switching power supply crisis - non-pedal

[blackieNYC]

Question and cautionary tale.  I work in a large audio facility.  In the last two years or so, we have been plagued with problems originating in external switching power supplies.  This place was built 9 years ago, so it is due time for some end-of-life issues.  Lump-on-A-line switching supplies (5, 12, 24V) have shown an ability to leak their internal 50-100kHz oscillator.  Fed back thru the power distribution, or airborne/proximity - can't say for sure. It seems that this leak confuses some digital audio equipment, which is always looking for 44-96kHz clock signal to latch on to. Some gear is unaffected, others will mute or spit out garbage.
Internal switching supplies can do this too - only one thus far.
We are of course considering preemptively replacing the external supplies.  My question is this:  Currently manufactured SMPS are of course going to meet FCC and other RFI standards, as did these 10 years ago, but are there supplies available that may have addressed this mode of longevity failure?  Frankly, I'd rather have them drop dead.  I need to note that in all cases thus far, the failure of the supply has not been severe enough to affect the connected equipment for which it provides power! The gear works fine, which is just incidental.  So a preemptive replacement would be wise, but to replace with a supply that drops dead at the first sign of deterioration would put us in a much better place.
sorry to be waaay off topic.

[anotherjim]

Are you sure you want the supplies to fail "sudden death"?
There are internal rack SMPS with "power good" status outputs and indicators. More to help locating a failed unit among many I think than a warning of degraded performance.
Perhaps a 3 monthly regime of preventative maintenance is in order? Scoping the supplies for output ripple and keeping records so you can see any downward trends might be useful. In equipment rooms/racks, keep an eye on any cooling fans & filters.

[blackieNYC]

Monitoring of the output voltage would be a great idea. But most of our problems are from external supplies, little plastic boxes.
A rackmount power supply that could provide power to an assortment of devices, instead of the wall warts and lumps, would be great.  You'd be distributing 5and 12v - maybe one rackmount unit per rack. With a network interface, traps, logging, etc. I've not heard of such a thing.
if a couple u of rack space could be allocated for a DC multiple output power supply, you could accommodate large caps, and possibly avoid SMPS altogether.

We will almost certainly sweep thru the place and replace power supplies, but if the new ones are also likely to leak high freq when it ages, I would rather have them die completely.  We have redundancies for that.  The random muting or noise throughout our system does not allow for an alternate audio routing backup plan.

[Rob Strand]

I'm assuming:
- early on there were no such issues and now there are  To a large degree, that means magnetic sources of coupling should not have changed.
- nothing new has been added to the system (which could be an indirect cause)

A 10 year old switch-mode running all day is likely to be approaching end of life.

A common fault mode of switch modes is the caps.  As they age the cap ESR increases and the ripple at the input supply and the output DC rails increase.   Often you will find the DC output voltage measures OK.  However, if you look at the DC output with a CRO you will see a lot of ripple.  In this state, the switch-mode feedback loop regulates the average voltage. A multimeter reads the average DC so you don't see the problem with a multimeter until the problem is severe.

I doubt all PSUs are dead so you should find some units are worse than others.

The solution is replace bad units with good units.    Good to keep records or date new units as the other old ones will fail soon.

You can also replace the caps.   You have to be careful about using caps with a long lifetime (and in some case an equivalent ESR to the original caps).

It is not uncommon for a cheap PC power supply to develop this problem within 2 to 5 years.   Higher reliability power supplies use longer life caps (IIRC they quote 100,000 hrs life).   Crap caps are rated at 2000 hrs and high-reliability ones upto 20000 hrs.   A cap's lifetime is rated at a specified temperature.  The lifetime increases beyond the specified hours as you increase T_rated - T_operating.  So choosing a 105C cap will last longer than an 85C cap, and keeping things cooler will also increase life.   You can extend the life of the caps this way but then you end-up with fan faliures.

Before assuming this is the reason for the problem, maybe you should swap PSU between good and bad units and see if you can transfer the problem with the "bad" PSU.

[anotherjim]

Such as these in the pdf. Several monitoring options are available. AC good, DC good, high temperature...

https://uk.tdk-lambda.com/KB/FPS-Datasheet.pdf

Wall warts, bricks... Well if you have spares, you could bench test with a dummy load. Measure ripple especially looking for switching frequency leakage   compare with previous results. Rotate the supplies so you always have one you've tested to swap in.

Agree with Rob that electro caps in SMPS are most likely to cause failures. Life will also reduce at high temperatures.

Actually, I had problems with Lambda supplies crapping out within a year. These were in coastal locations and the cooling fans were blowing the moist salty air all over the HT rectifier/cap block at the AC input. Solution -  reverse the fans!