May not be directly applicable but Mike Harrison makes professional LED installations and has a lot of wisdom and insight.

https://m.youtube.com/playlist?list=PL0KZLmPyL6AnkPSs5cGU437...

Being a home tinkerer that have a few meters of LED strips, I'd use multiple drivers.

How about measuring voltage across the strip and modeling the system resistance (per meter)?

You can measure the voltage where you are observing problems to understand your margins.

Your solutions are sound, you just need to throw in some math to optimize them to each application.

Chances are you'll need to start using different protocols (eg rs-485) to communicate over the long distance to numerous controllers that output the data to the LEDs.

Slow RS485 or CAN. The differential nature of that system sphysically removes most the the unwanted signals.

I’ve done multiple projects that use 20m+ of WS2812s.

I deliver the main power in segments from a single large PSU and run 5v signal (despite being designed for 3.3v signal most will handle 5v fine and works better for longer strips).

Running segments with connectors also makes it easier to swap out failed segments.

I've not quiet worked with that long LEDs, but i have experience with 10-20m length indoor installations.

For our standards, we supply the strips with power at both ends and at every 5m in-between. (So an LED is at most 2.5m away from a supply). Your supply needs may vary significantly depending on the type of LED you use, these numbers worked for us on WS2812B 60led/m at 5V rail voltage.

> For long runs, do you prefer distributed power injection vs multiple smaller PSUs vs a higher-voltage backbone + local regulation?

Either should work fine, but when you use multiple PSUs or regularors, you may need to match their output voltage suprisingly precisely to get good results. A pair of thick copper wires is often easier.

> For addressable strips (WS281x/SPI/DMX), what are your go-to fixes for signal integrity over distance (grounding, buffering, differential, level shifting)?

WS28xx has built-in "reclocking" as long as your longest distance between two adjacent leds is <1m, you are absolutely fine. Be aware though that the number of chained LEDs impact your maximum refresh rate, which is an issue for animating. For a 60Hz refresh rate, you can't have more than ~250leds.

SPI is a nightmare over any distance. DMX is great if you respect the standard in terms of cabling and unit loads, but it i'm yet to see a DMX-enabled individually addessable LED strip. A DMX universe would be limited to 170 rgb or 128rgbw leds anyway.

> engineering patterns that scale and don’t become a maintenance nightmare.

Use a projector or a TV instead :-D

If you’re already doing segments with separate power injection you may as well split it entirely.

I designed my led system at home in the basement which is not as big but same principle. It all depends on: 1. Distance 2. Type of LED: W, RGB, RGBW, RGBWCCT 3. Brand of LED manufacturer 4. Number of LED per meter aka amperage draw 5. 24v verse 12v

If this is a loop, like a rectangle, you can inject from beginning and end since that is the same point, otherwise, you'll need power injected at some distance. For long distances you'll need to calculate the power loss to figure that out. I used quinled.info site to get all this information and decided upon an RGBW LED setup using 24 volts and injecting at beginning and end of rectangle. This was for 20 meters, fyi, not 50 meters. RGBWCCT was too much amperage draw over distance and this was my first time doing it. Learn how to solder because using the snap-on connectors suck...I am redoing those connectors by soldering b/c two have failed so far.

Have you tried filters?

If you're getting flickers that's probably the best thing to do. This will cut your total power but you'll get a smoother signal. If it is visible then that is relatively low frequency but you might want to play around with it. The easiest thing to test (because you're more likely to have the parts accessible) is a pi-filter and then if that seems to be working you can either stick with that or use a better filter.

If you supply ground from one end, and +24V (or whatever) from the other end, brightness will be equal throughout the strip :)

But it would also be a good idea to measure voltage drop when it's powered on, and compensate accordingly, if your PSU allows for it.

What safety and regulatory requirements do you need to meet?

If you need NEC class 2, both Lutron and ERP make UL listed 24V class 2 supplies. You can calculate what length you can drive at 4A (96W) and drive that much. The Lutron one dims and the ERP does not, but you can chain an eldoLED or other “dimmer” between the ERP supply and the strip with excellent results. There is no perceptible flicker with either of these, and IIRC both Lutron and eldoLED even claim IEEE1789 compliance. (At least the Lutron one does leave sort of perceptible spots when dimmed when you saccade perpendicular to it, and you need to go to rather high frequency to avoid this. Of course, this is a nonissue at 100% duty cycle.). I have used both of these with strips from different power supplies butted up to each other with no perceptible brightness mismatch.

48V strips are a thing.

If you don’t need NEC class 2, you could try a much larger DC supply. You might consider circuit breakers to protect individual segments and wires. DIN-rail thermal-magnetic breakers rated for more than 24VDC are cheap, but they will buzz horribly if you apply PWM to them.

You’d be on your own in regard to regulation, but I bet most “120V driverless” strips work just fine at 120 times sqrt(2) DC volts, and they won’t flicker. You will get more than the design output and more heat, but no flicker. Test first and take appropriate care for safety.

Don’t use “driverless” strips with sine wave input. The 120Hz flicker is horrible.

Some LED strips contain internal constant current limiters that will give a roughly constant output across a (small) range of voltages. This may help.

Distributed power injection and "CC" strips. The CC chips in the segments provide constant current to the segments allowing the voltage drop along the strip. If you inject to infrequently, the drop will be high and the chips in the feed end will dissipate a lot of heat, so it's likely you will be servicing those ends. Feed more frequently and you'll avoid this. WS281x usually have seperate shift registers inside, so they usually have no signal degradation--theoretically. In practice, recommended you segments feed those signals, but can be much longer depends on the number of nodes and the size of your controller frame buffer.

There's many guides out there for exactly this, but this one from QuinLED is probably the best [0].

You can't really avoid power injection downstream. I'm not sure if you're using 24v LEDs based on your ask (you said rails). But even with 24v strips at 20m you'll need 4A edge injection with 720 LED/m. This is literally one of the most simplistic approaches without being under powered.

[0] https://quinled.info/the-ultimate-led-strip-power-injection-...

Did you learn nothing: https://news.ycombinator.com/item?id=46390875 ?

I've worked on several 20-50m long LED strip projects (including CV + addressable). The issues you described—"brighter at the beginning, dimmer at the end, white turning yellow, flickering under heavy load"—are basically problems with power distribution, connections, and signal links. Below is my long-term, stable approach (leaning towards an "engineering model," prioritizing maintainability over minimal cabling):

1) Power Topology: Separate "voltage transmission" and "load power supply" into layers.

I generally don't gamble on a single-ended path. A more reliable approach is: zoned power supply (every 5-10m section) + star/tree topology, with clear fuse/open circuit protection for each section, allowing for quick fault location.

If site conditions permit, I prefer a higher voltage backbone (e.g., 24V/48V) + zoned step-down/regulation: lower backbone current, lower line loss, localized voltage regulation/power supply at zoned ends, and more intuitive maintenance.

Regarding the choice between "multiple small PSUs vs. one large PSU," I prefer multiple small PSUs (or a large PSU with multi-branch protection). The core reason is: fault isolation and maintenance without affecting the overall system. However, this requires a unified grounding strategy to avoid signal/noise issues caused by ground loops.

2) Constant current/segmented adjustment: can significantly improve "brightness and color consistency," but don't treat it as a panacea.

Constant current/segmented adjustment can indeed reduce visual problems such as "darkening at the end and white drift," especially in scenarios where linear consistency is desired.

However, it cannot replace fundamental aspects: PSU margin, zoned protection, wire gauge, connection reliability, and heat dissipation. Many "flickering" issues are actually transient problems caused by PSU triggering protection/voltage sag/increased connector resistance.

3) Addressable Signal Integrity: Design "data" as a communication link.

SPI/WS281x Class: My commonly used combination is:

Keep data lines as short as possible (controller close to the first segment).

Add buffering/reshaping if necessary (re-drive every segment).

Use differential transmission (RS-485, etc.) for long-distance traces and then convert back at the end.

Use a unified ground (single-point grounding/partitioned grounding strategy must be clearly defined), and avoid bundling data lines too close to high-current harnesses.

By the way, this article focuses on the differences and selection of DMX vs SPI:https://suntechlite.com/dmx-led-strip-vs-spi-led-strip/