<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:</font><HR>Originally posted by jeff1500:
Very nice. I'll certainly post results if I build it.
How come four diodes all in a row for D2 to D5?
What do you think about this? It's a zener diode voltage regulator. Just by luck of the search engine, it looks like it's from Satcure.
http://www.netcentral.co.uk/satcure/design/tranex1.htm<HR></BLOCKQUOTE>
Hi again Jeff,
The four diodes in a row are used as a
voltage reference. This is a very
important part of the design because
when the output voltage rises too much,
the diodes conduct more heavily, turning
the transistor Q3 on harder, which serves
to regulate the output current in the long
run. As i noted, the diodes forward
voltage drop changes with temperature,
approx 2mv per degree C. Since Q3's BE
diode also changes that much, the total
change is about 4 diode drop changes
minus 1 diode drop change, or about
3 diode drop changes total, which comes out
to about 2mv*3 = .006 volt per degree C.
This means the average voltage across the
output capacitor (not the led's) changes by
about .006 volts per degree C change in
temperature.
Lets now look at a +20 degree change in temperature.
For that change, the output voltage changes
by .006*20= .120 volts. Since each 10 ohm
resistor drops about 0.22 volts while running
each LED at about 22ma and 3.5 volts, the output
voltage is about 3.5+.22=3.72 volts.
Now changing that voltage by -0.120 volts results in
an output voltage of 3.60 volts. Applying that
output voltage to the series combo of 10 ohm and
one white LED results in the current dropping
to about 18ma, which is about -10% regulation
with a temperature change of 20 degrees C.
Hmmm, this isnt quite as bad as i thought.
Worst case, with the combination of very very
low input voltage of 1.5 volts and a 45 degree
ambient temperature, the regulation would drop
to -20% (or about 16ma), but only at the extreme
of 1.5 volts input and 20 degrees C increase in
temperature.
The thing is, replacing the 4 series diode with
a 2.2 volt zener diode (anode towards Q3's base),
and it looks like the temperature stability of
the circuit improves by a factor of 10. This
could be the answer, instead of playing with
a thermistor. This would make the output
current drop by only about 1ma with a 20 degree
C change in temperature.
I'd have to check this out though,
because those very low voltage zener diode vi
curves arent very sharp. It would need actual
real life testing for temperature stability.
I checked out the zener regulator circuit, and
that may or may not be a circuit you would
want to use for LED drivers running off of
batteries. It will act as a constant current
regulator, but at a cost of low efficiency.
For example, if the input voltage is 6 volts
and the LED voltage is 3.5 volts at 22ma, then
the power in is 6*.022=.142 watts, and the
power out is 3.5*.022=.077, and the eff is
Pout/Pin=0.58, only 58%. This means for
every 100 batteries you buy for the light,
about 42 are wasted, and only 58 go to
powering the light. It is a solution
just the same

If your running a few LED's off the car battery,
with the engine turned off the eff drops to about
29%, but who cares in that app. If your running
off of a stand alone battery you may not like this,
because for every 100 batteries you buy, 71 are
wasted and only 29 actually go toward powering the
light. Instead, using a switching regulator
(down converter) you can get at least 85% eff.
I dont know if your following the thread on
the ZXSC300 chip, but in that thread we are
trying to implement this chip for sort of
general purpose LED driving. We hope to
get it to work for a wide range of outputs
and input voltages. You might find this of
interest also. Apparently the chips are getting
easier to get now too, but there are other
chips out there too which deserve consideration.
One of the things i think i realized was that
for driving heavy output currents using a
single cell may not be a good idea anyway, as
the current draw goes WAY up for multiple
LED's (like 10 or so). So perhaps two
separate general solutions are in order anyway:
one for less then maybe 5 LED's, and another
one for more then 5 LED's. Or maybe a few solutions,
depending on how many LED's and what kind of
battery you want to use (except for more then
5 LED's you would have to use at least two cells
in series).
I think i might breadboard that 4 transistor
circuit just to see how well it performs
relative to the simulations

Interestingly, because it uses a real error
integration scheme the individual gains
of the transistors arent as important as they
are in the pure brinkmann circuit. The only
adjustment should be that one resistor!
BTW, all the parts in that circuit are all low
cost, but did it look like it had too many parts
to be practical?
Well thanks for your input on these circuits
and for the other ideas too.
--Al