Linear Power Supply used 7 days 24 hours

[sdouble]

Hi folks,
I'm currently benchmarking a few low power linear power supply.
I'm actually after a 2-channel unit. It's supposed to provide me with clean rails (+9 and -5 V). Low power : only 4W on each rail.
My main concern (besides the noise level) is the fact that device should operate 24 /7 for a couple of years.
I wonder whether the electrolytic cap could sustain such a treatment.
Any experience on your side, guys ?
Thanks in advance,
J

[TimNJ]

I think electrolytic capacitor life is always a concern. At 4W, however, it doesn't seem like the unit will make significant heat...which is an electrolytic cap's biggest enemy.

If you are really concerned, you can check out what manufacturer/series of caps they used when the unit arrives. Or check online to see if anyone has pictures.

You can use the rule of thumb that a capacitor's life doubles for each 10 degrees Celsius it is operated below it's rated temperature. That is, if you operate a cap rated for 1000hr @ 105*C cap at 85*C, then its expected life is 4000 hours. If they use garbage capacitors, then maybe take the manufacturer's specs with a grain of salt.

If you are really in love with a certain power supply from a design/usability standpoint but it uses crap caps, then you can always swap out the caps preemptively. (Though, I'd be wary of the quality of the other components too.)

[mariush]

A polymer capacitor will have a much longer life, the formula used is

  Life = Original Lifetime x 10 ^ [ (Tcap_max - Tambient) / 20 ]  ...

so a 2000h @ 105c polymer running at 65c 24/7 in theory will last  2000h x 10^ [(105-65)/20]  = 2000 x 10 ^2  = 2000 x 100 = 200.000 hours or 83333 days or 8333 days or 22 years

Electrolytic capacitors ... yeah, you can approximate by doubling lifetime for every 10c under the max rating ...

If those regulators don't like low esr of polymer capacitors, maybe look into high temperature capacitors like the RX30 series from Rubycon which is rated for 2000..4000h @ -40c .. 130c : http://www.rubycon.co.jp/en/catalog/e_pdfs/aluminum/e_rx30.pdf

[David Hess]

My main concern (besides the noise level) is the fact that device should operate 24 /7 for a couple of years.

I wonder whether the electrolytic cap could sustain such a treatment.

The bulk input capacitor is sized to hold up the input voltage between power line cycles resulting in a massive ripple current derating so operating life is not a problem.  The output capacitor sees very little ripple current so its operating life is not a problem either.

You still have to pay attention to the operating life calculations but the application is not demanding.  Decades of operating life is feasible.

[retrolefty]

 I worked with large (up to 60 amps or so) linear regulated power supplies (5v to 24v) in a large refinery
that used several hundred or more and most worked 24/7/365 for several decades and then I retired from there.   

 Once or twice a year we might find a failed unit. Lamda brand supplies were the most used there but other brands were evident. I always attributed this reliability to the fact that they weren't powered up and down but left to run.

[james_s]

Linear power supplies are generally pretty gentle on caps, I have lots of 30+ year old stuff still running original electrolytics. Switching regulators wear them out a lot faster.

[David Hess]

For the oldest linear power supplies with can capacitors, the failure I usually see is leakage of oxygen and water through the rubber seals of the input capacitors causing one of the aluminum terminal connections to corrode through.  Slightly more modern supplies with "computer grade" aluminum electrolytics eventually lose enough capacitance due to drying out that the input voltage is not held above the dropout voltage of the linear regulator between power line cycles.  In both cases, this happens at 30+ years.

Of course if you stick the capacitor where it is operating at high temperature, it will dry out faster.  I have worked on passively cooled enclosed instruments where this resulted in all of the small output capacitors having one half of their rated capacitance after 20 years but everything still worked.

On the other hand, the input and output capacitors on my switching power supply based cordless drill charger dried up completely after just a couple years causing it to malfunction.  The same happened with my AA/AAA NiCd/NiMH charger.

[sdouble]

Thanks a lot guys for the info. that's pretty reassuring.
The rectifier bridge "sees" more ripple. Any special thing to look at in that direction ?

[IanMacdonald]

I'd have said that a greater problem in linear PSUs is under-rated rectifiers. Usually because power factor is overlooked. If you use good-brand caps it's unlikely they will give trouble.

[sdouble]

here is what the seller says :
"This power supply uses a Nichicon capacitor, MUR1560 rectifier diode, not a rectifier bridge. "

[sdouble]

here is the rectifier diode

[sdouble]

[sdouble]

here is what the seller says :
"This power supply uses a Nichicon capacitor, MUR1560 rectifier diode, not a rectifier bridge. "

MUR1560 for a 4W PSU is waaaaaay overkill even under any reasonable power factor an surge conditions.

certainly. It will run continuously (7/24) at 4 to 5 W but it's a 65 W power supply

[sdouble]

another pic of the beast :

[sdouble]

my only issue with that unit is that the seller can't remove its logo on the front plate. I can't re-brand it. That's clearly a problem...
BTW, thanks blueskull, i'm happy to learn that it won't kill me !

[sdouble]

does anyone know that transformer brand ?
Among the possible issues are transformers going bad.
I don't know those and wonder whether  those chinese transformers are reliable ?

[james_s]

Of course if you stick the capacitor where it is operating at high temperature, it will dry out faster.  I have worked on passively cooled enclosed instruments where this resulted in all of the small output capacitors having one half of their rated capacitance after 20 years but everything still worked.

One of the products a former employer made back before I worked there had a power supply with electrolytics right up next to a large heatsink that wrapped around and folded over part of the board. As one might expect capacitor failure was common after a few years of 24/7 service. I rarely find bad bulk filter caps but on those I encountered several that rattled like a maraca when shaken. Usually the failures would show up after a power outage, boxes would be running ok but after losing power they wouldn't start up again.

[floobydust]

another pic of the beast :

R-core transformers, ugh! 

[james_s]

What's wrong with the transformers? I've used some cores like that, they're a lot easier to wind by hand than a toroid, seemed to work fine for what I've used them for.

[mikeselectricstuff]

For electrolytics, it's ALL about temperature. Life doubles for every 10 deg.C reduction.
They can last decades if kept cool and well within their ripple current rating ( to reduce internal heating)

[sdouble]

dear all,
I just received the PSU.
It looks good, pretty heavy as expected. most of the weight is in the transformers.
The big output caps are Nichicon and rated 85 degrees C 35 V.
The low voltage wires are 16 awg, which is a bit low to ensure a constant voltage with varying loads
the noise level is satisfactory.
I will run it now for some weeks at 1.5 Amp on both rails.
Don't hesitate to comment on the design. This is still a prototype for me. If you see a mod which could improve the  reliability of the device, please mention it.
I intend to buy a bunch of those within 2 months if the burn in phase provide satisfactory results.

[Edison]

What metal casing and not GND 

[sdouble]

seems strange, I agree.
Safety is, together with reliability my main concerns for that device.
It's meant to be resold.

[Edison]

It's meant to be resold.

I'm worried - weak wires, GND, how are the transformers made - how they have done interlaces between the layers of  wire, pouring coolers - loss of the cooling surface .... etc

[sdouble]

I would not say that 16 AWG is a weak wire.