Read comments below...
GeorgeB wrote in message
. ..
Your concept is reasonable, but there are some significant problems.
The forward voltage varies with temperature,

Indeed it does, diodes are often used as the sense element for
temperature sensing, but it's a very small change, and you need to
amplify it when you do. Also, the self heating of the device will
swamp ambient temp effects I should guess.

and the power supply
LIKELY varies over time. As you have it, if the fwd voltage dropped
0.1v, and the supply did not change, you would have 0.7 vs 0.3 across
the resistor, for over 2x the current ... maybe a real problem.

Thats what makes it a self regulating circuit, as the current comes
up, the foward drop of the diodes go up as well, thus reducing the
drop across the resistor.

Now, let's have the alternator charging the battery, and have it at
15.5V or so ... now I have 3.8 volts across that current determining
resistor ... 12 times the "design". OUCH.

Same thing as before...

Now let's have the alternator fail, the battery voltage drop to 10.5V.
Your series string will draw no current and give no light ... and you
are in an emerency situation that is exactly when someone needs to see
you.

You can handle this case by dropping one diode off the string and
recalculating the resistor as before. Cuts your efficiency a little
but hey, some poeple drop over half the power delivered as heat into
their "current limiting device" for 28V applications.

What is the solution ...

There are "constant current" devices. I have used them, and they
work allowing operating this string with probably 3 LEDs over the
range at visually constant brightness.

You can do that, but for driving LEDs, since they form a nice self
regulating circuit with a single resistor I didn't feel it was
neccesary. Please share with the group which part you've had success
with as a "constant current" device.

You can design a pulse system turning the LED on for (maybe) 0.1ms
then off for maybe 5ms and PROBALBY not overdrive (into damage) the
LED and put "as many" as you want in parallel. The driver will likely
be a FET.

You could do this also but each LED would need its own current
limiting resistor in series because the forward drop of the LEDs vary
from part to part and with temperature as you've mentioned and the one
with the lowest drop would eat the most power without those resistors.
But again, with a pulse width modulation circuit, why so complex?

I was taught to allow about half the voltage for the resistor, half
for the LED string unless I had current control. In the "old days",
for current control we used an emitter resistor in a common emitter
circuit, 2 or 3 diodes to set bias (single vs darlington), and the
LEDs between collector and V+. There are other (better) ways, but
everyone understood this one.

So you're building a constant current supply from each group of 2 or 3
LEDS, thats pretty complex if a single resistor will work. What
you're suggesting is too complex for the average guy and I see no
practical benefit. Don't light bulbs vary in brightness with supply
voltage? Sure they do, and they vary more than the single resistor
method I've sketched out.

If you hook up your entire string backwards, no
harm will be done, but if you happen to solder one LED backwards, it
will likely be toasted on power up.

I disagree that there will be damage with any in backwards. The
reverse voltage will almost certainly be higher than the forward
voltage, so there won't be any current drawn. If there is, you still
would have less than correctly wired.

The reverse drop on the LEDS will be the supply divided by the number
of diodes. 12/4=3V. Last data sheet I looked at said the reverse
voltage limit was 5V. Thats why I also said that if you put one
backwards it will cook. It will see the full 12V.

What I've outlined is a simple method to build LEDs lights. Yes, you
could build a constant current supply, and the LEDS would see the
exact same current from 10V to 15V but your light bulbs will vary in
brightness (acnd color) over that range anyway more that my suggest
circuit due to the self limiting nature of a diode(s) in series with a
resistor.