I suppose I'll have to think about this more. I may simply add current limiting resistors because I do not want to over-complicate this build. ICR cells will help, as will the new thermal vias, if I drop the current a bit more with the resistors then I can help the nichias survive even longer lol.
Stupid question that I should know the answer to but it's too early to think: would switching to a smaller FET help? Or will 4 219's all pulling from a single 2502 (for example) simply blow the FET? Or will it limit each LED to around 1.1A (2502 is limited to around 4.3A IIRC).
- Matt
Blow the FET, probably. I’d go with the resistors instead I think - move the resistive losses away from the FET.
Deal. This is the last revision I did, though I'll add the resistors at some point this weekend before ordering new boards:
I removed the AMC chips and replaced them with IRLML2502 FETs (I like the look better and I'll get more light), evened up the nichia traces as much as possible, and quadrupled the thermal vias available to the quad array. Also added a few relevant icons lol.
As already mentioned I'm going to add current limiting resistors to the Nichia's (which might require a fairly significant revision to that part of the board), and possibly some small resistors to the XQ emitters as well.
- Matt
When I built the Tiny10 for my Texas Poker with the 2502 FET I was encouraged to check it’s potential with bigger cells. The most I saw it produce was 5.7A if I recall correctly, from a decent 18650. It’s now comfortably making 3.11A on the Efest IMR10440.
Cool idea, securityling is nice, I wonder about empty Supfire M6 hosts.
Matt, what about dumping the Nichias for XPG2s? It looks like you have the space. Then Comfychair’s trick with the copper wire looks like a real help with Oshpark boards.
Comfychair, what gauge wire is that?
I want the high CRI of the Nichia's. They're my go-to LEDs :)
I just realised I've never done direct drive with current limiting resistors before! Someone check my maths?
Assuming VIN = 4.2V (max) and the forward voltage of the Nichia's at 2A is 3.9V.
R = (4.2 - 3.9)/2 = 0.15ohm.
P = LED Current * Resistor Voltage Drop = 2A * 0.3 = 0.6W
So I need BATT+ > A 0.15Ohm, 0.6W Resistor > LED > FET > BATT-
If I use one resistor for 2 LEDs (because of space restrictions), my resistor value would be 0.75ohm and 1.2W resistor?
- Matt
Empty M6’s is what I want, but I don’t really see that happening at this price level. I doubt it would reduce cost by much. Even in the case that the M6 host was cheap to produce, if I was Supfire I wouldn’t sell it separately - I’d be worried about cheap-crap fakes built in my inexpensive hosts flooding the marketplace.
That’s exactly what I though about doing, but using a 50mm copper round from etsy
With solder filled vias to a solid copper back, you might not need to drill and drive in copper as you have increased the thermal path with the “liquid” reflowed solder vs a solid copper slug (copper will be better thermal conductor than solder but it’s benefit only shows up at VERY high currents)
Matt, when you make the 3D view, don’t drag over the GBP and GTP (paste layers) it will not show the “white” blocks on the solder pads…the paste layer from the cam processor is for those that want to up the build to OSHStencils for a good paste layer stencil 
Incredible build man, simply incredible
It’s past my bedtime, so I’d better check your current limiting resistor math another time ;). I’m not sure that switching to IRLML2502 is a good thing. I think you’re in the same position as Cereal_killer here you’re trying to DD a red emitter (<3v Vf…) from a lithium-ion battery. That will end in a poof as far as I know. I’d say you’re stuck with the 7135s as they are the most space efficient thing I’m aware of for this application.
EDIT:
- Still things are looking good. Progress!
- Also, FWIW, I thought the original design you showed off looked great. The circle of 7135s looks really good/fancy to me.
Very good point about the red emitter. I might have to switch back to the 7135 chips purely for this reason. The circle may or may not make a return lol.
Mattaus:
This is a really neat board, sort of like a LED development board, all kinds of things going on.
What size are your via’s under the thermal pads? Cree did a study or white paper on thermal paths with different via sizes and configurations. I know when I worked with the XML and OSHPARK minimums (at the time something like 12mil hole and a 7mil annular ring for a total of like 26mil), and I was getting 3 columns on the pad that way (what pad overlap ). Maybe you could get two columns for the XP/219 pad? Offset the holes for a tighter pack. Per the Cree paper, the further the via was from the thermal pad, the less effective they became.
EDIT: Oops, I meant mils or thousandths not mm on the hole sizes .
4 x Nichias:
I would argue that the majority of the big clump of vias on the top left would be better just being copper. Also the interior clump of vias could be substantially reduced, and striking a balance, for thicker LED+ tracks in my mind.
Looking at the Nichia data sheets I see 2A and Vf’s of:
219A - 4.1V
219B - 3.6V
I think 2A (per single led) on FR4 PCB is sketchy; the thermal resistance of the 219 is not that great plus the max package power dissipation is 5.1W per data sheet. More likely to survive at 1.5A on a long term basis?
I think it's 12AWG, it's just taken from leftover scraps of Romex. Actual size is .080-.081".
You can actually fit two pieces of .080" in a XML thermal pad, I've done that on single LED boards but for this one was a bit too much to bother with. Drilling into the FR4 PCB on top the drill wants to wander all over for a while before it breaks through and starts making a straight hole. One thing that helps is making a shallow starter hole with a round diamond bit first, since you can't really whack a PCB with a center punch like you can a piece of real metal.
Hey Matt,
This looks awesome, I’m happy to see it coming together. I was just thinking, if you’re doing a board rev, you might consider adding a thermistor on there at a strategic location. There are two pins available to use it, and there may be a firmware update down the line to use it…
whoa 
for a second i got stoned for what you did 0:)
Thermal protection and throttling sweet!
I need to look at what OSHPark's min drill sizes are and I'll work off that. FR4 is more or less direct copper bonding and if you fill the holes with solder then the performance is even better. I keep forgetting the copper part of the vias can overlap, so I can cram more in even if I use the existing via sizes.
OK, so 1.5A is probably a good idea too :)
Probably a very good idea! I'm assuming it'll basically be a variable voltage divider circuit? If you define the pins and tell me what thermistor/resistor combination you are using I will revise my board now in anticipation for the update. If the update doesn't eventuate then I just won't populate that part of the circuit.
- Matt
I don’t have an update yet, but might as well include the pads on the board in case we want to use them later. Make a divider from RC4 (pin 6 on TSSOP) to ground (47k NTC thermistor on top, 15k resistor on bottom). Connect the middle of the divider to RC3 (pin 7 on TSSOP). The reason we need both connections is because we’ll use C4 to power the divider only when we take a measurement (with C3), and it can be turned off in sleep mode so the divider doesn’t drain power all the time.
How does this look:
EDIT: Any chance the same functionality can be achieved but by using RA0 and RA1 instead? It's looking physically impossible that I can run the traces I need to run from RC3 and RC4. Of course this could mess your intentions up with your own hardware, so if that is the case then don't worry about it - I'll either get the traces working or just forgo the temperature sensing.
- Matt