This is a work in progress, but may have some useful info-
Testing Aluminum Electrolytic Capacitors
As usual, comments corrections and additions invited. Laughter and rotten fruit optional.
Testing Aluminum Electrolytic Capacitors
As usual, comments corrections and additions invited. Laughter and rotten fruit optional.
Excellent article, and as usual, Conrad and I are in perfect agreement.
I recently made some tests on NOS vintage caps, and I arrived at similar conclusions:
http://www.diyaudio.com/forums/parts/199046-oldies-but-goodies.html
The thread also includes the schematic of a cheap, vectorial esr-meter.
I recently made some tests on NOS vintage caps, and I arrived at similar conclusions:
http://www.diyaudio.com/forums/parts/199046-oldies-but-goodies.html
The thread also includes the schematic of a cheap, vectorial esr-meter.
Conrad,
Great stuff and very accessable explanations.
If I may give some positive remarks:
- could you perhaps lighten up the grey background a bit? Makes the black text easier to read.
- filling the text out between left and right sides would make it even look better
- for DC leakage test, it would be usefull to explain how to measure it.... An easy way is to connect a 10 kOhm resistor in series with the capacitor, and reading the voltage drop across the resistor to determine the leakage current. Drawn from lots of experience, I must say that 10 kOhm is hardly enough for large (>5000uF) capacitors if the resolution of your DVM is less then 10 mV (which is the avarage DVM at hand), and that small capacitors (<500 uF) have such a low leakage current that even a 100 kOhm will give a few mA to read out (reminding that the leakage current is in uA and proportional to capacitor size). Another issue with this resistor method (which is feasible for anybody), is that the 'leakage current at 2 min' is not feasible due to the resistor slowing the charging process. Hence, two solutions are possible:
1) just wait until the current (i.e. measured voltage drop across resistor) is stable for 10 or more minutes. This week I did just that for a large number of large power supply caps (5000 ~ 22000 uF) and this method showed usefull results but only after 2 to 3 hours, giving stable leakage current only after 12 hours, For small caps, it should be fine within 10 to 15 minutes
2) connect a forward biased diode (suitable rated to handle the inrush current for large PSU caps! ) in parallel with the resistor. The diode will ensure the quick charge of the capacitor and will switch off when the capacitor's positive lead is less then 0.6 volt from the PSU voltage, in which case the resistor takes it over. Again, 2 minute sample time is not a valuable approach with this method, give it some time to settle out.
- what about the actual applied voltage during leakage test? The datasheets give little guidance, but I would say at rated voltage. Thinking of it, I will test if the leakage current drops if the applied voltage is lower.
- perhaps I stressing too much here about leakage currnet, but I discovered once again this week that a cap which had an almost normal ESR and dissipation profile showed complete failure under leakage test (around 100x more than other caps), which then explained why its resonance frequency came earlier than for the others. For another (new) cap it demonstrated to be a counterfeit. Following this experience, I would almost always start with leakage current test for aging caps, albeit that for PCB mounted caps it is not an easy thing to do.
- in fact, the above leakage test method is also reforming the cap... hence an integral subject.
- and to complete the story: after reforming the caps (between 6 and 25 months old), the ESR lowered between 5 and 20%.
I think I should send you a PM since I'm doing a lot of cap testing myself
Good job !
Oscar
Great stuff and very accessable explanations.
If I may give some positive remarks:
- could you perhaps lighten up the grey background a bit? Makes the black text easier to read.
- filling the text out between left and right sides would make it even look better
- for DC leakage test, it would be usefull to explain how to measure it.... An easy way is to connect a 10 kOhm resistor in series with the capacitor, and reading the voltage drop across the resistor to determine the leakage current. Drawn from lots of experience, I must say that 10 kOhm is hardly enough for large (>5000uF) capacitors if the resolution of your DVM is less then 10 mV (which is the avarage DVM at hand), and that small capacitors (<500 uF) have such a low leakage current that even a 100 kOhm will give a few mA to read out (reminding that the leakage current is in uA and proportional to capacitor size). Another issue with this resistor method (which is feasible for anybody), is that the 'leakage current at 2 min' is not feasible due to the resistor slowing the charging process. Hence, two solutions are possible:
1) just wait until the current (i.e. measured voltage drop across resistor) is stable for 10 or more minutes. This week I did just that for a large number of large power supply caps (5000 ~ 22000 uF) and this method showed usefull results but only after 2 to 3 hours, giving stable leakage current only after 12 hours, For small caps, it should be fine within 10 to 15 minutes
2) connect a forward biased diode (suitable rated to handle the inrush current for large PSU caps! ) in parallel with the resistor. The diode will ensure the quick charge of the capacitor and will switch off when the capacitor's positive lead is less then 0.6 volt from the PSU voltage, in which case the resistor takes it over. Again, 2 minute sample time is not a valuable approach with this method, give it some time to settle out.
- what about the actual applied voltage during leakage test? The datasheets give little guidance, but I would say at rated voltage. Thinking of it, I will test if the leakage current drops if the applied voltage is lower.
- perhaps I stressing too much here about leakage currnet, but I discovered once again this week that a cap which had an almost normal ESR and dissipation profile showed complete failure under leakage test (around 100x more than other caps), which then explained why its resonance frequency came earlier than for the others. For another (new) cap it demonstrated to be a counterfeit. Following this experience, I would almost always start with leakage current test for aging caps, albeit that for PCB mounted caps it is not an easy thing to do.
- in fact, the above leakage test method is also reforming the cap... hence an integral subject.
- and to complete the story: after reforming the caps (between 6 and 25 months old), the ESR lowered between 5 and 20%.
I think I should send you a PM since I'm doing a lot of cap testing myself
Good job !
Oscar
Last edited:
Oscar, I'll see about addressing some of that stuff. I agree that leakage can be very important, but as you can see, measuring it to low levels can require a bunch of test equipment or DIY circuitry that many people won't have on hand. It's also dangerous if the caps are HV types. Still, including more detail on the procedure is a good idea.
I've put caps on the curve tracer and it's interesting, but also nearly impossible for most people to get a clear pass/fail answer on a marginal cap. The GR Bug Hound was essentially that, and an experienced user can work miracles with them, but I haven't seen one in use for years. Too odd and too expensive for most.
I've put caps on the curve tracer and it's interesting, but also nearly impossible for most people to get a clear pass/fail answer on a marginal cap. The GR Bug Hound was essentially that, and an experienced user can work miracles with them, but I haven't seen one in use for years. Too odd and too expensive for most.
Conrad,
I see your point. But for the average audio amp power supply cap in the 50~80 Volt range and above 4700uF the simple test with a 10k~50k resistor in series and only a reasonable DVM with 1 mV resolution will a be test possible for anybody (I think). And for many folks, the simple leakage current test for the power supply caps will be the most important question as replacement is often much more costly then the smaller caps (see lots of Q&A on various audio forums).
For example, with a 10 kOhm resistor in series I measured the following leakage currents after 2hr, 5hr, 12hr (they remained stable after 12 hours) for the following new caps:
NCC KMH 35V-22000uF: 26/22/1.4 mV (2.6/2.2/1.4 uA)
Pana TS-HA 25V-22000uF: 38/30/25 mV (3.8/3.0/2.5 uA)
Nichicon KG 25V-10000uF: 19/11/0.2 mV (1.9/1.1/0.0 uA)
and a fake NCC KMH 63V-10000uF: 3060/2800/2690 mV (309/283/271 uA) across 9910 Ohm resistor.
yep, Panasonic TS-HA ain't that good; ESR curves and phase shift aren't the best in class either. The Nippon Chemi-Con KMH are realy good capacitors, and the Nichicon KG are really fabulous.
hence, a simple test
I see your point. But for the average audio amp power supply cap in the 50~80 Volt range and above 4700uF the simple test with a 10k~50k resistor in series and only a reasonable DVM with 1 mV resolution will a be test possible for anybody (I think). And for many folks, the simple leakage current test for the power supply caps will be the most important question as replacement is often much more costly then the smaller caps (see lots of Q&A on various audio forums).
For example, with a 10 kOhm resistor in series I measured the following leakage currents after 2hr, 5hr, 12hr (they remained stable after 12 hours) for the following new caps:
NCC KMH 35V-22000uF: 26/22/1.4 mV (2.6/2.2/1.4 uA)
Pana TS-HA 25V-22000uF: 38/30/25 mV (3.8/3.0/2.5 uA)
Nichicon KG 25V-10000uF: 19/11/0.2 mV (1.9/1.1/0.0 uA)
and a fake NCC KMH 63V-10000uF: 3060/2800/2690 mV (309/283/271 uA) across 9910 Ohm resistor.
yep, Panasonic TS-HA ain't that good; ESR curves and phase shift aren't the best in class either. The Nippon Chemi-Con KMH are realy good capacitors, and the Nichicon KG are really fabulous.
hence, a simple test
Yes it's a very simple test.
But there are problems with interpreting the results.
The voltage drop across the resistor indicates that current is flowing.
That current is
1.) charging up the capacitor.
2.) reforming the dielectric layer (insulator).
3.) leaking through the capacitor.
I can't tell how much of each current makes up which proportion of the total current passing through the sensing resistor.
An example:
let's suppose the capacitor is nearly fully charged. That presumes the charging current is dropping to near zero.
Now for whatever reason, let the charging voltage increase very slightly.
The charging current will adopt a new higher value. And stabilises again at near zero pA
Let the charging voltage drop very slightly. The charging current flows in the opposite direction. The total current measured in the sensing resistor is a combination of +ve leakage current, plus +ve reforming current, minus the -ve discharging current.
I still don't know what proportion of each exists through the sensing resistor.
What I do know is that practical test procedures make it very difficult to estimate the leakage current.
As an example of looking for or at this problem.
Connect one sensing resistor to one cap. Repeat this for half a dozen other caps.
Now connect each combination to a common sensing resistor. Apply the charging voltage to the common resistor and reform/charge all the caps from the same emf.
Note how the currents vary hour by hour over the next 72hours.
You too will be surprised to see how the currents increase, decrease, reverse and they all appear to follow some kind of pattern but it is very confusing.
A good thinker/analyst among you might be able to develop a test procedure that allows the three types of current to be identified. I can't see that solution.
BTW,
I often find that electrolytic leakage current is 100 to 10,000 times less than specified in the datasheet.
So much so, that I thought I had the wrong units in the formula and doubted my interpretation of the datasheet. I sought confirmation on this Forum some years ago.
But there are problems with interpreting the results.
The voltage drop across the resistor indicates that current is flowing.
That current is
1.) charging up the capacitor.
2.) reforming the dielectric layer (insulator).
3.) leaking through the capacitor.
I can't tell how much of each current makes up which proportion of the total current passing through the sensing resistor.
An example:
let's suppose the capacitor is nearly fully charged. That presumes the charging current is dropping to near zero.
Now for whatever reason, let the charging voltage increase very slightly.
The charging current will adopt a new higher value. And stabilises again at near zero pA
Let the charging voltage drop very slightly. The charging current flows in the opposite direction. The total current measured in the sensing resistor is a combination of +ve leakage current, plus +ve reforming current, minus the -ve discharging current.
I still don't know what proportion of each exists through the sensing resistor.
What I do know is that practical test procedures make it very difficult to estimate the leakage current.
As an example of looking for or at this problem.
Connect one sensing resistor to one cap. Repeat this for half a dozen other caps.
Now connect each combination to a common sensing resistor. Apply the charging voltage to the common resistor and reform/charge all the caps from the same emf.
Note how the currents vary hour by hour over the next 72hours.
You too will be surprised to see how the currents increase, decrease, reverse and they all appear to follow some kind of pattern but it is very confusing.
A good thinker/analyst among you might be able to develop a test procedure that allows the three types of current to be identified. I can't see that solution.
BTW,
I often find that electrolytic leakage current is 100 to 10,000 times less than specified in the datasheet.
So much so, that I thought I had the wrong units in the formula and doubted my interpretation of the datasheet. I sought confirmation on this Forum some years ago.
Last edited:
Woody, unfortunately no. Electrical leakage is only one aspect and I have lots of bad caps that aren't overly leaky. I also haven't found that reforming improves Rs (esr) by much, if at all, though it's largely an issue with HV caps and I don't do that many of those.
IMO, a gross leakage test is certainly easy. It's only when you get down to the level of what I consider decent quality capacitors that it gets harder. Maybe the answer is to go with the data sheet values, even though most caps are far better. If it meets the sheet, it's probably fine.
There are all manner of ways to test caps, depending on what's available. It's easy to build a simple bridge that shows losses. Or, one could use a signal generator and resistor in series with the cap, then use an AC voltmeter to measure and calculate the value, or a scope to get both value and loss by looking at the phase angle. Or, use a big resistor, power supply and stopwatch!
I was trying to stick with the direct approach, using a proper bridge or esr meter, but maybe I should offer some alternatives. There is an article on my site for building a very simple bridge optimized for larger caps.
IMO, a gross leakage test is certainly easy. It's only when you get down to the level of what I consider decent quality capacitors that it gets harder. Maybe the answer is to go with the data sheet values, even though most caps are far better. If it meets the sheet, it's probably fine.
There are all manner of ways to test caps, depending on what's available. It's easy to build a simple bridge that shows losses. Or, one could use a signal generator and resistor in series with the cap, then use an AC voltmeter to measure and calculate the value, or a scope to get both value and loss by looking at the phase angle. Or, use a big resistor, power supply and stopwatch!
I was trying to stick with the direct approach, using a proper bridge or esr meter, but maybe I should offer some alternatives. There is an article on my site for building a very simple bridge optimized for larger caps.
@ Andrew
That's why I mentioned earlier to wait some prolongated time (>6 to 12 hours) to have a stable current flowing, which means that both the reforming and charging have ended, and that the measured current represent the leakage current. To minimise charging/reforming effects, the test with a diode in parallel with the resistor can be done. I agree with you that the test ain't holy and all-saying. It provides *some* information. Ideally it makes sense when comparing the leakage current at new and later condition.
@ Conrad
I will look into the bridge, as I would of course like to make better (more real) measurements whenever possible. Thanks for that. When I will build one, then I could compare it with the above measurements, see what comes up. Oh yeah, I did fully re-measure the 'reformed' caps again (new caps), and indeed the ESR (for 10mF and 22mF) dropped between 5~20%, depening on brand. Hence, this is not limited to HV caps, mine are rated 25V and 35V. The downside, however, is that the resonance frequency seems to shift backwards (to lower frequency) as well (depending on brand).
@ Woody
Well, at least you can compare two (or more) identical power supply caps and see if one (or more) has much higher leakage current, and draw *some* conclusion. Or if you have a batch of capacitors to choose from, it *might* help you to choose.
That's why I mentioned earlier to wait some prolongated time (>6 to 12 hours) to have a stable current flowing, which means that both the reforming and charging have ended, and that the measured current represent the leakage current. To minimise charging/reforming effects, the test with a diode in parallel with the resistor can be done. I agree with you that the test ain't holy and all-saying. It provides *some* information. Ideally it makes sense when comparing the leakage current at new and later condition.
@ Conrad
I will look into the bridge, as I would of course like to make better (more real) measurements whenever possible. Thanks for that. When I will build one, then I could compare it with the above measurements, see what comes up. Oh yeah, I did fully re-measure the 'reformed' caps again (new caps), and indeed the ESR (for 10mF and 22mF) dropped between 5~20%, depening on brand. Hence, this is not limited to HV caps, mine are rated 25V and 35V. The downside, however, is that the resonance frequency seems to shift backwards (to lower frequency) as well (depending on brand).
@ Woody
Well, at least you can compare two (or more) identical power supply caps and see if one (or more) has much higher leakage current, and draw *some* conclusion. Or if you have a batch of capacitors to choose from, it *might* help you to choose.
Last edited:
I have by now consulted quite a few manufacturers' datasheets on specifications and testing of their capacitors to check that they meet specifications.
Every manufacturers' datasheet that I have read states that the capacitor under test must be reformed before test and stipulate some time within which the specification test must be done after the reforming procedure.
I cannot see why there should be such universal agreement on this reform before test other than the manufacturers know that the results will be affected by the re-forming procedure.
Every manufacturers' datasheet that I have read states that the capacitor under test must be reformed before test and stipulate some time within which the specification test must be done after the reforming procedure.
I cannot see why there should be such universal agreement on this reform before test other than the manufacturers know that the results will be affected by the re-forming procedure.
I can understand the manufacturers being very specific as they have to be rigorous and consistent in their testing. My statements are based only on what I've seen. So many times I've hoped for some improvement in DF on older caps in my stock, but have never seen it. My guess is LV caps don't really un-form and rarely need reforming. HV caps are another story.
I've got an update underway but need to add a few more things before uploading it. Keep 'dem comments coming!
I've got an update underway but need to add a few more things before uploading it. Keep 'dem comments coming!
I also think that the whole storage (shelf) life numbers provided in the datasheets is very vague. Oh, more comments? Let me think
What about putting in the classic vector diagram showing ESR, |Z|, phi, etc, as to explain what is DF (tan delta), phi, ESR, Xc/Xl, |Z|, etc? Inspiration perhaps from page 6 of this Nichicon document? And a *little* bit from this website (extract what's relevant, at least they talk about phase shift)
This is some reference document from CDE.
Some reference on leakage current from BC.
What about putting in the classic vector diagram showing ESR, |Z|, phi, etc, as to explain what is DF (tan delta), phi, ESR, Xc/Xl, |Z|, etc? Inspiration perhaps from page 6 of this Nichicon document? And a *little* bit from this website (extract what's relevant, at least they talk about phase shift)
This is some reference document from CDE.
Some reference on leakage current from BC.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.