Over the years I’ve modded a few GTs, leaving me with 3 GT reflectors. Those, combined with a desire to learn FreeCAD and use my 3d printer, have led to the below build.
3 BLF GT reflectors
Texas Ace 46mm driver attached to a small heatsink with 3m 8810
3 SBT90.2 LEDs mounted on 90mm x 50mm heatsinks
14AWG wiring throughout
3 80mm 12v fans & step-up voltage converter controlled by toggle switch
3 mateminco 4x 18650 battery packs converted to 1s4p
3d printed body, centering rings, etc.
The light works, but there is room for improvement. I’ll get an initial lux measurement soon. Once the bugs are worked out, I’ll do a beamshot sesh with my 4m lux aspheric GT90, MT90 plus (2.5m lux), and MT70.3HI plus (1m lux).
Issues:
The driver overheats quickly causing the light to step down. I am working on a design revision that will allow some airflow around the driver heatsink fins and a larger heatsink for the driver.
The fans don’t run well when on turbo. I will replace the 12v fans with 5v fans and lose the step up converter.
One of the beams seems less bright than the other 2.
Hope you get the driver sorted, that’s a lot of current draw from 4 volts. I suspect the dim led has a higher vf than the others. Also you’re likely getting voltage droop at full power.
Maybe a 4s lion driver could be the solution? That’d save changing the fans as well. Having said that I’m not sure what options are available nowadays
Thanks for the positive feedback. I’m currently considering changing the battery arrangement to 3s4p and changing the driver to a TA 46mm LDO. The voltage would be just right for 3 series SBT90.2 LEDs. It would also drop the current going through the driver substantially, and the voltage would be good for 12v fans without a converter. That change “could” solve all of my problems. Time will tell.
It’ll be a while before that can happen. I’m going to start ordering stuff though.
I like the plan to convert the light to 12v, run the LEDs in series, and ditch the step-up converter for the fans. I have ordered boards and components to upgrade the driver to a TA 46mm LDO. I’m waiting on the boards at this point.
I’ve revised and re-printed the light engine.
Recessed fans completely. Now the switches are easier to access and there’s more air volume for the driver heatsink.
Added vent holes from the driver heatsink chamber to each LED heatsink area and increased pocket size. This will encourage air to travel over the driver heatsink.
Countersunk the screw holes to allow me to use the same mounting screws with increased depth and hide the screw heads.
Once the boards arrive, I’ll build the driver and assemble everything into the new light engine.
About the driver, if you ordered the same 3V board from OshPark in the .8mm 2oz version, and also used the SIR178DP fet which has ~.0005 Ohm resistance, that might be another possible solution. The same approach could be taken with the 12V driver, not that it’s necessary with the substantially lower current. Also, check out Pololu DC-DC regulators which are very versatile. Do you need a solution for the front windows? I can grind glass, and AR coated museum glass is fairly inexpensive. I can machine acrylic or polycarbonate too. There’s also flashlightlens.com which offers custom acrylic windows.
Thanks, Hoop. I did go with the .8mm thicc pour option, but used SIR404DP FET. I’ll pick up some SIR178s to use on the new LDO boards. Thanks for the tip, and for the regulator info as well.
I am using the original AR coated glass lenses from the GTs I modded.
I’d love any other feedback or tips that will help me improve this light. Thanks again!
Oh I couldn’t tell that there is glass on there, which is a good thing.
The SIR404DP FET being 60A rated and having ~.002 Ohm resistance, at 60A it drops 0.12V and has to dissipate ~7.2W of heat! SIR178DP would drop .03V and make <1.8W heat at 60A.
I don’t know if it would result in any benefit but you could connect all four electrical pads on the MCPCBs. Maybe jumping a thinner wire over from the thick wire’s connection point on the MCPCB would be the easiest approach. Or you could use multiple sets of thinner gauge wires coming from the driver.
Consider using a 3s RC lipo pack as the power source for less resistance and more capacity. If you need a hobby charger, the SKYRC B6AC V2 is a relatively inexpensive option and doesn’t need an external power supply.
De-glassing the SST90 would result in ~5% more output and possibly a better tint. This approach is talked about here. (post #12)
If spill is a problem you could make extended shrouds to push it down range a bit but they would have to be kind of long.
You could add a separate driver and flood optics and gain a dedicated flood channel.
Those are all of the ideas that come to mind at the moment!
Boards shipped yesterday, should arrive early next week.
I need a .hex file for NarsilM (or Anduril if it works on the TA driver) 3 channel, 3 series battery configuration. Does anyone know where I can find that? I’m not great at sorting through all the firmware options out there.
Gotta figure out the new forum UI for photos, but here’s a small update.
I swapped to a mtn FET driver and the new printed light engine. Unfortunately I requested the wrong voltage monitoring (4s instead of 3s.) Oops.
One of the spots is still significantly less bright than the other 2 on the wall even with the LEDs in series. That tells me the VF may not be the reason. My current working theory is a reflector mismatch. One of the reflectors is slightly different, from one of the very first GTs. The other 2 are from later models. The focal length may be different.
Since the series wiring did not address the spot mismatch, and I am stuck with no 3s driver firmware, I may go back to the 1s driver with the lower resistance FET, and swap the fans for 5v fans. Another option is to bypass the driver and run the entire light and fans in 3s/9v configuration off a single switch. I like the utility of having a driver with modes and control.
Things are very busy in my life right now with 2 kids and home renovations, so updates will be a bit sparse. My next initiative is to wire each LED separately on my bench and measure the Vf and Lux of each spot. I’ll also keep investigating the firmware solutions. This light has the potential to be an absolute beast, and I will take the time required to realize that potential.
