Talk about future projects and donation topic

The answer is simple:
Vista.

But the long story is that I was a huge Windows fan. I invested a lot of time into learning its internals.
Then Vista came.
I didn’t have the same complaints the most had.
I complained about not being able to load unsigned drivers of my own.
I complained about the existence of privileged processes. They are ones that can do anything with the machine, but even admin can’t touch them (unless they hack system protections, which isn’t hard really). The mechanism was created to facilitate Digital Restrictions Management. Who is the boss of my PC, me or media companies? For MS it was the latter.
I complained about changes to the UI that I liked.
I complained about dropped backwards-compatibility.

Then Windows 7 came. Technically, it was Vista SP1. It only fixed the most glaring problems people had. But after Vista, they loved it.
I didn’t. It didn’t fix any of the most important problems that I had. And I knew that no future version would.
Took me a year to decide that I couldn’t use XP forever.
I left the Windows world and I’m glad that I did.

I tried scaling my technique to use 0.1 mm / pixel. When drawing something simpler than lights, furniture. I couldn’t make it work efficiently.

Anyway, I use what I use. I’m not changing it immediately. I may learn CAD one day, but that’s not on the horizon yet.

WOW!
Looks like I’m looking in the mirror.
But I didn’t try the 7, I left before it arrived.

I have wanted to move to linux full time for years but sadly I need adobe, DXO, Sony vegas, Solidworks and a few other windows only programs and have not been able to figure out a way to use them on linux so far.

Although with the new system I am putting together now I am thinking about running linux and using a virtual machine for these programs since I should have the horsepower for it. Anyone tried this?

Here comes A3.
This light is based on A7. Very similar, yet quite different.

  • hot rod, but not as hot as A7 or D4
  • 3x dedomed LH351D
    • optionally: 3x Osram Q8WP for higher throw and higher output, but much lower efficiency
  • 21700
    • 4000+ lm OTF, (uncertain) 60+ kcd with fresh Samsung 30T
    • 3300+ lm OTF, (uncertain) 50+ kcd with fresh Samsung 48G
  • fully regulated to whatever max we choose?
    • driver would possibly overheat…if not for the LEDs overheating before that
  • Khatod PL60006 lens
    • Though Chun Kuang M312 lenses might be even better
    • So could be Ilenstech ILENS61-S34
  • hopefully, very EDCy beam profile
    • very roughly 15 cd/lm
    • bright, but not overpowering hotspot gradually weakening to flood
    • no tint shift
  • powerbank with magnetic connector
  • good cooling
    • very wide thermal paths (as wide as that of Q8)
    • unibody
    • many fins, as deep as possible without interfering with thermal path width
  • my dream EDC

Here it is (top) pictured next to A7 (middle) and a rough outline of D4 (bottom)

My work PC is Linux + VM. I need the VM to run MS Office. On a few occasions I needed some other software. This model works for me.

What VM do you use? Is it stable?

I really would like to put my Mom on a virtual machine as well, she would fall in love with the ability to save the machine state so she could work on separate projects and save them for later with all the needed documents already open. Same for me really.

I have been using virtual box but it seems to lack directX and does not allow some programs to install properly. I have heard VMware is better but not tried it personally myself.

Edit: I started a separate thread for linux / virtual machine talk. Building a New-to-me computer & want to move to Linux + virtual machine, any experiances?

OK, I decided to discuss some of the design decisions. I recognise that some that I mentioned here are totally uncontroversial and that others lack that luck. But they all have some tradeoffs involved and I wanted to bring them closer.

1. Triple

  • less output than quad
  • more peak cd than quad (though if the leds were driven to the same level they would be about equal)
  • with linear driver, marginally less sustained lm
  • more sustained cd
  • quad optics could be lower
  • lower cost
  • less board space taken
  • OK, I just wanted something throwier than a quad

2. Khatod TIR lens

  • unknown efficiency, probably lower than that of reflector
    • seeing the CRXed cups , I suspect that efficiency is quite low
  • it nevertheless gives much higher cd/lm than f.e. Ledil Cute
  • it is very low for the diameter, reducing the light’s size
  • nice beam profile
  • available in pebbled and frosty variants too
  • alternative: Chun Kuang M312
    • 0.9 mm shorter
    • smaller cups
    • less throw
    • not CRXed - better efficiency?
    • clear, pebbled, frosty - whatever
  • alternative: Ilenstech ILENS61-S34
    • 1.3 mm shorter
    • 1 mm smaller diameter
    • siamesed a lot - good throw at the expense of beam shape?
    • not CRXed - better efficiency?
    • clear, pebbled, frosty - whatever

3. Without glass in the front

  • more prone to scratches
    • but AR coating scratches too
    • anyway, it is replaceable
  • less prone to breaking
  • 3% higher output
  • 1.5 mm shorter and over 4g lighter
  • does any flashlight come with antistatic and/or hydrophobic coating? This possibility is lost.

4. 21700

  • alternative: 26700
    • would be smaller than the head too, but thicker, heavier, worse for EDC
    • for max power, compatible with 21700
  • alternative: 26650
    • as thick as 26700, marginally lighter
    • no 21700 support
    • with DQG tail it’s not any shorter than 26700
  • alternative: 18650
    • less comfortable to hold
    • a little less size, weight
    • would enable 18350 support
      • 18350 unibody + 18650 extension?
      • resign from unibody?
    • much smaller capacity

4. Dedomed LH351D

  • fairly cheap
  • single chip (good beam shape)
  • non-flipped chip (no tint shift, good throw when flattened)
  • the most efficient non-flipped single chip that we know
    • though XHP35 deserves a mention as optically it works as if it was a single chip. And at high power levels it’s more efficient than LH351D.
  • nice tints
  • available in wide range of colour temps and CRIs
  • high Vf
    • lower peak output, especially with weak battery
    • fairly high efficiency with linear driver
      • better sustained performance
    • falls out of regulation quickly
      • but thermals make it drop output quickly too
    • since it’s a triple, Vf is actually not that high
  • fairly small footprint for the performance

5. DQG tail

  • light length changes with battery length - never longer than needed
  • large battery contact - minimal resistance
  • no spring and related resistance
  • ~2 mm shorter than PCB + spring, several grams lighter too
  • some are worried of battery safety
    • I’ve heard about some problems with Tinys, none was battery-related
    • TorchLAB ZeroRez is somewhat similar
  • battery rattles slightly when physically locked-out

6. Texas Commander / LD4 style driver

  • low height
  • little board space taken
  • fairly low cost
  • OK efficiency
  • unmatched regulated power for the price
  • unmatched regulated power for the size
  • very high unregulated power

7. Driver + LEDs on a single board

  • smaller!
  • short thermal path from LEDs to MCU
    • though some drivers shorten the path with extra thermometers on the MCPCB
  • some driver components need good cooling too
  • but other need to be rated for high-temperature work
  • requires costlier MCPCB
  • hard to swap LEDs
  • very hard to swap driver

8. Magnetic connector

  • Type C wouldn’t fit!

How do you imagine the battery positive contact? Has to be isolated from the body (=negative).

Too bad I don’t have a lot of time right now but I would love to work something out.

Driver plus LEDs on a single board isn’t hard. Some manufacturers do it to make the lights more compact and/or for other reasons. I’ve even made a few drivers with LED pads on board. Isolating stuff is just a matter of “drawing it that way” when designing the PCB. If your design is very complex or needs to cram a lot into a tiny space, you can get extra routing room by using four-layer boards instead of the normal two-layer boards used by most drivers. It does cost significantly more, though.

A hole through the middle of the shelf. Press-fit plastic ring with a large brass button in the middle.
Fat wire (AWG 16? Gross overkill but why not?) from the button to the positive pad on the driver.
That’s just the first thought. There are surely other ways.

Here comes A1, a thrower sized between C8 and Emisar D1.
The host has the same basic architecture as A3. And the same major features.
It’s just 1 LED and a reflector instead of 3 LEDs and a TIR.

Reflector has exactly the same dimensions as Phoenix Electroforms PA19.01, so for the big budget users there’s an easy upgrade path.
This part costs $45 when buying 1 piece or $35 when buying 100. So manufacturer could offer it together with UCL2 window as a $50 option. Expensive…but I’m sure there would be buyers willing to pay so much for ~15% higher performance.

I got troubles picking LED for it, there are quite a few good options.

1. Dedomed LH351D:

  • Cheap, especially considering that it’s a part of A3 / A7 lineup already
  • High CRI throwers is untapped niche. LH351D enables the use of this LED
  • Available in warm white
  • untested throw
    • probably OK, but worse than any other mentioned here

2. XHP35 HI

  • very high efficiency at medium-high currents
  • possible high-cri
  • possible warm white
  • expensive
  • requires expensive driver
    • and actually a different driver, adding R&D costs
  • at peak power it’s going to oveheat quickly

3. Dedomed SST-40

  • quite cheap
  • good throw
  • quite efficient
  • only cool white

4. Osram Q8WP

  • moderately priced
  • extreme throw
  • low efficiency
  • only cool white

My choice would be to offer both Q8WP and LH351D. Q8WP is a really great thrower LED. LH351D covers the cases where it can’t do for minimal additional cost to the manufacturer.
Doing some performance analysis…let’s drive Q8WP to 7.3A to get round 1500 LED lm (note: this is Turbo mode, the light won’t sustain it). According to Enderman’s calculator with the upgraded reflector and UCL2-class lens this light gets 190 kcd and 1410 OTF lm. OK, minimal amount of dust and regular manufacturing tolerances will reduce it. Is 150 kcd and 1300 lm conservative enough?
With cheap alu reflector and regular AR coated glass calculations show 170 kcd and 1290 lm. Because of machining defects, the difference between calculations and reality is likely to be bigger for this one.

Dedomed LH351D should offer very roughly the same lm and half cd.

BTW, that may make no sense…
but here’s A1W. It’s just A1 with addition of thin rubber bag around reflector. The bag is filled with paraffin wax.

Today I drew an optional clicky tail switch, here presented with A1 but compatible with the other A* variants:

Technically the switch construction is based on GFS16 - FET (a necessity with 30T driving 7 LEDs), power indicator, its own li-ion battery.
This is not merely a mechanical lockout - UI should support changing modes with the clicky too.
If UI designer finds it beneficial to do a different UI for dual switch - it’s doable too as the driver can detect whether it’s powered with the switch or not.
How?
When turning on, the switch makes a quick pulse of power before actual full-on. This pulse can be detected by the driver (though this feature may somewhat increase the price).

A remote switch based on this one could be available as well.
Note: doing e-switch which would work like a clicky would make it like 4 mm shorter. But I guess dual-switch people don’t really want it.

I did some performance calculations for A1.
And received a quote for the electroformed reflector.
I’ll start with the latter…$45 one piece, $35 when buying 100. So manufacturer could offer it together with UCL2 window as a $50 option. Expensive…but I’m sure there would be buyers willing to pay so much for ~15% higher performance.

As to performance…let’s drive Q8WP to 7.3A to get round 1500 LED lm (note: this is Turbo mode, the light won’t sustain it). According to Enderman’s calculator with the upgraded reflector and UCL2-class lens this light gets 190 kcd and 1410 OTF lm. OK, minimal amount of dust and regular manufacturing tolerances will reduce it. Is 150 kcd and 1300 lm conservative enough?
With cheap alu reflector and regular AR coated glass calculations show 170 kcd and 1290 lm. Because of tiny machining defects, the difference between calculations and reality is likely to be bigger for this one.

Dedomed LH351D should offer very roughly the same lm and half cd. Hmm…maybe that option doesn’t really make sense?

I drew revision 2 of A7.
I didn’t like how it lacked the powerbank function, but was not sure if all components would fit on the MCPCB.
I did better calculations now and they should. But just to get extra headroom I increased the MCPCB size slightly.

Features (mostly repeated):

  • 7-up TIR lens
    • either clear or pebbled (if there’s demand, frosted is not a problem either)
  • 7 LEDs, XP sized.
    • Samsung LH351D
    • dedomed Samsung LH351D
    • Cree XP-L2 (pebbled lens only)
      • 10 000 lumens for 5 seconds
  • 21700
  • magnetic charger / powerbank
  • Texas-Commander / LD4 style driver integrated with the LED MCPCB regulating the light to 10+ amps. With Turbo above that.
  • Unibody
  • E-switch, ramping UI
  • DQG style tail cap
    • optional clicky tail switch. Also DQG style. :wink:

Thanks Agro.

I’ve been thinking about it for a while…and now I decided this is the way to go for the A family.

Let’s not make them 21700. But rather 21350.
Pros:

  • smaller by 1/3

Cons:

  • there are no 21350 batteries

So…it needs a little explanation.

First, there are no 21350 batteries now. But there are 18350 and 20350 now. And 21350 are likely to arrive eventually.
Why 21350 and not 18350 or 20350 then? Because it enables us to use extension tubes for 21700. And 21500 if they eventually arrive (18500, 20500 is available now).
21700 have 4600 mAh.
18350 and 20350 have 1100 mAh. And higher internal resistance which reduces performance.

Quarter capacity for 1/3 smaller size. Is that a good trade-off?
It depends.
If you don’t need more than 1100 mAh anyway - yes, it is. And seeing how many people EDC lights smaller than 18650 - I believe there are quite a few who don’t. On a twist, capacity reduction turns it from good powerbank to bad powerbank. So for those who EDC powerbanks and tiny lights - they would still prefer 21700.

So the real pros:

  • smaller by 1/3 for those who don’t need more than 1100 mAh
  • (optionally) smaller by 1/5 for those who don’t need more than 2000 mAh
  • just as large for those who do need more energy

Cons:

  • slightly costlier
  • worse thermal transfer to the tail
  • stylistically challenged
  • when using smaller batteries - slight to moderate performance reduction (depending on model)

I added a rough outline of Emisar D1 to the pic.

Is it just me or has this thread seem to have turned more into “agro’s future projects” rather than “BLF future projects” ?

No, it has not.

It has not because I neither have the capabilities to turn these drawings into real lights nor will have them in the future.
So these are just drawings and will remain so unless someone different from Agro will decide to make them.

Seeing how little response they generate I conclude there’s little interest. And find it extremely saddening. Because few times in my adulthood I was as convinced I did something right as with them. And if there’s no interest, they will never be made.

That said, I almost monopolized this thread. Again. The reason is that there are apparently few people willing to discuss the topic.
Is it right? Is it wrong? I don’t know, it is what it is.
I am willing to discuss and I intend to continue. Because even if I’m mostly discussing with myself - I want this to be public.