That device seems perfect. It covers our input voltage range, can run with only 1 capacitor, reverse battery protection, and ultra low shutdown. Most importantly it's consistently much cheaper across the board (Digikey, mouser, and element14). Nothing wrong with the one we have now, this one is just cheaper. It saves about $1 per board. I will include this in the next revision.
According to the simulations on the ST Micro site, with a VIN of 8.4V and Vout of 6V (simulating and MT-G2 on 2 fresh cells) the LED2001 IC is loosing around 1W in heat. The bigger the VIN/VOut gap, the worse it's going to get. Which brings me to...
For most BLF users? Yes. The AMC chips do a very good job for very cheap. But with a fresh battery and a single LED, where the VIN/OUT delta is around 1V, the Knucklehead is MUCH more efficient. Horses for courses though.
Anyway, I digress. Layout seems key here in order to dissipate heat. On reflection the layout can be improved somewhat, though I am saying that without having Eagle in front of me to play with. In it's current guise:
The ATTINY and LED2001 sit virtually on top of each other. The thermal pad on the bottom of the LED2K1 is basically dumping heat straight into the ATTINY. That being said there is a layer of PCB in the way...
The thermal pad of the LED2K1 is GND. Attaching the GND ring of the KnuckleHead to a proper heat sink should/could help
My suggestion for the interim is to throttle back the current. Remove the 0.025ohm resistor and drive the LED a little less hard. Also use one cell, or more closely match the LED Vf and VIN. Get a good firmware going and confirm what PWM frequency we can run up to. I know it's 2+ cells or bust, but this driver has the ability (When it's stable) of also doing (Where 'X' is number of parallel emitters):
2sXp/3sXp/4sXp XM/XP/Nichia from 2/3/4 cells
2sXp MTG-2 from 4 cells.
I will do a PCB revision. The changes will include:
Incorporating the Micrel regulator.
Rotate the bottom PCB components 90 degrees. This will pull the ATTINY and Vreg away from the primary heat emitting sources on the top of the PCB.
Add as many thermal vias as possible to the base of the LED2K1 IC, and leave the reverse side exposed. This will allow direct thermal path access to the IC.
Try and increase the thermal path thickness to the ground ring.
Please note that most of the above will be attempted. This board may be 17mm, but for this setup it's REALLY TIGHT. I will do my best.
Question…would a bit of epoxy hold it on w/o the need for the brace…just scuff the top of the IC and epoxy the bit down…I have one of those stepdown 3A regulators I did the same thing too…it get’s CRAZY hot when running at 3A but the epoxy doesn’t let go
Great post Matt. You put a lot in there. If the code can be worked out by the time Rich gets the kits ready then more of us can start to contribute to this effort.
I’m about gassed on this one. I spent yet more time this evening de-populating the MCU side so I could pull the regulator off and put a new one on. Then I re-populated the MCU side and put a nice big fat wedge of copper on top of the regulator, looks like an ancient sun dial with this 1/4” thick triangular wedge of copper glued on. I glued it to the inductor, and used more of the thermal adhesive under the edge between the via and ground ring and the copper. So it’s stuck on there pretty good.
And, I wired 2 XM-L’s in series.
And, it doesn’t work. It lights up, but in a very low mode only. That’s it, about .10 amps and nothing more. I’m re-using boards and the 8HSOP and the caps. And they’re as tired as I am, apparently.
So, Tag!
You’re it!
(until the re-enforcements arrive from DigiKey anyway)
When you re-design your PCB, I suggest you put a lot of 0.3mm thermal vias at the IC’s exposed pad helping the heat to transfer from the top layer to the heatsink at the bottom layer.
That's exactly the plan. It'll not only provide a direct interface for heatsinking, but will also add more PCB copper to the area as well as effectively implement a pseudo heat sink (lots of little holes with extra surface area).
In regards to kits from RMM - I'll be buying a few myself lol.
I’ve got 10 of the inductors from Coilcraft in hand, with 12 boards coming from OSHPark and more components on the way from DigiKey. I might be able to put together a few kits to get out if a few are interested in joining the fray.
I initially got into it for the one light that I was fixing for a friend. I’d like to have 2 or 3 to use if/when occasion arises but I don’t need all that I’ve got coming. So maybe 7 complete set-ups will be able to take a walk if someone’s interested.
To use the current board would it make any sense to use 2 boards spaced with wire pins joining the vias? Then you could drill through the buck pad and solder directly to a nice fat hunk of copper. As long as that hunk is isolated from any other pins/pads/vias it should work, right?
You mean to make use of the boards already fabbed at Oshpark? Frankly that sounds like a big pain and not worth the effort for what the new boards will cost. Maybe a better option would be to switch to a sense resistor which nets a smaller drive current.
Use “0.1 / current in amps” to figure out an appropriate sense resistor. For example 2Amps would be 0.1/2 = 0.05 Ohms.
You could make use of an ~2A drive current with those red XP-E or Rebel emitters.
The problem…the driver will be enclosed inside a hollow tube beneath an already hot emitter…the hunk of copper will only absorb the heat from the IC…it too will end up heating up…without some way of potting the entire thing to the pill thus a direct path transfer of heat…
I think the battery side large ground ring thru vias then solder to the pill on the battery side would be a very good idea…at least this way the heat has some place to migrate too
So I’d use 0.033 or whatever the closest available value is.
EDIT:
So this would probably be a decent choice: LRC-LR1206LF-01-R033F
Remember that we don’t actually know that reducing the current will keep these chips alive.
Yes, but in theory (a very large amount of educated guessing) it should. The voltage difference will be the same (actually slightly more) but seen as the current draw has a multiplying factor on the excess energy lost as heat, it should greatly reduce the heat the LED2K1 has to remove.
So if we decide to go with a “daily driver” multi cell buck driver of 3A into a XM-L (single emitter) maybe just maybe those poor chippies wont be splodin’ no more?
I’m not even thinking about putting this in a host yet.
I mean to allow those with boards in hand a way to use them for testing until the new ones are ready and it might take awhile for that. I’d just soon see us find out what we can before he does a redesign as it’s likely to affect the changes he makes rather than have him go through that every other day. It might be somewhat of a pain but waiting a few weeks for each new fab to get designed, produced, and shipped will slow us to a crawl(might save Dales finances though). By all means when a new one is ready we can shift over. In the end I don’t see how we’ll get to max output without a thermal pad and sink so why not implement what we can until then. Dual boards is one idea but I’m not stuck on it.
It’s along these lines that I offer to spread some of the testing out with the parts that I will soon have in hand. With the Coilcraft inductors already in hand and remaining components en-route, possibly here today, I’ll only be waiting on more boards. If I keep all that I ordered, I may end up with a new burial site for Knuckleheads. Since I seem to be the first to accumulate all the pieces parts, it’d be easy (and prudent) to spread some of those around and see if we can get the bugs ironed out.
I don’t know what I’m doing, obviously. I can nail the pieces together, but without an accurate set of blueprints there’s just no telling what kind of shack I’ll end up with. As has been seen. So, I plan on keeping 3 more sets, that will give me a total 6 attempts…if I can’t get it right with that many then I don’t need one. Which means I’ll have 7 sets to spread out.
How do I go about that? Obviously it would be of utmost importance to put some in the hands of someone who knows how to figure out what is going on. So, with that in mind, (and as I’m not a mind reader) please ask only if you are competent at electronics to the point of helping the team. I don’t have an unlimited amount of components on hand, nor boards, so this will have to be done with prudence. 2 sets each to 3 folks, with one set to someone else to dabble with.
How about it? Who’s getting these? (I will provide them for the sake of the team, shipped in the states…if out of the country I’ll ask you to pay shipping)
The boards were in fabrication 2 days ago, so I should be getting a shipped notification today or tomorrow…will probably have everything in hand by Monday the way it’s looking.
If someone was just dying to try 2A operation with what’s on hand (current PCB revision, current BOM) they could just use 2 of the 0.025Ohm sense resistors in series. You can make a little triangle with them like /\ to fit them on the board.
0.025 * 2 = 0.05 Ohms
0.1 / 0.05 = 2.0 Amps
If someone was just determined to heatsink the LED2001 with what’s on hand I’d recommend one of two options.
Doing what RBD already touched on, heatsinking pins with some solid wire. Pins 4 and 8 of the HSOP8 package are tied to the thermal pad on the bottom by a large PCB trace. The ground side of R1 is also on that same large trace. The ground ring itself is tied to that trace. I’d attach my 14-18AWG heatsinks to pins in that order of preference, but the ground ring presents a large solid area where you could really tie on some mass if you wanted. Obviously area of ground ring closest to the HSOP8 is best for this.
This isn’t my favorite thing in the world, and only works for bench testing. You could certainly hammer flat (ish) a chunk of 10-14AWG solid wire and solder that onto the thermal pad, with the tail sticking out the side. Tin the whole end. Setup your reflow skillet with the LED2001 stacked on top of that and make sure you aren’t shorting R1. Do not worry about the other pins on the HSOP8 package. Reflow. Then bend the pins down 1 by one and solder them individually. You could always try and do that in advance so that you could get all the joints done in one fell swoop with the reflow, but I think you’d be more prone to failure that way. You might need to anchor the HSOP8 to keep it from twisting, so either pre-bending 1 pin on each side or maybe some tweezers or a toothpick is in order.
I just realized how badly I was misinterpreting your most recent suggestion RBD! You mean to just lay two boards out side by side on the bench and use a fly wire / jumper wire to pass the PWM, BAT, and BAT- over from the “bottom” board over to the “top” board. I was imaging something crazy, sorry. Yes, I think that’s a good idea and there is no reason not to do it. As a practical consideration you would need to cut some traces when you did your drilling on the Buck board. There is a fat fill/trace which is connected to BAT there [eg right underneath the thermal pad on the opposite side of the board], you’d be drilling right through it to solder to something electrically connected to BAT-. Easy to deal with, just cut it and insulate or put a bevel/countersink on your drill hole and insulate. I think that this is much better than where I was going w/ #2.
Hang onto those sets, Dale. Just take a break and work on the Bigger Bullet for awhile. I’m eager to get some myself but I think the code needs to be adjusted and I can neither code nor flash and you can at least flash the mcu.
Wight- thanks, without having a board in front of me it’s hard to tell from the Oshpark rendering exactly what’s on each side but yes, for the purposes of testing only it might make sense to separate the two functional parts enough to allow proper heat sinking on the one part that seems to require it. I say “might” only because the data sheets make a big deal about keeping certain traces short and avoiding having others side by side and the physical separation of the two haves may cause more problems than it allows us to correct. If it helps in the short term, that would be great.
Are you thinking of any specific trace length considerations that you saw in the datasheets? I don’t think it should matter for the ones which connect the two sides, but I’d still like to know what you saw.
Since I seem to be the first to accumulate all the pieces parts, it’d be easy (and prudent) to spread some of those around and see if we can get the bugs ironed out.