DIYbanter - View Single Post - LED lamp sellers with 'equivalent' ordinary lamp wattages

Andrew Gabriel

In article ,
writes:
wrote:
Andrew Gabriel wrote:
In article ,
Jim K writes:
On Nov 27, 7:13 pm, (Andrew Gabriel)
wrote:

If you are able to manfacture your own, the component LED parts are
not expensive (at least, nowhere near as expensive as commercial
efficient LED lighting).

what's involved?

[snip long, detailed, description - thanks Andrew]

I want specifically to use them on my boat at 12 volts so I guess I
need to address the issue of providing efficient constant current
supplies from the 12 volt supply.

It *feels* like a cheap, single chip, constant current device should
be possible such that one chip per LED would be the way to go.

Raeding your post again it looks as if I could get two LEDs and the
SMPSU across 12 volts without too much trouble. So I need a way of
producing something like 500mA constant current *efficiently*.

This is the simplest SMPSU you can make.
You drive the LED in series with a MOSFET, a small ferrite inductor,
and a low value resistor to sense the current. At switch-on, the
current increases (limited by rate of change in the inductor).
When you detect the expected value across the sense resistor, you
switch the FET off. The inductor will generate a back-EMF so as to
keep the current flowing, and you use a schottky diode from ground
to the top side of the inductor so the back-EMF continues to generate
a forward current through the LED. This will decay and result in the
current measured in the LED falling below the expected value, and
this causes the FET to switch on again and the cycle repeats.

Left just like this, it will probably operate in the MHz range.
The FET won't disspate power during the on or off periods, but
each transition takes a little time, and power is dissipated during
the transitions. When operating in the MHz range, there are one
hell of a lot of on/off transitions, and this results in significant
power dissipation. Adding hysteresis or RC delay to the feedback can
drop the frequency to a few 10's kHz, i.e. ~100 times fewer on/off
transitions, and then the FET stays cold, dissipating no measurable
power. The only power dissipation will be in the current sense
resistor, which if setup to drop 0.6V at 700mA, is under half a watt.

This is almost certainly what the power driver modules from Rapid
will do. There's also a simple Velleman kit (K8071) which works this
way, except it doesn't drop the operating frequency down to the kHz
range, so it wastes power. There are some components which look like
they might have been intended to do this but the values are wrong.
Changing the value of a capacitor fixes it. Failing that, adding
some hysteresis (positive feedback) across the comparator would
also probably work, but I didn't try it.

What do they use in things like the MR16 LED lamps you can buy quite
easily nowadays, is it just a dropper resistor or is there some
electronics in there?

I don't know. I have one somewhere, and I observed that it only
works one way around from a 12VDC supply, so they didn't go overboard
with a bridge rectifier ;-).

--
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]