Real World Electrolytic Reliability And Aging
Electrolytic caps are a deep subject. You can find any number of scientific papers on the web written about their use, design, reliability, and so on. They are the only basic electronic part left that has not been improved to the point of virtually lasting forever. They have a definite and relatively short lifespan, with complicated relationships between the various parameters that affect their useful life and failure modes. If it wasn’t for electrolytic capacitors, it would be possible to design a piece of electronic equipment that would have an average lifespan almost completely determined by the wear of its electromechanical components.
For our purposes, we will limit this discussion to the two most potentially damaging parameters, heat and ripple current. These two are in fact related, because too much current through the capacitor and it’s equivalent series resistances causes heating, from simple ohms law. We can pretty much look at capacitor life and reliability from just the viewpoint of heat, because too much current makes it. . . . overheat.
As long as a capacitor is kept below its maximum voltage rating, lifespan is pretty much un-affected by whatever voltage is applied. Over voltage will kill an electrolytic quickly, as well as reverse biasing it, with catastrophic results.
Capacitors are rated for their worst case use, for temperature and ripple current. Most electrolytics that we encounter in audio gear are rated either for 85C or 105C maximum temperature. There are caps used mostly in military and medical equipment with a 125C rating. At max temperature and max ripple current, most caps have a useful life of 2000 hours. Today there are heavy duty caps made for high ripple current use, like in a switching power supplies, rated for 3000 hrs. I have also noticed that there are caps appearing on the market now, rated for much higher hrs.
For every 10C reduction in temperature, the life of the cap is effectively doubled. Bring the cap core temperature down to 30-40 degrees C, typical of most audio gear, and the number of hours becomes very large. An 85C cap, running at 35 degrees core temp, is 50C cooler than rated. That would be an increase in hours of 2 to the 5th power or 64,000 hours of use. At 4 hours a day, every day, that is over 43 years of use. We run into other problems before we get there, though.
Another factor comes into play, and that is simply the passage of time. The electrolyte, from which the cap gets its name, is a wet solution, and despite the best efforts of the manufacture to seal it in, evaporation occurs, and the caps dry out. Higher temps accelerate this process. This begins at the time of manufacture and gradually worsens as the plastic and rubber parts that seal the capacitor deteriorate with age. Most manufactures rate their caps for a 15 year life, beyond which they no longer take any responsibility for performance or failures.
This 15 year rating is mainly based on the reduction of rated capacitance, due to electrolyte evaporation. Most electrolytics are 20% tolerance, and so are considered to be out of spec when the capacitance has reduced by 20%. The cap will continue to function for a long time after the 15 year point, it just may not be in spec. A 10uf cap might become 8uf, and so on. Thus, the 15 year rating by the manufacturer is about the cap losing rated capacitance, not so much about failure.
So, you have this lifespan based on hours of use, and another lifespan that is simply based on how old it is. What about real world experience? We have changed out literally 1000′s of electrolytics. I have measured many, many caps and had the opportunity to observe the effects of older caps, and the effects of changing them out for new.
A year’s worth of caps from our recapping operations, probably about 4000 electrolytics. We have tested 100′s of them to find out what actually happens at 30-40 years of aging.
Most of the circuitry puts very little stress on the caps, except for the output section. Hi output audio pushes the power supply caps into a high ripple current mode.
If you haven’t already done so, you might want to take a look at these graphs from Elna, a supplier of some of the caps in the QRX’s, here. What we have found is that the bathtub failure curve seems to stay flat for around 30 years or so. At 30 years, circuit failure is very seldom caused by caps, and most units are still functioning, if not up to design standards. At 35 years, many units are starting to fail, and the failure can often be traced to capacitor faults. At 40 years and beyond, more units seem to be not working than are, and the electrolytic caps are the main culprits. Keep in mind that I am only talking about whether a piece of gear will function, not if it will meet specifications.
It seems, that at that 35-40 year point, when the electrolytics start to fail, the electrolyte dries out to the point that the cap eventually develops an internal short.
Component reliability is a matter of statistics and how it affects overall equipment reliability is also statistical in nature. About half of the smaller caps in that big tub of used electrolytics in the above picture, will check out fine, even though they might be 35 years or older. They will be in spec for capacitance, ESR, and leakage. Still completely usable. Caps used in low level circuitry, that are not stressed by high ripple current, can hold their specs for a very long time. It mostly depends on how well the cap has remained sealed, to prevent evaporation. Power supply caps are a different story.
Here’s a page from my notebook, based on capacitor testing I have done on old caps. What I attempted to do was to quantify, from experience, the actual years that go with the failure graphs that the manufacturers publish. I have no data on the actual numbers of failure rates. Click for larger.
We mostly work on gear that is around 35 years old and older and our rule of thumb is that we don’t troubleshoot anything until after the recap, if we are doing that. That is because 85% of the operational problems are fixed by the recap/restore work we do.
The Sansui QRX’s were produced starting in ’71-’72 and continued for the whole decade. Typically, the big power supply caps will have deteriorated to 50-70% of their rated capacitance, with a consequent rise in internal equivalent series resistance, or ESR and leakage. The top of a cap may be bulging, and sometimes they have vented, leaking out some fairly corrosive brown liquid– the electrolyte. At minimum, all this gear should have the power supply caps replaced, and this includes the larger caps spread around on the PC boards that have a voltage smoothing function.
You can verify the actual capacitance of one of these big power supply caps very simply. Since most audio hobbyists don’t have a capacitance meter, or many meters will not measure more than 100 or 200 uf, you can use the time constant formula to your advantage. With one terminal of the cap disconnected from the circuit, charge it up with a low voltage like a 12 volt supply, or a 9 volt battery. Hook a multimeter up to it, and note the exact voltage. Remove the source voltage and place a 10k ohm or 100k ohm resistor across the cap. Watch the meter. The time it takes the cap to discharge, in seconds, to 37% of it initial voltage will be a direct reading in 100′s of uf for the 10k resistor and 10′s of uf for the 100k resistor. This is how we measure big power supply caps all the way up to 20-30K uf. You need to use a meter with a high input impedance, but it works well.
For example, an old 10,000uf cap takes 66 seconds to discharge from 9.2 volts down to 3.40 volts, through a 10k resistor. 66 X 100 = 6,600uf. This is pretty typical for big power supply caps. This is why I cringe every time I see a set of power supply caps for sale on ebay, for a high asking price. I know from experience, that these old caps are probably only 60-70% of value, and are a waste of money.
If the cap has not been charged for a while, leave the voltage source attached for a few minutes to stabilize it, before starting the discharge. If you use 9-12 volts, there is no danger in touching the leads with your finger, it’s a safe test.
You will also likely run into bigger caps that have so much leakage, you cannot effectively make the measurement. I’ve seen them so bad that they self discharge fast without even adding the resistor. Trying to measure them is almost pointless.
It’s important to realize that there are two types of failure from the manufacture’s point of view. First, there is the capacitor losing value until it is out of spec. This is considered a failure by the manufacture, even though the capacitor will continue to work at a reduced spec. The second failure mode is when the cap would actually cause a circuit failure due to an internal short, or perhaps losing all its value and going completely open circuit. Keep in mind that as the capacitance reduces in value, many circuits will ultimately fail, as a reduced functioning capacitor can put much higher stress on other components in the circuit. This is why, the bad cap will often not be found, but a related part that blew or overheated will be found and changed out by a tech. The real solution of course, is to renew the electrolytics, in addition to doing the repair.
This gradual loss of capacitance value explains the most noticeable phenomena of doing a recap on audio circuitry. Since electrolytics are used throughout solid state audio circuitry for coupling between stages, when that coupling capacitance is reduced, low frequency response is rolled off. A recap brings back that full and rounded low end and is immediately noticed by practically everyone. It’s not subtle.
I suspect that caps being produced today will have a longer lifespan than those produced 35 years ago. It will take 35-50 years to make this determination, though. These days, we have another type of failure mode, which has to do with all the little ones and zeros getting lost, and the gear essentially becoming non-repairable. I imagine you feel the same way I do about this kind of stuff, or you wouldn’t be reading this site, devoted to analog quad gear from the ’70’s. LOL