Why the picture is black
A video chain that is correct on paper and produces nothing on the wall is almost never short of bandwidth. It has failed a negotiation — and the three negotiations it can fail all report the same symptom.
Before any pixels
The handshake nobody watches
Plugging a source into a display does not send a picture. It starts a conversation, and the picture is the last thing that happens in it. The display announces itself electrically. The source reads a block of data out of the display — the EDID — that lists the resolutions, refresh rates, colour depths and audio formats it will accept. The source picks a format from that list. If the content is protected, both ends then authenticate over HDCP. Only after all of that does anything appear.
Every one of those steps can fail, and every one of them fails the same way from where you are standing: a black screen. That is why "check the cable" is such a poor first move. The cable is the one part of the chain that has no opinion.
The failure mode that wastes the most time is subtler than a black screen. A splitter, a switch, an extender or a scaler in the middle usually presents its own EDID rather than passing the display’s through. So the source picks a format the middle box likes and the actual screen cannot show, and you get a picture that is the wrong resolution, or the wrong refresh, or that works on one output of a splitter and not the other.
- Testing source direct to screen, getting a picture, and concluding the chain is fine. You have just proved the two ends work and removed every device that was actually negotiating.
- One HDCP-incapable device anywhere in the path. It is a chain of trust: it fails at the weakest hop, and the symptom appears at the end.
- A long HDMI or HDBaseT run that carries 1080p60 and drops 4K60. That one really is bandwidth — 4K60 4:4:4 is roughly four times the data rate, and a marginal cable passes the smaller one.
- Hot-plugging in the middle of a show, which restarts the whole negotiation from step one on every downstream device.
Same time, same frame
Genlock, and why a seam shows what a screen hides
Once pictures exist, a second problem starts. Every output is running its own frame clock, and unless something ties them together those clocks drift. On one screen that is invisible — a frame starting four milliseconds late is still a frame. Across a join between two projectors or two panels of a wall, it is a tear: for a fraction of a second the two halves of the image are showing different moments.
Genlock fixes the reference. Historically that was a black-burst or tri-level sync signal distributed to every device, which locks their frame starts to a common timing. On a network it is PTP doing the same job with timestamps rather than a dedicated cable, which is why an ST 2110 or Dante-adjacent video plant cares so much about a grandmaster clock.
Genlock is about when a frame starts. It is not the same as frame rate matching, and it is not the same as latency. Two devices can be perfectly genlocked and still be four frames apart if one of them is doing more processing than the other — which is the difference between a tear and a lip-sync problem.
What a link actually carries
Bandwidth, and the three ways it gets reduced
A raw video link carries pixels per second times bits per pixel, and the number gets large quickly: 3840×2160 at 60 fps in 10-bit 4:4:4 is about 14.9 Gbit/s before any overhead. Every real-world link deals with that in one of three ways, and each one costs something different.
Chroma subsampling throws away colour resolution while keeping brightness resolution, on the grounds that the eye notices the second far more than the first. 4:2:2 halves the data; 4:2:0 more than halves it. It is nearly free on camera footage and visibly destructive on fine coloured text or a sharp graphic edge — which is exactly what a show usually puts on a wall.
Compression — from Display Stream Compression on a cable to H.264 or JPEG-XS on a network — buys much larger reductions and costs latency and, above some threshold, artefacts. And reducing the picture itself, in resolution or frame rate, is the honest option that people reach for last.
| Scheme | Data vs 4:4:4 | What it costs |
|---|---|---|
| 4:4:4 | 100% | Nothing. Full colour resolution on every pixel — what graphics and text need. |
| 4:2:2 | 67% | Half the horizontal colour resolution. Broadcast standard; safe on camera footage. |
| 4:2:0 | 50% | Half horizontally and vertically. Fine for delivery, poor for a wall showing coloured text. |
- Sending 4:2:0 to an LED wall and blaming the panels for soft-edged text.
- Adding a frame of latency at every hop and only measuring it at the end, when the sound is already locked.
- Assuming a "4K" input on a box means 4K at 60 in full colour. Very often it means 4K at 30, or 4K60 at 4:2:0.
Where the numbers live
The calculators behind this page
Frame budget turns a frame rate into the milliseconds you have to spend. Aspect fit shows what fitting one shape into another costs in bars or in crop. Video storage converts a bitrate into disk and a card into minutes. How a colour becomes a number has the subsampling grids and the bitrate arithmetic in full.
2 questions on the mechanism rather than the trivia. Nothing is scored and you can answer again — the explanation under a wrong answer is the point of it.
-
1A source connects direct to a screen and works. Put a switcher in the middle and the picture is black. What is the most likely cause?
-
Right. Testing direct removes every device that was doing the negotiating. The middle box usually presents its own EDID, and HDCP authenticates at every hop — either failure shows as black at the far end.
-
Worth checking, but the cable is the one part of the chain with no opinion. Three different negotiation failures all look like this.
-
A screen in standby usually says so. A failed handshake gives you nothing.
-
Possible, but it would usually show as sparkle or intermittent loss rather than nothing at all.
-
-
2Two projectors edge-blend into one image. The seam tears intermittently. What fixes it?
-
That fixes lip-sync against live sound. It does not align frame starts.
-
A wider blend hides a geometry mismatch, not a timing one.
-
Necessary but not sufficient. Two devices at the same rate still start their frames at different moments.
-
Right. Free-running outputs drift relative to each other. On one screen a late frame start is invisible; across a join it is a tear. Genlock fixes when the frame begins.
-
Answers are kept in this browser on this device. Nothing is sent anywhere.
- Why is the screen black?
- What is EDID doing?
- Why does the seam tear?
Where to go next
These pages are a chain rather than a library — the next one along usually answers the question this one raised.
Everything on this page, in the index
Each of these has its own entry with sources, ports, gotchas and an open gap where a field is missing.