Power Supply Issues

[Don Tillman]

   From: media at mail1.nai.net
   Date: Mon, 23 Mar 1998 22:12:50 -0500 (EST)

   With the extra cap a 317 has a ripple rejection of 80db.  Although I
   remember such a circuit, I do not know if ripple rejection can be improved
   by adding cap to the ground pin of a fixed regulator.

Only if you're using the fixed regulator as an adjustable device (ie.,
placing a voltage divider between the output and the ground pin).

   "... The PSU's  built out of a 723 with a big bypass tranny can be
   made so much more accurate it is foolish to use anything else."

Context...  The main useful feature of the 723 is that the big
transistor is physically seperated from the chip, so that as the big
transistor does it's job and heats up, the more delicate regulation
circuitry (the voltage reference and all) won't be affected.  

The single-chip regulators have temperature compensation circuitry
built in, sure, but the 723 approach removes the problem at the
very start.

   There must be a mathematical method for determining the best value for the
   capacitor.   Anyone have a formula??

Choose the cap for some specific amount of ripple under full load.
That depends on the application of course, but if you choose, say,
0.1v ripple, you could calculate the capacitance as:
   C = Iload / (0.1 V ripple * 120 Hz)

   Several people have mentioned using a transformer with a secondary
   rated for an RMS value lower than required input of the regulator.  On the
   one hand we have brownouts at the mains and diode drops across the
   rectifier, otoh, there is the peak voltage of the AC and the charge of the
   capacitor.  Using too high an AC input into the regulator is going to cause
   excess heat, to low a value will cause insufficient output.   Anyone have
   formulas for this??

I should have said something nasty to the guy who suggested that.  He
was really proposing a power supply that could just be strung together
with spit and duct tape to breadboard a circuit.  As opposed to a
reliable unit you can depend upon working flawlessly when you're
playing the instrument in front of an audience and it's time for your
big solo.

A real engineer designs for the worst case situation.  So you need to
include the worst case regulator drop-out voltage, cap ripple voltage
drop, diode voltage drop, compensate for a 10% transformer voltage
inaccuracy, brown out voltage drop, etc.  If there's a little extra
heat under less-than-extreme conditions, so be it; that's the price
you pay.

   If one were to desire a PSU better than a linear 7815/7915 supply,
   are there any reasons to build one yourself (besides the fact that you're
   on a diy list :) instead of buying one from say, Condor??

Low power situations, weird voltage requirements, that sort of thing.

  -- Don