I agree thefreeman, you’re an inspiration to the community with your persistence and attention to detail.
I’m jealous of the KR1/B35/Olga/driver setup.
Have you tested the spectrum of these B35s? What’s the Duv and Rg if so?
EDIT: Also curious what kind of intensity you are seeing from the setup. About 20kcd is my estimate… close?
Thanks, I have learned a lot of things in the process.
There is one thing I will change that is increasing the 0603 footprints lenght, current ones are a little short and I think this is going to be less reliable (shock resistance).
Edit : forgot the question… most high power boost converters use high side NFETs (with a bootstrap capacitor to get the required VGS). Check the diagram Page 16
But regardless, I don’t think it would works if the PFET or NFET was reversed, there would be a reverse current going through its body diode and the low side FET when the later is ON.
Here is a TM-30 report of 3500K No1 with an Olga RS :
Tint shift vs current :
Duv binning at 5mA :
The 4000K is the one in the KR1, unfortunately the AR coated lens adds +0.0015, (bad) AR coating is a blight for us tint snobs.
I’ll come back with an intensity measurement (need to check if the sitting height of the optic affects it)
This reminds me. We need to get on hank to start using a green reflecting AR coating. Going to comment in that thread. Hopefully others pile on and he gets the idea how important this is to so many of us.
And that 3500K report looks stellar! 102 and –0.0030 are just about perfect IMO. R9 of 84 isn’t bad either.
I don’t think he replied to messages me and Lux-Perpetua made about it unfortunately, but more support should help.
R9 is 84 but it’s slightly oversaturated (1% chroma shift on hue bin 1) anyway so it’s better than 100 IMO.
This shows the limitations of CRI, we can see it on the CIElab plot but not in the CRI numbers :
Hank can be quite stubborn. Hopefully his recent engagement with BLF is continued, but I’m not holding my breath. I left a comment just now…
As for the limitations of R9 and CRI, I am on the same page. The values I care about most are Rg and Duv. Very close second is Rf(skin). R9 is a ballpark figure, and about 75 is good in my book… Beyond that, I have to see the details like you point out here. Hue shift to a certain degree doesn’t bother me and quite frankly is tough to pin down what is “accurate” in real life, since there is no light source that is “perfect” and consistent to compare against. Even sunlight, by the time it’s filtered, reflected, and refracted through windows and of walls and furniture is not really much of a ideal reference anymore. Chroma however is very noticeable and I prefer a 2-3% oversaturation - and preferably most in the reds and blues, with neutral greens. This emitter is knocking most of this out of the park! I do prefer about 500-1000K cooler though for an EDC, but this seems like a splendid option for the more relaxed evenings and outdoor recreation.
That’s what I mean yes. That said I tried removing the retaining ring but haven’t had any success yet…
Post 163 - so this is what it looks like when enthusiasts flashlight history is being made.. nice! I must be feeling kind of what everyone was feeling a few years back when lexel was doing his thing too?
this is AWESOME. I cannot contribute anything but that. Thank you Freeman!
Do you mean that the resistor is underneath the button with the button led, rather than on the driver board? If this is the case, rather than trying to replace the resistor, I might ask Hank to use a bigger resistor in my next light. That would simplify things. Another idea is getting it with a deep red led. The one I have now is green, in retrospect not a great choice on my part.
I installed it in the D4v2 :
I cut a few traces and added a bridge to wire the MCPCB in 6V, I used sliced 2700K and 5000K (or 5700, not sure) LH351Ds :
I used thinner wires for the aux LEDs, I don’t know how they manage to fit such big wires originally.
They’re only for 110°C max though, probably OK but not for too much soldering, I think I’ll try to use enamelled wire like TTerev3 is using next time.
Quick video to show the ramping
Awesome!
Any idea where to get these stranded enameled wires commonly found in headphone cables or how they are called?
Freeman, you are new Lexel. That’s amazing!
TTerev uses magnet wire, which is a solid wire, but very thin, 34AWG IIRC, I think that’s better than stranded, I have some thin PTFE solid wire (32 or 34AWG I think) and the fact that it keeps the bend is IMO much better than the springy thin stranded silicone wire, the problem with PTFE is that the insulation is annoying to remove (and it can slide out of the whole wire).
Yesterday I reflowed a buck driver and tested it today :
6A, 1S to 1S, same 18mm clearance diameter and HDR of course, this one is 21.2mm for a H04.
It works very well and the efficiency is also very good for this small IC :
This is after I swapped the 0.9uH (8.8mΩ) inductor for a 0.47uH (4.6mΩ) one, but the efficiency was slightly higher with the 0.9uH at the two points I measured before swapping :
4Vin 1Aout : 97.2% vs 96.4%
4Vin 3Aout : 95.9% vs 95.5%
This is also with a quad SST-20 meaning a very low Vf and with a buck converter the higher the Voltage difference between the input and output the lower the efficiency.
The total resistance of the driver is ~37mΩ in dropout (100% duty cycle), not quite as low as a FET direct drive but still not too far from it.
This looks amazing! What is the value of the lowest current you tested? Is there a current level at which a limiting resistor (fixed or pwm) becomes more efficient? Wherever this crossover point is, is this above or below the floor of a “single stage” (non HDR) buck driver ?
40mA, take a look at the datasheet page 16 (PSM), the efficiency doesn’t decrease a lot at low current, here it decreases more due to the MCU’s consumption.
Thanks. I guess I’m not familiar with the typical range the MCU consumes. I’m thinking about ‘survival’ or ‘vampire’ types of runtime on super low outputs and if you could get better runtime using a bank of tinnyy fets and resistors and throw the mcu in a low power mode to stretch runtimes. Just thinking out loud. Sorry
I measured 2.4mA at VCC=2.8V and 10Mhz for the MCU alone, at moonlight levels the power consumption is mostly the MCU.
So yeah putting it into ULP would help, but then you have one output level per FET resistor channel, and if you put several you need a bunch of traces to route, something I don’t like very much 
I think two/three well selected firefly moonlight levels would be good enough. With two channels, you get three modes 10 01 and 11. This would make the hdr dual sense circuit redundant. You’d be giving up ramping ultra lows for possibly drastic extensions in runtime.
Oh right of course :person_facepalming:
But the ultra long runtime can be achieved with the buck converter, Zebralight does it and the Attiny1616 is capable of it by using the TCD for running the PWM in ULP, or the DAC, but I can’t implement that in Anduril, this is a job for the software guys/girls.
That way there is no need to give up the ramping and especially the firmware adjustable levels, personnaly I don’t want to have to swap resistors to change a level.
At the present time though I can increase the efficiency a bit by lowering the frequency to 2.5MHz (1.1mA) for low modes (dynamic underclocking), I need to check how much ripple the lower PWM frequency (2.5kHz) produces with the RC values used.