For future reference, this is the LDO mentioned in HQ’s thread:
Both this and the Micrel you used look like very good parts. Low drain, reverse polarity protection, good C2 compatibility. 10 uF C2 should be perfect for both. This one does something with pin 4, where it is N/C on the Micrel. N/C is fine for both.
The Micrel is specified to be no-load stable.
Looking at the LDO and general board layout nothing seems off…mystery.
Probably a stupid question, but is C1 rated for 8.4+ V?
Somewhat related, but on the total opposite end of the spectrum, this one is an SRK driver aimed at muggle use.
On a whim I tried this 4-emitter ‘Ultrafire’ SRK clone from Fasttech:
https://www.fasttech.com/products/1601/10010861/4494900
It took forever to ship and my sample should be fresh stock / current batch. The retaining ring for the button is red on mine and the spring board green. The rest seems to be the same as in the Fasttech photos.
My expectations were very low, but it turned out better than expected. No shelf behind the mcpcb, way off-center LEDs, aluminium bezel, whimpy driver and wires (26 AWG?!), but not bad for the rest. The host is a little different from the traditional, no fins around the button and a 6-lobe design on the tail cap, vs. the 3-lobe. Decent reflector, real XM-L2s, cool but not blue. Reflowing the LEDs to center properly was easy enough.
I considered to do the typical add-in shelf and using loose 16 mm dtp mcpcbs, but decided it is not worth the time. Running the light at 5 A for a while showed the thermals to be better than expected as well. The 3-4 mm rim below the mcpcb transfers heat surprisingly well. The head heats at about the same rate as the mcpcb. After 15 minutes everything was at around 55 C with the loose (and driver-less) head driven directly from a power supply set at 5 A CC.
So this will become a ceiling bouncer/cell holder for camping/power outages. No DD, but rather a simple 2-channel current-limited driver with space for up to sixteen 7135s:
Similar to the K2 driver above, there are resistors and pads to drive up to 2 indicator LEDs. These are not useful with the stock Ultrafire clone.
Ahh, not stupid Q at all. I dunno... Typcial 10 uF cap, probably the one Richard sells that comes from Mouser, or ones I bought from Arrow - not sure. Gotta check.
Ohh - just checked - 16V rated 10 uF caps, here: http://www.mouser.com/Search/ProductDetail.aspx?R=GRM188C81C106MA73Dvirtualkey64870000virtualkey81-GRM188C81C106MA3D
The first LDO I tried was a TPS76950, which simply did not let the light work properly.
Then I ordered other types but I’m not 100% sure what I ended using. I believe it’s the MCP1755T (I still have in my box LD2981 and LP2985 as well). All are fixed voltage 5V versions.
The driver board for cheaped-out SRK lights is shared. The pads for the 7135s have been up-sized to regular SOT-89 pads (most of the smaller diameter drivers use cut-down pads to save real estate on the PCB). To allow routing without vias the thermal pads are still not full size, but should be ample and are fully connected to the ground ring.
No FET, no DD. It should be a good match for the sample cheap SRK I have, even using the stock non-dtp mcpcb. There are probably worse ones out there that will still need thermal improvements for this to run well.
The pads for the lone 7135 on the K2 driver are now full-size as well (rev01).
Dang, this is a nice 5.2A SRK driver... Good match perhaps for the DBSAR SRK lantern project. For me, I'd be too tempted to stack 7135's though 
. But certainly based on this design, no need for these jumpers which really did pick up the amps on this design (and maybe because of stacked 7135's), but all these jumpers were not needed, just 1 or 2:
I also added the pry holes on the edges to help removing from a tight press fit.
Yes, this board should not need jumpers (stock build anyway). Generous tracks everywhere, a luxury not available when fitting 32 of those 7135s.
I actually did consider adding pry-holes and maybe the typical sand-to-fit side tabs. Both are easy enough to add if useful. Was thinking to use dabs of silicone caulk to hold the board. Access would then be from the front to push the driver out if needed. Not an option on some builds of course.
Ahh - funny you should mention that use of silicone caulk. On Tuesday I took your L4 board and expanded it's width a little using a 2 part clear epoxy because I really needed a 26mm board, but the L4 extended a little bit will fit snugly, and still get a good ground contact. I trimmed the epoxy to be flat aligned with the top and bottom of the PCB.
I'm using it in this classic: kaidomain.com-UranusFire-818
Nice looking host Tom, but that name… 
The share for the rev01 is up for the SRK-lite board. Only difference with rev00 is the board outline. Added some friction-fit tabs and pry holes:
Not sure which version I prefer, will leave both shares up.
I do like the tabs - sometimes you do need that extra material for a good press fit. For the pry holes, there could be a better solution. You need a good tool to use them without scratching up the threads (I've done this), then you could break thru the anodizing on those threads which makes intermittent electrical contact when screwing on the battery tube. Many of these cheap SRK's though have poor thread anodizing anyway, so maybe not so much an issue.
Not sure how to do this better - maybe couple enlarged ground thru holes, or maybe couple large holes little further in from the edge so it clears the underlying ledge? Really not sure what the best way is, but should match up with an appropriate tool. I've used a pry tool with a hook or bent end - seems to work best. Hard to avoid some light damage to the PCB. Most people won't be popping these in and out anyways.
Yes, the pry holes were useless on the Ultrafire host that started this. The board was jammed in tight and I think there was some superglue involved as well.
The pry holes on the OSH board above are oval/flat, to be used with a flat 90-deg pick. Or a cheap flat screwdriver with a bent over tip. Round picks would damage the PCB.
A nice option would be a hole, off the ledge as you say, with a small nut glued over it on the component side. The board can then be pulled easily with a bolt screwed in from the back when needed. Would be easy to change the center +vias to one large via for this purpose.
Very very nice features on that board.
Now I’ll be the one to start some borrowing
I really like the tabs. Less material to remove if not needed, but probably even better contact as it makes for 4 contact points. And: no extra cost because they are put diagonally and do neither increase width nor height.
Has anyone noticed how the tabs and the 7135 match with the alignment of the GND holes? I’m a big advocate of form follows function, but those details really get me.
Pry holes. Real pita when you don’t have them… Although even a present cutout makes it difficult to get the board out in an SRK. There’s simply not enough leverage. But 2 opposed cutouts… and long pointed pliers… and turning instead of levering… I need to test this.
Oh, and routing LED- on the inside of the 7135 instead of underneath them. I had this on my first draft of the BLF SRK V3, but decided against it for better routing of the PWM signal. But the more I look at your board the better I like it, even if PWM goes between 2 legs of the rightmost 7135.
Well done :+1:
Thanks HQ, happy if you find something useful in there! The 7135s are placed on 18 degree radials starting 45 degrees from east. The vias are on 9 degree radials.
There was a small bug on the version with tabs though. The top two and bottom two round segments of the outline had a radius that were a little off. Fixed now in rev01a.
The stock board had 2 round pry holes. I had no luck trying to twist them, but I did not have great pliers in that diameter either.
This version has a center ‘pull-via’, to be used with a glued nut as described above. Off-center would probably work better, but not really space elsewhere for it.
Thanks for the heads up on the L4 Tom, it is a pretty nice host! Got the boards in this week and put one together today.
The rev01 board is a pretty snug fit, no play. Still easy to get out, especially once the spring is on. Went with a 219C for the emitter. Using an SMD tactile meant for the Yezl board, but it works fine.
just ordered DDm17 rev04, can’t wait for them to show up! Haven’t decided which firmware to use and which host, but will probably put this into a ThorFire.
I know I am way late to the party but RE: 7135’s stuck on when run at 2s voltages.
I have had this same issue on several lights. I don’t even use 7135’s on 2S lights anymore, just not worth killing 5 of them to find one that works.
I now just use the FET alone, not a great moon mode but everything else works fine.
I killed like 25 7135’s trying to track down issues before giving up on them.
I can also confirm it is not heat that kills them, I tested them on the bench and the ones that would die, would die instantly. The ones that lived would work fine and get warm but not usually excessively hot. The output would not even step down if there was a good heat path to the host.
Ahh, ok - real good to know. In my 16X XHP50 light, I went with FET+1 in HQ's SRK driver, but the first 7135 died, and the 2nd didn't die but acts flaky - won't turn on. but ramps ok -- weird behavior. I'll drop the 7135 then. Need to generate a FET only ramping table for it. Hhmm using the new Narsil, wonder what temp I should use for thermal stepdown - maybe 80C? Yikes!
I had some act like that as well, particularly in my 8x xhp50 SRK. I had a lot fail so that they were stuck on a lot current and would not ramp but if I directly shorted the switch pin it would run at full power just fine. Took me forever to figure out that it was in fact a bad 7135 that was causing the stuck on light.
I created an FET only ramp table but not sure how good it is, the moon mode range is not very smooth but not much can be done there. The rest of the ramp is nice though.
Did you use TK’s ramping levels generator? I think that’s what Tom E has been using. Amiright Tom?
Some good data there TA.
Inductive spiking from FET PWM is now even higher on my suspect list. Seems we are OK with 1S voltages, but 2S pushes it over the edge.
Do you recall any detail on when or how fast the 7135s failed? Was it consistently after the FET has been PWMed?
Other big variables are wiring and current. How long were the wires to the LEDs and/or to the power supply? How much current?