I agree with AndrewT on this. If the specified life time expectation is enough, then theoretically no need for additional safety margin.
However for example 1000h is not a long time. ;-)
But at lower temperature the working hour specifications are increasing exponentially, so I would choose a lower operating temperature instead of a lower operating voltage. If needed and possible, of course.
I often do this, and only 5 elcos failed yet (as I know), but those were manufactured before 1988.
Actually when I use big old caps in unknown state close to its voltage rating (or over), I often test the leakage current. Not the value, but the tendency. If after some minutes it still decreases significantly at a specific voltage and temperature, then I consider it as a safe operating state. If it starts to increase, then it is definitely unsafe.
However for example 1000h is not a long time. ;-)
But at lower temperature the working hour specifications are increasing exponentially, so I would choose a lower operating temperature instead of a lower operating voltage. If needed and possible, of course.
I often do this, and only 5 elcos failed yet (as I know), but those were manufactured before 1988.
Actually when I use big old caps in unknown state close to its voltage rating (or over), I often test the leakage current. Not the value, but the tendency. If after some minutes it still decreases significantly at a specific voltage and temperature, then I consider it as a safe operating state. If it starts to increase, then it is definitely unsafe.
Last edited:
for ecaps, it is 80% or rated working voltage in tube amps....
you never know, someone might turn on a tube amp with
lots of the tubes not inserted....
in fact when i design my psu, i make sure that the maximum
rated voltage is not exceeded with all of the tubes out of sockets...
never lost a cap this way....
you never know, someone might turn on a tube amp with
lots of the tubes not inserted....
in fact when i design my psu, i make sure that the maximum
rated voltage is not exceeded with all of the tubes out of sockets...
never lost a cap this way....
63 V is irrelevant, as I stated in the the first post, baseline is 100 V.
This is actually a question where there is no consensus.
Ordinary electrolytic capacitors should behave fairly similarly but reputable manufacturers differ.
Some extend the estimated lifetime if the capacitor is used at less than rated volts.
Other manufacturers' formulas don't have any lifetime multiplier if the capacitor is derated.
I am inclined to derate a little, my 100 V specification capacitors will see 80 V worst case.
I suspect your suspicion doesn't apply here.
Metalized film or paper capacitors, as used for X safety rated applications, behave quite differently to electrolytics.
The film capacitors "self heal" to clear shorts but there is no such mechanism in heavy power supply electros.
Best wishes
David
Common and sound practice. I think the figures are, but I may be wrong, maximum voltage.
The Ericsson company use the 85-90% rule as standard.
The Ericsson company use the 85-90% rule as standard.
reforming a 63 volt rated 56000 ufd cap, i was getting 71 volts and the leakage current is still diminishing....
but i still used the cap at around 80% or around 50 volts...
that was about 20 rears ago and the amp where it was used on still lives..
I checked the Cornell Dublier data sheets and they advise that if capacitors are derated from max V, then lifetime is extended.
Their recommendation for typical power supply capacitors is ...80%😉
So thanks for the discussion to make me check that.
Best wishes
David
FWIW, I question the premise about ESR being the first indicator of a problem. I've seen caps that lost value, some by a large amount, but still showed low losses. I see DC leakage rise and suspect that's the first indicator of a problem, though the value and losses are still in spec. Failure criteria for the data sheet lifetime is for a highly stressed part. The failure we see due to age are probably different mechanisms- chloride contamination, PCB flux solvents getting past seals, material purity, etc. No rules of thumb- you have to test multiple parameters, and I've yet to see anything you can test on a new cap that predicts life in our typically low stress applications.
Back then (the 80's), when I was working at the RIFA plant (plastic and paper caps), we were (privately) using the electrolytics caps, knowing that in order to keep the specs, the factory had to set the production several notches higher resulting in grossly underrated caps. It was no problem applying voltages 10-20% higher that stated. However as the production process was tightened up, the spread in specs has dimished, given the caps the same specs with very little variation.
Hope that was understandable.
Hope that was understandable.
that is probably an underestimated parameter.
We typically get 50000hrs (11yrs @ 12hrs/day) to 200000hrs (23years) from capacitors that are specified to at least 1000hrs to 5000hrs.
Temperature is our biggest asset in low stressed capacitors.
Some have estimated a doubling of lifetimes for each 10C reduction in operating temperature.
If that number is about right, then an 80°C cap with a specification life of at least 2000hrs becomes a minimum of 64000hrs at 30°C
If we are allowed to double it for low voltage use, we arrive at 130000hrs (15yrs).
And that's a good reason to look for caps rated 105°C, at least for tube amps or if the cap is to be used in an heated environment, like a tight amp casing.
Yes, I agree.
Temperature seems to be the real killer of electros.
Get the temperature down and one gains enormously for life/reliability.
Then you see a layout designer putting electros right next to a hot heatsink to try to make a PCB look attractively small and aesthetically pleasing, rather than to suit the duty required !
A rather odd result came from a stress at RIFA when I was working there. A number of heat radiating components were placed around the object to be tested. However, it seemed that the heaters created a flow of air that cooled the test object instead of heating it. 😕
Ha, unintended consequences!
Something subtle that I see quite often is a dropping resistor for a cap, with a nice copper trace or wire connecting it. The amount of heat transmitted right to the guts of the cap is surprising, and cap life wold probably double if the wire had an extra loop, was a bit longer, was made of steel, had its own heat sink or various other things one can do to reduce the heat flow.
Even if you reduce the temperature, other things will get you in the end. Solvents past the seals and lack of purity in the foil or other parts will eventually kill the cap. Still, I've got some darned old caps that test just fine for every parameter. I've also got some only 10 years old that have obvious degradation. Naturally, those are the ones I've got a lot of!
Something subtle that I see quite often is a dropping resistor for a cap, with a nice copper trace or wire connecting it. The amount of heat transmitted right to the guts of the cap is surprising, and cap life wold probably double if the wire had an extra loop, was a bit longer, was made of steel, had its own heat sink or various other things one can do to reduce the heat flow.
Even if you reduce the temperature, other things will get you in the end. Solvents past the seals and lack of purity in the foil or other parts will eventually kill the cap. Still, I've got some darned old caps that test just fine for every parameter. I've also got some only 10 years old that have obvious degradation. Naturally, those are the ones I've got a lot of!
If your predominant interest is simple ESR measurement, and you want to put 1A at 100KHz through the DUT in order to resolve 1 milliohm reliably, then the answer seems simple enough. Many audio power amplifiers can safely deliver 8W into 8R at 100KHz. I wouldn't try this with conventional bipolar transistor output types, but good tube amps and some MOSFET amps fit the requirement. Put your DUT in series with an 8R dummy load and adjust gain for 8Vrms output. 8R is so high relative to milliohms that you effectively have a 1A constant current source and your voltmeter reads 1mV per milliohm at the test capacitor.
If you want to make a dedicated test fixture instead, take a look at Linear Technology's LT1210 power op-amp. I suspect it wouldn't be difficult to whomp up an amplitude-stabilized 100KHz power oscillator with a chip like that.
If you want to make a dedicated test fixture instead, take a look at Linear Technology's LT1210 power op-amp. I suspect it wouldn't be difficult to whomp up an amplitude-stabilized 100KHz power oscillator with a chip like that.
Naturally there are a variety of different failure modes.
But if the capacitor is conservatively rated for volts then the main one seems to be loss of electrolyte, and ESR seems to reflect this before capacitance starts to drop much.
Of course this assumes the capacitor is not mistreated with chlorine solvents, physically faulty etc.
105°C rated power supply capacitor are less common and quite a bit more expensive so I don't plan that option.
But it's a solid state amp and I plan to keep them cool.
As Andrew says, temperature is the main killer.
Best wishes
David
in my side of the pond, we are a dumping ground of used electronics from
first world countries, metal recyclers to recover precious metals....
105C caps are common and cheap, they also test good and used in
our builds....for almost zero cost...
first world countries, metal recyclers to recover precious metals....
105C caps are common and cheap, they also test good and used in
our builds....for almost zero cost...
- Status
- Not open for further replies.