Michael Rigby-Jones wrote:

> Performance wise, the LM317 current source isn't great,

  No, it certainly isn't.  Let's see, did we propose 350 mA charging
current for the caps?  At that current, the dropout voltage is about
1.8V at ambient temperature (but perhaps slightly less if the chip is
HOT) plus the reference voltage is 1.2V, so charging stops at 9V instead
of 12V, does it not?

> however the reason why I suggested it was firstly the minimal
> component count, i.e. one resistor and the IC, and the fact that they
> are pretty rugged devices with over temperature protection etc.
>  (although I wonder if that works OK in constant current mode).

  I should imagine it does.

>  Absolute accuracy and stability in this application aren't really
> important, as long as the caps can be charged within a certain time,
> then it should be fine.

  Well, the caps will be charged to 9V instead of 12.  Is that OK?

  The one-transistor current source has a drop-out of 1.3V, and the two-
transistor version, 0.8V, which makes it my all-time favourite.

  If however you are going to "gang" them, the one-transistor version
can use one resistor and the diodes in common (clever: Use a LED
instead) whilst you need two transistors for *each* source in the
two-transistor version.  Mind you, at least one transistor of those can
be a teeny SMD device.
--
  Cheers,
        Paul B.