L2 with dedomed XPL V6 1A can reach 280Kcd. Based on my measurements, the XHP35 HI could reach ~93% of that. If the dome slice is not very close, that could take away an additional 20% to make ~208Kcd.
I built an L6 with de-domed XP-G2 S2 0D and a Buck driver that makes 439Kcd. The driver is Richards 5.5A 27.5mm. Pretty nice in the L6.
I didn’t try to get the slice super close, like to protect the bond wires a bit ya know. 
Nice testing, I have an XHP35 on order to test myself. I want to see if you can drive it with a PWM from 4S cells and how that compares to a constant current in output.
The Texas buck driver is nearly done but frankly it is simply going to be too expensive for all but the most dedicated setups. So back to using a liner driver. Maybe have to rethink the whole op-amp thing.
I got the opportunity recently to mod a Lummi Short Orb. The original driver was non-functional and, if still available (which it isn’t) would be $50 for the driver itself. This little Lummi now has a 12mm FET driver with ramping firmware making from 10 to 1911 lumens. 
Some “proprietary” drivers get quite expensive!
Yeah, I am glad I decided to go ahead with the Texas Buck but it is looking like it will be $15 in parts alone all said and done. Not out of this world but 2-3X more expensive then most other drivers we use.
There is the possibility that price can be reduced some but not drastically while keeping the high current ability.
So that comes back to the Texas Avenger or an Op-Amp setup.
Texas_Ace, what is this op-amp setup you speak of?
Regarding your planned efficiency testing, you might check out these simple measurements I did comparing direct drive PWM to linear regulation.
Yeah, those are what made me want to do some testing of my own. Not that I don’t trust yours, I just like to do things for myself.
An op-amp is what the LD-2 uses, I have not wanted to really mess with it as I was hoping that a buck driver could be done for a reasonable price but that is not looking likely. Plus LED4power sells them and I didn’t want to hurt his sales with an open source version. That said if I was to use a driver like this myself I would want it to be open source.
Isn’t the LD-2 like $17 each? I’ve never bought one because of the cost and MOQ. If your high quality buck driver would be similar cost I’d rather have the buck (especially since I could use other FW’s). It would still need to be absolutely necessary to make the cost worth it, but I think the XHP35 is a special case.
Was thinking the same thing. Richards Buck driver is over $20. It just is what it is, component cost and complexity with dependability and reliability, well, you don’t shortcut when you’re counting on something to work, right?
Agreed. Most of my builds I do on the cheap because I want big numbers and wow factor, but then it sits on the shelf 99% of the time. For those I probably wouldn’t spend $15-16 on a driver.
If I had a scenario where the light was used often and reliability and efficiency were a must, $20 for a good buck that I can program myself isn’t a huge deal.
Yeah, thats the same reason I never ordered any LD-2’s. If an open source version was made it should be only slightly more then our normal FET+2 setups (actually could be the same price since we don’t need the 7135’s). Plus it would be able to run Bistro / Narsil.
The buck driver right now is only in SRK size, a smaller version is possible but not sure it would be worth it when the mtnmax is already open source and works fine. It would not be any cheaper, just a bit more updated with the latest components.
The op-amp is a good idea, the only issue with it is the heat it has to dissipate. Although if you combine the current adjustment with PWM you should be able to overcome the heat issue while still allowing it to limit current. The highest heat is generated in the mid range modes.
The integral is fine of course. The "problem" (not a big one) is that it assumes the LED is a Lambertian emitter. Just looking at typical output curves that certainly looks pretty good for HI by eye, and a bit off for HD. I checked one angle, cos 50 = 64%, looks dead on to within what I can read on their graph for the Hi version, which is what I'd kind of expect. (can use cos(60)=0.5 but there's no line at 60, still looks dead on though. There is free software that can "steal" numbers off graphs very precisely actually) The HD versions seems to be about 10% too high relative to lambertian for middle angles at least, so you'd understimate total lumens for an HD I think.
But if you really want to be accurate, that's why I said just normalize to CREE's own lumen numbers at lower output. I guess the angle profile doesn't change that much with power. Of course what you're really doing then is trusting cree to measure the absolute light output and you're just measuring the power curve and extending it out, but it's a good cross check at least, and really, I'd probably trust cree. Should I not? An intermediate way is to just trust their angle profile, use software like I mentioned and integrate it numerically (don't forget your sin(theta) factor). Of course that's more work, and you're probably within 10%, maybe quite a bit better.
I think I bought 4 of the first LD-1 run, then I bought 10 of the LD-2’s. I’m enjoying them and have 2 running the XHP-35 to good levels. There was some TINY schtuff to solder on that mod but in the end they work so it’s all good. I think I have one left. I’ve got LD-2’s in an X6, an Fenix TK22, and several other lights that I use seldom and forget what driver is in em, always get a bit flustered at 4 modes when I’m used to 7! lol
Same here PD, way too many hot rods on the shelves above me, just sitting there with charged cells waiting to go out and play but gathering dust all the same. With some 170 lights, I have a few favorites and rarely use the others. Need to sell a hundred and then some, let someone else have some fun I guess.
The MTN max buck is very different from the new buck. It converts down to a FIXED voltage (or current, I forgot which, probably voltage, or it wouldn't need those big output swings to do histeretic control).. and then it's as if you have a (very noisy, although non-sagging) battery at that voltage. The rest is done with PWM with its usual inefficiencies. That's just all that IC is capable of. So yeah, updated compontents I suppose coverers that, but one is a voltage converter with a PWM driver. The other is a buck driver. The voltage converter still has the advantage over a battery of not sagging, which can be a big deal, but is surely much less efficient than a current controlled buck driver like we're working on. To me this pwm "DAC" hack to the iadj pin and the reduced ripple voltage, are the two selling points.
To get again another generational leap, it would need to be synchronous and have a sense current amplifier, all more money, and probably won't fit in the small ones.
Wish I understood this stuff better.
Basically there are lots of ways to skin a cat.
But cooking it up so everybody likes the taste is another matter altogether…
That thought might gross some folks out, so let me say this. In the old days of this country the mountain men were hunters and trappers, had their choice of virtually every animal to eat. I’ve read in quite a few different places that the wiliest old mountain men preferred Cougar meat over all other.
I don't think that's a fair summary. It makes it seem as if it all gets to the same endpoint and that's not true.
DB custom I'll try to be a little more clear. The mtn-max basically produces one single output current (just checked). So imagine it makes 6A (you can buy for whatever you want). Then it uses a standard direct drive style control to PWM that, sending pulsed current to the LED for reduced modes. But the LED always sees 6A or nothing, just in pulses. an XML is brightest at 6A but it's not efficient at 6A, pulsed or not, it's still 6A and that's not efficient.
What you want is a buck where you can actually dynamically change the current setpoint. The LM3409 can do that but it requires an analog voltage input. What I proposed and we're doing, and I don't know maybe was already in the design too, certainly TA also always thought it should work that way, so I might not have added much except getting the cap value right for it and nudging him to get back on it, but anyway.. what this setup does is use PWM from the mcu and filter it through a cap to make a steady controlled voltage to control the IC current setpoint with. The buck should deliver a steady current for your selected mode (with a very low ripple voltage too at least in good circumstances).
That means in low mode, it's only pushing low currents through the LED, not high currents like the MTN MAX. For the very lowest modes it will probably still need some PWM to help it, but it will already be efficient at that point.
Well they do all reach the same end point, they all light up an LED without burning it up.
The same way you can deep fry, roast, broil, grill, smoke, boil a cougar to make edible meat.
Sure they will not taste remotely the same but they are still cooked meat in the end.
Most people would never know the difference between any of the drivers we are discussing from a purely technical standpoint. The only thing some might notice is the tint shift on the non-PWM versions. By eye you would not notice the difference in output with any of them.