Electrolytic Capacitors in Parallel and why?

[Laurie]

G'day people, I've searched around a bit on the net for answers regarding why Electrolytic caps are commonly used in parallel....is this a reliability thing?

I am asking this as I am currently working on a Class A Amplifier that requires 2 x 30,000 uF main filter capacitors....I would like to replace them with High ripple current low esr caps...But the original caps are only 80mm tall...

I am considering different options including building a bank of either 4 x 15,000uF or more likely 8 x 10,000uF

working through the circuits, I have also noticed high quality audio grade caps in parallel with general purpose caps...specifically Nichicon (M) in parallel with Nippon chem con NXA series...of which i can find no data sheet..

it has been easy for me to replace 2 x 470uF in parallel with 1 x 1000uf ...measuring the capacitance first which is close enough to an exact tolerance value...

So yes, would be interesting to hear your thoughts...?

[rs20]

When you're designing a circuit, you'll end up with a bunch of requirements for a given capacitor, e.g.:

-- 60,000uF capacitance
-- ESR < a certain value
-- Height < 80mm
-- etc.

You consider single capacitors, you consider pairs of capacitors, you consider groups of 10 capacitors in parallel... and you pick the cheapest permutation. Which may well turn out to be 2x 30,000uF. I can't imagine a reason why the designer would specifically *want* multiple capacitors, but it may turn out to be the cheapest solution (or the only one that exists, e.g. in the case of ultra-low ESR requirements), that's all. This is purely a sourcing/pricing issue stemming from real capacitors costing money and being big.

[T3sl4co1l]

Well, for a class A amplifier... there are probably miles of better options... like class D for one...

Most of those designs are written by people who don't really know electronics very well, but they do know superstition very well.  So they do illogical things like use exorbitantly oversized capacitors for filtering.  Which, besides being expensive and useless, compared to a proper switching supply -- they also tend to make things worse by drawing large peak currents, reducing input power factor.

In general terms, there are good reasons for doing this.  Some examples: ESR, ripple capacity, aspect ratio / physical shape, redundancy, etc.  Doing it, just to get the extreme capacitance value, is not a good reason.

Electrolytic capacitors have broadly similar ratings, so you don't gain much in electrical terms (the R*C product is roughly constant -- smaller caps have higher ESR), but you still get some help with physical inevitabilities like surface area.  Smaller caps (more in parallel) means more surface area, which means more power dissipation for a given value (capacitance and voltage rating), which means more ripple capacity for the same maximum (internal) temperature.

Obviously, in a project where you simply can't fit a 30mm tall capacitor, you must use wider (or more) 20mm tall caps.  Or even less.  Laptops and other mobile devices often have board requirements such that no component can be over 5mm high.  They use a lot of ceramic, tantalum and polymer type (chip package) capacitors in those applications.

Electrolytics generally have a predictable failure mechanism (i.e., too long at too high a temperature and the ESR runs away), so redundancy isn't really a thing, but it may be relevant for other types.  Ceramics can be connected in series in case one fails shorted (usually for safety purposes); film caps are available with similar internal construction (double dielectric layers, or floating internal electrodes) for improved ratings as well.

Tim

[EEVblog]

I can't imagine a reason why the designer would specifically *want* multiple capacitors

It does add some redundancy and spreads the heat (from the ESR) around, in theory increasing the overall product MTBF.

[Laurie]

Appreciate the replies...and yes it seems obvious to me, but there is most definitely many circuits with two or more caps in parallel where a single cap could be used and surely a cheaper option?...yes design of universal PCP's is common....but there are so many common values used in parallel for filtering...surely the cost savers would produce a common appropriate value? 

[SteveyG]

Well, for a class A amplifier... there are probably miles of better options... like class D for one...

I'm yet to see a class D amp that has THD specs even closely approaching that of a class A amp. The headline figure is usually close, but at some corner case. Class D amps are great for subwoofers and lower fidelity applications though.

[Kjelt]

I'm yet to see a class D amp that has THD specs even closely approaching that of a class A amp. The headline figure is usually close, but at some corner case. Class D amps are great for subwoofers and lower fidelity applications though.

Thats what they said about SMPS for audio applications ten years ago, today it suddenly is no longer a problem.
Look at companies like Chord (SMPS since ten years, top hifi brand but $$$$ ) and for instance Hypex for cheaper but really good class D amps, it is a matter of time.
But who knows 

[SteveyG]

I'm yet to see a class D amp that has THD specs even closely approaching that of a class A amp. The headline figure is usually close, but at some corner case. Class D amps are great for subwoofers and lower fidelity applications though.

Thats what they said about SMPS for audio applications ten years ago, today it suddenly is no longer a problem.
Look at companies like Chord (SMPS since ten years, top hifi brand but $$$$ ) and for instance Hypex for cheaper but really good class D amps, it is a matter of time.

I didn't say they weren't good, but they don't match the distortion specs of a class A amp. I have various class D amplifiers purchased or built and they're great for high power density designs, but usually the distortion is worse at average listening volumes. Now if someone came up with a resonant class D amp, we could be onto a winner.

[daqq]

Appreciate the replies...and yes it seems obvious to me, but there is most definitely many circuits with two or more caps in parallel where a single cap could be used and surely a cheaper option?...yes design of universal PCP's is common....but there are so many common values used in parallel for filtering...surely the cost savers would produce a common appropriate value?

Well, cheaper, yes, but how cheaper? Unless you're at some extreme requirement (space, cost) you can afford to have some extra capacity. Per rule of thumb more filtering/decoupling is better. If I need 100 uF and can get sligthly better results with 2x 100 uF, have some extra board space, 5c extra won't harm my budget that much. With the added benefit of distributing the load, higher reliability and all of the good stuff.

[SeanB]

Multiple parallel capacitors often work out cheaper if you need eg more than 4700uF, where most larger sizes tend to come in chassis mount variants with screw terminals, while the 2200uF is in a can that is soldered to the board. The screw mounted ones need a chassis clamp, screws and then wire with lugs and PCB mountings on it, while the solder in can just is placed in the board and soldered with the other components.30 seconds of placement as opposed to 30 minutes of labour along with the extra parts and tooling needed.

[PeterFW]

I didn't say they weren't good, but they don't match the distortion specs of a class A amp. I have various class D amplifiers purchased or built and they're great for high power density designs, but usually the distortion is worse at average listening volumes.

True A/B/X comparison with someone who dit the setup for you?
If not those comparisons are strongly biased, like allmost every time someone can hear a difference between amplifiers.

Please tell me otherwise if my information is out of date but as far as i know there are no people out there who can hear the difference between decent amplifiers, no matter what topology as long as they have a flat frequency response.

"X sounds better then Y" is mostly due to non linearity and user bias towards that.

[SeanB]

Class whatever, most of the time past a certain point you cannot hear a difference. The speakers are a very important part as well, I remember reading from JLH that a $500 speaker box might only have a $50 spend on all the speakers, cables and crossovers in it. With me I know it is hard to tell the difference between the original WAV file and the 256k MP3, or the 160k OGG. Thus the recordings I make are either at 320k MP3 or 256k OGG, or sometimes the 700Mb WAV file as is.

[Laurie]

Appreciate the replies...and yes it seems obvious to me, but there is most definitely many circuits with two or more caps in parallel where a single cap could be used and surely a cheaper option?...yes design of universal PCP's is common....but there are so many common values used in parallel for filtering...surely the cost savers would produce a common appropriate value?

Well i guess that pretty well much sums it up then....

thankyou.

[free_electron]

Paralleling 'real' capacitors ( not the theoretical crap ) has multiple effects. sometimes these effects can be desirable, sometimes not wanted.

- ESR comes in parallel yielding an overall lower esr
- the ripple current is spread across multiple caps so each cap will get a longer life.
- the current now comes from different sources as opposed to from one single trace. current flows in a straight line point to point. It doesn't matter how wide you make a trace , the electrons will run in a 'channel' as wide as needed. making the trace wider than the channel needed gives you nothing. having multiple caps , and placing them correctly  so the load sees them in parallel and not cascaded , gives some improvement, especially when dealing with peak currents or high crest factor currents.
- the single electrode wire of a single cap handles all the current : now the current is shared. so you drop the losses in the interconnects. P = i2r. drop I into half ( by doubling up to 2 caps ) while maintaining R ( the pin size and solder joint is the same ) and you have instant gain !  for example for a 5 amp current. and 0.1 ohm interconnect resistance  P would be 5x5x0.1 = 2.5 watts.

go to two caps , each half current 2.5 x 2.5 * 0.1 = 0.625 watt times two =  1.25 watt. you just gained 1.25 watt ! or, you shedded 1.25 watt lost in heat !

when dealing wiht pulsed currents you can stack tall caps.

[Laurie]

^ that's the kind of answer i was looking for,  thankyou

[f5r5e5d]

...
I'm yet to see a class D amp that has THD specs even closely approaching that of a class A amp. The headline figure is usually close, but at some corner case. Class D amps are great for subwoofers and lower fidelity applications though.

http://www.hypex.nl/docs/nc1200%20folder%20web.pdf I bet you can find worse Class A Audiophile amps

they do use parallel caps in their smps: http://www.hypex.nl/product/2012-11-23-13-42-14/smps3k.html

[SteveyG]

I didn't say they weren't good, but they don't match the distortion specs of a class A amp. I have various class D amplifiers purchased or built and they're great for high power density designs, but usually the distortion is worse at average listening volumes.

True A/B/X comparison with someone who dit the setup for you?
If not those comparisons are strongly biased, like allmost every time someone can hear a difference between amplifiers.

Please tell me otherwise if my information is out of date but as far as i know there are no people out there who can hear the difference between decent amplifiers, no matter what topology as long as they have a flat frequency response.

"X sounds better then Y" is mostly due to non linearity and user bias towards that.

This is my own testing at home, using various modules including the Hypex ones, my own projects and various hi-fi amplifiers which I seem to collect. I normally set up an automated switch matrix for testing my own designs against my Rotel amp.
This is not really the right topic to discuss though.

[Galenbo]

Paralleling 'real' capacitors ( not the theoretical crap ) has multiple effects. sometimes these effects can be desirable, sometimes not wanted...

There is another interesting aspect you mentioned in another tread.
Putting them so that one is turned 90°, in automotive applications.

[JuKu]

Also: In frequency domain, different caps are very different. I've often ended up using 10u, 220n an 47p in parallel (or similar). On first look it might seem pointless; with a spectrum analyzer, not at all.

[KD0CAC John]

Until the last 10 yrs. most of my electrical was in automotive , so I thought along the lines of FREELELECTRON's answer .
Much like multiple batteries , in series more volts , in parallel more current ?
Fitting into limited spaces adds more to combining , but was not implied in the question , but good to keep in mind .