful. Our product, the ohms measurement was pretty
consistent from unit to unit. So in our case, a dead short on 3.3V
to ground, would be different than the 3.0ohms you'd read when
the thing was fault-free. I didn't even attempt to justify
where that number comes from, as I had more on my plate to do than
that. Like, make it work. That test was intended as part of
incoming inspection, so you wouldn't waste time on the defective
ones, and only work on the ones that looked like good candidates.

It's also pretty hard to ohm something, with 1000uF of
capacitance across it. In any case, for us, it was simple
chance that we noticed our board (assembly) was easy to read,
and consistent from unit to unit. So we could throw it in as
a quick check. Before connecting power to it. That thing
probably drew 100-150W or so. And had externally applied
forced-air cooling.

Paul

Well, you're absolutely correct.
However, as a practical matter, the measurement at the power
supply terminals on the motherboard should not be ZERO. We
could discuss forever what it should be, but it ain't ZERO.
And if it is zero, the caps won't have any effect, unless one
or more of them is shorted and the cause for it measuring zero.

And if it's too low, but not zero, the power supply will probably
come up and something will catch fire.

I measured an old Dell Dimension 2550 board. Lowest number I measured
from any of the (non-ground) power supply pins to ground was 121 ohms.
I could easily tell that the measurement was not zero.

I maintain that an ohms measurement will give you useful data.

 

[Paul]

Well, you're absolutely correct.
However, as a practical matter, the measurement at the power
supply terminals on the motherboard should not be ZERO. We
could discuss forever what it should be, but it ain't ZERO.
And if it is zero, the caps won't have any effect, unless one
or more of them is shorted and the cause for it measuring zero.

And if it's too low, but not zero, the power supply will probably
come up and something will catch fire.

I measured an old Dell Dimension 2550 board. Lowest number I measured
from any of the (non-ground) power supply pins to ground was 121 ohms.
I could easily tell that the measurement was not zero.

I maintain that an ohms measurement will give you useful data.

If the value is exactly zero, I agree.

If the value is a bit higher though, and you only
have one of those motherboards, you don't really know
what is "normal". If you have 20 boards, you can
work out a mean value and a standard deviation.

On the board where the characteristic value was 3 ohms,
I was just a bit shocked when doing the measurement for
the first time. It seemed kinda low. And could have been
due to the excessive number of voltage divider termination
networks on the board. Since that board used a fair
amount of electricity, a low reading wasn't actually
a cause for concern.

Paul