[morzh]

Well, somehow (maybe because I am mostly solid state type of guy) while dealing with old radios I never asked myself that question: should I look closer to the ripple current rating of the caps we put to filter the full-wave rectifier?

After all tube radios even at huge power consumption overall draw very little from their rectifiers: tens of milliamps typically, with excursion into a couple of hundreds of those sometimes and even less frequent to more than that.
This is a little fraction of what I typically deal with. Transistors and solid state being low voltage devices will draw proportionally higher currents given the same power: in a simple Hi-Fi transistor amp one deals with several amps of load current at least. And in the switchers that I design mere 40 Watts make you handle several amps of ripple.

Anyway, I decided to take a look at this.
Usually when looking at the photos of people rebuilding the old aluminum caps or reading the description, or looking at the "kits" sold at DIY sites / eBay I mostly see General Purpose capacitors. They are used as long as microfarads/volts fit those of the capacitor being replaced.

Back in the days of yore when radios were, well, radios, the aluminum caps were simply aluminum caps. One did not have today's choice of "GP"/"low ESR"/"High ripple current"/"low leakage/"High Frequency" electrolytic caps.
Then again the caps were large: a then 10uF 450V cap was tens of times the size of today's one. This amongst other things accounted for better cooling.
------------------------------------------------------------------------
Let's take a hypothetical rectifier built around a tube. For simplicity let's make the load current 100mA. Let's make the peak rectified voltage 400V.
The LC filter typically employed by most radios will smooth the current and the voltage ripples making the radio more or less constant voltage / constant current load. But the very output cap sustains little ripple compared to the cap across the rectifier, so the latter is the one of interest.

So, first let's see what the voltage ripple is.
This one is dictated by the expression
Vr=Il/(2F*C)
Where the Vr - ripple voltage, Il - load current, F is the line frequency and C is our capacitance.
Let's pick a typical capacitance of 2x10uF which usually replaces that of 2x8uF.

So, our ripple voltage is:
0.1A/(120*0.00002)=83V.
Roughly sketching this it becomes visible that the discharge time equals approximately 3 times charging time. Which means that the average charging current that has to replenish the spent charge is 3 times that of the load (discharging) current.
So we have total charge of 3*Iload + I charge, or 6*Iload, and the average is 6*Iload/4=1.5*Iload.

So, our ripple is 150mA at the 100mA load.

Now let's look at the caps:

First the General Purpose:
A 10uF 450V cap by Nichicon, UVZ2W100MHD, has 65mA ripple rating.
Two of those in parallel will provide for 130mA.
A Panasonic ECA-2WHG100 has 75mA (two of them respectively 150mA.
So while having met the voltage/capacitance we are not meeting the ripple rating. This will at some point result in the premature failure of those caps.

But say a Panasonic EEU-EE2W100 cap will have 172 mA ripple current at 120HZ (the rating of 490mA is deceiving - one has to realize that the 120Hz ripple carries with it 0.35 multiplier), so in quantity of two that will give us nice enough 344mA ripple capacity.

-----------

A neat trick to go beyond these limitations is this: instead of 10uF caps let's use 22uF caps of the same or slightly lower voltage.
Why?
The same cap type usually shows sizeable increase in ripple current rating with increased capacitance.

So, the same type 22uF cap EEU-EE2W220S will have 331mA ripple rating at 120 Hz. At $0.96 a pop.
Better yet, EEU-EB2G220S has 410mA. It is at $1.05 a pop.
And best of all EEU-EB2W220(S) has 560mA. This is the one I bought. At $1.34 a pop.

Now let's serialize the two we will have 11uF cap with the same 410mA rating. Stuff them in one can.
Do the same for the second can.
Now once paralleled, we will have whopping 820mA ripple rating.
Same done to 560mA cap will give us 1120mA.

This will get us out of the danger zone with a very safe margin.

---------------

What kills aluminum caps? Pretty much (with the obvious exception of the voltage breakdown) one thing: temperature. The temperature in turn is usually the result of the ambient (which in tube radios is quite high) plus the heat due to the ripple current against the ESR.
At 20 degrees C (room temp) most caps will last from 2000 to 10000 hrs at full ripple current. Which is not that much.

Choosing 105C rated cap with a good margin of the ripple current will ensure long life.