Redundant power supply problem

[Tom Del Rosso]

Can anyone offer some insight into this? A server has 2 redundant power
supplies. After soft shutdown it won't boot again unless the PSUs are
alternately switched on and off several times. (All power comes on but no
sign of CPU life.) After that it runs fine on either supply.

I replaced the hot-swap chassis that switches between the supplies. No
effect.

Turning it off again could be risky so I need specific ideas before I
attempt that. I have some ideas but I need as many as possible so I can do
a lot of tests at once.

Of course there are backups, but the hard drives are ok anyway, and there's
no place to restore the data to if the server won't boot.

 

[Sam Goldwasser]

Can anyone offer some insight into this? A server has 2 redundant power
supplies. After soft shutdown it won't boot again unless the PSUs are
alternately switched on and off several times. (All power comes on but no
sign of CPU life.) After that it runs fine on either supply.

I replaced the hot-swap chassis that switches between the supplies. No
effect.

Turning it off again could be risky so I need specific ideas before I
attempt that. I have some ideas but I need as many as possible so I can do
a lot of tests at once.

Of course there are backups, but the hard drives are ok anyway, and there's
no place to restore the data to if the server won't boot.

Is there a RESET button? This sounds like the power on reset circuit isn't
being triggered - maybe the power is coming up too slowly or something like
that?

--- sam | Sci.Electronics.Repair FAQ Mirror: http://repairfaq.ece.drexel.edu/
Repair | Main Table of Contents: http://repairfaq.ece.drexel.edu/REPAIR/
+Lasers | Sam's Laser FAQ: http://repairfaq.ece.drexel.edu/sam/lasersam.htm
| Mirror Sites: http://repairfaq.ece.drexel.edu/REPAIR/F_mirror.html

Note: These links are hopefully temporary until we can sort out the excessive
traffic on Repairfaq.org.

Important: Anything sent to the email address in the message header above is
ignored unless my full name is included in the subject line. Or, you can
contact me via the Feedback Form in the FAQs.

 

[CJT]

Can anyone offer some insight into this? A server has 2 redundant power
supplies. After soft shutdown it won't boot again unless the PSUs are
alternately switched on and off several times. (All power comes on but no
sign of CPU life.) After that it runs fine on either supply.

I replaced the hot-swap chassis that switches between the supplies. No
effect.

Turning it off again could be risky so I need specific ideas before I
attempt that. I have some ideas but I need as many as possible so I can do
a lot of tests at once.

Of course there are backups, but the hard drives are ok anyway, and there's
no place to restore the data to if the server won't boot.

Does it have fan sensors that lock it out if they don't spin? Or might
the boot drive be spinning up too slowly?

What I'm driving at is something (probably mechanical) that's not
responding in time.

Is the "power good" signal coming up?

 

[Tom Del Rosso]

Is there a RESET button? This sounds like the power on reset circuit isn't
being triggered - maybe the power is coming up too slowly or something like
that?

There is, and pressing it did nothing.

There are a lot of things I didn't try, and once I got it to boot I left it
alone. What can I add to the to-try list?

voltages
voltage sequence? important?
Power-good
trying above again with hard drives and cards removed

 

[Tom Del Rosso]

Does it have fan sensors that lock it out if they don't spin? Or might
the boot drive be spinning up too slowly?

All the fans seem to be spinning. It only shows a blank screen, and if the
HDD was taking too long then it would at least do the POST.

What I'm driving at is something (probably mechanical) that's not
responding in time.

Is the "power good" signal coming up?

I haven't tested that. I can put a storage scope on it. Any idea what the
timing should be in relation to the voltages coming up?

I want to do a battery of tests next weekend (too sick now or I would have
done it this week). I don't want to cycle the power too many times, because
I think it got a little harder to make it come up each time I tried.

 

[Franc Zabkar]

I haven't tested that. I can put a storage scope on it. Any idea what the
timing should be in relation to the voltages coming up?

From Intel's "ATX / ATX12V Power Supply Design Guide":

PWR_OK delay (T3): 100 ms < T3 < 500 ms


- Franc Zabkar

 

[Tom Del Rosso]

From Intel's "ATX / ATX12V Power Supply Design Guide":

PWR_OK delay (T3): 100 ms < T3 < 500 ms

Thanks. Does it say what voltage level T3 is referenced from? I mean 5V
must pass a threshold before PWR_OK should be true.

 

[Franc Zabkar]

Thanks. Does it say what voltage level T3 is referenced from? I mean 5V
must pass a threshold before PWR_OK should be true.

The undervoltage thresholds are -5% for all three major rails.

See http://www.formfactors.org/developer/specs/ATX12V PSDG2.01.pdf

"PWR_OK is a “power good” signal. It should be asserted high by the
power supply to indicate that the +12 VDC, +5VDC, and +3.3VDC outputs
are above the undervoltage thresholds listed in Section 3.2.1 and that
sufficient mains energy is stored by the converter to guarantee
continuous power operation within specification for at least the
duration specified in Section 3.2.11, “Voltage Hold-up Time.”
Conversely, PWR_OK should be deasserted to a low state when any of the
+12 VDC, +5 VDC, or +3.3 VDC output voltages falls below its
undervoltage threshold, or when mains power has been removed for a
time sufficiently long such that power supply operation cannot be
guaranteed beyond the powerdown warning time."


- Franc Zabkar