I'm wondering why are you using an external 4Mhz crystal instead of
16LF628's internal oscillator? Or the LF version doesn't have one? That
could save some power and reduce circuit complexity.

-----Original Message-----
From: pic microcontroller discussion list
[mailto:PICLIST@MITVMA.MIT.EDU] On Behalf Of Olin Lathrop
Sent: Saturday, September 07, 2002 7:37 AM
To: PICLIST@MITVMA.MIT.EDU
Subject: Re: [PIC:] [EE:] Driving a 15W LED?

>         Any thoughts on using the PIC (which will also be in the
circuit)
for
> regulating the LED current?

Yup, done that.  See the schematic at
http://www.embedinc.com/pic/light.pdf.
This was something I whipped up quickly the day before I was heading off
to
the woods of Maine for a week, so the design was limited to parts on
hand.

The purpose of this circuit is to run a constant 20mA thru the string of
4
LEDs D2 - D5.  These are bright white, and need a bit over 12V total.
The
battery is a rechargeable NiMH pack with a useful voltage range from 4V
down
to about 2.5V.  The PIC is a 16LF628 because that was the smallest L
part I
had around.

The basic algorithm is to do a single fixed pulse if the current is too
low,
the voltage not too high, and the previous pulse is done.  The port B
passive pullups are enabled.  The B-E voltage of Q2 is used as a cheap
reference with R4 the current sense resistor.  Q2 pulls RB1 low if the
LED
current is above the regulation threshold.  Q5 works similarly except
that
it senses the LED power voltage.  The purpose of this is to keep the
circuit
from destoying itself in case there is ever a break in the LED current
path.
Q5 triggers at about 15V, which is above the 13V or so needed to run the
LEDs, but low enough to keep things from getting fried.  Therefore RB1
high
means another pulse is desired.

Q3 turns off during the flyback phase of a pulse, allowing RB0 to go
high.
RB0 is set to catch rising edges, so the INTF bit gets set when the
previous
pulse is over.  However, this mechanism only works when the LED voltage
is
700mV or more higher than the battery voltage, so it is not reliable
during
startup.  The firmware therefore also uses a timeout to decide that the
previous pulse is over, which should only kick in during initial power
up.

When all the conditions are met, a pulse is started by raising RB7 for a
fixed time.  This is directly tied to the gate of Q1.  These are amazing
little FETs that have only 23mOhm on resistance with 2.8V gate drive,
even
better with more.  Inductor L1 charges while Q1 is on, and dumps its
energy
onto the LED supply thru D1 when Q1 is turned off.  The firmware then
waits
for INTF to get set indicating the pulse is over, then waits for RB1 to
be
high to do another pulse.

Efficiency was only 74%, but there are a number of improvements I didn't
have time to make or didn't have parts in stock.  I think in this case I
made the Q1 pulses too long, thereby increasing I*R losses in the
inductor
and increasing the losses in D1.

The thing worked great, see the attached picture.  I built the circuit
one-off on a piece of perfboard.  Unfortunately you can't see the
circuit
itself because it's all wrapped up in electrical tape.  I also used the
tape
to make a pocket for the battery.  That's the green thing with the wire
coming from it.  I epoxied the LEDs to the perfboard you see and rigged
up a
mount with two paper clips to clip it to the visor of a hat.  A used
margarine container gave its life for the white bracket.  For now the
cord
runs thru another paper clip behind the right ear and then down into the
shirt to a pocket somewhere.  The circuit is small and light enough that
I
think it could be mounted on the back of the hat somehow.  I'd like to
make
a few real circuit boards, but haven't found (haven't looke real hard
either
yet) a nice way of holding the battery in place.  Anyway, it seemed to
work
at least as well as any other hiker's headlamp.  I had no problem
walking
around the woods with this thing at night.  One fully charged battery
pack
lasts for about 4 hours.  These things are small and light (about 3
AAA), so
it's easy to carry several of them.

I've been thinking of similar circuits where a little more accuracy is
needed.  I haven't done this yet, but I'm thinking of using a 12F629 and
using its built in comparator and an external reference (LM385-1.2,
1.23V @
20uA).

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listserv@mitvma.mit.edu with SET PICList DIGEST in the body