Do you know what bins are tested in that spreadsheet, there is no mention, all of the CCTs come in 3 luminous flux bins, for example D180, D200, D220.
You can easily use Nichia datasheets basically, at least that is what I have done, and could easily figure out it cannot be anywhere over 2500lm using D220 (220lm-240lm) bin.
Usually datasheets are ignored even if they can prove very useful, especially for the E21 they are pretty good.
The rating is for 700mA . Then you can see what multiplier to use for that 700mA with the graph blow.
Multiple that 220lm by 0.65x and get a relative 143lm * 16pcs = 2288lm
Or the max rage of 240lm by 0.65x and get 156lm *16pcs = 2496lm
So D220 for the 4500K?
That is why I was calculating with D220 ( min 220lm and max 240lm at 700mA) for a 4500K.
5000K is D240 at IOS, so with that the results are 9% higher (increase from min 220lm to min 240lm at 700mA) for the minimum and 8.33% higher for the maximum (increase from max 240lm to max 260lm at 700mA)
It sounds to me like 1x7135 + 5A CC (+ FET?) driver might be a means to achieve this.
I don’t think Loneoceans’ lume1 inductor would fit in the shallow driver cavity of the D4, nor do I know how low that driver can go, but the buck-boost + FET approach seems appealing, too.
A 7135 chip does 350 mA. To make this work well, it would probably need to be significantly smaller… and ideally a nice power-of-two ratio to the higher power channel, like 5A * 1/64 = 78.125 mA.
It uses 10-bit PWM, or 1.95 lm per step. But 1.95 lm is pretty coarse resolution at the bottom end of the ramp… so the idea is to add a second power channel to fix that. It’d make better moon levels, and make the bottom of the ramp smoother instead of having visible steps.
If the second channel used a 7135 chip, it would effectively be an upgrade from 10 bits to ~11.5 bits. It’s not nothing, but it’s not a big upgrade. The effective resolution ends up at about 0.6 lm per step. And the math gets really awkward since it’s not very close to a power of two ratio between the power channels.
So I’m hoping for something smaller with a cleaner ratio.
If the second channel has a power-of-two ratio to the main channel, we can treat it like it’s just extra resolution on the main channel. Like, let’s say we run the main channel in 8-bit mode. It then has 256 steps, and each step is 7.8 lm. Then add a second channel with a 1/64th ratio. Now instead of 8 bits, we basically have 14 bits… and each step is 0.12 lm.
Or it could even go as far as 16 bits… 5A for the main channel and 20mA for the lower channel. This gives 16-bit resolution where each step is 0.03 lm. And it would still run at 16 kHz so there wouldn’t be any visible PWM.
Basically, if 7135 chips or other linear regulators come in a 20mA variety, that would probably be about right.
Output tested using Texas Ace Lumen Tube calibrated with Maukka lights at 2 seconds. Light quality measured using Sekonic C-800-U. The E21A 4500K is just as rosy as the 219B sw45K but with cooler tint. It is a very beautiful tint but I think mixing it with 8x4500K with 8x3500K will result in somewhere near 4300K, which would be more idea. Also the E21A 4500k does not reach 9080.
The E21A 2200K is much rosier than I expected. The 2200K batch that Clemence had was on the BBL. This batch is nicer IMO and exceeded my expectations. Because of the negative duv, it doesn't take long for your eyes to adapt because it doesn't look very yellow. The 2200K tint is similar to a pinkish sunset whereas the E21A 2000K that Clemence sells look like real candle light.
Also pay attention to how much output the bezel kills on the mule lights. Even the color of the bezel (stainless steel vs dark gray) makes a big difference in output. I think Hank should use white mcpcb for the mules to reflect as much light as possible to minimize output loss.
The D4v2 TI with the 16x E21A 2200K is the only light with a loud electronic whining noise on turbo and it smokes as if there's a fire burning. At first I thought it might be moisture that its burning but the smoke doesn't stop and it sure doesn't smell like water. I wonder what is the cause. I don't get this on the KR4 16x E21A 4500K mule or any of the other Hank lights. Anyone have this experience?
D4v2TI 16x E21A Mule Warm White Switch (Aux Light)
(disclaimer: I do not know electronics well so I could be wrong)
A very simple 20mA channel could be no more than a resistor inline with battery and led, switched by a tiny FET. The current (and so the light output levels) would not be truly constant but vary somewhat with battery voltage but maybe that is not such a big deal.
Thanks for all these measurements. The output loss due to the bezel is significant, I did find it weird that the LEDs were packed to the edge of the MCPCB , I think they should be centered with the wires and screws on the edge instead.
The color measurements are not integrated but at the center of the beam right ? E21As do have quite a bit of tint and CCT shift so that explains the low duv and higher CCT.
You’re not wrong. That is one way to do it… and it’s basically how the aux LEDs work. But as noted, the output is not constant; it varies with battery voltage. That’s not really an issue for aux LEDs, but it’s kind of a problem for the main LEDs.
I hope this image explains the idea, and why I want the 20mA channel to be regulated instead of just a resistor. It’s meant to smooth out the ramp between steps, but the ramp will only be smooth if the extra channel has a consistent slope.
Basically, the point is to make it look like the “4.0V” bottom center image… no matter what voltage the battery is.
It’s a different method than how FET+7135 or FET+5A works. With those, the lower power channel ramps up to full power and then stays on while the higher power channel ramps on top. But I’m thinking of doing a totally different method where the lowest power channel does miniature ramps between each step of the higher power channel.
I’m not sure if it’s a good idea, because it’s more difficult to do correctly… but it sure would be nice to have a truly smooth and stable ramp all the way down to firefly levels.
Also not sure if I should suggest it, because it means I’d have to rewrite a bunch of code. Would be nice once it’s done though.
Ok, I think I understand what you’re suggesting - the low channel tops out around the lowest stable level of the 5A channel.
I was thinking of using a 1x7135 channel full range, then either switching on the 5A channel in parallel as is done in 1+N x 7135 drivers. Alternately, if the different driver types don’t play well in parallel, then once the first channel reaches 350mA, the next step turns off the 7135 and turns on the 5A driver at the next logical step above 350mA.
But I hadn’t thought through what that would imply for the ramp resolution or scaling.
SKV89, thank you for sharing all those measurements.
I was under the assumption that 16x E21A mules would be limited to 7.5A CC only, and there would be no direct drive, because it isn’t safe. I am surprised that direct drive is enabled. Are you sure it is safe to use Molicel P26A for this?
I cleaned the contacts, and actually very lightly sanded them to make sure there’s nothing blocking the connection surface, but the flicker is still there. Now, when I say flicker, it’s very subtle. It’s only really noticeable when you’re not moving the beam.
I tried varying the floor from 1/150 to 3/150, and the flicker is evident on 1/150 and 2/150, as expected.
When moving it to 3/150, though, it’s as if the very low level flicker from 1/150 and 2/150 is still there, but is masked by the higher brightness of 3/150. I hope that makes sense.
It’s tolerable, I guess, but it’s also annoying when I notice it, and just knowing that one of my favorite lights isn’t stable at moonlight drives me mildly nuts since I use it every night.
Do you think there’s anything on the code side that could be done to eliminate this? I’m not familiar with the coding at a hardware level.
Thanks, again.
Edit: Would increasing the clock speed at level 3/150 make a difference? Or, maybe just disabling dynamic underclocking would be a reasonable test.