High current linear PSU capacitor requirements

Question

Electrolytic caps have rated max ripple current (usually about 2A or so, even the "low ESR" ones). So, if I want to build a high current power supply (say 100A at 12V), I should use a lot of capacitors (more than 50 caps 2A each) in parallel so as not to overload the capacitors. Is this correct?

Does this only apply to the mains frequency power supplies? I have not seen a PC power supply that uses enough caps to satisfy the current requirement.

I want to build a simple unregulated power supply - just a transformer, bridge rectifier and smoothing caps if it turns out that this is cheaper than buying a (good) ATX or similar PSU.

Answer 1

To answer in your question about the “PC power supply”. Briefly the switching mode ps does not have a good transient response in load changes and also they are generates lot of EMI. So you can not use in ,lets say, inductance dominated loads or in high end audio applications. It is common practice to banking (parallelize) capacitors, since this allows to sum their capacity as well as reduce their series resistance. So we can handle bigger currents with less dissipation and heating, since ripple current per microfarad is often increased by reducing capacitor sizes. Regarding max ripple current, do not forget to include the "ripple current multiplier" given at the operating temperature, an d also keep in mind that temperature of the capacitor core is hotter 3~5 oC /watt of ripple power compared with the case temperature. Usual heat dissipation factor is 0,00093W/oC/cm^2. So in a 4700μF/35V capacitor, a 10oC temperature rise limits the current to 2A In case of capacitors banking, main task is to keep them in low temperature (i.e do not place close to high heat generated or dissipated components, or use forced cooling methods) Bonding each cap even with hot glue to a cooler base helps allot. Also protect capacitors from high frequencies (i.e using of very fast and exceptionally soft recovery power diodes, or RC snubers) since elcap does not respond well in hf and ESR decrease more than twice at resonance. The type of capacitors that you have (temperature rating as well as geometrical size) plays important role to the dissipation factor (or tanδ) stated in data sheets (there is also a temperature and frequency derating). For banking we often use large diameter and short length capacitors. My experience limits to 20x1000μF/35V bank, but your is 50x4700μF.... Assume that your capacitors are uniform, or if you mix them keep in mind that each capacitor have a different discharging time. As already stated above, rectifier bridge plus heatsink sizing should carefully design because on this VA rates, the equivalent series resistance of transformer it is very low and now comparable to the total capacitor ESR. So it is very possible to need an external power resistor.

Regarding the comparison cost of the two: do you have price of a good torroidal transformer 2500VA that you need?

Answer 2

Electrolytic caps have rated max ripple current (usually about 2A or so, even the "low ESR" ones).

I don't think so. Take a look at these: -

On standard AC supplies using a bridge rectifier, the ripple current can be higher than what is implied by the load current. Consider that the cap discharges (due to load) over nearly half a cycle then the energy is returned back to the cap as the incoming voltage peaks near maximum. The time period of the re-charge can be as low as a couple of milliseconds and it may have to push 10 x load current into the caps for this limited period. This is also why you'll probably have to rate your bridge diodes as 1000A peak.

But please do a simulation to see what happens. This is important as your power supply requirement is more than 1kW.

Answer 3

A switching power supply requires significantly less ripple capability (at higher frequency) in its output capacitors because it's typical to keep the ripple current to a small fraction of the total DC output current (i.e. 10A of ripple current for a 100A rail). ATX and other switchers fall into this category, so your statement about not meeting the requirements is incorrect.

Even for a linear supply, the ripple current should be a fraction of the load current else you'll be deeply discharging your capacitors during the mains off-time (i.e. not regulating voltage very well).