Capacitors, Burn-in and mythology

There are many contentious issues in this world of high fidelity. Supports, cables, tweakery and various forms of black magic (or accusations thereof) are among them.

I wish to address capacitors and burn-in.

I have recently performed some experiments. In brief, these experiments consisted of measuring the performance of an amplifier output stage, then replacing the capacitors with new ones of the same type, and re-measuring. I then replaced the capacitors again with new ones of a different type. The same output stage was used, and had been long run-in before the experiments were performed.

The measurement taken was a simple THD sweep from full power down to fractions of a Watt into an 8ohm resistive load at 1kHz. At this frequency, the resistive load is a reasonable approximation to the behaviour of a loudspeaker, unlike at very low or very high frequencies. I do not intend to get into this debate.

I regret that I am unable, for various reasons, to publish the actual results I measured. However, a brief summary...

At very low powers, no considerable difference was observed between the capacitors.

At the 1W power level, there was starting to show a difference. The worst capacitor was the new one of the original type, with the "burned-in" capacitor showing a small but measurable improvement. The other new capacitor performed better still, but this was a more expensive capacitor (I will not say what it was, but the type has been mentioned here on more than one occasion).

The more expensive capacitor maintained a performance advantage through higher powers up to about the 40W level. Of the other two, the burned-in capacitor maintained an advantage over the new capacitor of the same, cheaper type.

At higher power levels still, things changed. The two new capacitor types started rapidly increasing in distortion whilst the burned-in items remained lower (this is before the onset of clipping).

Going even further, just before the onset of clipping, the story changed again. Now it was the brand new cheap capacitor that gave the best result, with the expensive capacitor poorest and the burned-in cap just in front.

Conclusions...

1) There are measurable differences between different types of electrolytic capacitor
2) Burn in is a very real and measurable effect (The amplifier had been characterised when first built, and that characterisation matched that measured with the new capacitors).

Explanations

So, what changes? There is some misunderstanding about how electrolytic capacitors work. An electrolyte, as anyone who has done GCSE chemistry knows, is a solution that contains a large number of charged atoms (ions) and is very conductive. Clearly this is not the dielectric material. The dielectric in an electrolytic capacitor is, in fact, a layer of aluminium oxide on the surface of the foil that forms the plates of the capacitor. This layer gets damaged and made uneven manufacture, and before it leaves the factory, the capacitor undergoes a process called "reforming", where a larger than specified voltage is applied to the capacitor in an elevated temperature environment, in order to recreate this oxide layer (or "reform" it). This process is not entirely complete, and the capacitor may spend a year or two in various warehouses before it is actually used in a production item. Thus, when equipment is first put into use, and the capacitor is given a voltage and elevated temperature, the process is slowly completed. This is the burn-in that people have anecdotally refered to. The completion of the reforming process results in a drop in ESR, increase in ripple current and general slightly improved performance in the cap.

It is also extremely well known that electrolytic capacitors have a limited lifetime. If you have a piece of kit that is 10 or more years old and contains electrolytic capacitors, you may find an improvement in performance simply by replacing the capacitors with new ones of the same type, let alone higher performance items.

Although I accept that there could be measurable differences between capacitors, I question whether these differences are audible. If, for example, distortion goes from 0.001% to 0.01%, that's a tenfold increase, but neither is audible.

Furthermore, electrolytic capacitors do take time to reform, I measured the leakage of one electrolytic that had laid in my drawer for over 30 years and this reduced substantially over a 48 hour period, but then electrolytics are generally used in circuit positions where leakage isn't important, like decoupling.

I suggest that just about everything makes a measurable difference (even cables!) but very little makes an audible difference that isn't very easily explained.

S
 
I-S. This is an excellent read since it's not a very easy job to tediously monitor the effects of voltage and current overrating on capacitor that are working in a circuit. It has always been the case that the electrolytic capacitors have a very limited lifetime, in most cases it is just three to four years depending upon the conditions they are used in. The conditions such as humidity plays an important role in determining that. It was a very interesting read and cheers for that.

Electrolytic capacitors are rated at a certain number of hours at a rated voltage and temperature. A rule-of-thumb is that lifetime doubles for every 10° temperature difference, so a 85° capacitor with a rated life of 5000 hours, will, in a normal domestic environment, say at 40° internal temperature, will last 80000 hours, or over 100 years, used day in day out for two hours a day. I have a number of vintage amplifiers, all over 40 years old working to their original specification and with original capacitors.

In most cases, recapping amplifiers 20 years old is pointless. Yes, some amplifiers might have been abused, overheated and left on 24/7 for 20 years, but most domestic HiFi isn't.

S
 
I can vouch for this, tested a load of old large caps that were thrown from a repair shop that closed down. They were still within tolerance despite their age.
 
When i service old kit the caps get thrown into a large box - must be many hundreds in there over the last few years.
Everything from 1uf couplers to 10kuf PSU caps. I recently pulled 50 at random and measured them for ESR, capacitance and leakage, Of the 50, only 1 was outside spec!

This isn't always the case and I've noted some brands of axial non-polar caps used in old crossovers can drift a fair bit, but generally I'm finding quality branded polarised caps from the 70s and 80s kit I tend to work on are perfectly fine.

Heat, high ripple and running caps near the voltage limits do the most damage.
 
Non-polar electrolytics as used in most crossovers are normally made from two conventional electrolytics back-to-back inside one envelope. This means that none of the electrolytics in the pair have any polarising voltage across them, which isn't how electrolytics are meant to work. That may account for why they lose value over time, whereas normal polarised electrolytics that spend their life with their design DC voltage across then last very much longer.

Non-electrolytic large value non polarised capacitors are a lot bigger and more expensive which is why they're rarely used in commercial crossovers.

S.
 
Interesting to know, thanks Serge! I didn't realise a lack of DC would make them die quicker.

To be fair, a lot of speakers use film caps. At least the high-end stuff people here have. Only when a really large value is needed say more than 47uf will electrolytic usually be found.
 
Interestingly, I've found in case that where old non-polar caps in crossovers have drifted its been upwards - often 50-100% high uf than spec.
One of the reasons many old loudspeakers can sound 'off'.
 
I haven't seen any properly conducted test results on electrolytic capacitor performance Vs voltage, so my suggestion above is based on supposition, knowing how electrolytic capacitors work. In particular, electrolytic capacitors take a small amount of time to form, and will only form in the presence of a DC potential across them. In back-to-back non-polar electrolytics, there is never that polarising voltage, so they will never form properly. That will also make leakage higher, but as loudspeaker crossovers are inherently low impedance devices, the fact that there's leakage in the kohms region is pretty irrelevant.

Why the capacitance drifts upwards, I can't suggest a mechanism, but certainly, if the capacitance doubles, then the crossover point will be way off the design frequency. Film caps are a lot larger and more expensive than non-polar electrolytics, so whilst I can see them being used in larger, more expensive loudspeakers, at the budget end, which also tend to be smaller, electrolytics get used.

S.
 
NP electrolytics tend not to follow the expected filter function either, IME. If I model a 2nd order xover and use an electrolytic, the filter tends to go okay for the first ~20dB then the slope either becomes more or less steep. It messes with the phase so the result needs tweaking without the aid of simulation. Film caps are a lot easier, lol.

As you say though, the size and expense means there are not always alternatives. I have some 100uF polyprop caps here that are the size of a Coke can!
 

Latest posts

Back
Top