Hi all,
i am building a new dual rail, linear power supply for an amplifier. My power transformer has a 90v CT secondary (2 x 45V rms), which gives close to 64V DC after the rectifiers and smoothing (i have seen it climb to 65 volts when unloaded).
Now the problem is that i already have 8 pieces of 10k μF / 63V electrolytics that i intented to use, but the voltage ended up on the high side.
I know of a couple of ways to drop some volts (thanks diyaudio), but i don't really want to go down that path.
Now, as a test, i connected each capacitor to a 130V DC power supply (full winding of my secondary) through a 220kΩ resistor, in an attemp, not only to reform them, but also see where the voltage would climb to before leakage became too high (i was constantly monitoring the capacitor voltage).
After a lot of SLOW charging time
, and to my surpise, the voltage easily and without any visible obstruction climbed to 73 volts on ALL caps (10% overvoltage), at which point i decided to disconnect them. Now, according to my calculations, the leakage current at 73 volts should be less than 250μΑ, which is the charging current at that voltage (since the voltage was still climbing).
I know the usual tradition about the safety margin in capacitor voltage, but 100 or even 80 volt capacitors are disproportionaly expensive (at least around here, and i don't trust ebay sellers with suspiciously cheap capacitors)!
I also know that commercial products almost always have higher voltage rated caps, and i say "almost" because i have seen an old Kenwood amp with 50V/4700μF capacitors used on 51V (idle) rails!
Now i feel like there shouldn't be a problem in my case, the capacitors don't seem to mind a slight short term overvoltage that much. Also, according to my calculations, the ripple current should not be a problem for use in the power supply smoothing duty.
But what about long term reliability? Any experience on a case like mine? By the way, said capacitors are cheap YAGEO LH 85C (there exist better quality, properly rated capacitors downstream, onboard, as rail decoupling / bypassing etc...)
i am building a new dual rail, linear power supply for an amplifier. My power transformer has a 90v CT secondary (2 x 45V rms), which gives close to 64V DC after the rectifiers and smoothing (i have seen it climb to 65 volts when unloaded).
Now the problem is that i already have 8 pieces of 10k μF / 63V electrolytics that i intented to use, but the voltage ended up on the high side.
I know of a couple of ways to drop some volts (thanks diyaudio), but i don't really want to go down that path.
Now, as a test, i connected each capacitor to a 130V DC power supply (full winding of my secondary) through a 220kΩ resistor, in an attemp, not only to reform them, but also see where the voltage would climb to before leakage became too high (i was constantly monitoring the capacitor voltage).
After a lot of SLOW charging time
, and to my surpise, the voltage easily and without any visible obstruction climbed to 73 volts on ALL caps (10% overvoltage), at which point i decided to disconnect them. Now, according to my calculations, the leakage current at 73 volts should be less than 250μΑ, which is the charging current at that voltage (since the voltage was still climbing).I know the usual tradition about the safety margin in capacitor voltage, but 100 or even 80 volt capacitors are disproportionaly expensive (at least around here, and i don't trust ebay sellers with suspiciously cheap capacitors)!
I also know that commercial products almost always have higher voltage rated caps, and i say "almost" because i have seen an old Kenwood amp with 50V/4700μF capacitors used on 51V (idle) rails!
Now i feel like there shouldn't be a problem in my case, the capacitors don't seem to mind a slight short term overvoltage that much. Also, according to my calculations, the ripple current should not be a problem for use in the power supply smoothing duty.
But what about long term reliability? Any experience on a case like mine? By the way, said capacitors are cheap YAGEO LH 85C (there exist better quality, properly rated capacitors downstream, onboard, as rail decoupling / bypassing etc...)
One must ask why the manufacturers rated the voltage as 63. I am sure they didn't do it because they are happy at 75v. I have seen the carnage that occurs with aluminium foil shorting out other components when they go bang! Wire them in series parallel pairs with stabilising resistors to stop any unwanted issues. 40,000uF is adequate for almost anything.
Nothing magically happens to a 63V cap when the voltage exceeds 63V. The cap was not carefully constructed for exactly 63V, so it works fine below 63V and blows up above it. It was actually formed at the factory for something a bit more than 63V, as the OP found in his experiments. As I said, there is a curve (which some datasheets or 'white papers' may give): higher voltage means shorter life so underrate for longer life. My guess is that 1.5% overvoltage is well within the build tolerance so running a 63V cap on 64V will give approximately the same life on average as running it at 62V. If you were building thousands of these items then you might notice that the 64V ones needed more frequent cap replacements than the 62V ones. It is probably much more important to keep it cool: a cool cap at 64V will last longer than a hot cap at 62V.
The manufacturer will make a batch of prototype capacitors.
They will measure the failure voltage when excessive DC is applied.
There will be variation. The will apply some statistics and determine the average failure voltage, the mean failure voltage, the standard deviation voltage and probably a few others.
They will then determine what voltage the whole batch (100%) will pass at and what voltage they get 0.1% failures. Or some other failure percentage that they think suits their customers and the costs of warranty returns.
Let's suppose their model T 63V caps have an average failure voltage of 69.5Vdc and they know that their failures at 63Vdc will be ~0.1%
They go into production and constantly monitor the averages from samples and standard deviations from those samples. Make adjustments to production to ensure they meet their chosen failure rate and maximise their profit.
Now to life.
A polar electrolytic presented with 99.9% of it's rated DC voltage and kept cool will last almost for ever. We have anecdotal evidence that well cared for electrolytics over 40years old still seem to perform to specification. I have a little bit of evidence as a result of reforming, that electrolytics reformed slowly to overvoltage do not suddenly leak worse, or blow up. The rate of leakage does go up as rated voltage is approached and passed.
Now apply some spikes on the applied voltage. This will lead to deterioration of the foil and insulation and electrolyte.
In time the effect of absorbing the energy of the spikes will make the cool capacitor fail, either in capacitance, or in leakage, or some other non compliant parameter. These spikes that exceed the deterioration level might repeat @ one hundred per second.
Reduce the DC voltage to 50% of rated. The spikes on the supply now need to exceed the level that causes deterioration. Maybe none of the spikes cause any deterioration. Or maybe one spike every second exceeds the deterioration level. Eventually the capacitor fails because enough spikes have deteriorated the capacitor sufficiently to not meet it's specified parameters.
The low DC + spikes last a lot longer than the very high DC + spikes. I'll guess and suggest that running at 50% of rated voltage operates correctly for 50times longer than running at 99.9% of rated.
It is my view that it's the spikes that cause much of the failures after taking account of temperature.
They will measure the failure voltage when excessive DC is applied.
There will be variation. The will apply some statistics and determine the average failure voltage, the mean failure voltage, the standard deviation voltage and probably a few others.
They will then determine what voltage the whole batch (100%) will pass at and what voltage they get 0.1% failures. Or some other failure percentage that they think suits their customers and the costs of warranty returns.
Let's suppose their model T 63V caps have an average failure voltage of 69.5Vdc and they know that their failures at 63Vdc will be ~0.1%
They go into production and constantly monitor the averages from samples and standard deviations from those samples. Make adjustments to production to ensure they meet their chosen failure rate and maximise their profit.
Now to life.
A polar electrolytic presented with 99.9% of it's rated DC voltage and kept cool will last almost for ever. We have anecdotal evidence that well cared for electrolytics over 40years old still seem to perform to specification. I have a little bit of evidence as a result of reforming, that electrolytics reformed slowly to overvoltage do not suddenly leak worse, or blow up. The rate of leakage does go up as rated voltage is approached and passed.
Now apply some spikes on the applied voltage. This will lead to deterioration of the foil and insulation and electrolyte.
In time the effect of absorbing the energy of the spikes will make the cool capacitor fail, either in capacitance, or in leakage, or some other non compliant parameter. These spikes that exceed the deterioration level might repeat @ one hundred per second.
Reduce the DC voltage to 50% of rated. The spikes on the supply now need to exceed the level that causes deterioration. Maybe none of the spikes cause any deterioration. Or maybe one spike every second exceeds the deterioration level. Eventually the capacitor fails because enough spikes have deteriorated the capacitor sufficiently to not meet it's specified parameters.
The low DC + spikes last a lot longer than the very high DC + spikes. I'll guess and suggest that running at 50% of rated voltage operates correctly for 50times longer than running at 99.9% of rated.
It is my view that it's the spikes that cause much of the failures after taking account of temperature.
Last edited:
More likely that a batch will be made with known anode and cathode foils etc. then factory formed up to a particular voltage, then marked with 80-90% of that voltage.
I doubt if spikes do much harm to electrolytics. The high capacitance absorbs short spikes. Drying out due to heat is the main problem. This can be external heat or internal (due to leakage current). ESR rises, and electrolytic action is no longer able to maintain the oxide dielectric so leakage rises too.
I doubt if spikes do much harm to electrolytics. The high capacitance absorbs short spikes. Drying out due to heat is the main problem. This can be external heat or internal (due to leakage current). ESR rises, and electrolytic action is no longer able to maintain the oxide dielectric so leakage rises too.
> 10k µF ....... through a 220k
220K ? ? ?
Caps that big could easily throw-off a Watt of heat from leakage. Not in my living room, but as a test. At 75V, 13mA. Or acts-like about 5K.
Picture 100V supply, 2K series resistor. Worst-case will be if cap breaks-down at 50V. This is also 50V across the 2K. So 25mA. 50V*0.025A is 1.25 Watts. A fat Watt in a can that size will heat very slowly. Hours to burst. Hopefully the cap goes soft above 63V, current is less power is less. And you will know the approximate voltage current and Power from your voltmeter readings.
I say 220K is about 100X bigger than you want for testing caps this large and this voltage scale.
What does the cap data-sheet say about leakage? (Or a sheet for a similar part.) I would expect 10,000uFd to be allow 10mA leakage. In 220K you would need 283V supply to reach this point. In 2k only 83V per-spec, and likely 100V with <spec leakage and >spec breakdown voltage.
If you have ever been around an exploding soup-can cap, you know to wear glasses, an old tough shirt, and keep a heavy TupperWare over the cap. If you are less experienced, be warned. (I have heard of shrapnel in the ceiling but have not done it myself, yet. I have filled an amplifier with gooey paper-fluff.)
I had decent 35V rated caps running at 37V for decades, but short-term. The need was urgent and the Web was not invented, so I did what I could. Now that 5 clicks will bring over-rated caps to my porch for low bucks, I would not play the almost-enough game. A recent build, I used 25W resistors for a 3W job. I hate avoidable repairs.
220K ? ? ?
Caps that big could easily throw-off a Watt of heat from leakage. Not in my living room, but as a test. At 75V, 13mA. Or acts-like about 5K.
Picture 100V supply, 2K series resistor. Worst-case will be if cap breaks-down at 50V. This is also 50V across the 2K. So 25mA. 50V*0.025A is 1.25 Watts. A fat Watt in a can that size will heat very slowly. Hours to burst. Hopefully the cap goes soft above 63V, current is less power is less. And you will know the approximate voltage current and Power from your voltmeter readings.
I say 220K is about 100X bigger than you want for testing caps this large and this voltage scale.
What does the cap data-sheet say about leakage? (Or a sheet for a similar part.) I would expect 10,000uFd to be allow 10mA leakage. In 220K you would need 283V supply to reach this point. In 2k only 83V per-spec, and likely 100V with <spec leakage and >spec breakdown voltage.
If you have ever been around an exploding soup-can cap, you know to wear glasses, an old tough shirt, and keep a heavy TupperWare over the cap. If you are less experienced, be warned. (I have heard of shrapnel in the ceiling but have not done it myself, yet. I have filled an amplifier with gooey paper-fluff.)
I had decent 35V rated caps running at 37V for decades, but short-term. The need was urgent and the Web was not invented, so I did what I could. Now that 5 clicks will bring over-rated caps to my porch for low bucks, I would not play the almost-enough game. A recent build, I used 25W resistors for a 3W job. I hate avoidable repairs.
> What does the cap data-sheet say about leakage?
Nichicon LGU data says 10,000uFd 63V will leak 2.38mA after 5 minutes of rated voltage.
220K is clearly too much.
(Also of course explains "a lot of SLOW charging". Half-HOUR time constant.)
Nichicon LGU data says 10,000uFd 63V will leak 2.38mA after 5 minutes of rated voltage.
220K is clearly too much.
(Also of course explains "a lot of SLOW charging". Half-HOUR time constant.)
2.38mA @ 63Vdc is equivalent to ~27k
That is within 5minutes of starting the applied voltage.
Once fully reformed expect the leakage current to drop to lower than 10% and possibly as low as 1% of the specified 5minutes value.
That fully reformed gives an equivalent leakage resistance of 270K to 2M7
That is within 5minutes of starting the applied voltage.
Once fully reformed expect the leakage current to drop to lower than 10% and possibly as low as 1% of the specified 5minutes value.
That fully reformed gives an equivalent leakage resistance of 270K to 2M7
130V via a 220k resistor when combined with voltage monitoring seems to me to be a very sensible method for checking what the caps can stand. It limits the maximum possible current to little more than 0.5mA so no chance of any explosions. As the OP says, the fact that the voltage was still rising when he stopped the test shows that the cap was still charging so leakage current was still low. Of course, this test makes the assumption that good caps have leakage much lower than the datasheet says - but this is a good assumption, and proved valid in this test.PRR said:
I suspect if you read it carefully you will find that it says that the cap will leak no more than 2.38mA. In most cases it will leak much less, maybe just 10's of uA.
> will leak no more than 2.38mA. In most cases it will leak much less, maybe just 10's of uA.
Yes, and thanks for pointing it out.
But the thread is "long term". Leakage will rise over time.
I'd have no qualms about testing for 10mA. ** In the garage, with protection. ** New, it is likely a "63V" cap will go well over 90V and run warm. If it takes that for an hour, I'd be happy, set that one abused cap aside for play, use its brothers.
I used to do this with surplus caps. I was not over-volting, but some had been in storage too long (or the "surplus" was really rejects). About 8 in 10 of 300uFd 350V would take full voltage through a few K all day long; two would not come up and got hot.
I do think that today's cap prices do not encourage "marginal" uses. If you need 65V, 100V caps are not expensive.
Yes, and thanks for pointing it out.
But the thread is "long term". Leakage will rise over time.
I'd have no qualms about testing for 10mA. ** In the garage, with protection. ** New, it is likely a "63V" cap will go well over 90V and run warm. If it takes that for an hour, I'd be happy, set that one abused cap aside for play, use its brothers.
I used to do this with surplus caps. I was not over-volting, but some had been in storage too long (or the "surplus" was really rejects). About 8 in 10 of 300uFd 350V would take full voltage through a few K all day long; two would not come up and got hot.
I do think that today's cap prices do not encourage "marginal" uses. If you need 65V, 100V caps are not expensive.
I have disassembled quite a few amplifiers and receivers where the iddle-voltage on the caps was a little above the rated voltage. Never seen a failed* cap in one. (*Did not measure capacitance, but no sign of physical damage).
So yes, a higher voltage will not be good, but depending on the use, it will probably work.
Will the amp be ON 24/7? or 2hours every other day? - that will make a difference.
Kind Regards TroelsM
So yes, a higher voltage will not be good, but depending on the use, it will probably work.
Will the amp be ON 24/7? or 2hours every other day? - that will make a difference.
Kind Regards TroelsM
Maybe. Maybe not. I suspect that a small overvoltage will do less harm long-term than leaving the cap on a shelf with no polarising voltage.PRR said:
Yes, feeding 10mA into an electrolytic to see what voltage it can stand is abuse, so quite correct to discard the cap afterwards. Feeding under 0.5mA is legitimate testing and reforming so will do no harm.
Full voltage via a few k is not careful reforming but destructive testing. If you tried careful reforming you might find that one of the two rejects might be fine after treatment. If you are lucky both might be OK. Full voltage via 470k would reform those which needed it and could benefit from it.
For some reason there seems to be a popular myth that old electrolytics fall into two camps: doesn't really need reforming, or won't reform. A variant myth is: can reform quickly, or won't reform. In either case the believer in the myth will apply something like full voltage (with or without a low value resistor) and then judge from cap temperature whether the cap is OK. There are actually three camps: doesn't really need reforming, can be reformed, or won't reform. People applying brute force will never see the 'can be reformed' camp, so will conclude that such caps are in the 'won't reform' camp. Note that in this context 'old' means unused for more than a few years, especially for high voltage caps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.