Ok comfychair, another guess with whats wrong. You’ve turned this thread into a lesson in Chinese as I dont understand a thing your talking about. I do love your soldering ability though as i can only dream that my soldering is that good. Even though I dont understand I’m liking what your doing.
LOL, any of the soldering that looks nice only does so because it was done with hot air and lots of flux, which makes for really pretty flowing solder. I don't even have an iron with enough torque to solder that crimped lug for the negative leads, had to use the butane torch.
aieeeee, halp! an octopus!
pft, anybody can see that’s a pentapus! Why didn’t you just solder the LEDs together in parallel then solder some 14g solid copper wire from the LEDs to the driver? It’s how I’ve seen parallel setups done before and it’s a lot easier 
The pentapus wiring will balance the emitters better, should work OK
I’m curious to know how much total current you get with this setup.
Well, that's basically what I ended up with as of now. I would make it just straight up direct drive, but I can't figure a way to do it and still switch modes and/or use the momentary switch. So it's got to keep the FETs, at least.
Adding the stacked FETs did something somehow that disabled the low mode, so I took those off, and that part is back to normal. After I corrected the mistake with the PWM connection to the wrong pin and retested, it did go back to using the resistor pack to control the current. But even with the 3 000s in place output was only around 780mA (measured inline with one LED, other 4 LEDs connected as normal). So replaced the resistors with a 1/4" wide copper bus bar. Running 1.1 amps to each LED. Jumpering LED- to ground to bypass any and all parts of the regulation only increases it to 1.3A, so it's pretty damn close to being direct drive.
And I think it might put out less heat if it were actually on fire. I think I have a sunburn on my arm.
Oh, and the 4p 18650 battery box I'm using for bench testing only has 22AWG wire about 8" long, and it gets HOT. So it'll do better (worse?) after it's all installed in the light.
I can’t see how the wiring set up will affect the relative draw of each LED in a parallel set up - I thought that was more a function of individual temperature and Vf differences? If anything, having long lengths of (somewhat) resistive wire going to each LED would add another variable (voltage drop along the wire) to the mix.
I was referring to variations in path resistance. If all the wires go to the same point, are the same length, and same size, then the current variation is less affected by it. Wiring them in a parallel daisy-chain fashion can mean some have longer wire paths.
Regarding emitter Vf, it does have a big affect, but that is something we cannot control.
I think the 5x design would benefit from a 5x driver but comfy is going for a just make it work approach. Putting five 105c drivers (one master, four slaves) in here would add to the cost.
Actually I've experimented with using 7135s as slaves a few times and they never completely turn off the LEDs when signal on the Vdd is removed but input voltage is still present (as it is with a momentary switch). Is there some trick for that I'm missing?
These are Panasonic NCR 2900s, they should be able to deliver way more than 1.3 x 5 = 6.5A, 1.6A per cell... the skinny wires from the battery box is the only thing left that could be limiting it. I'll rewire the box tonight and see if it changes the non-direct drive numbers.
I think it’s 6 of one and 1/2 a dozen of the other to be honest - I’d go with whatever’s easiest to wire up.
This has been sitting in the 'think about it for a while' pile, but I finally got motivated to polish the reflector. The baseplate is joined to the threaded base, and machined for clearance (so the baseplate contacts the step in the head for best heat transfer before the threaded ring bottoms out). All that's left to do is drill the wire passthrough holes in the baseplate, and JBWeld the sinkpads into place and give that time to cure (wish I didn't have to, but I've put it together once without the sinkpads fixed in place and keeping them all aligned while placing the reflector requires more hands than I have access to and I really don't care to try it again because last time it took about an hour, maybe longer). Then it's just soldering all the wires and basic reassembly.
Its sounding good comfychair. Where still watching.
Looks good! I have a feeling that will be the brightest SRK seen on this forum…
Comfychair, you should join in on the “build from scratch” contest.
Noooo, my brain doesn't work like that. I'm really really good at picking apart an existing design in minute excruciating detail and spotting areas for improvement. But starting from a blank page my brain usually stays as blank as the piece of paper.
-----
I'm not sure why, but I'm only reading around 5.8A at the tail, using 4 fully charged Pana 2900s. I don't know if the current's being limited by voltage sag butting up against the LEDs' Vfs, or if it's the driver itself doing the limiting. I would have to test it running from a power supply at a constant 4.5v or so to pin down which it is.
Did you add copper braid to the tail springs?
I was too lazy to read back and check 
For the tailcap test I used 3 strips of copper in an 'H' pattern to get good contact with all 4 cells, so the rear plate wasn't in the picture. Even so, I'm now making a new rear plate out of .040" copper. No reason to use a silly PCB there with all of them connecting to the same tube.
Front face of the reflector is a matte finish, wet buffed with green scotchbrite.
The poor thing is nearly useless outdoors, between the way it makes all the air in the surrounding area glow and the rather disturbing amount of heat after 2-3 minutes on high. It sure feels like it's running at more than 6 amps.