Looking for bus controlled LED driver

[Medium]

Hi EEVB!

I'm a long time viewer, but not an EE. I'm an IT guy and watch this out of pure theoretical interest. But it finally happened that I want to do a project myself. The final thing would be 50x50cm tiles (for hanging ceillings) with an RGB LED strip around the edges, and a few higher power white LEDs in the middle of the tile. The panel might be diffuse acrylic with a thin, white wood plate on top. I'd like to be able to control each of these like a pixel, but with 4 channels. (RGBW so to speak.)
I would have my home computer do all the hard work (or substitute a PI or something later on), and basically only need some way of adressing each tile on a bus that would run across my ceilling, connecting all tiles. I imagine having the bus-master coupled via USB to my PC, and send off tiny packets with a payload of 4 byte to set each channel's intensity. Basically any bus would work, there are USB-bridges to aything out there. Prefered would be Modbus RTU or DMX, since I've worked with these before.

I tried to find a ready made product that would come close to what I have in mind, and all I found was a unit that cost 100$ each. Considering that I'm probably going to need around 150 tiles for the finished project, that's absolutely no option. If I get to roughly 50€ per complete tile, I'd still barely consider it viable. Anything too far over that, and I'd have to re-think scale or entirely. But I have at least 1.5 years to finish this, more likely 2.5, since I won't have built the house I'd like to have that in until then anyways So there is plenty of time to try things out and source parts from good deals.

My biggest problem is that I have close to no experience in the EE field, and couldn't design and build such a thing all by myself. I'm thus looking for either of these things:

• Ready made products that fit my purpose and cost a LOT less than 100$
• Open-source hardware I could use. I have access to a programmer board for Atmel ICs, so I could in theory load them. I don't have any experience in actually programming them myself, though. But I know C and BASIC from desktop-development well enough to be confident to try, but would have not the slightest idea how to implement bus communications in them in any way. On the PC, most USB-bridges offer a virtual COM port I just send ASCII text over. That's basically where it ends.
• A chip that does most of the job, and some advice on what I'd need to put alongside the chip to make it work

Since each pixel would only have 2m of RGB strip and some higher power LEDs for white normal room lighting, I imagine I wouldn't need overly complicated or expensive things. But I simply lack the skill to do it all by myself. But I would like to take this as an opportunity to learn, if the easy way of cheap available products isn't available or it could be made much cheaper. (I currently have little time on my hands, but I hope this passes. I'd just like to start early by gathering information and seeing if this is a viable idea to begin with.)

In the hopes of having chosen the perfect place to ask such things, and many thanks in advance for any insight you might be able to provide,
Medium

[ovnr]

Well, if we ignore the cost of building the actual tile - that's entirely dependent on how you actually want to accomplish it - the needed hardware would not be overly pricey.

All you really need is a microcontroller, preferably with four PWM peripherals. You might be able to get by with one slowish I2C bus per room, but it's not really intended for large-scale installations. I'd probably go with RS485 myself; if you do, all you really need to do is use a packet format that includes an address byte, to which the slave devices respond. Also, I doubt you need two-way comms. (Of course, you also need some transistors and misc. passives.)

Since LED strips tend to be 12V with resistors for current "regulation", you can simply PWM them without any issue. Your plan to have some higher-power LEDs in the middle may require a bit more effort (as in a proper LED driver); what kind of LED did you intend to use? Also, you may have issues with getting even illumination if you use LEDs in the center. Edge-illuminated panels seem to be preferable.

Also, I should mention that Lumin Lighting does supply some RGB and RGBW panels as well as some (supposedly) RGBYW panels. You may want to ask them for a quote; also ask them if they can supply the panels without drivers, so you can use something that works better for your intended use case (the included drivers are probably unsuitable for a centrally managed system).

Lastly, feel free to shoot me a PM; since this is a pretty basic system, I'd be happy to help out with the design.

[Medium]

Hi ovnr,

thanks for your input. I have asked for a quotation at the place, since the panels do look very interesting. It's strange: Normally I would be super interested in actually building the thing, but in this case I somehow care more for the result than the way there. The less soldering I have to do, the better. (I know that this won't make me many friends here )
I also got a tip from a friend to look at certain shops on Aliexpress, and did actually find DMX PWM drivers there that don't cost an arm and a leg. I could get away with about 7-8$ per channel, when taking drivers that are strong enough to power 2m of RGBW strips. I also considered using 1 RGB strip and a separate CW/WW strip, but I would easily exceed 50W per panel then. I almost believe that this would be way too bright in the end. (But I'm a bit unsure about how good of a warm white I could get out of a RGBW strip.) I would also exceed 512 channels then, and would need a second DMX universe. Might be too expensive, all in all.

I'll wait and see how much they want for driver-less panels and come back! Thanks!

[Jeroen3]

Get in touch with mikeselectricstuff
He's into all sorts of fancy lighting projects.

[Medium]

Heh! That's exactly the "friend" I was talking about . I didn't realize he was active here and just wrote an email. He kindly suggested a store on Aliexpress that has all sorts of LED strips and pixels, and a generous amount of different drivers that come directly with a DMX PWM decoder. With the stuff on there, making what I want should almost be easy. I just have to be carefull to choose the right strips, and have to determine how many drivers with how many channels and enough power I need to combine in which way to minimize installation effort and cost.
My biggest problem at the moment is, that I have not the slightest idea of how many watts each panel should have so that I don't need any other lighting in the room when the whole ceiling is full of the panels. It's easy to get 2m worth of strip (=1 panel) with 30-50W, but is that enough? Too much? I have no clue. Sadly, the shipping to Germany from China is quite expensive (~20$), which makes trying a bunch of them out a slightly pricey issue.

[Rerouter]

Suppose thats the perk of PWM dimming, if its too bright you can just scale back the final brightness,

[ovnr]

I did some quickish math, and you could get custom driver boards for about $8 in parts, including a proper LED driver for white power LEDs. $2 less if you use strips for everything. Of course, assembly is on top of that; expect $2-$5/board in China.

That driver would offer 3x 16-bit PWM channels (~490 Hz) at 2-3 amps, plus a 1-2A constant current channel (also 16-bit).
Comms would be RS485; you'd basically have an unlimited amount of addresses. For signal integrity I'd split the bus into sections - one per room perhaps, maybe two if you need >30 panels per room.
It also includes certain "convenience" features, like pluggable screw terminal strips, etc.

On top of this you'd need the bus splitter boards (~$5/pop), a USB->RS485 converter (probably less than $10), and suitable 12V power supplies.


When it comes to power levels: If you want a normal indoor illumination level of ~50 lux, you should only require around 10W of white light in a normal-sized room, or to be generous, about 1-2W per m². Since you'd seemingly have four panels per m², that yields 0.5W per panel.

This immediately makes the following question spring to mind: Are you actually intending to fully cover your ceilings in illumination tiles? Nothing wrong with that, but if you are, you really do not need much power per tile.

[Medium]

@Rerouter: In theory yes, but I have slight reservations because I also operate a video projector in the same room that uses a color wheel. I hope this won't cause me notable interference flickering with PWMed LEDs.

@ovnr: The pricing is tempting, thank you for putting that together! I'm just still a bit cautious because of the need to program it all myself. (Sooo many pitfalls!) Btw: It is all just one big room that spans the kitchen, dinner table and living room with home cinema and computer station. And yes, every single tile as a light. The idea is to be able to throw one HELL of a disco party when the table is removed (Also to set different moods and gradually adapt the lighting to the time of day.) The final room size isn't fixed yet, since I am still planning the house. But I'll land at about 70-80m² of a mostly rectangular shape.

I'll probably go for all strips. I initially thought that they would not be bright enough for room lighting, and a "spot" in the center of each tile wouldn't look half bad. But if what you say is the case, then I could actually go for the really cheap sparsley populated strips (30 leds/m or so), and even get away with 2x0.5m white + 2x0.5m rgb with pairs on adjacent edges! (I'd only have to watch orientation when putting them in then.)
How about using 24V strips but only a 12V supply? That would get me to ~10W per tile with the most types of RGBW strips, if halving the voltage translates well enough to halving the output. Could help with LED survival and brightness stability over time as well. A viable thought?

[Kjelt]

Do you want the tile to be an even colour or do you want individual pixels in the tile.
If you for instance make each tile out of 4x4 = 16 individual addressable pixels you can even make animations on your ceiling (video).

[ovnr]

Flicker: It should not conflict with a projector. Also, you will generally find that pretty much every LED dimmer is PWMing the outputs.

24V strips: It would be advantageous to use them anyway - you get away with half the cable thickness. But no, you cannot run them on a substantially lower voltage - at <20v they'll be close to completely dark.
Edit: Also, if you do use a 16-bit controller, you still get useful resolution when dimming below 10%; having slightly overpowered strips isn't really an issue.

Programming: Really, it's quite simple software. I'd be quite happy to give you a hand with it too. What kind of refresh rate would you like to see on your slightly low-res display? 60fps? That will result in a moderate data rate from a PC (~1 MByte/s if not overly compact); you may want to increase the complexity of the controller software (so it does fades and stuff), since a PC doesn't really do super well at constant streaming due to OS issues.

[Medium]

Do you want the tile to be an even colour or do you want individual pixels in the tile.
If you for instance make each tile out of 4x4 = 16 individual addressable pixels you can even make animations on your ceiling (video).

It would be pretty sweet, but no. That would amount to way too much cost at the quantity I would need it. I mean, yes, I could use addressable strips and try to build controllers that multiplex a primary bus to SPI for addressing everything (would be ~2500 pixels á 4 channels total in your example), but it still comes down to quite substancially more cost and work. A pixel per tile is funky enough for me.

@ovnr:
Thank you for your very kind offer! I've put together a parts list yesterday night to see what it would cost me to make it from available parts. I landed at ~$25 per tile for the cheapest option, where ~$14 is already spent on acrylic panes, alu-U-profiles and back plates. (The acrylic is the worst offender here.) The entire thing (excluding wires) would amount to roughly $4500, which is still not cheap, but hopefully doable.
I have partially factored in the PSU and DMX controllers per tile, since I'd use one device for several tiles. I also already accounted for prices at the final volume I'd need - buying the non-electrics for just one tile would be 25$ on it's own. The DMX controllers come in at ~$1.20 per channel if I recall correctly, and there are ones with a multiple of four number of channels for ease of use. (I forgot to send the file to my office.)
I did control for a 48 channel dimmer pack via USB-DMX once, and got quite decent and stable results out of it. It didn't run on a fixed frame rate, but adopted its step-sizes based on how fast stuff got sent out over the virtual COM port. The lights would be static most of the time anyways. I think I could get away with PC controlled DMX here. Hopefully (Again, thanks a LOT for your offer, but I fear the overall cost I could shave off by doing it myself wouldn't offset the potential troubles and work enough. Unless you tell me that it could be done for $1 per tile (4ch) or less!)

I have a few questions about the power supply.
1) I have chosen 24V PSUs, but 12V strips. (I found no good RGBW strips for 24V.) If I'm not mistaken, I can just wire 2 strips up in parallel and pour 24V into that. Right?
2) How much "slack" should I give the PSUs? I have 28.8W per tile worth of strip, and used 30W for calculating how many tiles I could power with one PSU. (0..20+(0.2*panelcount) Watt slack per PSU, depending on the total wattage.) Do I need more or is that okay?
3) The most cost effective option I found is a 250W PSU for 8 panels. I'd group them into rectangles of 2x4 panels (1x2m) with the PSU at the center of the group. How is this for cable length? What gauge is appropiate here?
4) Would they survive if I strobed a tile from 0 to full at 10Hz or so? How about if I did this with the whole ceiling? Potential for desaster or okay? (Would be ~550W @ 230V, 85% efficiency factored in, if I calculated this correctly.)

Thanks!!

[ovnr]

Dimmers/controllers: Are you sure your dimmers are appropriate? $1 per four channels sounds very cheap. Remember that you need a dimmer that is fine with a common anode configuration. Also, what PWM frequency does it run at? If it hard-switches all the channels at a low-ish frequency (100hz or so), you may get significant flickering issues. Examples would be having objects not appearing to move smoothly, being unable to record video with a rolling shutter camera (without it being shockingly terrible), etc. I find that I'm rather sensitive to flickering light, and find it rather infuriating.

PSU:

1) No, you cannot use 24V PSUs. The strips are always either common anode (most common) or common cathode; this means that you cannot put them in series. Putting them in parallel does not affect the voltage drop, and you would be destroying them in very short order.

2) As long as you don't plan to run all the tiles on a PSU on full blast all day long, it should be fine. Think about the most common lighting configuration for normal use - will all panels be evenly illuminated, or will you want certain zones with higher brightness? Over, say, a kitchen countertop? If so, consider staggering the PSUs so that multiple PSUs will share the load in the areas of interest. You can also do a software map on the host controller that is aware of the PSU connections, and limits the permissible power. Lastly, consider that these PSUs will (I assume) be suspended inside a dusty ceiling with little ventilation, and will be powered on all day long. Are you picking cheap chinese supplies that may catch on fire?

3) You're looking at ~2.5A per panel; if all panels are close together, I'd probably go for AWG 20 wire (0.5 mm²). It offers a reasonable compromise between voltage drop (0.64V for 8 meters of cable worst case - remember, the power goes to the panel and back) and price.

4) If your PSUs are not terrible, then yes, that would be fine. I'd suggest doing a fast (1-2 ms) fade instead of completely hard switching, but it is unlikely that your dimmer will allow this.


Also, if you run 150 panels @ 30W, that's 4.5 kW. This is quite a lot, and I would strongly suggest limiting the power in software - it's going to be absolutely blindingly bright. I suggest watching ; it shows a very nice way to accomplish it.


Lastly, you should link the products you're looking at (PSU, controllers, etc). There may be small details you're missing.

[Medium]

1) No, you cannot use 24V PSUs. The strips are always either common anode (most common) or common cathode; this means that you cannot put them in series. Putting them in parallel does not affect the voltage drop, and you would be destroying them in very short order.

Aw shucks. But what about the DMX modules? If I had one with, say 8 channels, and put 2 in parallel and 2 of these pairs in series, totalling 4*8 channels for 8 panels. (The DMX modules are usually for 5-24V.)

2) [..] Lastly, consider that these PSUs will (I assume) be suspended inside a dusty ceiling with little ventilation, and will be powered on all day long. Are you picking cheap chinese supplies that may catch on fire?

Considering that the market almost only offers either cheap chinese stuff or horribly expensive super quality, I'll probably fall into the first category. But I heared that they're not all bad! Links to examples below. Software load balancing will probably occur, though. Already thought about that a bit.

3) You're looking at ~2.5A per panel; if all panels are close together, I'd probably go for AWG 20 wire (0.5 mm²). It offers a reasonable compromise between voltage drop (0.64V for 8 meters of cable worst case - remember, the power goes to the panel and back) and price.

Thanks!

4) If your PSUs are not terrible, then yes, that would be fine. I'd suggest doing a fast (1-2 ms) fade instead of completely hard switching, but it is unlikely that your dimmer will allow this.

Thanks!

I suggest watching [video]

That is easily my most favorite project from Mike. Beautifully engineered. But probably multiple times above my price and skill range.

Here are parts that I, as the layman that I truly am, thought might work well (all from the store Mike suggested):
LED Strip: http://www.aliexpress.com/store/product/DC12V-5M-300-LEDs-60led-m-5050-RGBW-RGB-Warm-White-LED-Strip-waterproof-in-silicon/701799_32300792556.html ($8.00/tile)
DMX module: http://www.aliexpress.com/store/product/8CH-easy-dmx-constant-voltage-decoder-DC5-24V-input-3A-8channel-output-L72-W55-H15mm/701799_2011417927.html ($1.56/tile)
PSU 1: http://www.aliexpress.com/store/product/250W-24V-10-5A-Small-Volume-Switching-power-supply-for-LED-Strip-light-LED-module-etc/701799_32269539247.html ($1.94/tile)
PSU 2: http://www.aliexpress.com/store/product/24V-250W-LED-switch-power-supply-size-199-98-50mm/701799_736620061.html ($2.00/tile)
They also offer Meanwell PSUs, but these clock in at ~$5/tile, adding almost a whopping $500 to the whole project. Mhh...

I, of course, have no clue about the details and what they might mean in the long run. I'd be super thankfull if you could comment on these parts, that would help a lot!

Edit: I would absolutely first order small quantities and whip up a prototype before going full scale to see if everything is working like it should.

[ovnr]

No, you cannot put your DMX modules in series. You must use the appropriate PSUs (12V) here.

PSUs: I'd personally go for the Meanwells. At least they're a (mostly) reputable brand.

LED strips: Seem fine. Don't buy the dipped-in-plastic waterproof ones (the ones you linked seem to just be inside a tube of some sort which is fine) - I've had issues with the plastic used being incorrect, and corroding both the LEDs and the flex.


I'm mostly worried about the DMX dimmers, to be honest. They're most likely using software PWM, which means the frequency is either going to be really low, or they have very distinct steps. And since you will likely only use the lower 10-15% of the range, you'd end up with only around 30-40 distinct dimming steps (assuming 8bit PWM). And that's linear (power) dimming; you will want to do gamma compensation to get it to look right, further reducing your range.

Edit: Also, each of those DMX dimmers will only handle two tiles.

Edit 2: I changed my design around a little, and I ended up at around $10/pop (parts only) for a 16-channel dimmer that does PWM (same specs as before) only. Say $5 in assembly costs, and you end up at $3.75 per tile. You could always move up to a 32-channel dimmer, but there isn't much to save in doing so.

Edit 3 because why not: ~$3.5/tile in parts for a somewhat improved 32-channel dimmer, saving you the hassle of having a pile of bus splitters and shit. Just attach it to a 250W PSU, and hook 8 tiles to it.

[Medium]

No, you cannot put your DMX modules in series. You must use the appropriate PSUs (12V) here.

Ok. It was worth a try

PSUs: I'd personally go for the Meanwells. At least they're a (mostly) reputable brand.

I feel similar, but need to watch cost closely. The smallest increase becomes real money at that volume.
I thought about only using each 2nd tile as a light, in a checker board fashion. But I fear that then the effect will suffer a lot. But I could maybe go for the 40% transmittant acrylic panes instead of the more expensive 88% LED optimized ones and save a bit there, since I have too much power anyways.
By the way: Do you know of some cheap sources for these and maybe even something I could use on the back-side? I only really found German manufacturers, and they usually tend to not be lowest in price. The back plate especially has basically only the requirement to be white on at least one face, and not warp too much when placed on it's edges only. (~1-2cm width per edge)

LED strips: Seem fine. Don't buy the dipped-in-plastic waterproof ones (the ones you linked seem to just be inside a tube of some sort which is fine) - I've had issues with the plastic used being incorrect, and corroding both the LEDs and the flex.

Yeah, I have one of these around my bath room, and the "silicone" slowly turns yellow and becomes brittle. I would have picked non-enclosed strips, but they oddly were more expensive then these.

I'm mostly worried about the DMX dimmers, to be honest. They're most likely using software PWM, which means the frequency is either going to be really low, or they have very distinct steps. And since you will likely only use the lower 10-15% of the range, you'd end up with only around 30-40 distinct dimming steps (assuming 8bit PWM). And that's linear (power) dimming; you will want to do gamma compensation to get it to look right, further reducing your range.

Hmm, I understand. Really do. I haven't thought that far.

Edit: Also, each of those DMX dimmers will only handle two tiles.

Since I calculated price per channel (or price per tile), this was accounted for. It would only mean a bit more wiring effort. (And easier replacement, actually!)

Edit 2: I changed my design around a little, and I ended up at around $10/pop (parts only) for a 16-channel dimmer that does PWM (same specs as before) only. Say $5 in assembly costs, and you end up at $3.75 per tile. You could always move up to a 32-channel dimmer, but there isn't much to save in doing so.

You're starting to convince me here . Would you actually share your design? Keep in mind: I have never ever designed a PCB, let alone have the tools for it. What steps would I have to do to make your solution work? Who would even make these? Do they do everything: PCB, soldering and supplying the parts? What about programming the ICs? Do I do that after assembly? I don't know how this works at all!

I can't thank you enough for your support!!

[ovnr]

Tile construction: Can't really help you much there (never made any tiles myself), beyond recommending that you purchase samples of various materials for testing. For the actual construction, I was of the impression that the backing foil is placed directly on the light guide, and supported by that. You really do need to do a fair bit of testing to ensure you get an even result.

Power supply: Keep in mind that PSUs at this power level tend to have fans. A ceiling full of loud fans sounds like a bad idea. Anyhow, if you can accept not being able to run all tiles at full blast at once, I'd go with a 120W PSU per 8 tiles. You can then run up to four tiles at full blast, or the whole pile at 50%. And 120W of LED light in a 1x2 meter area is quite enough.


Dimmer design: It's really just a shopping cart full of parts and a

session to determine the PCB size (spoiler: 15x9 cm). It's not cost-optimized either, this is just ballpark stuff (for instance, if you buy the screw terminals from China, you'd save $60). That's the 32-channel design, by the way.


Honestly, given your lack of experience, I think the best solution would be that I finished the design, and that you either had a contract manufacturer produce it, or had me do so - it's just 20 boards after all. I'd obviously charge some for it, but nothing excessive - it'd still be less than or around the price of the AliExpress DMX dimmer, just less terrible. I'd be supplying functional software too, but without any fancy extras. You can then build on the code yourself if you want to implement hardware fades and effects, for instance.


First off I do suggest you buy some LED strips and light guides and such, and see if you can get a single tile going. If you do, I can whip up a prototype single-tile dimmer (in a week or two) that you can experiment with. There's little point in putting a lot of time and effort into it if the tiles themselves prove problematic to construct.

[Medium]

You're quite generous, and I'd of course happily compensate you for any trouble I cause. Out of the question. I'm just deeply happy that this project more and more looks like it could go somewhere. (I had similar fantasies from childhood on, but of course a lot less knowledge and money.)

I initially thought I'd have the back plates against the acrylic, too, but since reasonably priced acrylic that is thick enough does not really exist, that suddenly changed. Since the strips are usually already 10-12mm wide and use 5050 SMDs, I'd probably need something with 7-8mm thickness to not cover the LEDs if I want the edges to be flush. (I do.) Since I realized that, I had the idea to to it like this:

Code: [Select]

BBBBBBBBBBBBBB BB BB B B B  B
+--------
|xx
|xx
|xx
|AAAAAAAAAAAAA AA AA A A A  A
+--------
B = Backplane (3-5mm)
x = LED strip
A = Acrylic pane (3mm)
| = Aluminum C-profile
Everything would be glued together with silicone or acrylic goo, and small slits at the corners for cables and pressure balancing. Total height would be around 20-25mm. (I might even put the acrylic unter the C-profile, could look better. Would also mean less "ugly-prone" while sawing the corners.) I luckily have a suspended ceiling available at work that I have the permission to use for for real world tests. I just hope that the acrylic won't bulge at 50x50cm...

A thought that just flashed by: The PICs wouldn't be programmable in circuit via the RS485 bus, wouldn't they? But even if not: Since I already entertained the thought to do most stuff on the PC, that wouldn't be too bad. Does the bus actually need a splitter? I've used Modbus via RS485 before, and there you could just string everything together (even branch out). You just had to terminate each loose end properly.

I'm currently heavily involved (12h shifts and upwards in the office) at the moment because my own business suddenly got busier than I thought. Therefore it could take a good while to finish the build-tests (that definitely come first). Also: The house I want to use them in isn't even properly planned yet, and construction may start in 2y earliest. I just like to have a sizable head start to have enough time to make it well, and catch good deals for materials. I'd be very happy if you would PM me your e-mail, so that I may contact you again easily once it's all certain and on it's way. HUGE help! Many thanks so far already!!

[ovnr]

No worries; it's a somewhat interesting project, after all.

PICs: Would be reprogrammable via the bus - at least the controllers. The initial software load would be entirely tedious though, as that needs to be done manually - unless you want to pay a $100 premium for pre-programmed chips. Splitters would only be used if you ran single-tile dimmers; RS485 buses don't really like having 150+ devices hanging off of them. The 32ch dimmer I showed would sort it all out internally, though - you'd simply hook it up with some beefy wires to the power supply, then daisy-chain them all together with some thinner wire (think cat5 ethernet or something) for signalling.

Construction: I do have some doubts on whether it'd work well. I've got some Lumin Lighting panels lying around, so I took one apart - turns out they have a 4 mm light guide, then a 0.5-1 mm diffuser plate in front of that. I doubt it'll distribute the light properly across the entire diffuser without a light guide (and just with an empty space); but hey, just need to experiment a bit. You may find that having the back plate sag a bit in the middle is actually beneficial, as it means there will be more light reflected downward in the middle. In any event, the light guide only needs to be around 4 mm thick to be effective enough.

Also, did a quick simulation, and a 500x500x3mm acrylic plate should sag around 0.3 mm in the center if supported on all edges. So no worries there.


I'd still try to see if it'd be viable to get ready-made panels from someone like Lumin. It'd be a lot of work to make them yourself.

[Medium]

No worries; it's a somewhat interesting project, after all.

I'm glad I'm not the only one who thinks so.

PICs: Would be reprogrammable via the bus - at least the controllers. The initial software load would be entirely tedious though, as that needs to be done manually

Sweet! What other than the controllers would I want to reprogram? And since it's not THAT many pieces, the initial load couldn't be that bad. Speaking as someone who has never done it, of course. I understand that this would fall to me? What hardware would I require to do the initial load? (As long as it's less than the pre-programmed premium, I'm willing to do the task. I'll have a chance to take some days off, eventually.)

@Splitters:
Thanks for the heads up. This might eventually come up with one of my customers, who runs a plant-wide energy statistics system on Modbus. (We have lines >50m via tiny telephone wires operating there. But data rates are at a measly 11500 baud, since we don't need more.)
Side-question: Beefy as in 1.5mm²? Or even more? Solid wire or litz preferred?

@Construction:
Wow. You're amazing. The back-plate sagging really makes a lot of sense, too! But if I added another layer of a quite specialized material (as in light guide), I'm certain the whole project falls through due to cost explosion. I know I initially said that ~50€ per panel would be still in range, but at the time it didn't really sink in that I'd be spending a good 8000€ on "just" lighting one (admittedly large) room, easily amassing to 2% of the total cost of the whole building! (If it stays at the current size. I may add another story to rent out.)
That being said: I'd be quite okay with a solution that isn't perfectly evenly lit. It might actually add some charme. The only thing I really fear is to have tiny stalagtites every 50cm. That would look silly at best. I've looked again at some offers for acrylic panes, and practically everything 3mm thick costs about 12-15€ per cutout of 48x48cm at a quantity of one. (8.50-10€ at 150pcs.) I've also looked at the 2mm options, which would save about 1.50-2€ per tile at 150pcs. How did you calculate the sag? I'd be interested if going 2mm would make much of a difference. If push comes to shove, I may even consider adding a small supporting aluminium strip across the middle of the panel for support. (Still less than what I could save on the 2mm panes!) Also: The panes would be glued to the supporting frame of aluminium profiles, which should in theory give a little bit of tension.
Heck, if everything else falls though, I may even just tack some thin paper or cloth to the bottom of the panels! (Which on second thought might not even be THAT bad... mhhhh... just need some more tension... mhhhh...)

I'm also still torn between a white and a mirroring back-plate. I'll try this out. Would be easy: Aluminium foil to the rescue!

I'd still try to see if it'd be viable to get ready-made panels from someone like Lumin. It'd be a lot of work to make them yourself.

Like i said: Probably 2yrs until I'd be able to think about actually installing the things. I'll find the time to make them somewhere. All readily made solutions I found so far were priced wayyyyy over what I could fathom spending. It's also a bit of fun, actually. And I at least have the tools for this kind of work at hand. I really hope I get around to make a few prototypes this fall. (Summer is taken up by actual work for paying the bills, as I currently see it ) I'm still waiting for Lumins' quote, though. Maybe they'd be able to convince me otherwise.

Take care, fellow imgurian!

[ovnr]

Well, my design involved 8x microcontrollers doing four PWM channels a pop and nothing else, and one larger chip to potentially do more fancy stuff - maximum power limiting, gamma correction, color temperature calibration, fades, mixing, etc. Since the PWM micros don't do anything but listen for new values, there would be little to no need to reprogram them. I'd be doing the loading, since I'd need to test the boards anyway.

Wiring: Well, if you used a 240W power limit, the wires between the dimmer and the PSU would see 20 amps. So I'd suggest 2.5mm² or above. Normal insulated copper wire is fine. The wires to the tiles can be 0.5mm² as mentioned before.

Sagging: I just modeled it in Solidworks, because it's quick and easy. 2mm would be worse, but not shockingly much so.

Back side reflection: You will need to use a white reflector. I accidentally found some white film - the kind used in professional units as backing "paper". $136 for a 1x30m roll. Spray glue that onto some MDF or something. $2.26/tile, so not super cheap, but it beats paint or printer paper...

Front side diffusion: I remembered that Rosco does diffusion films; while it certainly is not going to be the cheapest option, I did find 1.2m by 7.62m rolls for $93, or $3.8/tile. I'd imagine you could find better deals on this sort of thing though. I did also find pre-made 2mm diffuser sheets (these would not do anything as a light guide) at around $10/pop for 60x60cm sheets. For experimenting, I suggest finding a suitably sized broken LCD TV; it will have all the relevant films inside the panel.

Evenness: Any issues will show up as the edges of the tile being brighter than the center. Worst case you will see the individual LEDs shine through. Also, just plain flat acrylic is not going to do very much on its own.


Also, you may want to take a look at the suppliers here - the minimum orders are hilarious, but there may be suppliers willing to work with you.  Also, for pricing it, you need to look at 100+ unit pricing to get a realistic idea, given that you're making a lot of units.


Lastly, just checked and I paid $35/pop for my Lumin panels (30x30 cm, 4200k, including a fancy driver - probably $20-$25 without it). I'd expect the price to drop 30% or more if you placed an order for 100 panels.


Edit: Just found these LED strips, with all four colors in one LED package. Would make diffusing it easier, plus they make 24V versions, which is substantially nicer to work with.

[Medium]

You have an interesting set of knowledge and tools. May I ask what you do for a living? Sounds fun.

Thanks for "accidentally" finding these interesting materials! I will do a variety of tests involving different sheets and panes and backplates to get a sense of how bad it would really look without the light guides. I would be quite okay with brighter edges if it's not all too bad. Seeing individual LEDs would not be pretty, but if it's just a hint of slightly brigther "splats" around the edges, that'd be fine, too. I'm pretty sure professionally made panels cost as much as they do exactly because they try to even the light out as well as possible. I would see this as necessary for monitors, but I'm less picky with lamps.
I'll try to get hold of a distributor that's willing to supply in slightly smaller units than several tons minimum. Hilarious is the right word here. As for the Lumin panels: Keep in mind that going from 30x30 to 50x50 is going up by a factor of ~2.8 in area. Since almost all the expensive materials are priced per m², I estimate $60-$70 per panel, $50 at best with discount. And that'd be without controllers or power. Way too much for the amount I need.

I had considered RGBW strips, but liked that the RGB+W type also come with a warm-white option. I'm not sure which way aorund I want to go: Neutral/cold white and adjust temperature with RGB when in "cozy mode", or warm white and RGB adjust to cooler when more of a workplace lighting is desired. I will have to test that as well. But yes, I should look for 24V options in both cases.
I'm so hyped to go try things out now, but that won't happen for at least another week or two. Arghs!

[ovnr]

How tall can the tiles be without it being an issue? If you can accept 5+ cm, you may be able to get away with attaching the strips to the backing plate instead of trying to edge-illuminate it. You may get some "splotchiness" but it would probably be less objectionable. If you find a good diffuser, it should give fairly good results while being noticeably less expensive - and if you need to add another meter of LED strip to even out the light more, they're not overly expensive.

Also, here is a supplier that seems to have a lot of what you need at mostly reasonable prices (linked: set of diffuser samples).


Warm white: Well, a lot (most?) high-CRI warm white LED replacement bulbs that I've seen use neutral or cold white LEDs + some red LEDs, because it's likely cheaper than proper LEDs. As you know, running pure RGB light doesn't yield very good results, but having to supplement it by using the RGB LEDs is not really a big concern. You're also probably better off by adding some red to a cold light, than adding blue to a warm one. White LEDs already spit out more than enough blue light, while getting a phosphor to produce deep red is a bit harder.

[Medium]

I wanted to keep the suspension height between 10-15cm, so 5cm is getting close to not being able to safely operate fanless PSUs. (I might add some active ventilation to the gap through a tile with a fan or through a wall, but nothing more than what would result in slight whiffs to counter too warm build ups of hot pockets. 2 or 4 slow turning 12cm fans or so.)
I'll just have to make and see different builds. I find it hard to imagine how different setups really look like. The try-out package you linked to is a neat idea for a first impression, but I'd need a full size panel in the end to get a proper impression. But great for first "in-hand" sampling with a keyring LED torch or so!

CRI is exactly what I wondered about. The warm LEDs in my bath for example have an ever so slight greenish tint. Barely noticable, but it makes the light feel "empty" and colder than the percieved color actually looks. I noticed that especially the reds are underrepresented, which results in a strange skin tone. Not too nice in the bathroom. I also personally experienced that pure RGB is terrible. I once tried to make some accents with two short strips in my bedroom, and when only they were on (and switched to white), it looked horrible. Some form of RGBW ist mandatory. (I even entertained the idea to use one RGB and one CW/WW strip, but that upped the number of channels noticably.) I'll try CW first then!

[ovnr]

Well, if you do go for a 24V system, the currents are low enough that you could move the PSUs elsewhere (and just trim the voltage up a bit). That also reduces the heat production and fire hazard inside the ceiling. Put a large electric enclosure (think fusebox) with some temperature-controlled fans in an adjacent room.

Color tints are also easy to deal with - if it's too green, add red & blue. If it's too violet, add green. Too cold, add red. Too warm, add blue. Etc. There is no need to have a switchable color temperature strip once you also have RGB LEDs.

[Medium]

The color temp issue was with a white-only strip. But yes, one reason for the RGB in the tiles is to be able to compensate for poor white / adjust mood.

Could I go for 10m+ long lines with 24V? Is it THAT much of a difference? (I'd probably have to make all wires the same length, though.) Having the power somewhere else would be very much preferable indeed.

[ovnr]

10m 2.5mm² cable to the dimmer (effectively 20m): voltage drop of 1.3V, at 120W per dimmer @ 24V. Not a problem, and you can tweak the PSU to output a little higher voltage too - most have a trimpot. There's no problem at all. You could go with AWG 22 wire to the panels too (0.3mm²) with acceptable voltage drop (0.4v), saving some money.

I did see some... questionably genuine Meanwell PSUs that may be of interest. Fan-cooled, but will handle 40 tiles each (at least). Only $2.2 per tile.

[Medium]

"questionably genuine" I like that phrase.

The dimmer modules you designed were for 8 tiles each, correct? So I could in theory plop 5 of these into one of the PSUs, and would make sure I limit my total average output over these 40 tiles to 50%. That would be correct? The price is quite nice. How do you find these things? Aliexpress is super confusing for me to navigate. Getting to the relevant sections alone is difficult, let alone spotting suitable parts in the masses. Thanks again!

[ovnr]

Yep, five 8-tile dimmers per PSU. To be honest, you'd probably be fine running the whole setup on two 600W PSUs; for day-to-day use you will probably use less than 50W total. Hell, I'd consider buying a nice 100W PSU, then just having a beefy contactor switch over to the cheapo PSUs when you feel the need to set fire to your eyes.

Also, the nice thing about keeping the PSUs separate is that you can just re-wire the power supply cabinet if you find that you need more sustained power.

Do note that you will need two 2.5mm² wires per dimmer, or a total of 38 wires (oh, and no, you do not need to match the lengths - the dimmers will support compensating for the input voltage). If you did need 10m, that is quite a lot of wire and will cost a fair amount.


Aliexpress: I generally just search, never even tried browsing it...

[kolbep]

Use WS2811 LED Driver Chips,
or WS2812 LED Chips with Built in Drivers
A lot of people use them for the fancy christmas displays

you get them in Strips of Pixels (LED Striplights), individual LED's or in Strings with Pixels

Pair it up with a PC/RPI and a SanDevices E682 DMX to Pixel Controller and have a blast.

This lightshow uses lots of them (Eg in the Arches they just have the 3m or so of the LED strips.
You can individually address each pixels RGB and Brightness. If you use the Vixen software it is even easier to design 'shows'

[arobincaron]

Use WS2811 LED Driver Chips,
or WS2812 LED Chips with Built in Drivers

I read this thread with the intent to suggest use the same.  I've used WS2812 and it's pretty amazing what they are capable of.

WS2812 might be perfect though you don't get to pick the LED/casing and WS2811 would allow doing anything as far as LED.

[ovnr]

WS2812: I've certainly thought about them, but:

A) You need RGBW, not just RGB. There are SK6812-based addressable RGBW strips, but they're substantially more expensive ($10/meter vs $4.5/meter), so even adding in the cost of a separate driver means the RGBW strips are cheaper.

B) I don't quite think you've contemplated the scale of this project. It'd involve 300 meters of LED strip, and 18 000 LEDs. You'd still need drivers for every single piece of WS2812 strip, because daisy-chaining several thousand is not going to work. Also, there is zero gain in using addressable LEDs, as you'd always be setting a whole set of 120 LEDs to the same value anyway.

C) They run on 5V. This means wiring costs would explode (you need 5x as much copper - and the whole mess would reach something on the order of 1100 amps peak), plus there will be way stronger brightness changes across a strip.

D) They only offer 256 levels. That may sound like a lot, but you will lose all control at lower dimming levels, since under normal operation the tiles will only run at 10-15% power.

As for buying WS2811 driver chips: Since these are constant current devices, you can't really use them to drive whole strips. They just don't offer enough current capacity, and you can't easily extend it. And even if you did... well, they're still $0.13/pop, vs the $0.74/pop PIC micros (and you'd need output transistors in either case). Not a massive saving; even if it worked, it'd be on the order of $100.


It's entirely reasonable for a smaller project, but by the end of the day, you'd end up wasting well over $1000 by going for addressable strips.

[Medium]

Yep, five 8-tile dimmers per PSU. To be honest, you'd probably be fine running the whole setup on two 600W PSUs; for day-to-day use you will probably use less than 50W total. Hell, I'd consider buying a nice 100W PSU, then just having a beefy contactor switch over to the cheapo PSUs when you feel the need to set fire to your eyes.

I'm still not convinced that they would be THAT bright, but my tests will show. My scepticism is because I would by far not have a build that is aimed at squeezing every last photon out of the panels into the room, since I'd be missing the light guides and have a generally simpler construction in mind. But that's one of the things I'll defenitely find out.

Wire costs aren't too much of a concern. A friend of mine works for a company that wires large industrial plants, and they usually consider anything <50m at these gauges as "rest". I'll have the chance to get some cables quite cheaply there after they had the right kind of customer. Time is on my side here.
Voltage compensation in the controllers is a mighty fine feature! Cool.

Aliexpress: I generally just search, never even tried browsing it...

I guess it comes in very handy to roughly know what exists and what you want for this. A luxury I'll have to learn. (And currently do!)

Regarding the addressable strips:
I also looked at them in the beginning thinking "oh, handy", but then realized that being able to control every LED is wasted functionality in my case, and also figured that it would mean a LOT more expensive controllers. They are amazing for actual LED screens made from strips, but I am better off with regular "dumb" strips I suppose. (I also need to watch data rates at some point, since I'd rather not have to offload too much work to the controllers, since I'm a lot more comfy programming for PCs.)

I have a weekend at the office, so I sadly won't get too much further with this project. But I thought: Even if the tiles project is a good 2 years into the future, I could benefit from fiddling around with the controllers much earlier to make myself more acquainted with the tech. My current place has terribly expensive lighting (consumption-wise), and I might be able to whip up something small scale to play around with. I'll try to think of an interesting mid-term solution. (Problem is, that it's basically one big 9x4m room, completely under a roof. Close to 5m high in the middle, and I'm afraid of heights enough to loathe ladders even if only 2m high...) But I already have something vague in mind.

[ovnr]

Well, if you did need to run it on a several hundred watts power level, you might find that you didn't need much heating in the room, haha.

As for (PC) programming: What languages do you prefer?


If you do need controllers earlier, I'm sure that can be arranged too.

[ovnr]

So I happened to have some time to kill, and decided to dust off V-Ray.

tl;dr: Side-mounted strips are not going to work very well - at least not better than top-mounted strips. You should get satisfactory results with a 70mm tall panel, but a 100mm tall panel is better. Note that this is all a fair bit of guesstimating, but it should be close enough.

. Note that the noise level is solely dependent on how long I left it cooking, and my patience was distinctly lower with the bad solutions...

Also, this was with fully lit white LEDs and partially lit red LEDs, hence the somewhat displeasing color. It assumes strips with 60 LEDs per meter.

[Medium]

Wow! Thanks a lot! How well does V-Ray emulate actual physically correct rendering? I tried to model something with Cinema 4D and use its Physical Renderer, but there, of course, are no ways to know at what settings I'd have the equivalent of a 500mW LED worth of light, or if my diffusor is smoked glass or just has a frosted surface. With that much uncertainties, I figured that any result from this could only be misleading. I however lack the experience and never actually compared the real thing to a rendering.

I hope I get around to visit the hardware store tomorrow and see what materials I could obtain in small quantities locally. One thing to consider with 100mm thick plates also is, that the materials to make them will get notably more expensive. Especially because 100mm wide profiles are relatively non-standard. I really hope to get somewhat further this weekend. But I'll have to order the strips anyways, so actually building something functional won't happen this soon.

I thought about actually making the panels like I want them for the full project, but drill small holes into the sides and string them onto two steel cables that go across my current room. Would winding the wires to the panels around a steel cable cause trouble?

[ovnr]

Render results: It's not perfect, but it's within reason and matches my expectations. It only simulates surface interactions (because SSS would make it even slower), but on a thin diffuser, they will dominate. Also actual light brightness doesn't matter for the diffusion results.

Construction: I'd honestly consider just making the sides out of 10mm plywood. Also, top-mounted strips will be a bit more labor-intensive as you have to solder wires between the 4-5 strips, as opposed to run one length uninterrupted.

Steel cables and running wires on them: Not a problem.

Strips: These seem to be the cheapest I could find.

[Medium]

Ah okay, I hoped you could model actual physically correct simulations. But I guess that's near impossible anyways, considering that getting a proper impression of light on a screen is never going to work - even on photos.

I'll go to the hardware store today and see what options they have at what prices for construction. Wood tends to be a lot more expensive than expected - at least in my experience. And I will defenitely try the strips you linked, although they're with separate white LEDs. The diffusor is still a rather diffuse topic though. Local stores won't really have anything like it, and small volumes aren't highly available either at reasonable prices. (I didn't want to spend $200+ just for trying out a few materials.) But I haven't searched intensively yet, that'll be done this evening as well.

Taking delivery times into consideration, I should be able to make some prototypes in a month or so. I found connectors for 4-pin strips somewhere, but can't remember how expensive they were though. But these might make connecting the strip pieces very easy.

Another thing that came to mind: Some strips are common annode, some are common cathode. I can't say which way is the more available one. Would it add a lot to the controllers to make them switchable between these modes? (If that's even possible!) Or would that be too much because both types are widely available anyways?

[ovnr]

You can model it, but getting the parameters right is going to be next to impossible anyhow. Giving it a go anyhow. Again, surface effects will dominate on a thin diffuser (most photons won't travel very far laterally in a 1-2mm thick plate after all).

I did look at local wood prices, and ended up at around $4-$5/tile in total. The issue really is the diffuser, which seems to end up at >$10/tile no matter what I do. Total tile cost around $33-$35. If you could magically get some acrylic sheets for cheap, that'd solve it, but plastics are expensive.

LED strips: Didn't your original strips have separate white LEDs too? In any event, if you get it diffused enough that you can't see each strip, you certainly won't be able to see each LED. As for connectors, I've never seen five-pin strip connectors myself; they'd certainly make life easier tho. Also, you will want common anode strips; common cathode will involve an extra set of resistors and transistors per channel bumping the price up a fair bit, and also make it harder to control the switching times. You ideally want somewhat soft switching for EMI reasons.

Making a switchable controller is going to add a substantial amount of money in any event; it's not worth it.


Prototypes: Do you at least have things like a lab PSU, soldering iron, and other normal electronics kit?

[Medium]

Don't waste too much time on simulations. It was pure curiosity on my side.

The wood prices depend a lot on volume. For small scale testing, it'll be quite expensive. For 150+ tiles, not so much. With regard to the diffusors, I may just have to look out for something not purpose-made for light diffusion that still yields decent results. Anything that is made for lighting somehow ends up at least twice as expensive as one would assume based on the materials used. Buying normal ceiling lamps alone is often a quite shocking experience.

The separate white LEDs aren't much of a problem, you just mentioned earlier that combined chips could yield a more even result. But I am still considering being okay with uneven lighting within one tile, and having the edges brighter may even be a somewhat appealing effect after all. I'll have to see the actual thing though. So no, I'm not really going for the optimal fully professional looking solution, but a compromise between cost and looks.

The switchable controllers are off the table then! Especially since you make a very good argument for common annode.

I actually have a well equipped workshop I can use at work. PSU, stand drill, motorized miter saw, soldering iron, thread tapers, multimeter and all sorts of small stuff is available. We have an electrician that builds our switching cabinets (we do industrial automation), but he already admitted to being next to useless when it comes to anything that fits on a PCB. He only deals and has experience with the big boy's stuff.

[ovnr]

Well, for diffusers, if you could get hold of cheap acrylic, a single layer of diffuser film should be enough. Also, I imagine the prices would drop reasonably much if you talk to the manufacturers about ordering 100+ units.

And in general you will end up spending substantially more for prototype quantity stuff.


Combined chips: Again, turns out it's not needed, at least if you have a diffuser that works at all. Also, you will find that they have a somewhat different power level; around 4W/m of white compared to 5.7W/m for the separate WLED/RGBLED strips. And of course, less RGB power (4W vs 2W, respectively). This is because the combined chips only have 60 white emitters per meter, vs 90 for the separate (as each chip contains 3 dies). And 30 colored emitters vs 60, of course.

If that was confusing, here's roughly how it's set up:
Combined: [RGBW]   [RGBW]   [RGBW]   [RGBW]
Split: [RGB]   [WWW]   [RGB]   [WWW]


Also, how do you intend to mount the tiles?

[Medium]

I have ordered the strips and a preliminary RGBW controller to play and test around with so far. Sadly, delivery times for the items is up to 4 weeks. After that I'll first play around with a few options made from cloth, wrapped around using velcro to "mount" it. (Enabling me to wash it from time to time.) I'm not overly convinced that this ends well, but my mom would like to see if her idea has merit. Who am I to deny my creator? (And it may even be a cheap alternative after all!)

More white power (dear Americans, it's not what YOU read in these words!) is very nice. I recently changed the plan for the house again, which would have upped the tile count to 234. Unacceptable. I'd do half of that. But that also means only 2 instead of 4 tiles per square meter. (I may have to use bigger PSUs and limit at more than 50% output. But total light output still is an open point to test for.)

Mounting would be super simple: The tiles would be sized to fit a standard suspended ceiling, like you often see in offices or so. The benefits of such a ceiling are aplenty: Easy to run cables trough it, damaged / dirty tiles are easily replaced, keeps heat in/out better than naked concrete, and of course easy installation of fitting light fixtures. They would just loosely lay on the frame like the normal tiles.

[ovnr]

Mounting: Thank god, I was worried you were going to screw each tile into the ceiling or something.

Tile count: I'd consider letting the row/column closest to the wall be blank - light shining straight down onto a wall isn't terribly attractive in any event. That gets you to around 176 tiles, if it was a square room.


On controllers: Would you like to have final hardware (that is, the 32-channel design I was poking) in hand for prototyping? It's of course going to be substantially more expensive than the final units due to tooling costs and a lack of quantity discounts, but nothing completely shocking.

[Medium]

As for (PC) programming: What languages do you prefer?

I fortgot to answer this, but it'll probably not make much of a difference anyways. I'd either go for C# or Delphi, since I'm most familiar with these and there is a good amount of usable third party stuff out there.

Mounting: Thank god, I was worried you were going to screw each tile into the ceiling or something.

Good lord no! I cherrish a good challenge, but I'm not a masochist

Tile count: I'd consider letting the row/column closest to the wall be blank - light shining straight down onto a wall isn't terribly attractive in any event. That gets you to around 176 tiles, if it was a square room.

That's a very good idea. I'll have to model this out in C4D first to see how much I save by this, since the room wouldn't be square at all. Could be divided into 4 rectangles with the current plans, which will change anyways a lot until the house gets started. I have to wait for the rent of the occupants of the currently existing building there to run out, and since I will do this in cooperation with my parents (owners of the ground), they need to pay off an older loan before they can shoulder their part of the project. (Read: The banks play along.) I hope we can get there in <2 years. I'm quite eager to get this thing started.

As for the controllers: Once I found a design I'd like to use for my current place to play around with, a full 32 channel job would fit nicely. It would also provide a nice "real-world" test phase to iron out whatever issues may come up. (Or rather: It would be good for me to learn with the real thing.) What would you estimate for a prototype?

[ovnr]

C# is nice. Delphi... well.


Controllers: US$ 160 for two full controllers (not including assembly cost); due to minimums and NREs, more is always cheaper. Each additional controller (up to 5; haven't done the math for larger quantities) is around $50. You can also do a cut-down version without full output drivers which would work well for software development (it would have eight onboard RGBW LEDs) for $25/pop, assuming you buy full controllers too. Basically just the boards and the stencil is $50 down the drain before you even begin looking at parts.

The controllers do have some convenience features that add some cost, like dual pluggable power connectors (so you can daisy-chain several off of one PSU more easily), pluggable output connectors, etc. Replacing these with non-pluggable terminal blocks would drop the price around $7-$10 per driver; note that they're not present on the cut-down dev edition. They're nice to have; whether they're worth it is up to you. Spade (Faston) connectors are another option, but are less foolproof - and getting the polarity wrong will end badly.

[Medium]

Delphi as a language is fairly nice, and I find it more pleasing to look at than C-likes. It's just sad what it's developer is doing to it from a tools- quality- and pricing-perspective for the last few years. I'm actually just at the brink of phasing it out in my job after they were bought out yet again a few days ago, but legacy stuff will ensure that I'll get to work with it plenty, still. Porting a codebase that grew for a good 15 years isn't an easy thing though. I'll basically re-rwite everything, including UIs (which are my focus as the guy responsible for process visualization). But there are many subtle bits and pieces where I see quite well that C# (the language itself) was mainly developed by the very same guy that made Pascal great back in the days.

For the controllers: 2 are even better, leaves more options! Also the nice to haves don't sound overly over engineered but rather sensible. Making pure dev boards would be a waste, though. So, $200 for two fully assembled units, and $50 for each that follows? Do I smell a deal? (I can't yet say how many exactly I'm going to need in the end. Depends on what the room looks like then, and what my brightness-tests yield as a result.)

[ovnr]

Well, the reason why I said "... each additional controller (up to 5)..." was because you'd have already paid for the PCBs. Minimum orders and all that shit.

Anyhow, component cost for a 15-board order later on would be $43 per controller; yes, including PCBs. Do note that each controller contains 137 parts, and I'm afraid I wouldn't do the volume assembly for free. That's a discussion for another day (and place) however.

Dev edition: The idea is that you can have one sitting in a drawer, and can push a firmware update to it to check the functionality before pushing it to the whole damn ceiling - without having to have a 24V source and LED strips dangling off of it. If you somehow were to brick the bootloader, it'd save you the effort of taking down every controller and manually re-flashing them.

I would in general strongly suggest keeping a couple spares too, once you get the full install up. Getting new ones manufactured could easily involve substantial lead times.

[Medium]

137!? Sweet baby Jesus! That is a LOT more than I expected it would take. But alas, if it's well engineered, it may as well have a certain complexity to it.
I wouldn't dare to ask for free assembly. That's why I rounded the figure you stated excluding assembly up to 200, but I have to admit that I can't estimate how much work two boards may already be, at all. I hope I didn't insult you there! You make a very good point for a dev board. Mhhh. Well then, I guess I'd like to have 1 dev and 2 full pieces then
I'm not too shabby with a soldering iron, but when it comes to finer pitches on some chips, and at that quatitiy, I'd probably save me a lot of time and potential trouble, if someone who is more apt at it could do it. (I understand that you need to order more than 3 PCBs in a batch, which makes the pre-assembled options very front-loaded cost-wise, especially because the final count isn't yet determined.) We'll figure something out.
For the time being, please let me know what you need for 1 dev + 2 full boards, fully assembled and ready to play around with. (I'd need a brief spec of the protocols for programming and light-data, too, of course. Something short is okay, since I'm not entirely new to the topic of RS485 comms.)