Both help to suppress the inductive spike seen at FET (and even 7135) turn-off. They do not seem to be necessary for 1S lights, but both you and Tom had failed 7135s with 2S.
So C3 is in addition to C1 or just a replacement?
C4 is interesting, I will have to try that on my next 2S build.
I think that the 7135’s has gone down lately. As some of the ones I use work fine with no problem and other die instantly. There is no pattern I have been able to detect.
C3 would be in addition to C1. We need C1 for MCU stability. C3 goes before R5, C1 after.
I just got a 7135 order in from Aliex. Tried a new source since reports of Fasttech not consistently shipping the ‘good’ ones. Received 20 ‘saber-tooths’ and 80 ‘crappy sail boats’, in spite of the product photo showing a saber-tooth. Hopefully the saber-tooth supply is not drying up. I have another Aliex dealer lined-up for a try.
Ah, I see whats going on now. Yeah, that sounds very reasonable if it is indeed the spikes that are killing them. Which makes sense.
Personally I would like to see 7135 use come to an end. They are so bulky and far from ideal way to do things. The only advantage they offer is the ability to spread the heat over a wide area.
They do work though, so till a better option is found I am afraid we are stuck with them.
Would you mind showing photos? I have no idea how to see the difference, would be nice to know.
Don’t mind at all, but excellent photos here and the relevant discussion:
Actually I see it is ‘Raptor Claw’, not ‘saber tooth’…at least I got the crappy sailboat part right 
All credit to tterev3 for the naming and investigation.
Thanks! I had missed that thread entirely.
Just ordered a set of your 5.6m's - dunno why it took this long... It's perfect for my 12X SRK that I modded with tint mixed XM-L2's, but still on the stock MCPCB.
12x XM-L2 sharing 5.6 amps? Well, that’s still almost 3000 lumens, so yeah, okay, sure. :+1:
Just got a second set of those in yesterday as well Tom. Using them on 3x, 4x SRKs, for which I think it is a good match. Also using the stock MCPCBs.
Rev01 had the sand-down tabs which I finally found unnecessary. So the second set are rev02, without tabs. Just shared it on OSH if anyone has a use for it (mechanical fit is not tested yet). Only other change is full size pads for the 0603 parts (IPC compliant, since we have the space). There is also a C4 which, like C3, is not really needed on these ‘low’ power boards.
All good! I still have to get back to the 16X light to test the C3 and C4, because there's still some issues. Not look'n forward to it though to open that light up - spaghetti wiring, working from the top down, etc... So many things goin on right now and not much time.
Oh btw, think I've posted bout it, but the HaikeLite models MT03 (triple flooder) and MT07 (single thrower) have a perfect SRK size driver, spring mounted though on the driver for the carrier. So, a FET based LDO 46 mm driver is ideal, but needing a C3 and C4. Last night I ordered HQ's SRK driver to do this.
The MT03 has a copper MCPCB where each LED has it's own pair of wires - perfect for parallel operation, which is how the stock driver is configured. Actually the stock driver has 3 separate output circuits (inductors, etc.).
I already upgraded the MT07 with a XHP70.2 and have 3 more for the MT03, and I got 2 spare HQ drivers, earlier v3.2. The challenge wit the XHP70.2's is taming the amps - insane #'s of 17-18A.
One of those low-power SRK boards above is built and running, good fit in the new-generation SRK clones.
Also finally ordered some board sets for the K2 lights. Got one running and it is also a good mechanical fit. In the end I went with 5+1 7135, as opposed to FET+7135. So only 2.1 A, but good for the purpose (around the house, emergency tail-standing light). It is running Tom’s latest NarsilM, with ramping table adjusted for the 2.1 A range.
Interesting note on the board: First time OSHpark gave me full copper flood right up to the edge.
Thats really nice work DEL.
So without the Fet you have to modify the firmware so the driver will work?
Thanks MRsDNF! It is still a 2-channel driver, so Narsil works out-of-the-box. But to get a smooth ramp the ramp values need to be recalculated to suit the 2.1 A range, rather than the expected 10+ amp range (I used the latest NarsilM, configured by default for the Q8). Otherwise there is a noticeable pause/plateau around where the higher-power channel kicks in.
The K2 has two indicator LEDs behind the button, I still need to add firmware support for the second (red) LED.
Thanks for the explanation DEL. Tom will tell you and I agree with him that I’m totally clueless on how this all works. :person_facepalming:
Just ordered the K2 board sets - nice DEL! Over the weekend I actually was able to crack open the thread lock on the K2 tailcaps - all 3 I have 
. Also got the bezels off of a NiteCore MH20 and Astrolux S41 open.
DEL - finally assembled an HQ BLF SRK v3.4 driver with your C3 and C4 mods. Goin in a triple XHP70.2 Haikelite MT03. Hopefully will start testing soon.
Edit/Results: it worked for a little while, while ON though, noticed something like a flicker. From that point on, it gets stuck at about 100 lumens, not quite a full 7135, but suspect the 7135 got blown.
I'm using the Hyperion FET, this one: https://www.arrow.com/en/products/bsc009ne2ls5atma1/infineon-technologies-ag. It's the first time I've tried it, was hoping it's as good as the SIR404DP. I am using the raptor claw 7135, and I used 10 uF caps for C1 and C2, the good LDO, and used a value if 4.7 uF for C3, and 0.1 uF for C4. C3 is between Batt+ and ground, before the R5. C4 is situated between LED- and ground.
I think we were hoping the C3 and C4 mods would prevent blowing the 7135 by reducing spikes, but maybe the FET here is the unknown. Maybe heat is a concern.
So, this has 3 XHP70.2's. When it blew, I was using fully charged VTC5A's (best high amp 18650 cell there is), and read just over 16,000 lumens but have no clue what the amps were. Since I know ~18A for one XHP70.2 ia bout 9,000 lumens @start, 16,000 lumens is "only" 5,350 lumens per LED.
I'm using fairly long 20 AWG wires to each LED, no psinrg bypasses but the driver spring is doubled (stock springs), and the battery tube has thick springs, but no bypasses.
Hi Tom, will have to catch up a little with this. Your part selections seem good though.
The 7135 seems to do OK with the inductive spike abuse in 1 S, so with the addition of C3 and C4 I would still expect them to survive in 2 S.
The long wires for sure does not help. And then there is just the sheer power (amps) in these builds.
add C3, C4. Is this right? For 2S
Steel_1024: Not my board, but from the photo C4 looks good. C3 does not.
Tom, for the lone 7135: Was wondering if the 7135 is more vulnerable when its gate is high. You might want to try keeping the 7135 off while the FET is working.