There is a comfort in putting up colour bars. The vectorscope draws its six neat dots, the boxes land where the overlay says they should, and everyone agrees the chain is calibrated. I want to talk about what that agreement is actually worth, because I spent a week last year chasing a “broken” monitor wall that every test card in the building swore was fine.
Bars assert one narrow thing: that this signal, through this path, hits these voltages. That assertion is real and worth having. The lie is in what we let it imply — that because the test signal survives the chain, the pictures will too. But bars are a saturated, static, full-level pattern, and almost nothing about a real picture is saturated, static, or full-level. The monitor wall that started this was clipping just below peak white with a gamma bend in the near-blacks — a region the 75% bars never visit. Every card passed. Every camera feed looked crushed.
The deeper problem is that a modern chain is not one gamma but a negotiation between five. The camera has a transfer curve, the mezzanine codec has assumptions, the graphics stack tags one colour space and renders in another, the monitor applies its own idea of contrast, and the browser-based multiviewer in the middle — my corner of this industry — does whatever the GPU compositor felt like that driver version. Each layer is individually defensible. Nobody owns the product of all five. A test signal generated after layer two tells you nothing about layers zero and one, which is where the cameras live.
What actually found the fault was a ramp — a boring grey staircase, plus a colleague with twenty years of eyeball who said “the third step is wrong” before the scope agreed with her. Which points at the useful half of the tradition: test cards were never only for machines. They existed so a human could learn what right looks like and notice its absence. That skill transfers; the overlay does not.
So: keep the bars. Automate them, run them continuously, alarm on drift — they are free regression tests for the path. Just refuse the implication. The chain is calibrated when real pictures survive it end to end, and the only test card for that is the one with the shadows in it.