Capacitors in series - pitfalls?

[dinkydi]

Thanks for that GMB, I think we are in agreement.

Electrolytic capacitors are certainly robust. It has not been mentioned that they have surge-voltage ratings which allow their maximum working voltage to be exceeded by about 50%, for a short time! Also, they can be run continually at high temperatures with a core in the vicinity of 100 degrees C or more. The considerable surface area of the larger sizes means they can dissipate substantial power. While the leakage currents of electrolytics are normally low, circuit designs must nevertheless be able to cope with possible high leakage currents of 10s of mA for medium size, to 100s of mA for very large capacitors.

If you understand this robust nature you can appreciate why those using new-old-stock or recycled electrolytics, and who think that cap oxide reforming is a lot of nonsense, can get away with soldering them straight into a circuit without reforming. Most circuits by default limit the reforming current and the electrolytic is soon purring after reforming itself in situ. However, an occasional exception is in a high-current circuit, directly after a rectifier, for example. If you don't bother with reforming, check the temperature of the just-installed cap with your finger after a few minutes running (don't forget some caps may be at high voltage!). If the cap is warm keep monitoring it until it settles in.

Peter

[dinkydi]

Hi Craig, Hope I haven't misinterpreted your post, but you appear to disagree with my assertion that the analysis you provide assumes that capacitor voltages should be kept equal.

These tiny quotes from your document are revealing: "the ideal value of V/2" page 2 after equation 6, and "the ideal mid point voltage V/2" page 2 after equation 7. (Other references exist.)

Another assumption concerns the advisability of even using balancing resistors - no technical justification is given. As you point out, the problem of the methodology is the difficulty of estimating leakage-current mismatch. However, other problems also exist.

The source of equation 9 (with its uncombined constants!) is not given, and we are not told of its applicability or limitations. I remain to be convinced that the equation is even of the right form. This does not inspire confidence to rely on this analysis in your circuit design.

According to the "Revision to TD001" there was originally a mistake in the analysis (it is now "corrected") - the definition of "difference in leakage currents" in equations 8 & 9 wasn't identical. To "fix" this a factor of 2 was added to equation 9. However, the factor should really be a variable. This simplification compromises the analysis, I believe.

The analysis is not a "worst-case" analysis, and neither is the "Capacitors in banks" analysis. This means that designs based on these analyses may fail, though hopefully infrequently.

The theoretical problem of capacitors in series seems so simple (well at first, anyway) you would think it must have all been sorted out long ago, however, no one seems to have uncovered a technical analysis that withstands scrutiny and is without controversial assumptions. If you want to know the size of balancing resistors for electrolytic capacitors in series, there is no accepted method of calculation, unfortunately.

Peter

[Andy Day]

Oh dear, I did stir up a hornets nest didn't I?

On a sane note, the Philips 462A with several series parallel capacitor networks inside waxy casings is in spite of frequent use still healthy and beautiful, and occupies pride of place in my sitting room. It's been there since 2007!, so perhaps pragmatism is the best approach.

[Craig Sawyers]

Originally Posted by Andy Day

Oh dear, I did stir up a hornets nest didn't I?

Bzzzz

Lots of apps notes about voltage sharing - not limited to the one I originally referred to.

For example http://www.coilgun.info/theorycapacitors/AEappGUIDE.pdf
Or http://www.lintronicstech.com/index%..._resistors.pdf

Everyone seems to use different formulae. But then again there are a dozen or more formulae for the maximum heart rate as a function of age - and for me at age 56 they all predict the same +/- 1 beat per minute.

And if you are (say) using two 350V capacitors in series across a 600V supply, do you need balancing resistors? I would say so. They also handily discharge the capacitors after switch-off.

Craig

[techiesteve]

Personally I would never try to achieve matching a 0.05uF capacitor with 4 in series parallel, but simply use a 0.047uF of adequate voltage. Not as common as they used to be, but can still be found.

Re electrolyics in series, I would always use adequate balancing resistors. What is often overlooked is the voltage rating of a resistor, this can vary and may be unknown if using a junk box part. I worked repairing SM PSU's for many years, and a common failure was the inverter transistor, sometimes FET, failing, due to one of the series connected reservoir caps. One of the 2 balancing resistors will have failed and the cap sometimes had the top vents bulging and was low or open. Two 200V capacitors in series with 330 - 340V across doesn't leave much margin.

[emeritus]

Hopefully no users of this forum would make the pitfall of calculating the values of unequal series capacitors in the way given for unequal parallel resistors that is set out in the attached, which is from the 1992 reprint of the 1990 edition of a Longman Revise Guide for GCSE CDT-Technology.

I don't know what I find more depressing, the thought that pupils who did know how to do it properly, would have been marked down, or the fact that the author could have made such a fundamental error that was not picked up before the reprint was made. These guides were written by "... the people who set GCSE papers and mark them" , and this particular guide was written by a "... Chief Examiner in GCSE CDT...for a major examining group"!

[Station X]

By that reasoning a quarter plus a quarter is 2/16 or 1/8.

For two components I always use "product over sum" and as a check make sure the answer is lower than the lowest value cap or resistor.

[robjkmannering]

The text also states that the caps are in parallel when the diagram clearly shows them in series.

In the series resistor network, the first resistor is not shown with a value yet it is taken to be 100 ohms in the answer.

As you say, very poor. Would probably get you an A* grade at GCSE level nowadays though!

Rob

[Station X]

It's a case of the heading "Capacitors in series" being missing.

[kalee20]

Oops!

There's also the 'Resistors in Series' example at the top of the page - don't Longmans know that the k in kΩ is supposed to be a small k? (Interestingly the 2.2kΩ resistor has this correct - zero marks for consistency!)

[Station X]

Yes, but SI prefixes are not very consistent. We have G for giga, M for mega (bigger than the basic unit) and c for centi, m for milli (smaller than the basic unit). By that logic I'd expect the prefix for kilo to be K rather than k. Whether you use kohm or Kohm, there's no confusion. That's not the case m and M though.

[G8HQP Dave]

SI is very consistent, unless you try to impose your own ideas about consistency. It is best to stick with 'k' for 1000, as then that leaves 'K' for 1024 (=2^10) as used in IT. (There is an ambiguity then for 'M' which SI uses for 10^6 and IT uses for 2^20, similarly 'G').

The revision guide is appalling, but it demonstrates the modern problem of rigid marking schemes: being brighter than the examiner can lose you marks.

[Station X]

SI prefixes less than 10^3 are always lower case. I wonder why the committee or whatever, who wrote the standard in about 1960, decided on that though?

[kalee20]

The official abbreviations for SI prefixes that are multiples of the basic unit are capital letters, except kilo which is small k. (Prefixes for submultiples are always small letters, including µ, the only Greek one).

Capital K is used as abbreviation for Kelvin - so arguably there could be scope for confusion, though it's unlikely in practice.

I also use the product over sum formula for two capacitors in series - and yes a quick sanity check is that the result should be less than either. And, other things being equal, if the capacitors have different values the voltage divides between the capacitors in inverse proportion to their values, until leakage resistances redistribute this...

[cannissolis]

@Brymac, I didn't see a reply to your question relating to the dishwasher caps,
These are plastic film caps and will be great in power supplies, they will have an AC voltage rating so will be good on DC to at least peak value (ACx1.414) and probably good to nearly pk-pk value.
The leakiest 'lytic in series will have a slight tendency to further form and may even reduce its capacitance marginally, both actions will have a self balancing effect but the bleed resistors shoud still be included and have a value that will swamp the leakage current.

[Billy T]

Originally Posted by boxdoctor

Just a small point on the above subject - If you connect electrolytics in series, you will also have to bear in mind that the E.S.R. will probably be materially larger than that of a single unit rated to do the same job. If you use two identical caps , then the E.S.R. would be twice that of one of them. This probably wont matter in valve equipment, but would need to be borne in mind in the case of a S.M.P.S. or similar application. Tony

I agree, but the ESR of modern caps is very low (this I know from testing, not assumption) while the ESR of the original caps would probably not be up to current production standards. However, for an SMPS repair I would always use a single cap of the correct spec and of reputable brand, regardless of cost. With over 700 SMPS repairs under my belt, I know that cheap electros (ex China) are not advisable for SMPS use, and I will only use known brands of good repute.

Cheers

Billy

[GMB]

I agree, but the ESR of modern caps is very low (this I know from testing, not assumption)

That's odd because I would say the opposite.

Time and again I have assumed what everyone says that modern parts have better ESR and time and again I am disappointed that the old parts I removed turn out to be better.

The spec. sheets of modern parts don't thrill me either. The quoted life expectancy for electrolytics is usually terrible (OK, it is mainly about staying within tolerance rather than going pop, but even so).

[KeithsTV]

The quoted life of an electrolytic is at maximum ripple current and maximum temperature. The life doubles for each 10 degree drop in temperature. A 105C capacitor with a rated life of 2000 hours (83 days) has a life of over 14 years when run at 45C. In practice the life would be longer as the ripple current would normally be a lot less than the maximum.

Keith

[Herald1360]

Originally Posted by GMB

Time and again I have assumed what everyone says that modern parts have better ESR and time and again I am disappointed that the old parts I removed turn out to be better.

Maybe possible that the size of older parts relative to new ones has some bearing on this? Greater cross sectional area would suggest lower resistance in a conductor, anyway.