some comments on linear power supplies: was : RE: [sdiy] Power Supply Design Questions

ASSI Stromeko at compuserve.de
Thu Mar 27 23:34:40 CET 2003


Some more comments...

To the question of how the load capacitor should be dimensioned: simply 
design for ripple voltage (Upeak-Ulow) at maximum load current and cut 
a few corners on the way. The ripple frequency is known (mains 
frequency for half and double that for full wave rectification) and for 
simplicity assume that the cap supplies the current for the full 
period. Then chose C=Imax*T/Uripple and select the next standard cap 
size. The conduction angle is approx. the arc cos of the normalized 
droop (Ulow/Upeak) for half wave rectification and twice that for full 
wave rectification. You normalize to 2pi or pi and take the inverse and 
that tells you the factor with which to multiply the load current to 
get the inrush current for the cap. Now check the specs of your 
capacitor, rectifier and anything else upstream carefully to see if 
they're designed for that.

So, for 1A load current and a full wave rectifier, 10V peak and 10% 
ripple we get a cap of  1000µF, which sounds about right. The inrush 
current to the cap is around 7A. For ripples of 1% and 0.1% we arrive 
at 10mF/22A and 100mF/70A respectively. Now you know why these caps get 
hot and why severely oversizing them is a bad idea, as is designing for 
exceedingly low ripple. Adding chokes to the load path limits the peak 
current by extending the conduction angle and makes for a more quickly 
decaying spectrum of the ripple hum, but their use is limited by 
various factors in practice. If the hum is the problem, then shunting 
the known frequencies after the cap is a more effective solution.

Having said all that, I really think a primary switched power supply 
with PFC and slow start as the raw DC source plus point of use 
regulation is a better solution for large currents. If you have an 
assortment of wall-warts and individually powered gear, then the best 
solution to slow start is power sequencing, i.e. never switch more than 
one at the same time. I've got some distribution sockets from Conrad 
with a switch on each outlet and that works fine for me. There are more 
fancy things that sequence slaves after a master has been switched on 
or even *gasp* switch power outlets on demand from MIDI.


Achim.
-- +<[ Q+ & Matrix-12 & WAVE#46 & microQkb Omega sonic heaven ]>+ --

Q series MIDI Implementation & additional documentation:
http://homepages.compuserve.de/Stromeko#WaldorfQDocs



More information about the Synth-diy mailing list