Changing display resolution

pjp wrote:


Perhaps it's something to do with power supply? Though that seems
unlikely seeing as you indicated it hangs at Splash Screen which
suggests some hardware problem else it'd at least give you the "No OS"
message. Check cables, reseat ram etc. might work. I had to do that with
one came by my way same way as yours, e.g. pulled out of recycle bin.

I've seen Ken's bug too.

I tried the same experiment, set the DPI to a customized value
and things froze up on me.

It's a Windows 10 issue.

*******

But I can't seem to find the menu I used for
Safe Mode. (I don't like spelling out some
obscure method of triggering Safe Mode, and
I'm not sure that holding shift key or whatever,
works any more. There is one menu that offers
"Startup Options", and that's the Safe Mode menu.)

As an example of a "Get Out Of Jail Free" for Windows 10,
I like adding this to the boot sequence (with bcdedit).
Note that this doesn't work, if multibooting certain
combinations of Windows OSes... Should work for
single boot situations.

https://i.postimg.cc/T3WkrRrn/win10-safe-mode-menu.gif

With a little effort, if you have that installed *before*
there is trouble, you can be like Houdini.

Paul

Ken Springer wrote:
On 6/22/19 12:43 PM, pjp wrote:

Perhaps it's something to do with power supply? Though that seems
unlikely seeing as you indicated it hangs at Splash Screen which
suggests some hardware problem else it'd at least give you the "No OS"
message. Check cables, reseat ram etc. might work. I had to do that with
one came by my way same way as yours, e.g. pulled out of recycle bin.

I'd bet power supply, but the system is too old, and slow for efficient
use of W10, to make the repairs worth the time. It's also not a generic
power supply, AFAIK this supply is only used in this series of computers.

What kind of CPU did this thing have ?

*******

There are a variety of standard and semi-standard power
supplies.

https://www.tomshardware.co.uk/power...w-32338-5.html

Modern Form Factors: EPS, TFX, CFX, LFX, And Flex ATX

A small model year range of Dell supplies had non-standard
pinouts. Using a pin-mover tool, you can re-arrange the
cabling of a standard supply, to match what Dell did.
The only tough one, might be some setup using the
Aux six pin box connector.

And there are units with "pure custom" LxWxH, designs
from Hell, that you're right, there won't be an easy
replacement. You can re-package recalcitrant computers
in larger packaging and make standard parts work with them.

Some motherboards, on purpose, are mis-shapen. They might be
L-shaped, with the panel header on the slim part of the L.
And it would seem the intention there, is so they won't
fit a microATX or ATX case. The wealth of Dells in BTX form,
with "blower near the front of the machine", those would
be a chore to re-package as well.

But in 2019, repacking would be a Smithsonian project,
not a practical one. Like, sometimes, you can find a
whole used machine, for the price some twirp wants
for the power supply alone. It's quite possible to
acquire two whole boxes "for parts" and cobble together
one working machine from it. There was a time when the
pricing wasn't quite that "compressed".

Paul

On 6/22/19 2:38 PM, Paul wrote:
Ken Springer wrote:
On 6/22/19 12:43 PM, pjp wrote:

Perhaps it's something to do with power supply? Though that seems
unlikely seeing as you indicated it hangs at Splash Screen which
suggests some hardware problem else it'd at least give you the "No OS"
message. Check cables, reseat ram etc. might work. I had to do that with
one came by my way same way as yours, e.g. pulled out of recycle bin.

I'd bet power supply, but the system is too old, and slow for efficient
use of W10, to make the repairs worth the time. It's also not a generic
power supply, AFAIK this supply is only used in this series of computers.

What kind of CPU did this thing have ?

According to the label on the side, AMD Phenom X4 9350e

HP Pavilion Slimline s3650f So far, I haven't been able to find
anything on this, not even on HP's site. But I'm sure, years ago, via a
web search I was able to locate an HP speck page.

*******

There are a variety of standard and semi-standard power
supplies.

https://www.tomshardware.co.uk/power...w-32338-5.html

Modern Form Factors: EPS, TFX, CFX, LFX, And Flex ATX

The power supply is a Kentek-FATX250. As soon as I saw that, I
remembered the original supply was dead, and I found this one on Amazon.
There appears to be a 270 and 300 watt model also.

Most likely, when I searched for the original supply, it's likely I
cross referenced to this model based on other units it would fit in.

Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?

If it is dead, I'm not adverse at this point to ordering a replacement,
even of a larger wattage. They aren't very expensive, and I've got
other things in my Amazon shopping cart to order anyway.

snip

Some motherboards, on purpose, are mis-shapen. They might be
L-shaped, with the panel header on the slim part of the L.
And it would seem the intention there, is so they won't
fit a microATX or ATX case. The wealth of Dells in BTX form,
with "blower near the front of the machine", those would
be a chore to re-package as well.

This is definitely a custom made board that only fits these series.
I've got another one of these designs but different model I had to
replace the motherboard. Linux Mint 19.1 is on it,


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

On 6/22/19 12:59 PM, Paul wrote:
pjp wrote:


Perhaps it's something to do with power supply? Though that seems
unlikely seeing as you indicated it hangs at Splash Screen which
suggests some hardware problem else it'd at least give you the "No OS"
message. Check cables, reseat ram etc. might work. I had to do that with
one came by my way same way as yours, e.g. pulled out of recycle bin.

I've seen Ken's bug too.

I tried the same experiment, set the DPI to a customized value
and things froze up on me.

It's a Windows 10 issue.

How far back in Windows time did you try this?

But I can't seem to find the menu I used for
Safe Mode. (I don't like spelling out some
obscure method of triggering Safe Mode, and
I'm not sure that holding shift key or whatever,
works any more. There is one menu that offers
"Startup Options", and that's the Safe Mode menu.)

But, if the HD isn't even powering up, nothing like this should work.
Ergo, my willingness in another message to buy a new power supply as
they aren't that expensive.

As an example of a "Get Out Of Jail Free" for Windows 10,
I like adding this to the boot sequence (with bcdedit).
Note that this doesn't work, if multibooting certain
combinations of Windows OSes... Should work for
single boot situations.

https://i.postimg.cc/T3WkrRrn/win10-safe-mode-menu.gif

With a little effort, if you have that installed *before*
there is trouble, you can be like Houdini.

If we get this thing running, I'll try to remember to do this. I was
going to do it in W8, but it's such a little used system here, I never
did it.


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?


You need to pull PS_ON# low (ground) to switch
on the main part of the supply. Since PS_ON# is
wired-OR logic, it's safe to ground it while it is
in-circuit.

However, if you do that, it defeats PROCHOT. If the
processor is regularly in the habit of overheating
(takes only two seconds if the heatsink falls off), then
grounding PS_ON# would torture/roast the poor CPU.
It's not recommended then, to try to force the PSU
on while in circuit, if you suspect any such fault.

The power supply has two halves. +5VSB is used for
supervisory circuitry. The front panel POWER button
won't work, unless the PSU makes +5VSB to bootstrap the
process. The +5VSB powers the IC that pulls PS_ON# low
to start the supply. That voltage also hold up DIMMs while
the machine is sleeping.

There are three ATX specs from the Intel formfactors
site, with pinouts. You can also use the playtool site
for info on connectors.

http://www.playtool.com/pages/psucon...onnectors.html

"The 24 pin main power connector was added in ATX12V 2.0" == table below this

Pin 20 in the 24 pin connector is the (missing) -5V pin.
Supplies no longer have -5V. The -12V is there to power
RS232 serial ports. You can key off things like which side
is the green PS_ON# wire on, in the table. The latch is
there to prevent thermal walkout, and the latch must
be depressed to pull the connector.

OK, so what could we do:

1) Check for +5VSB on the purple wire.
That's the supervisory voltage that must
be present for later steps.

2) Ground the green wire. It's held to logic 1 by
a pullup resistor. If you grounded the signal with
an ammeter, you might see 5mA of current flow or so.
(Normally you don't need to check this. It's just
a cheesy way to ground PS_ON#.)

The cooling fan should spin, if the PSU has not
latched off on failure.

As soon as you remove the grounding wire on PS_ON#, the
main portion of the ATX supply stops. The fan stops spinning.

3) Check PWR_OK (grey). It will be a logic one
(3 to 5V or so) if the power supply is convinced
3,3V, 5V, and 12V are at full voltage. It will measure
0.7V or slightly lower, if the supply is not confident
things are OK.

4) Do a "lap of the pins" and check the output voltages.
3.3V, 5V, 12V, -12V, +5VSB, all within 5% of nominal.

I have a home-made dummy load for testing supplies
and it provides a "light load" of 100W or so. The idea
of loading the supply, is to keep the supply in regulation.
Modern supplies don't normally have a problem with this
issue. Some ATX supplies have *two* rows in the spec
table. One row is "absolute max", while the second
row is "absolute min". The worst supply I ever saw,
required that you draw at least 25% of the rated power
at all times (which modern motherboards simply cannot
do - if the motherboard uses power saving states,
the voltages on such a supply could climb a bit on
the high side. Modern supplies only have absolute max,
implying the minimum load is "zero" - the unit will
properly regulate if no load is present.

These docs are only if you're curious. The Playtool table might
be enough to start. The same color-coded table is here.
(Note - "modern" supplies use all-black wire for the main
connector, which makes it very very inconvenient to check
green and purple wires...)

# This is 20 pin, back when -5V was still on the connector (pin 18 present).
http://web.archive.org/web/200304240...12V_PS_1_1.pdf

# This is 20 pin, with the -5V pin pulled from the connector (pin 18 missing)
https://web.archive.org/web/20040731...X12V_1_3dg.pdf

# This covers 24 pin versions of ATX PSU (pin 20 missing)
https://web.archive.org/web/20070103...public_br2.pdf

You can shove a multimeter tip into the back of the shell
while the connector is plugged in. The ground lead on the
multimeter can be clipped to an I/O screw in the I/O plate
area (if the supply remains connected to the motherboard).

In general, I *do not* recommend ammeter measurements
of full load currents on machines like this. I use a
clamp-on DC ammeter for that, which can cost $200 to $300
bucks. Even though I don't work on other peoples kit
(not running a computer store), I've still got plenty
of usage out of it. For example, when the starter went
on my previous car, I could use that clamp-on DC ammeter
in peak hold mode, to tell me the starter was drawing
150 amps from the battery. And I could make the measurement
in complete safety, because that ammeter needs no ohmic
connection. It measures magnetic field instead, to get
current.

Most of my verification with a multimeter is on ohms and
volts. Much of the time, DC volts. You can check resistors
in circuit, with the "low power ohms" setting. Check the manual
for what ranges are low power ohms, if you need to do this
for some reason.

Paul

On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

Ken Springer wrote:
On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.



If the power supply will run, you can easily do the
lap of the pins bit.

It's simple. Can't get simpler.

With side off PC, switch on at the back (if it has a switch),
push the front power button. Make sure the panel header
(the 2x4 cable inside the case, that goes to the front LEDs)
is connected.

If the fans come on, now, take the multimeter and poke with the
red probe, each "hole" where the wire goes into the white plastic
shell. There is sufficient metal exposed from the back side of
the connector, that you can make measurements with the system
running.

Have an assistant handy, or even a tape recorder will do, go sequentially
around all the pins, measuring voltage (20V full scale will do for
all measurements). Either write down the values or speak the
values into your tape recorder.

When finished, go back and correlate with the diagram.

*******

If the supply won't start, check +5VSB first. Then you
can ground PS_ON# and see if the supply starts.

Sometimes the motherboard driver for PS_ON# isn't functioning
properly. I don't know why so many of these fail to function,
but quite a few times, this style of testing hints at
the mobo not asserting PS_ON#, rather than the PSU
being shot.

If you are a "chicken of the sea", use a PSU tester.

Now, this isn't a tester, but it will assert PS_ON# for you.
You might use this, if you just bought an ATX supply and wanted
to make it switch on, without any computer present.

https://www.amazon.com/24-Pin-Supply.../dp/B07K8F6QP7

The actual PSU testers have multiple connectors on them and
can make measurements. They range up to around $35 or so.
With those, you make sure the connectors are in the
correct holes, before plugging in the AC plug on the supply.
That's because I don't see a switch on those, to turn the
supply on and off.

Paul

Ken Springer wrote:
On 6/22/19 12:59 PM, Paul wrote:
pjp wrote:


Perhaps it's something to do with power supply? Though that seems
unlikely seeing as you indicated it hangs at Splash Screen which
suggests some hardware problem else it'd at least give you the "No OS"
message. Check cables, reseat ram etc. might work. I had to do that with
one came by my way same way as yours, e.g. pulled out of recycle bin.

I've seen Ken's bug too.

I tried the same experiment, set the DPI to a customized value
and things froze up on me.

It's a Windows 10 issue.

How far back in Windows time did you try this?

But I can't seem to find the menu I used for
Safe Mode. (I don't like spelling out some
obscure method of triggering Safe Mode, and
I'm not sure that holding shift key or whatever,
works any more. There is one menu that offers
"Startup Options", and that's the Safe Mode menu.)

But, if the HD isn't even powering up, nothing like this should work.
Ergo, my willingness in another message to buy a new power supply as
they aren't that expensive.

As an example of a "Get Out Of Jail Free" for Windows 10,
I like adding this to the boot sequence (with bcdedit).
Note that this doesn't work, if multibooting certain
combinations of Windows OSes... Should work for
single boot situations.

https://i.postimg.cc/T3WkrRrn/win10-safe-mode-menu.gif

With a little effort, if you have that installed *before*
there is trouble, you can be like Houdini.

If we get this thing running, I'll try to remember to do this. I was
going to do it in W8, but it's such a little used system here, I never
did it.

Cycle the AC power on the unit, before giving up.

Sometimes a supply "latches off" on a fault. To
clear the fault, remove AC and try again.

And don't do like one poster did. He posted in,
that he worked the switch on the PSU really fast
and "turned it on 50 times" and "it popped" :-/
When you use the switch on the back of the computer,
you're supposed to wait 30 seconds before turning it
on again. This allows the inrush limiter to cool
off and it's then ready for another cycle.

Some supplies will latch off, if they've "become weak".
Sometimes, after clearing the fault, you might get the
cooling fan to "twitch". And that tells you the power
supply is trying to drove too much of a load. IT's
either a partial short, or the supply is weak.

My very first PSU, it was so weak, it could run 1 cooling
fan, but would shut off when powering 2 cooling fans.
It's not really "dead", but it might as well be.

Paul

Ken Springer wrote:
On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.

Since you saw a splash screen, what about booting
a Linux DVD and see if it comes up ?

You can prepare Linux on a USB stick too. Hybrid ISO
images can be "dd" transferred directly to a USB stick,
and will preserve their hybrid characteristic (boot either
legacy BIOS mode or UEFI mode, not that it matters for a
quick check like this).

If hardware is unstable in Windows, Linux will also wobble.
But the wobbling motion is a bit different. On Linux, as
the OS dies, you'll see some icons disappear, before
the kernel crashes.

If the Linux desktop remains responsive, and you can
click stuff and it opens and runs, then it could be
some problem with the Windows content on the disk drive
(bad disk maybe). Linux has SmartMonTools, as an example
of a program you could run.

If you've done any amount of booting of alternatives,
this should be a familiar test for you.

The first time I booted Linux for differential analysis,
was on a Win98 system. Win08 would keep tipping over
if there was too much RAM. But it wasn't *that* problem,
it was an actual hardware design issue (440BX). I booted Linux,
and it was *exactly* as flaky as Win98, and that's when
it occurred to me, that this is a hardware problem for sure.
And not the Win98 address space issue.

Paul

On 6/22/19 9:21 PM, Paul wrote:
Ken Springer wrote:
On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to text a
line or two out of the power supply to see if it's "dead' in some place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.

Since you saw a splash screen, what about booting
a Linux DVD and see if it comes up ?

OOooooohhhhhhkay...

Someone else mentioned me seeing a splash screen, and it looks like I
should have clarified that reply. No clarification on my part is my
mistake, my apologies.

At the first power button off/on reset, there was a splash screen and a
reboot. At the second, one the splash screen. Now... nothing. Black
screen, no blinking cursor, so signs of life.


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

Ken Springer wrote:
On 6/22/19 9:21 PM, Paul wrote:
Ken Springer wrote:
On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to
text a
line or two out of the power supply to see if it's "dead' in some
place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.

Since you saw a splash screen, what about booting
a Linux DVD and see if it comes up ?

OOooooohhhhhhkay...

Someone else mentioned me seeing a splash screen, and it looks like I
should have clarified that reply. No clarification on my part is my
mistake, my apologies.

At the first power button off/on reset, there was a splash screen and a
reboot. At the second, one the splash screen. Now... nothing. Black
screen, no blinking cursor, so signs of life.

But the fans are spinning, I take it...

The power supply hasn't latched off and the fans stay spinning.

Now, you could check for PWR_GOOD with your multimeter.

*******

There's a couple ways to make a black screen:

1) No PWR_GOOD prevents system from coming out of reset.
2) System has a RESET button and... RESET button is
crushed in the ON position. There are lots of computer
buttons, cheaply made momentary contact, where when they
fail, the button crushes in the ON position. This prevents
the system from coming out of RESET.

If VCore fails, it may not stay running long enough
to provide much in the way of symptoms to work on. If you
unplugged the ATX12V 2x2, plug it back in.

http://www.playtool.com/pages/psucon...s.html#atx12v4

If you had a Port 80 debugger card in a PCI slot or
a PCI Express slot, the two or four digit hex display
shows BIOS "progress" codes, and you can tell from
that, before video lights up, whether the BIOS is
"progressing" or not. Port 80 cards vary in price
from $10 to $100, depending on where you buy one.

*******

There are no beep codes ?

If the BIOS code is getting loaded, the system will
make beep codes before POST is finished.

When the system first comes up, the DIMMs are not running.
They need to be programmed, by doing a "special cycle"
and writing the CAS value to the DIMM.

So the first code in the BIOS, uses nothing but registers
and L1/L2 cache to do stuff. And that code can function
even if the DIMMs haven't been programmed yet.

Once the DIMMs are programmed, now you can have a frame
buffer declaration for video.

In terms of beep codes then, a RAM beep is more likely
to be detected first, before a video beep. You could have,
say, two failures, but only the first gets reported.

*******

If the system has more than one DIMM, you can *completely*
power off the system, and try the DIMMs one at a time.
Sometimes a duff DIMM prevents a good response
at startup.

Don't change DIMMs while +5VSB is running. Unplugging the
computer is one way to remove power. Wait 30 seconds
to 60 seconds, for the power to drain, before moving
DIMMs. Don't forget your antistatic precautions (bring
yourself to chassis potential when adding or removing
DIMMs, store DIMMs in disk drive bag if you have one
- tin foil is a lousy substitute for a disk drive bag
so don't do that).

Make a note of the slots the DIMMs are in, as eventually
you'll want to put the DIMMs back. (Some systems have
four slots, and DIMMs go in matched pairs of slots.)

https://www.cnet.com/products/hp-pav...z-4-gb-640-gb/

The chipset is NVIDIA GeForce 6150 SE / nForce 430.

That means graphics are in the Northbridge, the 430 part
is the Southbridge (a few SATA ports), and the main chip
has a heatsink on top. If the heatsink has a 40mm cooler,
verify it's still spinning. It doesn't need a video card,
because of the "built-in" graphics.

The labeling on this chip, tells the story. It has the
61xx number as well as the 4xx number. In other words,
this one chip, is both the Northbridge and Southbridge,
all in one chip. And the heat from the graphics is why
there is a heatsink (and possibly a cooler fan).

https://www.ixbt.com/mainboard/image...-nforce405.jpg

Anyway, try the DIMMs one at a time. But remember to
remove all power from the system, let it drain, and use
antistatic precautions. (Store the pulled DIMM in a disk
drive bag, and touch the inside of the bag before
adding or removing a DIMM from the bag.)

The CNET press release says the unit has a video card.
The internal graphics would likely be disabled if the
video card is detected. It means you have *two* video
connectors you can test with the system, not one. The
video signal could be on the other connector, if the
9500 video card is still plugged into the system.

But it really should beep, if the BIOS is running
(and the ATX12V 2x2 connector is plugged in).

http://www.playtool.com/pages/psucon...s.html#atx12v4

A single beep, confirms the piezo works. A repetitive beep
pattern means the BIOS is running, the CPU works,
but the video or RAM is bad.

There's a few things to do, before running off and
buying power supplies. It all depends on what
equipment you have on hand.

Paul

On 6/23/19 6:53 AM, Paul wrote:
Ken Springer wrote:
On 6/22/19 9:21 PM, Paul wrote:
Ken Springer wrote:
On 6/22/19 5:27 PM, Paul wrote:
Ken Springer wrote:


Since I've got this thiing torn apart this far, is there a way to
text a
line or two out of the power supply to see if it's "dead' in some
place
it shouldn't be, using a multimeter?


snip

Paul, I truly appreciate all the work you've put into this on my behalf,
but the entire post is way over my head. :-(

Now that I've got this thing torn apart this far, I think I'd rather
gamble the money and purchase a new power supply. And I may as well
replace the thermal paste at the same time.

Since you saw a splash screen, what about booting
a Linux DVD and see if it comes up ?

OOooooohhhhhhkay...

Someone else mentioned me seeing a splash screen, and it looks like I
should have clarified that reply. No clarification on my part is my
mistake, my apologies.

At the first power button off/on reset, there was a splash screen and a
reboot. At the second, one the splash screen. Now... nothing. Black
screen, no blinking cursor, so signs of life.

But the fans are spinning, I take it...

No, they weren't. No lights at all, except possibly the power on light.
I can't see that light unless I stand up and look down on the unit.

To check some of the previous items, I just about had to completely
disassemble the unit, and to double check the fans spinning, I partially
reassembled it.

To my surprise, fans ran, HD light lit up even though the HD was not
installed, and all the beeps telling you something was wrong.

Hooked up the HD and a monitor, mouse, and keyboard, and it booted just
fine.

A plus out of all of this... Last Thursday, I used this computer to do
a presentation to a small group. But I couldn't get the PS2 mouse to
work. So, used a USB mouse, Now, the PS2 mouse works.

I'm going to reassemble it, see what happens.

The power supply hasn't latched off and the fans stay spinning..

Simple terms, please, what does "latched off" mean?



snip


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

Ken Springer wrote:


The power supply hasn't latched off and the fans stay spinning..

Simple terms, please, what does "latched off" mean?

Latching off means that an item

* Sensed a problem was present. OVP, OCP, temperature, whatever

* The power supply switches off for safety. On ATX, this
means the switching action stops. (In this diagram, Q1 and Q2
stop alternating on the upper right of the diagram, then there
is no drive pumping the transformer T3.)

http://www.pavouk.org/hw/en_atxps.html

* The unit "remembers" (latches) the status.
This prevents it starting a second time, or getting into
a loop where (maybe) it could catch fire. It doesn't want
to start any more, because it knows "a bad thing happened".
This is good if you leave a computer running when on vacation,
the power drops, and on restoral of the power, something
happens to your PC. This will help it stay off, rather than
catch fire.

* The "memory" of the event, stays with the unit, because
some power is being extracted from AC Mains to power
any kind of digital circuit that can remember the flaw.

* Pulling the power cord, or using the switch on the back,
removes all digital memory. Allow sufficient time for the
inrush limiter in the PSU to cool off. If you turn
off a PC, wait 30 seconds before turning it on (at the back)
again.

The "soft power" switch on the front of the PC, does
not cause stress. The rear switch (or even, plugging in the ATX
PSU with the switch already in the ON position), stresses the
rectifiers as they charge the main capacitor.

Even with the inrush limiter, the nameplace rating may mention
a "40 amp" or "80 amp" peak current. If you have incandescent
bulbs, sometimes you can see a "glitch" from the bulb. Or, if
you have poorly regulated stereo power, the stereo could emit
a noise. The inrush limiter slows the filling of the main
capacitor enough, so the rectifiers don't blow.

A 1KW PSU that Anandtech tested, it was missing the inrush limiter,
and every time they actuated the switch on the back, it would
blow their lab breaker :-)

Another example of this, was years ago. The motherboard has the
VCore circuit, which sits near the CPU socket. The switching
converter that is VCore, it senses overcurrent, and modern
ones even "police" the power level. If VCore senses the PSU is
slow to make +12V on the output at startup (which should
take about 35 milliseconds), the VCore circuit can also latch
off, requiring the user to kill the AC feeding the PSU. The
"memory" on the motherboard in this case, is based on +5VSB
power, so the +5VSB is what needs removal, and switch-off
at the back, does it for you. This particular case, was
an interaction between a specific Asus motherboard and
a specific Antec PSU. (The owner gets to return one of the two products,
and select something else, to fix it. If it was me, I'd change
out the Antec PSU :-) )

One thing to know about switching power devices, is they
*disable* the overcurrent for a fixed interval at startup.
The 35 millisecond delay I reference there, the ATX PSU
doesn't check for overcurrent, until the output caps are
charged up. Then it switches on the OCP current limiting
protection and monitors things. The Asus VCore was doing
the same thing. But, if the PSU is slow, the VCore will also
be slow drawing its slug of current, and charging the
capacitance on the motherboard. By the time the PSU power
is fully present, the VCore OCP is already armed, and
it trips off because now it sees the charge-up current
flowing.

*******

While we've discussed the (inherently safe) latchoff method
above, back in the Pentium III era, they used "putt putt" mode.
On sensing overcurrent, the output folds back on the motherboard,
but, it retries a fraction of a second later. If you had
a speaker connected to that power source, it would
sound like a motor boat, as VCore tries over and over
again, to start.

One of the great things about the putt putt era, is
users never knew that every time their systems started,
the thing was going overcurrent. But, once the caps
get charged up, it stops going "putt putt" and settles
down. It's not wise to use this method on things
like CPUs that draw 400W... It's a good method for
a single phase 35W VCore.

Paul

On 6/23/19 1:08 PM, Paul wrote:
Ken Springer wrote:


The power supply hasn't latched off and the fans stay spinning..

Simple terms, please, what does "latched off" mean?


OK, this appears to have been the problem. I hope I can remember this,
should something like this ever happen again.

Everything is reassembled, and back up and running, including any updates.

At least I don't have to look at new hardware at this time. But, if you
would, what do you think of mini units like this? ...
https://www.amazon.com/Windows-Compu...TVPDKIK X0DER

Since I've no plans to use W10 as my main system, I'm thinking this
might be all I'd need. I have zero feel as to the speed of the Atom
processors.

And, I think this could be a solution for all the W7 latptop owners who
don't haul their systems around anymore. Or, laptop owners who
*thought* they would be hauling them around but don't.



snip


--
Ken
MacOS 10.14.5
Firefox 67.0.4
Thunderbird 60.7
"My brain is like lightning, a quick flash
and it's gone!"

Ken Springer wrote:
On 6/23/19 1:08 PM, Paul wrote:
Ken Springer wrote:


The power supply hasn't latched off and the fans stay spinning..

Simple terms, please, what does "latched off" mean?


OK, this appears to have been the problem. I hope I can remember this,
should something like this ever happen again.

Everything is reassembled, and back up and running, including any updates.

At least I don't have to look at new hardware at this time. But, if you
would, what do you think of mini units like this? ...
https://www.amazon.com/Windows-Compu...TVPDKIK X0DER


Since I've no plans to use W10 as my main system, I'm thinking this
might be all I'd need. I have zero feel as to the speed of the Atom
processors.

And, I think this could be a solution for all the W7 latptop owners who
don't haul their systems around anymore. Or, laptop owners who
*thought* they would be hauling them around but don't.

You're not really done.

When it happens again, examine the capacitors (aluminum cylinders)
near the CPU socket. Look at the tops. Look for rusty brown deposits.
Sometimes the same fluid appears near the base, near their two legs.
These are leaking capacitors, from the "capacitor plague" era.

https://en.wikipedia.org/wiki/Capacitor_plague

The power supply (that Flex ATX of yours) could have leaking
caps inside. Typically (at least a few people have mentioned
this), it seems to be the +5V filter caps. Don't know why those
go first. You'd think all the output filter caps would have
equal failure probability.

A leaking cap can make a tiny puff of smoke at startup. Once
the cap warms up a bit, the smoke stops. Caps that are
that close to failure, last around three weeks or so.
(They're getting pretty dry at that point, and there
is little liquid electrolyte left.) From the time you see
smoke, you've got about three weeks to purchase a
replacement supply and change it out. If you're lucky,
the computer will start crashing at BIOS level, before
the power supply blows or something.

Normally nothing bad happens, if the label on the supply
has a long list of protections (OVP, OCP, etc). A $20 supply might
have no practical protection at all, and those afford more
risk of blowing a motherboard and/or drives. The all-time
champ for this, was a certain model number of Bestek, which
would fail on +5V, the +5V would shoot up to 8 or 9 volts,
blowing drives and other stuff fed directly from +5V. Maybe
the DIMMs and CPU would survive, but the age of those machines
meant saving the left-overs was not sufficient to make
for a "cheap" repair. (These days, such a machine is a
write-off.) And with those, it was just a matter of time
until it fails.

There might have been one or two Dell models, with a 95%
field failure rate on mobo (leaking caps). You don't dare
risk even buying a replacement motherboard off Ebay, because
it could be ripe too, for those.

I don't know that's your root cause, but keep
an eye pealed.

Paul