> After setting up ports etc, the PIC sets up the pin
> that drives the PNP transistor to pulse on and off 
> continuously, say 50%
> duty-cyle for now.  That effectively chops the input 
> voltage in half, with
> the filter cap selected to ensure that the voltage at the 
> input of the LDO
> does not drop below 5V+dropout.  With a 0.5V dropout, 
> that's 5.5V required,
> and 50% duty-cycle means I need at least 11V at the input, 
> which I have.
> With the PNP in front, I should be able to eliminate the 
> input rectifier,
> which is there for reverse-voltage protection.  And at 14V 
> max continuous,
> the LDO would dissipate 0.4W average, and 0.5W peak.  BTW, 
> I'm proposing that
> the PNP would chop at 50% all the time, with no feedback 
> into the PIC to
> regulate it.  The LDO would just dissipate the difference 
> as heat.  Also, the
> duty-cycle could be (and would probably need to be) set to 
> say 55%.

> Here, parts cost is about $1.25 to $1.50 (depending on 
> filter cap required),
> which I can live with.  But will it work?  Or is there 
> something else I need
> to be concerned with?

Yes, alas.
It won't work as expected.

Any system that does not use "energy transformation" must 
instead use energy dissipation :-) :-).

In this case there is an implicit (and very real) resistance 
in the feed to the filter cap via the chopper. You WILL get 
I^2R losses in this resistance. If you make the traces etc 
very solid the chopper switch (FET etc) will form the 
resistance. There is no way of not dissipating the energy 
resistively somewhere.

Energy converters (otherwise SMPS etc) act like an 
electronic gearbox.
The norm is inductive storage (as in a classic buck 
regulator).

An alternative less seen is to use capacitors charged in 
series and discharged in parallel. if the ripple on the caps 
is kept small then the energy losses can be low. Doing this 
with discrete parts usually gets messy and it's normal to 
use an IC based solution. In this case the Vin is > 2 x Vout 
so you may have enough headroom for switch losses etc. At 
lower current you could use eg a CD4066 but this demand is 
in excess of that.

At the 75% efficiency that you've mentioned as acceptable my 
GSR would be a possible solution depending on how precise 
the 5V needed to be.  The "easiest" solution would be the 
most economic commercial buck regulator IC available 
operating at 1MHz or so. What has been mentioned looks 
good - BUT and external switch FET will add little to size 
and will probably be cheaper overall and allow more 
flexibility.

It's easy to miss key points when design requirements are 
conveyed in the informal manner that this sort of 
interchange encourages. (It's an excellent way to explore 
things but sooner or later a hard spec is needed.) Given all 
design requirements I believe that a solution that met specs 
and cost less than has been mentioned so far would be 
possible.


        Russell




 

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist