[morzh]

Ron,

You've summed it up pretty much.

I will add this: there are reasons we have different types of capacitors, and even amongst the Al Can Lytics of the same capacitance and voltage one will find several types. And this is even if we only consider one reputable manufacturer, let's say Panasonic or Rubicon, or Nichicon, or Chemi-con.

There are general purpose caps. If one wants to filter voltage where no significant current is drawn (we could suggest as an exampe the SG filtering or plate load filtering RC where the R is in 10s of K and the cap is about 10uF or so), a GP cap will do just fine. It works with huge RC timing, the ripple current is pretty much non-existent and at these parameters its ESR, or ESL are immaterial. This is where a $0.20-0.30 GP aluminum cap of an appropriate voltage will do just fine and there is no reason to buy anything more expensive.

Then there is a rectifier filter. This is where there is a lot of ripple, as the cap handles constant quick charge-discharge at a significant rate. This is where one might want a high ripple current rated aluminum cap (yes, a film cap of the same capacitance and voltage will also do but if the space is of importance, Panasonic makes some of the best high ripple caps).

There are also Low ESR caps. Low ESR naturally goes with high ripple rating, but there are caps optimised for ripple, or for ESR. Often interchangeable, if the reuirements are especially stringent, it is then best to look for the parameter in question.
Low ESR is often needed where low ripple voltage should be provided at high ripple current, such as at a Switching DCDC converter power supply output. The reason is the inductance at the output provides for triangular (sawtooth) shaped ripple current and the ripple voltage is simply Ir*ESR, so the lower the ESR the less the voltage ripple. This is why the low voltage output caps in DCDC supplies are usually exceptionally good with ESR down to 3-4 mOhms, and then parallelled if even better result is desired. So this is not the capacitance on its own that provides for the low ripple output but the low ESR.

Now, when speaking of Voltage rating vs Capacitance, as Ron has explained, the capacitor that directly filters a tube rectifier's output is limited in capacitance by the tube's specifications. Often this is below 40uF.
Also, increasing the capacitance of this particular cap even withit specs against the original value of the schematic will increase the output voltage making the tubes run hotter. Which we do not want.
To add to this, larger capacitance dictates the size increase.
So does higher voltage.

Higher voltage is preferable as it increases reliability. Also for the same type capacitor larger voltage corelates with better ripple current rating.
But due to the size increase it is not always possible to increase the voltage.

What Ron said I also confirm: 22uF in place of 20uF is just fine. One reason being, this is within the tolerance of 20uF caps and the schematic is created to tolerate the cap with +/-25% capacitance as those were (and still often are) the factory specs for most caps.
And the higher voltage....well, I just had addressed that.

In the end, the electrolytics are by design a limited lifetime component, and if the size and space allow, it is good to replace them with good film capacitors as they have comparable or better ripple ratings and are practically immortal and even self-healing. Unfortunately, this is 1) not always possible (especially if one rebuilds a multiple section electrolytic having to stuff more than one cap inside the can) and 2) fairly expensive, a good film cap of a few microfarads capacitance going for a few dollars apiece with the electrolytics, even high ripple ones, being under a buck.

Well, pretty much it.