Smoothing capacitor question

[steve30]

Quick question about smoothing capacitors for a linear power supply.

I figured I will need a smoothing capacitor in the region of 47mF, at a voltage of higher than 40V.

Would I be better off having one large 47mF capacitor, or should I go for several smaller ones in parallel?

Thanks

steve30.

[Simon]

In theory yes, in reality for 50 Hz you'll be fine with one.

[AndyC_772]

Several smaller caps in parallel might give you lower ESR.

[EEMarc]

Capacitors are not just capacitors. They have resistance and inductance usually referred to Equivalent Series Resistance (ESR) and Equivalent Series Inductance (ESL).

Larger capacitors tend to have higher ESR and ESL. One giant capacitor will have a large effect on lower slower frequencies, but will have very limited effect on higher frequencies. Smaller capacitors have lower ESR/ESL but don't have the capacitance to effect the lower frequencies.

Often a set of capacitors of various capacitance are used. Large bulk capacitors handle the slower frequencies such as smoothing the 50 hz. Medium sized capacitors handle the medium frequencies. Then small low ESR/ESL capacitors handle the higher frequency and very small capacitors handle the very high frequencies if needed. Then the effective ESR/ESL of the array is much smaller. The more capacitor values used, the better bypassing you will get over the full frequency spectrum.

[SeanB]

Often using a few smaller units is more efficient space wise, as they can be made to fit in a space, rather than the other way around.

[T4P]

47000uF is kind of big O_o
Attached is 10mF 35V so you get the idea ...

[steve30]

Thanks for the replies .

For now I decided it would be better to go with several small ones in parallel.

Are higher frequencies likely to be a problem in a linear supply?

Also, I specified about 1 Volt pk-pk of ripple voltage when deciding on the capacitor values. Does this sound a reasonable value, or am I being too generous/too big?

[poptones]

1V ripple is huge. When you say 47mF, do you mean 47 MICROfarad? Because that's what it's starting to sound like... 1V ripple, 47mF. If it's 47 uF then that's a small enough value to just use the one - but if you're really designing for a volt of ripple, you probably want to go higher, like 470uF. A volt of ripple out of a high voltage line might be nothing, but a volt on most transistor circuits is going to be like ten percent or more ripple.

It depends tremendously on the application and the exact cap, but generally a 470uF cap bypassed with a .1uF film or ceramic will do just fine. Even a 4700uF would likely be just fine bypassed by a .1uF. Cheap caps, might want to put a 10uF across there as well.

[amspire]

1V ripple is huge. When you say 47mF, do you mean 47 MICROfarad? Because that's what it's starting to sound like... 1V ripple, 47mF. If it's 47 uF then that's a small enough value to just use the one - but if you're really designing for a volt of ripple, you probably want to go higher, like 470uF. A volt of ripple out of a high voltage line might be nothing, but a volt on most transistor circuits is going to be like ten percent or more ripple.

1V ripple is not huge.

It only takes a bit over 5A of  load on a 47,000uF capacitor connected to an 50Hz full wave rectifier to cause about 1V p-p ripple.

[poptones]

But again, we're in the realm of application. How many amps and what's the service?

[steve30]

It is certainly application dependent.

In this case, I was looking at 4A at around 30V.

But this PSU build is more of an exercise in designing and building something useful partially using spare parts that I had lying around. Therefore I am quite flexible with the specification .

I think for now I'll aim for a maximum of just 1 or 2 amps. I probably won't need too much current for this PSU, but I would just like the flexibility to have more if I want.

[T4P]

4700uF is enough for 4amps ...
47000uF if i read that correctly is overkill

[Simon]

It would help if you explain exactly what your trying to build. If it's a switch mode supply you want to look for low esr capacitors and if the value is large break it down into smaller ones.

[steve30]

Its a linear power supply, not switch mode.

Eventually I will build a lab type supply, but at the moment, I'm just playing around with ideas.

I have ordered a couple of 10mF Panasonic capacitors to use for now. That will do for my experimenting at the moment. I can always add more or replace them later.

[T4P]

10,000uF? I might say still overkill might hey, your call man 

[M. András]

last time i checked the high capacity caps higher than 16 volts are damn expensive

[amspire]

4700uF is enough for 4amps ...
47000uF if i read that correctly is overkill

I think you are missing a number of considerations. For a start, most 4700uF caps cannot handle 4A continuous, but even if it can, that is for a new capacitor and they degrade with age.

If you use a 4A rated capacitor in a 4A supply, you are designing a bit of gear that you only expect to last a year or two.  Put a bigger capacitor in and it can easily last 30 years without a problem.

Next issue. 4A from a 4700uF cap means over 8v p-p ripple. A 47000uF capacitor has less then 1V p-p ripple, so you can get an extra 7 volts reliably from the supply.

If you take a 30V 4A supply that uses a 47000uF capacitor, and replace the capacitor with a 4700uF one, you now have a 23 volt power supply.

Another consideration is the supply output ripple. Say your supply regulator has a 80dB ripple rejection from the input. An 8V p-p input ripple would give a 0.8mV output ripple. A 0.8V p-p input ripple gives a 0.08mV output ripple.

Given all that, selecting a large capacitor may easily be worth the extra price.

Richard.

[steve30]

I'm with amspire on this.

It will depend on how much ripple you can tolerate. I tested this on a linear supply which is on my bench right now and that has rather small caps. Its ripple is huge, but the voltage on the caps in still MUCH higher than the output from the regulator, so in this case, it doesn't matter.

But in my case, I want to take as much advantage of the transformer output voltage as possible.

[AndyC_772]

I can definitely feel a simulation coming on - there's no easier way to see what the ripple current and voltage are for different values of capacitor. There's really no need for all the guesswork when we have LTSpice for free.

[Rerouter]

then you start feeding in your parasitics for your massive caps and you reach a whole new world of pain,

[T4P]

Maybe i was thinking otherwise, but my 4700uF figure was going to be not so continous usage and as the filtering cap for a regulator

[BravoV]

4700uF is enough for 4amps ...
47000uF if i read that correctly is overkill

I think you are missing a number of considerations. For a start, most 4700uF caps cannot handle 4A continuous, but even if it can, that is for a new capacitor and they degrade with age.

The notorious crap poster strikes again ... 

[amspire]

I can definitely feel a simulation coming on - there's no easier way to see what the ripple current and voltage are for different values of capacitor. There's really no need for all the guesswork when we have LTSpice for free.

There is no way to get the exact ripple without testing as it depends on the transformer behaviour, and you will never get enough specs on a transformer for an accurate LTSpice simulation.

But you can easily calculate the worse case ripple without any computer.

The theoretical worse case peak to peak ripple from a fullwave recitfier / capacitor combination is just

V_pp =  I / (2 * Mains_Frequency * C).

Typically, the ripple may be smaller by perhaps 25% due to effective series resistance fromt he transformer. That is close enough for an adequate estimate. A worse case figure is always great to use when you are designing, as it is impossible for the ripple to be higher then the worse case figure.

There is no need or point to using simulation to get more accuracy. It is not as if you are buying 1% accurate electrolytic capacitors.

Richard.

[AndyC_772]

For the 5 mins it would take to run the simulation, I'd personally still bother.

Even if you have to model the transformer as a sine wave oscillator and a resistor, you can still run useful back-to-back comparisons between different caps. Build in some rough estimate of the cap's ESR and you might get a better idea of how much you can tolerate before it starts to dominate output ripple. You can see the surge currents that flow when the input voltage is near its peak, work out what the rms value is, then use that to determine the heating effect in the cap. That gives an idea of how long the circuit might last, and whether the magic smoke is likely to escape from somewhere unexpected once the supply has been on for 5 minutes.

[EEMarc]

One aspect that I forgot to mention is the rated ripple current. This is the AC RMS current through the capacitor. This is a big factor because having too much current will significantly shorten the life of the capacitor. Having multiple capacitors of a fraction of the capacitance often allows for higher ripple current capacity.