Just off the CNC - Update: It's Alve!

I continue to be amazed at the ingenuity that I see here. If you were in DK or even Europe I'd have to go see you real life to co-develop some ideas I have for industrial LED lighting in EX zones on ships. That would be awesome! Maybe not as awesome as this but still...

Keep up the good work Pilot.

That is a hell of a nice compliment. I'm not sure it's deserved yet, but truly, Thank you.

I'd love to work with you on something like that - it sounds very interesting. Ship lighting; Awesome! :)

PPtk

It works, and it works quite great, in fact. Well, it did work anyway :) I smoked 3 XM-Ls, but it was my fault - not the electronics. Everything was working great, and 6 XM-Ls were burning away at 1.5Amps each - looking good...

So I started cranking up the input voltage to see how the DSP handled the larger Delta-V from input to output.. Got to 60V and it was doing great - The DSP kept backing off the Buck converter keeping a nice steady 1.5 Amps going to the emitters.. Here's me, wanting to look at the signals - Output of MOSFET Gate Driver - beautiful. FAST rise and fall times. Output of Current-Sense Op-Amp - Really Nice.. Clean.. And then.. Slip, and I shorted the high side input of the op-amp to ground with the scope probe... The output of the op-amp went low (as it should), The DSP's internal comparator went low (as it should), the DSP instantly ramped the buck-converter up to 100% duty cycle (as it should), and the XML-s (near) instantly saw 60V accross the group of 3.. 20V each.. BAM, POP, Smoke, and I'm still seeing spots. For about 1 micro-second, XML-s are REALLY bright at 20V :) .. The perils of poking around with an oscilloscope..

So, I've got a perfectly functioning board! All tests out perfectly. I placed the order for 24 brand-new XM-L emitters (going Neutral White T6 instead of Cool White U2) and they should be here Friday. I'm also going to swap out the 15uH inductors for 22uH since they're back in stock - 22uH is really what I wanted to begin with, but CoilCraft was out of stock on them, so I went with the 15s. For this load 15uH is just not quite enough - the current rises and falls a bit too quickly. It doesn't take more than 10 minutes to swap out the 8 inductors, so I'm going to get the ones I really want on there..

Other than the little mishap, I couldn't be happier. If I can keep my grubby little scope probe out of places it doesn't belong, this board will be making more than 20,000 lumens by the weekend.

Now on the the control software in the DSP. All it does right now is turn the LEDs on or off - time to starting coding up the communications routines (Bluetooth) and the dimming routines. Luckily, now that the DSP is properly controlling current (the hard part), there's nothing I can really do to "blow things up".

I really can't explain how bright those three emitters were for a fraction of a second - it was like someone cut off a small piece of the sun and dropped it on the desk in my lab. I wasn't looking anywhere near the emitters and I was still absolutely blinded. I don't recommend feeding XM-L LEDs 20V - it's truly dangerous to the eyes.

PPtk

Swapping emitters is nothing, the main thing is, it works. Crossed a huge threshold. Always wondered what 20 volts would do to an XM-L . . .

nowiknowFoy

Stop wondering, and don't try it. It probably wouldn't have been quite as bad if I had been using 5 18650's in series - They would somewhat self current-limit at somewhere between 4 and 10 amps.. The supply I was hooked to though can deliver 100 Amps - and I'm confident that it did for the fraction of a second before the bond wires vaporized. Because of the Voltage-Current curve on LED's (diodes), at 20V they would conduct essentially infinite current. I'm just glad the bond-wires in the emitter vaporized before the traces on the circuit board did! I'm tired of soldering 64 pin fine-pitch QFN's by hand :)

PPtk

OMG this thing is seriously hawt!.

I can really really see the design and thought that has gone into this design!, something that is very rarely seen these days. I have many ideas in my head of what I want to be achieving, but my gosh I don't have anywhere near your ability!, such as Hand soldering QFN packages (OMG!), enough said!

I don't have a distinct need for one, but my gosh, if you do down the line consider putting another one together, I'm very much interested! Even if its just to appreciate the engineering and thought that has gone into this!

Haha.. Hawt :) I like that :)

Thanks very much - your comments and compliments are very much appreciated. There has definately been a lot of thought put into this thing - and thus far, it's been a lot of fun. I never get the chance to design something for myself, and I finally decided it was time.

Surprisingly, soldering qfn's by hand isn't really that 'difficult' - it's just time consuming and tedious. Maybe I'll post a video of how to do it some day - Most people could do it, it just takes a bit of patience.

If I ever decide to build more - I'll make sure to drop you a note.

Thanks again for the comments!

PPtk

This is like a work of art. I can't wait to see it in action.

And if the optics are removable still (appears they should be), then you have a customisable beam pattern too!

Yes, the optics can easily be switched even after the whole thing is fully assembled. Remove the front bezel and pop the optics right out.

To my knowledge, the only thing that will fit my emitter arrangement is the CUTE-3 Optics from LEDiL, so I'm stuck with those, but I could Use Wide, Medium and Narrow is different ratios.

PPtk

yeah, that should do the job :) I didn't even realise that you can't get Cute-3 elliptical optics, which kind of makes sense as you'd have to use the triple in a certain orientation for it to work, but it's still a bit of a shame because those wides are going to be putting a lot of their light up into the sky (and mediums and narrows to lesser and lesser extents). I'm more used to 20mm optics where there's a bit more variety.

that's a really neat bit of engineering. Varying current to different triples would be handy if you wanted to vary the beam within a certain thermal envelope too.

I don't doubt that you've done your modelling and the thermal paths look exceptional, I'm only wondering if the efficiency losses from running the LEDs at higher currents are worth it, when you could get a similar light output at a lower current and LED temperature. A guy called Troutie on the mtbr DIY light forum has done some tests of light output vs. temperature for the XM-L and it was pretty instructive, even if I can't find it any more :(

Both of my lights (~10W and ~20W, so a shadow of yours) have their thermal trip set at 60C, so that the LEDs should never run at more than 90C, although that is more to protect the driver. Their real world temperature is most likely a lot lower than that. I just think that a trip point set so high doesn't give you much of a margin, that's all.

I would be really happy if the CUTE-3 Optics were available in an elliptical, but sadly, they're not. Thankfully, although I'm going to be pointing a bit of the beam up into the sky and down into the dirt, I'm pretty confident that I have enough raw lumens that it just won't matter.

Not only does it help keeping thermals under control, but because I have the different optics, it also allows me to vary the beam pattern. I can turn on All flood, All Spot, or any combination that I want.

I have little doubt that you're quite right - Almost as many lumens will be produced at 2 amps as will be at 3 because of thermal sag. That's why the dimming function exists. Most of the time, I can run at 66% (2 Amps) which will be more than adequate for most situations - when I really need max power though - it's available. Unless my modeling is way off, the 100C trip point is still 20C shy of being destructive to the emitters. For me, that's plenty of margin. Keep in mind also that my thermal model suggests that 100C would be almost impossible to reach - 78C is max expected even under really extreme circumstances (90F Ambient with practically zero air-flow). 100C is the Safety Shut-off point, it has nothing to do with lumen efficiency. (edited - It's actually not a "Shut-Off" point, it's a back off point. The light will not go off, it will just self-dim to maintain < 100C)

PPtk

Dimming Code Finished. Small demo - The bluetooth communications routines are not done yet, so this isn't being 'controlled' by anything - the DSP is auto-looping from max bright to zero and back again. Right now, the current limit is set to 1.2Amps instead of the full 3 because I don't have any heat-sinking behind the circuit board.

PPtk

Already looks bright as heck in the video.

Foy

It is - It was really difficult to video without the video camera going into saturation. I had to turn all the lights in the room on, and you can still see how dark my desk-pad gets as the camera closes the aperture.

PPtk

I don't doubt that last bit at all :)

I think that's one of the coolest things I've heard of. If bike lights could be bigger, this is one of the things I'd hope to see.

you're right, I was somewhat mixing the two things (thermal sag and safety cut off) - one is about efficiency, which is somewhat moot if your alternator can handle it, and the other is about protecting the electronics, which you seem to have perfectly under control. Being able to control power levels is a huge boon and might even help with visibility in different conditions (dusty, foggy, highly reflective terrain) where less light could mean better sight. I almost always run my bar light at ~1.8A as there isn't a huge difference between that and 3A (although I'm told the difference is more marked the more LEDs you're running) and the difference in battery life is considerable, but it's nice to have that little extra when I need it or when I want to wind up my mate in front of me :)

Well, I've been really busy again - and I've not had much time to invest in this project.. But I did get some more XM-L's soldered down (4 sets of 3 installed now) and There are no interaction problems. The sets can run quite happily in conjunction with one another. The problem at this point is with taking measurements and looking at signals - it's just SOOO bright in and around the circuit board that I can't even see the components. I finally put some diffusion film that I had laying around over the groups of emitters, and that helps a lot.

Since the rise/fall times of the FET Gates was a problem, I figured you all might be interested in how well the new gate drivers work.

This is a capture of the scope taking measurements of (top) the gate of the fet and (bottom) the output of the DSP / Input of the Gate Driver. These parts work incredibly well.

As can be seen, there is 47.7 nano-seconds of delay from input at the gate-driver to when the gate driver begins driving the gate. Once that happens, these things slam the gate voltage from zero to 5.5V (well above Gate-Turn-On / rdsOn) in a frightening 6 nano-seconds. Incredible. 53.7 Nano-Seconds for the whole turn-on. The turn-off has roughly the same delay, and an even shorter fall of about 4 nano-seconds. Remember that a nano-second is 0.000000001 Seconds. We're talking FAST.

Going well, just slow because of my limited time.. But forward slowly is a hell of a lot better than backward quickly :)

PPtk

Where in Michigan are you from? I am originally from Michigan, lived there until I graduated from school and moved to NC for a job, I really miss Michigan though to be honest.

South East. The Northern Suburbs of Detroit.

I've tried to leave twice. Never works. Hard place to get out of your blood..

PPtk

Ah, I was from the West Coast. Spent 5 years up at Michigan Tech though before moving down here, like I said, I miss it.