Automotive turn signal

I've been thinking about this but not making any real progress.

My Probe GT's front turn signals use a 194 wedge bulb and an 1157 amber dual-filament bulb. The 194 and 1157 low are both on as parking/marker lights, and I believe when flashing the 1157 high filament turns on in addition to the other two. According to http://www.lighting.philips.com/nam/product_database/pdf/mini_seal.pdf
this means I should see 50 lumens normally, and 351 lumens with the turn signal flashing.

(25 lumens for 194, 25/301 for 1157NA.)

I've seen these "drop in replacement" LED bulbs for sale but they're a total joke. The 194 replacements use 4 LEDs, and most 5mm LEDs can't even output 1 lumen per LED. (Using the formula lumens = candela * 0.00024 * beam angle^2.) Likewise, the 1156/1157 replacements use 15-30 LEDs, but still that means only 30 lumens, max.

It seems to me to do this right, for my amber front turn signals, will require 10 LXHL-ML1D amber Luxeon Stars for 360 lumens at full brightness. I then need a way to run at only 15.4% output to produce the 50 lumens of normal operation. For a pair of turn signal lamps this means $300 for the 20 LEDs. Seems kind of ridiculous. I just paid $400 for a pair of HID headlamps, and I thought *that* was expensive...

Anyway, assuming cost is not an issue, let's continue the discussion - what is the easiest (smallest part count) way to drive this 10 LED array from the 12-15V power supply in the car, and achieve the low/hi brightness levels required?

It seems to me that I could run the 10 LEDs in two strings of 5, without needing any current limiting resistors, for full brightness. Obviously the output will be a little dimmer at 12.8V (engine off) than when the alternator is running, but I don't think that's too important. The question is how to underdrive these LEDs to deliver the 50 lumens for running lights.

I've pored thru dozens of PWM regulator datasheets and just gotten no useful conclusions. I think the SuperTex HV9904 or HV9906 sound pretty good. Then it seems to me that I may as well run a single string of all 10 LEDs in series; this way they will all run with exactly the same current, so the brightness should be uniform over the entire array.
I'd need a couple of switches to reconfigure the output depending on whether the flash signal is active or not.

Another thought was to reconfigure the array instead; run 6 or 7 in series for low-beam operation, and then two strings of 5 for high/flash operation.

I've been playing with a couple of white and red LSs and driving them from various batteries (my cell phone battery, some AAA NiMH, etc.) to see how low they can go. With a white and red in series, driven from my cell phone battery, both LEDs light very weakly and only 0.05mA current is flowing. The red LED is at 1.65V and the white at 2.44V in this setup. (I have so far been unable to get hold of an amber LED to experiment, but the red's electrical specs look identical.)

With the two AAA NiMH batteries and the red LED I saw about 70mA at 2.5V and then I tried a pair of more heavily discharged batteries, getting 7mA at 1.8V, with still a very bright light output. This got me thinking that just reconfiguring the circuit may work. However, I haven't found any SPDT analog switches that can handle 350mA, so reconfiguring the circuit means adding some additional transistors to route the power.

Any suggestions?

The biggest problem with replacing turn signal bulbs with LED is that the current will be wrong and affect the flashing frequency. Your car appears to have 21/5W dual filament bulbs in the indicstor lights, very like English cars of the 80's used to have. The turn siignal bulbs I have seen are normally replacements for sngle contact 21W bulbs. Try www.theledlight.com or www.ledtronics.com and check out thier based led's.

The LED replacements may not have the same lumen output but they are much more directional than the standard bulb. The ones I have seen appear much brighter than regular bulbs.

The relevant SAE specs for turn signals seems to require the flashing bulb to be 10x brighter than the normal running-light mode. None of the existing LED replacements meet this requirement.
I've already wasted a couple hundred $$ on a set of bulb replacements from www.wixtools.com... That very tight directionality is part of the problem; a turn signal bulb is supposed to illuminate the entire lamp so that the glow is visible, it doesn't need to project a focused beam down the road. In the daytime the LEDs are visible, but only as point sources. At night, with the running lights on, the added intensity of the "flash" is barely detectable.

To do this job right with LEDs requires a large number of the spaced evenly through the lamp housing, so that the whole lamp is illuminated, and to get rid of the point-source effect. I guess using 351 5mm LEDs would best accomplish that but I don't feel like soldering so many leads together. The 10 Luxeons feels about right, spaced about 1 inch apart they should fill this turn signal lamp very nicely.

There are no amber 5W Luxeons yet, but if there were, I wouldn't use them because even though you'd only need 2 or 3 to get the right lumen output, you'd still have that point-source effect.

As an aside, my 3.6V LiIon cellphone batteries actually measure about 4.10V on a full charge. And powering a single white LS with this battery is definitely overdriving it, the slug gets too hot to touch and the white light turns bluish. My pocket multimeter only measures up to 200mA so I haven't got a reading for the actual current flowing, but it must be a lot more than 350mA.

I wonder what those Tadiran Lithium-ThionylChloride 3.6V AAs would do, they seem to have a very high internal resistance.