Screen Tester

This is worth stating plainly, because it was not a copy problem. The gradient pattern was built as a black-to-white ramp with a second layer over the top: rgba(16,185,129,0.35), the accent green used across this site. The contrast pattern had the same green laid over its black and white bars at 35% opacity.

Those are precisely the two patterns whose whole job is to be neutral. A banding test works by showing you an even ramp of greys, so any unevenness belongs to the panel; wash it in green and every grey has a cast, and you cannot tell a tinted step from a tinted screen. A black-and-white contrast test works because black is black — with a green overlay the darkest bar is no longer near-black at all, so clipping in the shadows cannot show up.

Both patterns are now pure. No brand colour appears anywhere inside the preview area, the gradient is a straight #000000 to #ffffff ramp, and the shadow and highlight wedges step through real near-black and near-white values. The grid also moved from 32px to 64px squares, since a 32px cell on a 2× display is 64 device pixels and the pattern was measuring nothing you could reason about.

Each pixel on an LCD is three subpixels — red, green and blue — with their own transistors. A fault is usually in one subpixel rather than the whole pixel, and that is what makes a single-colour test misleading.

A subpixel stuck off is invisible until you ask it to light up. A red subpixel that has died shows nothing on a black screen, nothing on a blue screen, and a dark dot on a red one. A subpixel stuck on is the mirror image: it hides on white, where everything is lit anyway, and shows as a coloured dot on black. So a white pass finds dead subpixels, a black pass finds stuck ones, and the primaries tell you which channel is at fault.

The secondaries earn their place for the same reason. Cyan is green plus blue with red switched off, so a red subpixel stuck on appears as a warm dot in a field that should have no red in it at all — a fault that is genuinely hard to see against white. Yellow does the same for blue, magenta for green.

Mid grey is the one most people skip and the most revealing single screen. Uniformity problems and colour temperature drift show up in the middle of the range rather than at the extremes, so a panel whose left side is cooler than its right will look fine on white and obvious on grey.

Clean the screen before you start. A surprising share of reported dead pixels are dust, and a fingerprint smear on a black screen is indistinguishable from clouding until you wipe it. If a dot moves when you tilt your head, it is on the glass, not in the panel.

Backlight bleed and IPS glow get confused constantly, and they are different things. Glow is a broad, even brightening of dark areas that changes as you move your head, and it is a characteristic of IPS panels rather than a defect. Bleed is localised, usually at a corner or along an edge, and does not shift with viewing angle. Both show on black, so look from straight on and then from an angle before deciding which one you have.

Ambient light will change your verdict. Judge a dark screen in a dark room, because in daylight a panel with poor black levels looks fine and in a dark room a good one can look grey. Judge uniformity in even lighting instead, since a lamp to one side produces exactly the left-to-right gradient you would blame the panel for.

The sharpness pattern is the quickest way to catch a display not running at its native resolution. Alternating one-pixel black and white lines should read as crisp lines; if they shimmer, blur into grey, or show a moiré pattern that drifts across the screen, the signal is being scaled somewhere between the operating system and the panel. Compare the device-pixel figure in the capability panel against the resolution the panel advertises — a mismatch there is the same fault stated numerically.

Photograph it. A dead pixel is nearly impossible to describe over a support chat and trivial to show, and a photo taken at the time is worth more than a description given a month later. Shoot the solid colour that makes the fault most visible, straight on, with the room lights off.

Then find your own model’s pixel policy rather than a number from a forum. How many faulty pixels are permitted before a panel qualifies for replacement varies by manufacturer, by panel size and by product line, and some premium monitor ranges carry a zero-bright-pixel promise that ordinary ranges do not. This page deliberately does not print a threshold, because a wrong figure would either cost you a replacement you were owed or send you into a support queue with nothing to stand on.

Test on a second machine or cable if you can, particularly for anything that looks like a line or a whole-column fault rather than a single dot. A failing cable or port produces artefacts that look like panel damage, and ruling that out first is a five-minute job that occasionally saves an entire return.

It cannot measure anything. Brightness in nits, a contrast ratio, colour accuracy as a Delta E figure, gamma — all of those need a colorimeter held against the glass, and no web page has access to that. What you get here is your own eyes on a clean pattern, which is enough to find faults and not enough to grade a panel.

The reported colour depth is not a measurement. Browsers return 24-bit almost regardless of the hardware, so a 10-bit panel and an 8-bit panel usually look identical through this API. Judge banding from the gradient rather than from that number.

Dynamic range reflects what the browser and operating system currently report, not what the panel is capable of. On Windows especially, a display marketed as HDR will report standard range whenever the HDR toggle is off, so a standard reading is a statement about your current configuration as much as your hardware.

Response time, overshoot and motion blur are outside what a static pattern can show, and the moving-object tests that claim to measure them are limited by the same frame-timing noise that makes the refresh reading approximate. Nothing here says anything about how the panel handles motion.

Everything runs in your browser. The patterns are CSS, the capability values come from media queries on your own device, and nothing about your display is uploaded or stored between visits.