[R-390] Power supply capacitors

[2002tii]

Ken wrote:

>Gord Hayward wrote:
>
>>IYou want the smallest capacitor that will give the minimum
>>voltage for the regulator to work.  A bigger capacitor gives a 
>>higher minimum voltage (less ripple) but the extra energy just shows up as
>>heat in the regulator.  Bigger capacity also increases the inrush 
>>current - the diodes only conduct for a short time so the instantaneous
>>transformer current  is a lot larger than the average DC load current.
>
>That's really interesting what you say Gord about the extra energy 
>showing itself as heat in the regulator. That didn't even occur to 
>me. My particular circuit uses a 2200uF and LM317 (among other 
>things) to produce a stable and regulated 12.6 volts for the 
>filaments of the BFO and PTO tubes and the regulator DOES run 
>hot...even with a good sized heatsink it will heat up the heatsink 
>quite nicely. I had a nagging feeling 2200uF was too big....so I'm 
>going to start playing with it until I feel I have a good balance 
>between acceptable performance and long term reliability.

Before everybody goes nuts putting smaller capacitors in their power 
supplies, here is another view.  Don't forget that electronic voltage 
regulators don't eliminate ripple, they just reduce it by the loop 
gain of the regulator.  Reducing the ripple in the raw supply will 
reduce the ripple in the regulated supply, and lower ripple is pretty 
much always better.  If a regulator runs hot, the thing to do is to 
fix the inadequate thermal design, not put more ripple in the raw 
power supply.  I can't imagine designing a low-voltage power supply 
to deliver hundreds of mA with as little as 2200 uF primary 
capacitance, much less thinking about reducing it below that.  As far 
as inrush current goes, if you aren't killing your rectifiers, 
there's no problem -- they are the weak link in the power supply 
chain.  Ground loops shouldn't be a problem if you think about the 
current loops and keep them compact.  RF switching noise won't be a 
problem if you snub the diodes.  And if you're really worried about 
either of these, build a C-R-C "poor man's pi filter."

Best regards,

Don