[Info-vax] rx2620 power supplies / VRM voltage low

[Scott Dorsey]

George Cornelius <cornelius at eisner.decus.org> wrote:
>Anyone else had problems with rx2620 power supplies?
>
>We have been helping someone in-house who has a laboratory
>instrument with attached rx2620 and no hardware contract.
>
>We are not hardware support people but got him around his
>last two hardware problems - bad scsi cable, (soft?) ecc
>errors that appeared to be interfering with boots - but
>he now has chronic power issues.
>
>The iLo MP shows a message like "VRM non-recoverable voltage
>low - check all boards at this address".  VRM's - voltage
>regulator modules - are on the motherboard and generate
>regulated voltages from the output of the power supplies.
>
>Swapping out the power supply resolves this.  Over and over
>again.  We get about a month out of a supply.

Where are you getting these supplies?

>Just recently tried inserting two supplies (these are used/
>refurbed, with 90 day warranty).  One of these died within
>a day or so - different error, indicating defunct supply.

Who is refurbing these?
Are they actually refurbing them, or are they just fixing whatever is
broken?  Are all the electrolytic capacitors in them new?

>Soon after - replacement supply not yet in - VRM voltage
>low error returns.
>
>Ideas?

Check ALL the power supply rails.  See if one of them actually is low.
Swap the "bad" power supply into another machine and see if they give the
same error.

When supply rails are pulled down, it's almost always because of bad
electrolytic decoupling capacitors.  If you buy a "refurbed" supply, it 
should have new electrolytics throughout, and they should be high grade
high temperature ones.

If you have a supply that "tests good" but has 20-year old capacitors
throughout, it will not last very long before another cap fails in it.

Now, you might have bad electrolytics on the CPU board as well... and I
would certainly inspect the board for bulging or leaking ones even if
you don't have an ESR tester to check them in-circuit.

Does the customer have an electronics tech on staff who can check caps?

>Don't ask why they won't pay for support.  Something
>about the instrument being due for replacement. Research
>funding goes through cycles and sometimes they're
>pinching their pennies.

That's how it goes.

>Oh - and did I tell you we have no Itanium experience?

This is a good way to get some!
--scott

-- 
"C'est un Nagra. C'est suisse, et tres, tres precis."