If the far end of a white run looks pink, you do not have bad pixels. Pink is what running out of voltage looks like.

Pink means voltage, not hardware

Set the display to full white and compare the first pixel of a string with the last. The community documentation states the tell directly: pixels may look pink instead of white when turned fully white, and a first pixel that differs in color from the last is voltage drop in the string and the wires supplying it. This is worth internalizing because the instinct when the end of a run misbehaves is to suspect the pixels there, and replacing them changes nothing. The fix is power injection, meaning voltage fed in again at or near the far end rather than asked to travel the whole way from the controller.

Join the negatives. Never join the positives

Two rules that sound like they should be symmetrical and are exactly opposite, and the community wiki puts both in capitals: do not tie together positive outputs from different power supplies, and always tie together negative outputs from different power supplies and pixel control signal feeds. The reason the ground rule matters is not power at all, it is data. Spiker Lights explain it plainly: without a common ground, data will go out but cannot complete the circuit, and the pixels will not function as expected. So a missing common ground presents as a data fault and sends people hunting through controller settings. Tying positives together, meanwhile, sets two supplies fighting each other. Only one power supply feeds any given pixel string.

Know the rule-of-thumb numbers, and know they are not specifications

With appropriately sized wire, the widely used figures are roughly 50 pixels at 5 volts and roughly 100 at 12 volts before power injection is needed. Both correspond to about a fifteen percent voltage loss, which is the edge of the operating window rather than a cliff. They are community rules of thumb derived from voltage drop math, not published limits, and the real answer depends on pixel wattage, wire gauge, distance from the supply, the actual output voltage of that supply, and the brightness the show runs at. Treat them as a starting point for a plan you then verify with the pink test.

Brightness changes the limit, and almost nobody accounts for it

This is the least known variable and it is a large one. Current draw scales with brightness, so a display run at thirty to fifty percent, which is what many hobbyists actually use, draws far less than the same display at full white. At those levels two hundred pixels or more before injection is described as customary rather than optimistic. The trap is planning at one brightness and running at another: a display sized comfortably for fifty percent will show voltage drop the first time somebody pushes an effect to full white. Decide the brightness the show will run at, plan for it, and then test at full white anyway so you know where the headroom ends.

Distance to the first pixel is a data problem, not a power one

The run between the controller and the first pixel of a string is where data signals degrade, and it fails differently from everything above. Builders report managing roughly fifteen to twenty feet before needing help, with one documented build needing a null pixel every ten feet on eighteen gauge wire, and AusChristmasLighting noting that distances beyond about five meters can present problems. The solution is a null pixel: an ordinary pixel spliced into the line purely to receive and retransmit the signal, with the controller told how many to ignore so they stay dark. If a whole string is unresponsive or showing nonsense while its power looks fine, the lead-in distance is the first place to look.

Do not crank the power supply more than a third of a volt

Power supplies often have an adjustment that lets you raise the output slightly to offset losses in wiring, and it is a legitimate tool. It also has a hard ceiling that the community documentation states explicitly: do not adjust the supply more than 0.3 volts DC above nominal, or you risk damaging the electronics in the pixel controllers and the pixels themselves. The temptation when a run goes pink is to dial the supply up until it looks right, and past a third of a volt that stops being a fix and becomes a slow way of destroying the expensive parts. Inject power instead.

Size supplies from watts per pixel, and fuse the positives

Sizing is arithmetic rather than guesswork. Watts per pixel is current multiplied by voltage: a WS2811 pixel cannot exceed 0.0555 amps at full white, which at 12 volts is 0.666 watts, and at eighty percent brightness about 0.533. Divide the supply wattage by that and a 350 watt supply comes out at roughly 656 pixels. Do not load a supply to its full rating. Then fuse it: fuses belong on the positives leaving the power supplies, and it is worth knowing that the wire from supply to controller carries the highest current anywhere in the system, which makes it the wire to size generously and protect properly.

Use the fattest wire you can, or double it up

Since the whole problem is voltage lost in wire, the wire is the lever. Use the largest diameter you have, remembering the numbering runs backwards so 18 AWG is thicker than 20. Where thicker wire is not available, running two or three pairs in parallel for the positive and the ground is a legitimate way to reduce drop, and it is a technique the community documentation specifically endorses. It matters as much for injection feeds as for the main run, since an injection point fed through thin wire over a long distance arrives with the same problem you were trying to solve.