Get ya a can of that expanding foam, blow it between the layers and let it expand in to fill. 
Today I did some warm LED swaps.
One was an Jaxman E2L that I took out my shaved SST-20 4000K and put in SST-20 3000K 95CRI, looked like too much star burst pattern with smooth narrow optic so I went with frosted narrow and it seems much nicer. I liked the tint better of the 4K shaved one LEDs but it was just too much current/heat for the output I was getting and the beam shape wasn’t great either. I will probably put one of those in a small reflector light or something (BLF 348?).
On to the more interesting mod. My first LED swap which didn’t require solder. In fact I only used a screwdriver, wire cutters, and tweezers:
I recently ordered some batteries from Shockli’s aliexpress store and they threw in some freebies with my order. One was this cheapo keychain light. It has a 5mm straw hat LED that is press to activate and has a tiny switch to “lock” it on.
I bent and trimmed the legs of my Yuji 95 CRI 3200K LED
We have light!
I just now am realizing it had 2 stacked CR2016 batteries so I am probably giving this thing way too much voltage? Maybe a 2032 battery would have been more appropriate, who knows. I guess if it dies i’m out 55 cents…
Ironically that would be also be modding it into a better cooler at the same time. It may suck as an integrating sphere but it will keep a 6 pack cold longer than anyone else’s.
Actually there is reason why you want to limit the amount of reflections that a photon undergoes. You need a certain amount of reflections to obtain a useful level of integration but not many more. If the reflecting surface does not have a flat reflectivity for the complete spectrum (no surface has that, and a DIY surface is worse), with every reflection the spectrum changes a bit. Too good reflectivity results in larger spectrum changes which lowers the accuracy of the measurement.
I have a small experimental sphere with my best BaSO4-PVD coating so far and that one warms up 5100K 95CRI light to 4500K, as measured in the exit hole.
My most used 46cm sphere warms up 5100K light to 4375K.
But if I enlarge the entrance hole from 30mm to 80mm (it has an insert), which makes the sphere loose more light and therefore decreases the average amount of reflections that a foton undergoes before hitting the detector, the warming up is 50K less, to 4425K. Integration is still more than fine.
But if it’s on the outside the light will pass through the sphere surface twice if it does get reflected, so it’s even more diffused than the light simply reflecting off the inside.
Obviously putting foil on the inside would be a bad idea, but on the outside it would just reduce the leakage.
And with all that said djozz, my PVC P Trap light box is less than 10% off as compared to Maukka’s lumisphere…
Light that spoils good integration is direct light or light from the first reflection, they do not come to the detector from a random direction. Direct light from source to detector is blocked by a baffle, while light from the first reflection must be diluted by the light from all subsequent reflections. If the contribution of the light from the second and further reflections is enough to get the dilution done, there is no need to re-use that few percent that makes it through the wall.
That can very well be, and I’m really glad that the pipe thing works so well .
What I try to find out, build, and sometimes annoy you with, has little to do anymore with the practice of measuring and comparing flashlights, it is a personal nerdish obsession to get light measuring as correct as I can, which is quite off-topic for a flashlight forum actually. 
Lol, yeah, I hear ya. My obsession is to see how far they can be pushed, irregardless of what color or how correct the light is.
But still, I bought the calibration lights from Maukka and they came in today. My light box reads 9.3% high according to these 2 calibration lights. Well within standard deviation of emitters, cells, lenses, springs, whatever.
Edit: While 9.3% doesn’t seem like much when looking at a BLF 348 56 lumen light, on the other end of the spectrum it looks horrible when subtracting it from a 17 emitter monster… like 2400 lumens of loss horrible! Of course, on the mainstream lights that we normally deal with it’s not such a big deal. lol (about 600 lumens on a Q8 out of the box, for example… if my .345 multiplier shows 5400 lumens then the correction factor reduces it to 4836) This could even be attributed to my charger pushing a 30Q to 4.21-4.22V compared to 4.19-4.20 at Maukka’s charger. I don’t know. Not a lot of difference though, 7 lumens on the BLF 348 and less than 30 on the S2. I used a brand new, never before charged 30Q fresh off the charger in the S2… used a brand new Duracell AAA as well as an Imedion LSD NiMH AAA to test the small light.
Top is shoebox “lumens” and bottom is bobbersphere “lux” with the conversion between the two below each. I used mostly constant current modes/lights. Some throw some flood. Different CCTs:
As an early test i’m surprised how similar they are. My shoe box is white lined using a Samsung Galaxy S6 and ceilingbounce app.
The bobbersphere has the cheapo HS1010A lux meter. No baffle but a short few inch long PVC with the inside end of it covered in parchment paper. Settled with .054 multiplier until I can convince someone to mail me their fancy lights
Nice looking light Dale! I see you are still busy. Looks like your lathe skills have progressed nicely.
Since I've been away for so long, I guess I should share my latest meager mod. As a few of you know, a light that is near and dear to my heart is the Streamlight SL20X. It came from the factory with a whopping 300 lumens of crook catching power and ran off of a 6v NiCad battery. I made one previous attempt that I shared here which turned into a machining fest. Pretty much a new head screwed onto the old one.
Anyway, I decided that with some of the 6v emitters out, I could direct drive one of them with the ol NiCad and see how it goes. I'm pleased. I finally got what I think is a good drop (or screw) in solution for the venerable SL20X.
Added Beam Shots
Both pics taken at same exposure. Incan first.
Good to see your machining skills again Bucket. Welcome back. 
Thanks! It’s good to be back. I see a lot of catching up to do. It seems that Dale has broken a few laws of physics and a whole slew of other drivers etc I need to read up on.
Good to see you Bucket, still cherishing my M8 you fixed up for me, one of my all time favorites!
Wiil yak atcha l8r, kicked in the teeth at the moment… it is what it is…
I finished modding my Astrolux S43S.
- Anduril on original driver.
- Replaced the brass disc on the positive terminal with a cut-down Emisar spring to make it bump resistant.
- 20 AWG wires.
- Swapped in SST-20 3000K 95CRI emitters.
- Cut slots in the switch hole so the switch/LED PCB can be slid through without removing the switch and LED leads.
- Tore a slit in the silicone switch boot
- Stripped the protective lacquer on the copper with alcohol.
I damaged the vias for the brass posts that connect the top and bottom boards by clumsily dissasembling too many times. If I do another (I don’t plan to) I’ll use some individual test clips to connect up to the MCU without having to take the thing apart.
I think the copper and the rich, warm emitters are a nice pairing.
Dale, I guess you need to update https://budgetlightforum.com/t/-/46051….
It’s no longer a 10klum light with the battery tube it has. 
Eagle Eye X1R + Noctigon 3XP 20 mm + 3 x Nichia 219BT - V1 D220 CRI92 5700K + Carclo 10511 + alluminium pill
Any chance to open up the bezel some more to reduce light loss?
Very slightly.
The internal diameter of the bezel is 18.1 mm, the glass is about 20 mm, the fitting diameter for glass is 20.4 mm
Think, try to comprehend …