SaabCentral Forums banner

1 - 1 of 1 Posts

56 Posts
Discussion Starter #1

I got tired of burning through Sylvania filament bulbs in my low beams, and fancying myself a bit of an amateur EE, why not try putting in some cheap LED bulbs?

I was aware of a few things going in:

1) Drop-in LED bulbs (especially cheap ones) have a reputation for quickly burning out due to the fact that stock headlights made for filament bulbs were not designed for the heat dissipation LEDs need to have at these kinds of intensities, let alone making the drive circuit, heat management system, and LEDs fit in the H7 form factor.

2) My '07 9-3 Sport-combi tells me if bulbs are out, meaning the LED drive circuit might not be able to sufficiently mimic a filament bulb which might have to require a resistor to avoid the "low beam failure" warnings.

3) Dust covers might not fit since these bulbs tend to come with extra beefy heat sinks.

So, these were issues I expected to encounter, here's how it went.

First, bulb selection. After looking at few bulbs that advertised compatibility to the H7 bulbs, I settled on a pair of "Win Power" bulbs for 30 bucks on Amazon. I chose these specifically because they had external drivers, which meant it won't get the bulb hot from the driver, just the current running through the LEDs, good thing too cause those drivers get real toasty. Also it gives me the ability to put in my own drivers in case they're too bright. M̶y̶ ̶h̶e̶a̶d̶l̶i̶g̶h̶t̶s̶ ̶w̶e̶r̶e̶ ̶b̶a̶d̶l̶y̶ ̶f̶o̶g̶g̶e̶d̶ ̶o̶v̶e̶r̶ ̶a̶n̶y̶w̶a̶y̶ ̶s̶o̶ ̶t̶h̶a̶t̶ ̶w̶a̶s̶n̶'̶t̶ ̶a̶n̶ ̶i̶s̶s̶u̶e̶,̶ ̶s̶i̶m̶i̶l̶a̶r̶ ̶b̶r̶i̶g̶h̶t̶n̶e̶s̶s̶ ̶t̶o̶ ̶f̶i̶l̶a̶m̶e̶n̶t̶ ̶b̶u̶l̶b̶s̶ ̶a̶c̶t̶u̶a̶l̶l̶y̶.

After two days, they arrived and I immediately ran into a problem, the ring that twist-locks into the bulb socket was too square to fit the LED bulb, you could probably cut it to fit, but I opted to spend the whole day designing one in the 3D printer, I used some fiber-infused filament which made it slightly more resistant to bending when heated, which worked, the fit was nice and tight... before you drive the car around for 20 minutes, they then heat up enough to deform the 3D printed plastic rings a bit, not enough for it to fall out but when you go in to remove them after you let it cool, they're pretty loose, haven't had them fall out yet though because of this.

Now, the 3D printed ring wasn't the real problem, plugging in the bulb, and turning on the car, it works, no warning comes on that the low beam is out, sweet, I turn off the car however, and the bulb flickers... when it's off... continuously... w̶t̶f̶ ̶G̶M̶ ̶w̶h̶a̶t̶'̶s̶ ̶t̶h̶e̶ ̶p̶o̶i̶n̶t̶ ̶o̶f̶ ̶t̶h̶a̶t̶?̶ ̶T̶h̶e̶f̶t̶ ̶p̶r̶o̶t̶e̶c̶t̶i̶o̶n̶ ̶o̶f̶ ̶t̶h̶e̶ ̶b̶u̶l̶b̶s̶?̶ Bit of surfing reveals that this is a common test signal that the filament bulbs don't flicker to cause they require a lot more energy to actually light, LEDs aren't as thirsty though and happily blink to such small and infrequent pulses. Probably illegal to have them "strobe" when off so if I wanna drive at night, basically have to unplug the bulb when not driving, screw that.

I didn't give up though, Win Power also happened to sell an "Upgraded CAN AI Decoder" that when placed in-line with the bulb is supposed to stop all flickering, another 30 bucks on Amazon. Get it, put it in, doesn't work... Crap.

I still didn't give up though, determined to get it to work, I opened up the "AI Decoder" to see if I could solder something in to work. After reverse-engineering the "Decoder," it's obvious that it's just a parallel resistor and 4300uF of capacitors to try and kill the pulses by loading them to crap, can be seen in the third image with the power resistor. Didn't work though cause the it was inductively coupled. So, the "Decoders" ended up being 30 dollar cases. Of which, I made my own circuit that basically uses a resistor divider, capacitor and MOSFET to only allow the bulb to get power if the voltage going to the headlight is continuously on, and it worked, can be seen in the fourth image with the green tint. I do stream some of these projects to Twitch though under the same username and the "decoders" are archived for full detail viewing of assembly and testing.

So, my solution worked, side effect is the bulbs don't turn on immediately, but fade on. It's also quite surprising how much free space is in the headlight cause that "Decoder" case was really large. Anyway, got the modded decoder in t̶h̶o̶u̶g̶h̶ ̶t̶e̶c̶h̶n̶i̶c̶a̶l̶l̶y̶ ̶i̶t̶'̶s̶ ̶a̶ ̶p̶r̶e̶-̶c̶h̶a̶r̶g̶e̶ ̶c̶i̶r̶c̶u̶i̶t̶, the driver and the bulb in, managed to get the dust cover on too which was a surprise, and had no issues... except on the other headlight.

For some reason, the second headlight, had on-state flickering, as if it was getting dirty power, or something I did in the second decoder was wrong, and it was fading on slower than the first headlight. I took it out and opened it up and everything was fine, but there was evidence that it was getting hot. The only explanation is the MOSFET was only turning on to the saturation state instead of the ohmic state, similar to having a valve half-open and hearing a lot of noise come from it. But the values of the parts were the exact same as the first headlight, so the only explanation is the right headlight in the 9-3s get a lower voltage than the left headlight w̶t̶f̶,̶ ̶G̶M̶. Solution? Changed the resistor divider value to have a higher gate voltage, specifically changed the 1400 ohm on the top to 1100 ohm, works, fades on a little faster than the left headlight, might have to redo that one too soon, but it works, been driving around for a week and no issues so far.

Just wanted to share my experience, will update if something happens.

Thanks for reading!


1 - 1 of 1 Posts