On Sun, 17 Jan 2010 22:55:31 -0500, Charlie Smith
wrote:

flipper wrote:
On Sun, 17 Jan 2010 11:43:13 -0800, John Larkin
wrote:

On Sun, 17 Jan 2010 11:22:23 -0800, Archimedes' Lever
wrote:

On Sun, 17 Jan 2010 10:23:00 -0800, John Larkin
wrote:

On 17 Jan 2010 16:54:58 GMT, mick wrote:

On Sun, 17 Jan 2010 08:35:12 -0800, John Larkin wrote:

On 17 Jan 2010 16:01:35 GMT, mick wrote:

On Fri, 15 Jan 2010 23:02:08 -0800, John Larkin wrote:

snip
There's a pot on the linear regulator; it's in plain sight. And that's
all this thing needs. The PWM accomplishes nothing.


PWM dimming is more efficient than linear dimming - by a long way. It
all depends on how much heat the OP is willing to let the dimmer
dissipate.
PWM is no more efficient than resistive dimming the way he did it. There
are no inductors in his circuit. All this sort of PWM can do is move the
heat around.


I couldn't see his circuit on my server, but I assumed that he was
varying the mark/space ratio. In that case power dissipation in the
output device is always low, depending on Vsat.
That just moves the heat into the series resistors. Overall efficiency
is always the same for a dissipative (inductor-free) regulator.

With a linear regulator and properly chosen series resistors, you can
balance the regulator heat distribution versus dimming level.
Regulator power dissipation versus output voltage is sort of
parabolic... low at low illumination, low at max illumination, peaking
somewhere between.

It's easier to heatsink a voltage regulator than a lot of small
resistors.

Another advantage of a voltage regulator is that it regulates.

John

You take a few of these and drive four sets of LED pairs per chip.

That's $0.50 per LED for constant current, precise control.

Might as well go all out. Beter than any use specific "LED driver" chip
out there.

http://www.edn.com/article/CA6702709.html

http://a330.g.akamai.net/7/330/2540/...6702709XLG.jpg
Yikes! It's horrendously complex, needs an I2C interface to set
brightness, and is constant-voltage, not constant current. And it will
generate lots of EMI, bad news in an airplane.

All he needs is an LM317, a pot, and a few more passives.

John


After looking at all the suggestions made I tend to agree. Put some
load dump protection on the front end and be done with it.

The only difference is I tend to favor distributing heat to the limit
resistors but it works just fine either way.


Sorry to see the conversation has degenerated to name calling. Its the
primary reason I dislike usenet and I really wish you wouldn't.

Thanks for the suggestions but I think we need to refocus a bit. The
question here is not how to dim the lights. That decision is made; it
will be with PWM as I have yet to come across a single reference that
would suggest that dimming more than one LED by changing voltage is good
practice. They are constant current devices and must be treated that
way. And I mean by dimming to suggest variable dimming as you would
have with a normal car instrument panel lights. Not dimmed to one
constant, less than full-on level.

Your PWM thing will work, although it doesn't provide any regulation
against 14-volt changes, and it's more complex than necessary.

An LM317 or LM1117 regulator, with a pot, could generate a regulated
voltage from, say, 4 to 10. That could drive a number of strings, each
one resistor and two LEDs. It would regulate and continuously dim, and
would be simple and reliable.

John