How do I drive LEDs without flicker ?

[DeadlyChemist]

if you put 5v directly across a 10R resistor (ignore MOSFETs or other things) you can have a maximum of 500mA, that's just ohms law!
to calculate current you should divide whatever voltage is at the non-inverting pin (+) of the op-amp by the current sensing resistor.

i have tried other resistances than 10 ohms as well (said it above?), i did put 2 and 3 10 ohm resistors together, but anyways...

DO NOT order PCBs if you don't have a working circuit, trust me on that one. if you're worried about the breadboard causing issues, solder the components with short wires (for SMD parts you can buy breakout boards similar to this: https://www.ebay.co.uk/itm/362262676455)

got some, might try that, would need to buy some wires first or might just cut up old lan cable

Edit: If you can give me a schematic of exactly what you tried, that would be helpful.

will probably re-test everything, the test was very quick and dirty, veeery unoptimal, i do have a picture i took, but not sure you can make anything out lol, tbh, better don't look, you'll question your life, my work-bench is full of boxes
https://cdn.discordapp.com/attachments/718850705363828748/991711381621583962/20220629_162635.jpg

The only extra components you will need to solve oscillations are resistors and capacitors (I assume you have some common values on hand to test, for example, 100ohm, 1kohm, and 100nF). If you really want to jump straight to PCB design, it's a good idea to include the extra components and not populate them later on.

i actually dont, resistors i do, but i have no caps, i usually just order them as i need them or desolder from junk (because i dont make things like you all, i fix/mod things once a while)
actually found 2x 220µF caps
the pcb would probably just be a final test
because whatever we come up with, i dont believe the first prototype will be the actual PCB i'll use lol

[OM222O]

one problem you'll have is that the digital pots ICs you bought have a 2 bit address (A0 and A1 pins), meaning you can only have 4 of them on a I2C bus (total of 16 actual pots because each single chip has 4 pots). What MCU are you using? does it have multiple I2C ports or do you have to use 2 MCUs as well?

I can start designing the schematics tomorrow as well.

[DeadlyChemist]

one problem you'll have is that the digital pots ICs you bought have a 2 bit address (A0 and A1 pins), meaning you can only have 4 of them on a I2C bus (total of 16 actual pots because each single chip has 4 pots). What MCU are you using? does it have multiple I2C ports or do you have to use 2 MCUs as well?

I can start designing the schematics tomorrow as well.

thought of that, there is a i2c multiplexer, would use that
yeah, if you know where to add fuses/protection/anything, that be nice, so far i only thought of 24v input fuse, but nothing else

for mcu, im most likely gonna use Raspberry Pi Pico W (launched today hehe) high overkill, but for 6€ you can't beat it
EDIT: the dev time in python would be minimal, so i like it

[alligatorblues]

Wow, all this work. I wanted to build some studio lights. I purchased five 100W square led arrays about 1.5" square, maybe 16x16 leds. I purchased the drivers from the same seller, which were 12VDC to 40VDC converters enclosed. I took some heat sink fan cpu coolers, tapped some holes and affixed the leds to them with acrtic silver thermal paste interface and 2 screws! I took some wallwarts I had laying around for the 12VDC IN, and cut off one side of (4) PC PSU cords.

I soldered the cord leads to the prongs on the wall warts. So, now they plug into the wall! I made 4 lights, three 100W and 1 200W. I think the  brightness of leds compared to incandscent is about 25:1 per W. So, the 100W are equivalent to about 2,500W incandescent, and the 200W about 5,000W incandescent! So, I required a dimmer, because these things are surreal they're so bright!

I just used a variac I had laying around for a dimmer, and it works pretty well. So, I think my total bill was ab out 150.00USD because I had a lot of little stuff I needed already! My basement is a electronics salvage yard. There's some pretty good stuff in there too! If I was disciplined enough to sell it all I'd probably make 50,000.00USD!

But I've got plans for almost everything down there! I just picked up a ESI SR104 10k resistor, Fluke 732B DC STANDARD, (2) 10Mhz. rubidium standards, 6 HP 735A 1.0VDC Standards with +-0-1000uV by 1.0uV control, 2 HP 34401A 6.5 DIGIT DMMs, Fluke 8842A DMM, and a huge box of miscellaneous passives, actives, and I don't know what all else, for a C-NOTE, at a yard sale. The lady was older than dirt! Her husband had just died. And, she was having a sale to downsize. 

The husband had been in electronics somehow, and she was really thankful for all the work I did carrying all the stuff to my UTE, all by myself! Always carry cash. She wanted cash! I said, all I have is a $100 bill. She said, Oh, I cwn give you change back! I said, No, you've just had a significant loss in life! You can just keep the whole thing. She said, Oh, Bless your heart!

So, now I've got a little project testing that stuff. Oh, and there were (2) 10A Staco variacs, and a bunch of leads.

It pays to shop yard sales. She thought she was going to have to leave the stuff at the curb for collection, because no one else even looked at any of it!

So, anyway, that's the story of 9500k studio lights. They just get a little warm. So, the fans on the heatsinks must help a lot.

[DeadlyChemist]

Wow, all this work. I wanted to build some studio lights. I purchased five 100W square led arrays about 1.5" square, maybe 16x16 leds. I purchased the drivers from the same seller, which were 12VDC to 40VDC converters enclosed. I took some heat sink fan cpu coolers, tapped some holes and affixed the leds to them with acrtic silver thermal paste interface and 2 screws! I took some wallwarts I had laying around for the 12VDC IN, and cut off one side of (4) PC PSU cords.

I soldered the cord leads to the prongs on the wall warts. So, now they plug into the wall! I made 4 lights, three 100W and 1 200W. I think the  brightness of leds compared to incandscent is about 25:1 per W. So, the 100W are equivalent to about 2,500W incandescent, and the 200W about 5,000W incandescent! So, I required a dimmer, because these things are surreal they're so bright!

the light quality will be eh, but depends what you doing
as for flicker, this will definetly flicker, set your camera fps as higha s you can, and lower the shutter speed as low as you can

at a yard sale. The lady was older than dirt! Her husband had just died. And, she was having a sale to downsize.

USA i guess
the best things i bought at a "yard sale" basically we have a flohmarkt (german) every couple of months... you go, rent a spot and offer to sell your junk
so best things i bought were 6 mercury thermometers (some high temperature) for like 10€
no one here makes yard sales like in movies/youtube flip series

[DeadlyChemist]

did some math and thinking

accuracy:
my resistors are this: https://www.digikey.de/de/products/detail/stackpole-electronics-inc/RNCP1206FTD1R00/2240302
1 OHM 1% 1/2W 1206 so 4 of them (ignoring gaussian distribution) would be 3.96  - 4.04

and my transistor https://www.digikey.de/de/products/detail/onsemi/FQD7N10LTM/1053477
has a resistance of 0.275  (typ) to 0.350  (max) sadly does not say minimum, but i can just assume 0.200  so basically 0.200  to 0.350  (check if im correct)

together that would be maximum of 4.16  - 4.39  worst case possible

putting that into simulation gives me 246.133mA - 233.238mA basically 5.53% difference, considering everything is gonna be at very similar temperature, from the same box and datasheet values are usually in typ range instead of min/max everything should be within 2% or so which is pretty good, but 5% is also fine, i assume the leds will also have some variation.

Current measurment:
thinking to integrate this or similar
https://www.aliexpress.com/item/1005003671166977.html?
ina3221 tripple current sensor "Up to ±1.638A current measurement"
basically could split warm and cold and get some proportions/feedback, could be useful, they cost pennies and Rpi Pico W has few pins i can dedicate to that (or do you know of a better chip?)
but because it's "upto 1.63A" and i don't trust aliexpress so 4 channels/1 input (4x 250mA) x3 boards to get upto 36 leds
one board is like 3€ so might include it

other features:
I'm looking to detect the power switch (230V) beeing turned on and off quickly, basically some leds have the feature that if you double tap the power switch you get another mode, im looking to implement that, any ideas?


I'm unable to test anything so currently just looking thru things i can implement

[OM222O]

The resistance of the MOSFETs doesn't change the accuracy, only the voltage reference, the input offset voltage of the op-amp, and the resistor value determine the output current.

Current measurement shouldn't be necessary, as long as the circuit is verified to work correctly, then whatever you set in the software, is what you get on the output. It could be useful for fault detection and protection if somehow the MOSFET fails short, but adds a lot of costs.
If you really want to monitor the current, there are tons of I2C chips available, just need to find one that has enough gain to read the small currents you're working with. I can have a look if you want. This uses 0 extra pins which should be ideal for your situation. If you want to search for yourself, you can have a look here: https://www.ti.com/amplifier-circuit/current-sense/power-current-monitors/products.html

I'm not sure how they do it in commercial lights, but there are many methods for detecting a double press. I think the easiest solution is to add some sort of backup power (probably super capacitors + boost converter). Then you need to monitor the "high voltage side" (probably monitor the 24v output of the PSU rather than 230v mains) with a comparator of some sort. If the high side voltage drops below the output of the boost converter, then you know the switch has been turned off. Then you can write the logic for what happens after power down / power up.

[DeadlyChemist]

Current measurement shouldn't be necessary, as long as the circuit is verified to work correctly, then whatever you set in the software, is what you get on the output. It could be useful for fault detection and protection if somehow the MOSFET fails short, but adds a lot of costs.
If you really want to monitor the current, there are tons of I2C chips available, just need to find one that has enough gain to read the small currents you're working with. I can have a look if you want. This uses 0 extra pins which should be ideal for your situation. If you want to search for yourself, you can have a look here: https://www.ti.com/amplifier-circuit/current-sense/power-current-monitors/products.html

yeah, i think the aliexpress listing i showed above with ina3221 should be perfect tbh, the board is already made, and it's cheap enough.. unless you have better idea?, surely the alie board will be crappy quality but should be fine.

I'm not sure how they do it in commercial lights, but there are many methods for detecting a double press. I think the easiest solution is to add some sort of backup power (probably super capacitors + boost converter). Then you need to monitor the "high voltage side" (probably monitor the 24v output of the PSU rather than 230v mains) with a comparator of some sort. If the high side voltage drops below the output of the boost converter, then you know the switch has been turned off. Then you can write the logic for what happens after power down / power up.

yeah, but what if the leds are really low power, i flick off-on the switch, but the caps in the psu/boards still are basically supplying the power, so it won't detect anything im worried...
https://www.aliexpress.com/item/1005003739343987.html?
i dont know what it outputs, but maybe have the micro powered by big cap with diode or something, and use that to detect if main power is on or off or even secondary chip/comparator ? or instead just put a hall effect sensor next to the main wire?

i prefer to deal with 230V instead of 24v in this case, because it should be more reliable

[OM222O]

those modules have a ton of extra components that you don't need and add a sense resistor that you already have. If you want to use a current sense amplifier, the best bet would be to buy the IC rather than a module.

those current clamps are way too much of an overkill (definitely not used in commercial circuits), but they do work. If you have one of the commercial units that do this, might be a good idea to take it apart and reverse engineer it, but I have to question the use of this though, especially the fact that you mentioned using the Pico W. Why not host a small web page on the Pico to change the settings/mode etc. ?

Edit: if the current is low (LEDs on very low brightness), the current clamps will also have trouble detecting it. I still think it's best to monitor the voltage of the PSU and see when it drops below let's say 20v or 22v or something similar. With any reasonable load, time taken for the output voltage to drop from 24v to something below that should be small enough that it'll be detectable.

[DeadlyChemist]

those modules have a ton of extra components that you don't need and add a sense resistor that you already have. If you want to use a current sense amplifier, the best bet would be to buy the IC rather than a module.

true, but then how much would it cost to add 32 current measurments? because if im gonna use the 4R as current shunt, i have 32 of them?

but I have to question the use of this though, especially the fact that you mentioned using the Pico W. Why not host a small web page on the Pico to change the settings/mode etc. ?

i'll have that as well, but since i'm adding RGBYWU leds as well (havent bothered about it at all for now)
RGB for accent with white/RGB party lol?
White for well, not used
Yellow (kinda monochromatic) for accent or generally just yellow for that low pressure sodium lamp look, or if i want to work with film/light sensitive stuff
Uv (405nm, safe) for party mode i guess? or when i would firm some glow in the dark stuff, could be useful

Also to anwser the question:
i want to add "sleep mode" basically double hit the switch, the leds turn on very low on Red (because red does not kill human nighvision apperantly) and after a minute or whatever but still beeing on, so i can turn in on from my phone/clapping?/whatever
or maybe double cycle (so 4 clicks) to get into safe mode so, yellow only untill changed in settings or whatever, honestly, not 100% sure, but having the ability is soo useful

Edit: if the current is low (LEDs on very low brightness), the current clamps will also have trouble detecting it. I still think it's best to monitor the voltage of the PSU and see when it drops below let's say 20v or 22v or something similar. With any reasonable load, time taken for the output voltage to drop from 24v to something below that should be small enough that it'll be detectable.

fair, but i could just use the like power lines detector that goes thru walls and looks for active cables, some meters have it built in
basically this, whatever you call it https://www.aliexpress.com/item/33036736656.html?

might be overkill on the stuff i wanna add/implement even if i don't end up using them, but having possibilities is nice

[OM222O]

those probes use a very high gain amplifier to detect ANY electric/magnetic fields, won't give you a good measurement of how much current is in the wire (few pA or uA as leakage vs few mA for a device that's turned on). Again, I really don't know how they do it in commercial units, so if you really want this feature, you may wanna reverse engineer one.

Funny enough I had a look at TI'sparts and the ina3221 is the only part that can monitor 3 lines at once, but even in massive quantities (10k+ parts) it still costs around 1.5 pounds, but the module you showed is only 1.7! I have no idea how that is possible unless they're using fake ICs? I still don't think you need current monitoring. Even if the MOSFET fails short, you will only have (24-22v) / 4 which is 500mA. This can be easily detected with a comparator since you said the max current you plan to use is less than 300. This will be much cheaper and require no programming. If you bought extra op-amps, you can use them as comparators, if not, comparators are dirt cheap these days.

Either way, I think you need to think about those features later on. let's focus on the actual circuit first.

[DeadlyChemist]

I have no idea how that is possible unless they're using fake ICs?

Either way, I think you need to think about those features later on. let's focus on the actual circuit first.

most likely and fair enough.

[OM222O]

I actually remembered that I made a very similar thing about 3 years ago for a robotics competition. I've put it up on wordpress and as you can see, I did include two 1k resistors and a 100nF capacitor to stop the op-amp oscillation. The brightness change isn't super visible on the camera (only visible when the LED has low intensity).

https://manisprojects.wordpress.com/2022/07/03/eurobot-lighthouse/

I've been really busy so haven't had time to start a new design for you, but I think I'll have time after this week. Other than using a digital pot instead of a manual one, I think the rest can be copied without any changes. let me know if there's anything else you want to add (other than current sensing).

[DeadlyChemist]

https://manisprojects.wordpress.com/2022/07/03/eurobot-lighthouse/

damm, that's a nice thing !

I've been really busy so haven't had time to start a new design for you, but I think I'll have time after this week. Other than using a digital pot instead of a manual one, I think the rest can be copied without any changes. let me know if there's anything else you want to add (other than current sensing).

no rush, i basically only want to get the "driver" pcb/design done so op amps, maybe pots, transistors, etc

the board and final schematic with the micro, sensors and whatever else i come up with and or voltage reference? i'll do probably after i assembled the drivers and tested everything.

also can't think of anything (except sensors) to add that changes the design

so i think making the drivers is the first step i guess?

[OM222O]

I have started designing the schematic and PCB for the project: https://oshwlab.com/theepicn008/eevblog-cc-led-driver

A really neat solution would be to have a stackable design where the control board sits at the top and provides the 1.024V reference to the driver modules and you can stack the driver modules on the bottom (you can stack up to 4 driver boards (16 output channels) because of the I2C address issues). If you have an I2C multiplexer let me know; It can be included in the control board.

[DeadlyChemist]

My health is back (ish)

I have started designing the schematic and PCB for the project: https://oshwlab.com/theepicn008/eevblog-cc-led-driver

I took a look at the schematic and looks like basically a done driver board, good job
Should I make an account there so you can add me? or how does that work? later in the build process i'm looking to dump images/videos/stuff on there so it's more or less easy to reproduce, always used github for that stuff, but i guess that website works as well.

A really neat solution would be to have a stackable design where the control board sits at the top and provides the 1.024V reference to the driver modules and you can stack the driver modules on the bottom (you can stack up to 4 driver boards (16 output channels) because of the I2C address issues). If you have an I2C multiplexer let me know; It can be included in the control board.

the multiplexer would be https://www.adafruit.com/product/2717 or similar TCA9548A based one (they cost 3-5€ on ebay)
yeah i guess that's the best flow, not sure if "stackable" like litteraly, is what I want (because im worried the stack might cause a shadow + heat buildup), but yeah, 4x driver board and 1 main board or so. what size are we even talking for the driver board? i prefer a long skinny one so i can put it on the edges.
The main board (layout) is last worry for now until i test it and do the heatsink

I began writing driver (made 0 progress, but technically began) for the 4 channel digital potentiometer in micropython. I'll update once a while if i'm stuck/done

Will also buy a large piece of paper and draw the heatsink/leds/whole layout (can't find software that does not suck)
most likely will do with 400x400 heatsink (16x 10x10 heatsink) i only technically need 300x300 but meh

[OM222O]

I meant literally stackable. 1 control board, 8 driver boards. Heat won't be an issue as you can use a fan to blow fresh air on the stack (highly doubt that'll be required) and what do you mean by shadow? the board will be placed above the light fixture (aluminum angles/plate). Do you have drawings / 3D models of what the final fixture would look like? that'd be helpful.

Driver boards are not completely finished as I still need to add comparators, a latch circuit, and P channel MOSFETs to disconnect the 24v input in case of overcurrent detection.

For the digital pot, I have never used that specific IC but had to write my own library for a similar microchip pot, let me know if you need help.

other than that, I think I can start designing the control board as well.
To join the projects, you need to create an account on easyeda.com and send me your email/user ID so I can add you.

[DeadlyChemist]

I meant literally stackable. 1 control board, 8 driver boards. Heat won't be an issue as you can use a fan to blow fresh air on the stack (highly doubt that'll be required) and what do you mean by shadow? the board will be placed above the light fixture (aluminum angles/plate). Do you have drawings / 3D models of what the final fixture would look like? that'd be helpful.

i have 2 plans in my head for mounting/heat
https://prnt.sc/JugNYgPaiaAT
or
https://prnt.sc/Fx0IrxL2DE3G

first of all, i don't want any fans, i did the math, and a 3x3 grid of heatsinks (100x100x15) (green) with 4x4 grid of warm-cold leds should be enough to keep them under 60C
the red would be the electronics and blue would be RGBYWV leds (i just briefly talked about), i have 5 or 6 of them, and not sure of the locations, might go with more, not sure as of now, will worry later as the voltages are unlucky, not super good with 24v, lots of energy to burn

the 2nd plan would be to super overspec and make it bigger, what i like, the price would be 50€ vs 100€ for heatsinks (9 vs 16)
the heatsinks would be in 4x4 grid and everything should be ice cold basically (most likely will go with that, even tho it's wasteful)
but it should be basically like a tank that can disappate a lot of power

i spent about 2 days looking for cheapest, best heatsinks for this project (30cm - 50cm) and welding up 10cm heatsinks is sadly the best way i found

the 24v powersupply will be external

will prob mount it 3-6cm from the ceilling to have airflow/convection (should be enough) using a bracket or something.
the "case" for the light will be just a square of stainless steel with a frosted plexiglass mounted to it, not sure, spent many days looking at existing designs and don't like a single one (suggestions open)

the light will be super thicc and ugly (well, i think stainless looks nice) but that was never the goal.

other than that, I think I can start designing the control board as well.
To join the projects, you need to create an account on easyeda.com and send me your email/user ID so I can add you.

will do, might take me few days (moving and stuff)

also have done 0 experiments regarding light uniformity, like how far away the led needs to be from the frosted glass to look uniform, also how close the leds need to be to each other

i guess once i built the driver i can just experiment with 4 leds and see what looks how
the 4x4 heatsink might put the leds too far apart i'm worried, i'm not sure how to calculate that without actually testing



here is my terrible explanation of the shadow im worried about

https://prnt.sc/2spWGymy4zcu

basically, let's say the electronics stack is white
and red would be the frosted glas, i'm worried that the light might hit the electronics on a weird angle where one side/corner is just slightly dimmer than the other ones, basically less uniform

yeah, i'm probably just paranoid, but i feel like placing the electronics all around is better than i 1 place anyways

unless you have a better design/idea

[OM222O]

hmmm ... it is possible to design the modules to be daisy chained sideways instead of stackable on top of each other, but due to the increased distance, there might be a lot of noise/error in the 1.024v reference, especially on the furthest modules. adding capacitors can somewhat overcome that, but it's not ideal. Since you are so unsure about the final dimensions/design, I think it'd be a good idea to order some prototype PCBs and once you were happy with the test results, finalize the PCBs, but that requires desoldering parts from the prototype which can be a bit annoying.

Do you have access to fusion360 or other cad packages (Solidworks, etc.)?

For the RGB LEDs, you can add extra resistors to dissipate the heat without overloading the MOSFETs. since you bought a lot of extra 1 ohm resistors, this can be a nice solution.

[DeadlyChemist]

I think it'd be a good idea to order some prototype PCBs and once you were happy with the test results, finalize the PCBs, but that requires desoldering parts from the prototype which can be a bit annoying.

dont have hot air, buying isn't a problem but space is, so currently can't really desolder SMD

i'm not that worried about reference, i can just run more individual wires to it instead of daisy chaining, but then you probably know better

Do you have access to fusion360 or other cad packages (Solidworks, etc.)?

i have autodesk fusion360 but im very terrible in it (can't get anything done)
so i have been using blender for everything (including modifiying 3d prints...) but it's not ideal for accurate parts as you know

For the RGB LEDs, you can add extra resistors to dissipate the heat without overloading the MOSFETs. since you bought a lot of extra 1 ohm resistors, this can be a nice solution.

probably, but again, first i need to buy one and test the actual voltage/current like i did with the white ones

[OM222O]

You will ABSOLUTELY need to buy a hot air station for this project. Soldering that many SMD parts by hand is both very slow and unreliable (high chance of bad joints).

I'll leave the PCB design for the end since you don't even know the physical dimensions of what you're trying to build. I made a customizable fustion360 light fixture which you can make from 4 angled pieces (any specific reasons for choosing stainless? aluminum is both cheaper and easier to weld. Anodized aluminum also looks better in my opinion):


the frosted glass sits on top of the metal (I recommend adding some sort of foam or felt under it. You can easily buy self-adhesive foam tape in different sizes) and you can add a layer of light diffuser sheet+ fresnel lens (common construction in most LCD panels to make the backlight look uniform).


Finally, the LEDs + heat sinks sit on top of that assembly with a bit of distance (using some sort of spacer, probably wood or 3d printed). Do you know how you want to mount it to the ceiling?

Edit: I personally bought a YIHUA 862D+ from amazon for about 60 GPB and have been pretty happy with it, I'm sure you can find similar cheap soldering stations if you have a look.

[DeadlyChemist]

You will ABSOLUTELY need to buy a hot air station for this project. Soldering that many SMD parts by hand is both very slow and unreliable (high chance of bad joints).
Edit: I personally bought a YIHUA 862D+ from amazon for about 60 GPB and have been pretty happy with it, I'm sure you can find similar cheap soldering stations if you have a look.

hm, probably will get the same one, heard great things about it somewhere
but will solder bulk by hand anyways, it's relaxing for me to solder for hours and hours, hotair is meh, not satysfying and loud
only worried about reliability tho, as hotair is just better i agree

(any specific reasons for choosing stainless? aluminum is both cheaper and easier to weld. Anodized aluminum also looks better in my opinion):

my friend can weld whichever material i want, and stainless feels premium?, might go with aluminium but for now it does not matter that much
i would go with bright pink anodized aluminium but sadly it would cast a pink cast on everything + welding anodized isn't really possible

the frosted glass sits on top of the metal (I recommend adding some sort of foam or felt under it. You can easily buy self-adhesive foam tape in different sizes) and you can add a layer of light diffuser sheet+ fresnel lens (common construction in most LCD panels to make the backlight look uniform).

looks good i guess
foam is easy to get, but don't get why i need it in the first place? is direct contact bad? (there wont be any shaking or anything)
no clue where i would get fresnel lens or difuser sheet, except broken tv's or something? but that's not really repeatable

Finally, the LEDs + heat sinks sit on top of that assembly with a bit of distance (using some sort of spacer, probably wood or 3d printed). Do you know how you want to mount it to the ceiling?

i thought the "frame" will have hooks (L bracket or whatever) on top to mount to the big heatsink, so basically the frame will be a spacer?
as for mounting to the ceiling?, no idea, my plan was to go to a hardware store, look at the parts/what exists and figure something out
unless you have a "ready" solution

also great job on the 3d cad model, i would have need weeks for it, days in blender...

[OM222O]

welding anodized isn't really possible

You first weld it into shape and then anodize it.

foam is easy to get, but don't get why i need it in the first place? is direct contact bad? (there wont be any shaking or anything)
no clue where i would get fresnel lens or difuser sheet, except broken tv's or something? but that's not really repeatable

I personally don't like the idea of glass on metal (too many broken PC side panels and those are tempered glass), but if you're ok with it, then just leave that part out.
Diffuser sheets are pretty common in photographing industry, you can buy huge sheets for relatively cheap:
https://www.amazon.co.uk/Diffuser-Diffusion-Seamless-Modifier-Photography/dp/B085FRNY1H
same with fresnel lenses:
https://www.amazon.co.uk/DarwinUK-Large-Rigid-Fresnel-395mm/dp/B01MU4HR20

or like you said, you can take apart old broken TVs.

i thought the "frame" will have hooks (L bracket or whatever) on top to mount to the big heatsink, so basically the frame will be a spacer?
as for mounting to the ceiling?, no idea, my plan was to go to a hardware store, look at the parts/what exists and figure something out
unless you have a "ready" solution

using L brackets sounds like a good idea, but then you need to figure out a way for mounting the heat sink to the ceiling.

I think the first step is to figure out exactly how many LED modules you want to use and in what configuration (4x4 only white? larger array with RGB?) and the final spacing and dimensions of the LEDs. This lets you finalize the dimensions of the frame and tells you how many driver boards you will need. I still think a stackable design works much better if you're willing to mount the frame a bit further from the ceiling (maximum of about 10cm if I had to guess). All the electronics will be mounted above the heat sink (you can just use double-sided tape if you want something easy or you can 3D print a small enclosure for it if you want to be more professional).

Let me know when you decided on the LED arrays and their spacing.

[DeadlyChemist]

Diffuser sheets are pretty common in photographing industry, you can buy huge sheets for relatively cheap:
https://www.amazon.co.uk/Diffuser-Diffusion-Seamless-Modifier-Photography/dp/B085FRNY1H
same with fresnel lenses:
https://www.amazon.co.uk/DarwinUK-Large-Rigid-Fresnel-395mm/dp/B01MU4HR20

thanks for the link

I think the first step is to figure out exactly how many LED modules you want to use and in what configuration (4x4 only white? larger array with RGB?) and the final spacing and dimensions of the LEDs. This lets you finalize the dimensions of the frame and tells you how many driver boards you will need. I still think a stackable design works much better if you're willing to mount the frame a bit further from the ceiling (maximum of about 10cm if I had to guess). All the electronics will be mounted above the heat sink (you can just use double-sided tape if you want something easy or you can 3D print a small enclosure for it if you want to be more professional).

Let me know when you decided on the LED arrays and their spacing.

4x4 (16) of this https://www.aliexpress.com/item/1005003210480731.html? (white)
the spacing, im unsure what looks good on a 400x400 grid (more below)

as for this https://www.aliexpress.com/item/1005001267331243.html?
i can see the voltage range is 2.2V - 3.6V so technically i can fit 6 to get 21,6V (but i haven't tested the leds at all so i don't know) the current is 500mA but i'll go with tiny bit less (-50mA?) so maybe 7 leds can be fit into 24V?
i know the RGBWYV are crappy quality but that's fine. i only really care about yellow and red
I'll have to buy 10 of them for 50€ but oh well, unless you know of an EASY way how to use like 10 of them? without making my lamp into a bigger space heater than it already is
also i dont know where/how to fit 6-7 or all 10 leds into a 4x4 grid that look nicely

The heatsink will be this https://www.conrad.de/de/p/fischer-elektronik-sk-81-100-sa-strangkuehlkoerper-2-k-w-l-x-b-x-h-100-x-100-x-15-mm-188557.html (can't find anything better/cheaper)
in a 4x4 array welded together (400mm x 400mm x 15mm) i think they are anodized, but one acid bath is enough.
my friend can weld aluminium, worst case i'll use brazing rods and a dozen of gas burners.

I'm still thinking of mounting the electronics on top as you said. with 400x400 heatsink it should be plenty enough cooling and space

[OM222O]

the easiest way to deal with 10 is to break them into two groups of 5 LEDs, but since these are small COB modules, I think you can always add them later. Let's focus on the 4x4 white LEDs for now.

You don't need to weld the heat sinks together; it's easier to weld some straight pieces of steel/aluminum (whatever the frame material is) to the top of the frame. You can then drill some holes and screw the heat sinks onto those bars. This way you won't need extra L brackets either (bars act as L brackets). Then you can use 1 heat sink per LED (400x400mm) which should be enough cooling. I'll post an update when I finish the design.