At least Chinese roulette is less dangerous than Russian roulette.
I have had to drill out broken off bolts, in the past. It starts with centering a very small drill bit as close as possible, followed by sucessively larger drill bits, up to just shy of the actual thread diameter. I’ve taken it so close to the threads (from the inside) that what remained was nothing more than the bolt theads themselves. Only then has most “easyouts” ever worked, for me. I’ve even peeled the remaining bolt threads out with needle nose pliers before. It’s a pain in the axx but with patience it’s do-able.
I had the same problem.
Drilled it with 5mm drill bit. Just drill it until head goes off.
Then you can remove reflector ant it should be easy to remove whats left of screw as it is long enough.
Funny thing is I removed remaining part of the screw with my fingers, looks like it was stuck only on the head part.
Ahh - ok. Will give that a go. I was concerned where the binding is - against the MCPCB or did it bottom out the hole in the reflector and bind there. I had 2 other reflector screws mangled and was able to get them out way before it went this far. I'll try that though this eve.
I believe mine was stuck at the MCPCB as I removed whats left of screw with fingers from reflector.
Even if it’s stuck in reflector you should have no problem removing it once reflector is out because you can use pliers easy and screw should be about 4-5mm out from reflector.
Just be careful when drilling and drill only screw head, not MCPCB 
Tom, adding e.g. 100uF does not cause a “spike” “every turn on”. Lets knock this on the head now please.
There will be one brief surge when cells are first fitted. Limited by R6, 4.99 Ohms to a peak of 840mA, over in less than half a millisecond.
There is no such surge thereafter, when the torch is operated using the E-switch.
DEL is confident that it is within the surge rating of this type of Schottky Diode, as am I.
Flashy Mike has even been using double this value of capacitance, with no limiting resistance, in clicky switch torches where the (un-limited) surge happens every time the switch is pressed, with no failures.
It is a non-issue. I raised it as a minor concern initially, and was completely satisfied by DEL and Flashy Mike’s feedback. Consider it put to bed.
Now the question of whether 100uF is the “best” value. I have asked someone privately for their knowledge of actual MCU current drain so that I (for my personal interest) can estimate the lower range of a suitable values, so far without reply. Toykeeper has mentioned figures of e.g 3-4mA seen in other applications.
Once I get my own Q8 I will measure it. Meanwhile perhaps you could. Just disconnect the main LEDs and measure the current taken by the driver when turned on, operating in a mode where the FET is being driven with PWM. If necessary use a small LED with high series resistor instead of the main LEDs, as an indicator, so you can see what mode it is in.
I very much doubt that an expensive capacitor will be necessary in this application, probably almost anything will do for hold-up purposes, maybe already on hand, or even something salvaged from some electronic junk.
Turn on meant power turn on, 1A spike is what DEL told me - I passed along what DEL emailed me, but you know you know better, so I'm done here.
Don’t remove the original capacitor, it is a critical component for “decoupling” the MCU, it’s short direct connection to the MCU power pins, and it’s high frequency dielectric properties are very important for this.
The added “holdup” or “bulk-decoupler” should be connected in parallel, the method is probably not critical, a small leaded component may even be easier for those without fine SMD soldering skills.
As long as the positive connection (only important for a polarised cap. e.g. electrolytic or tantalum.) is to D3/C2 junction, and the negative is to ground by shortest route, say C1/C2 junction.
Can you show a picture head-on of the location to be modified? Or a drawing?
“for those without fine SMD soldering skills”
WILCO once I have mine, and am more confident that component selection or layout is not critical to results. Not anticipating a surprise, but one never knows until tried ones-self.
Even perfect program verification can only establish that a program meets its specification. […] Much of the essence of building a program is in fact the debugging of the specification.
Fred Brooks (1986)
Thanks for your reply Tom Tom. Is this a suitable capacitor? http://www.ebay.com/itm/10pcs-1206-107Z-100uF-16V-chip-SMD-Ceramic-capacitor-/292207189605?epid=2163346082&hash=item4408e7de65:g:AZEAAOSwcB5ZHHEp
I’m not too sure where to even begin with the size and the voltage but the picture looks similar If I order the capacitors now it’ll hopefully give enough time for others to take some additional pictures so it’s easier to follow. I wouldn’t want to fry it accidentally by shorting it out. Thanks for everyone’s help!
> bicycling
Bumpy ground causes jerks forward/back as well as up/down
and you certainly want the light to stay on if you crash your bike in the dark and get thrown over the handlebars, for example.
Stuff happens.
https://www.google.com/search?q=define+jerk+physics
https://en.wikipedia.org/wiki/Jerk\_(physics):
In physics, jerk, also known as jolt, surge, or lurch, is the rate of change of acceleration; that is, the derivative of acceleration with respect to time, and as such the second derivative of velocity, or the third derivative of position.
You see the cap beside the 3 wires. I removed the protection over the ground trace of the board and soldered the cap vertically with one end to ground. Then connected the other end of the cap with the “meeting point” of C2 and D3.
Here is a schematic of the driver:
On the face of it it, it looks a suitable physical size and value. 1206 is bigger than 0805 (but the minimum for this sort of capacitance value I think).
It won’t just stack on top. e.g. Flashy Mike showed his stacked on one end, and a fine wire link to complete the connection. Fine SMD soldering skills.
However I don’t recognise that dielectric (107Z ?), which might be sub-optimal for stability. But we don’t yet know the minimum that will be a solid mod. It could be a lot less.
Suck it and see is where we are at the moment, I think FM is still the only pioneer yet.
I would give it a go, knowing too much I can anticipate all sorts of esoteric issues that are probably irrelevant. If it isn’t perfect, stack on a second one. For $1.50 you have ten to play with 
A named class 2 dielectric like X5R or X7R would give me more confidence.
Holy simultaneous post
Edit: You could even stack it on one of the un-used LED- pads and wire it over to D3/C2 junction as a test, but that might not be best, depending on how the ground plane is laid out.
I don’t think so. LED- is not connected to ground, it’s switched by FET and 7135.
You are quite correct. Sorry for stupidity. Is there any other large Gnd pad exposed, e.g the ones around the 7135? or is scratching off solder mask essential ?
There is a pad directly at the 7135 which appears to be connected to GND, don’t remember why I haven’t used it.
That’ll do