Reading a Pixel Response Chart Without the Marketing

Two monitors quote the same response time and behave nothing alike, which is the single most common source of confusion in display buying. The usual explanation is that one manufacturer is lying, but the more accurate explanation is that the figure is measured under conditions the listing does not state. Once you know which conditions matter, the numbers start to make sense.

We measured several panels with the same quoted figure and logged the full transition curves. The quoted values were all defensible under some definition, and the curves were clearly different. The difference lives in the transition chosen, the thresholds used, and whether overshoot was included.

What transition is being measured?

A pixel can move between many pairs of values, and the speed varies enormously between them. A transition between two dark shades is typically slower than one involving a bright value, and the slowest transitions are usually in the dark part of the range. Manufacturers quote the fastest.

That choice alone can account for a large gap between the quoted figure and the experience. A panel whose worst transitions are slow will smear on dark content while posting an excellent headline number, and the listing will never mention it.

What thresholds define the start and stop?

A transition is a curve, not an event, so any time figure needs two thresholds. The common convention measures between ten and ninety per cent of the change, which excludes the slow tail where the pixel finishes settling. A stricter definition would produce a larger, less flattering number.

Some measurements also exclude the initial delay before the pixel begins to move at all, which is a genuine part of what the eye sees. If the clock starts after that delay, the figure describes only part of the transition.

Is overshoot included in the number?

Almost never, and this is the most important omission. A pixel driven hard reaches the ninety per cent threshold quickly but continues past its target before settling, and the time to actually settle can be much longer than the quoted figure. The number is technically correct and practically misleading.

A complete chart shows the whole curve, including any excursion beyond the target. From that curve you can read both the speed and the artefact, which is exactly the combination you need to choose an overdrive setting.

Why does the average hide the problem?

Because the transitions that cause visible artefacts are the slow or badly behaved ones, and averaging pulls them toward the well-behaved majority. A panel with a few terrible transitions and many excellent ones produces a good average and a poor experience.

Charts that show the worst case alongside the average are far more useful, and they are rare. When a source publishes only an average, the honest reading is that the worst case is unknown rather than good.

What does a rise and fall asymmetry tell you?

That the panel behaves differently depending on the direction of the change, which is normal and varies by panel type. A large asymmetry means some content will look clean and other content will smear, on the same monitor with the same settings.

We logged both directions on every panel, and the asymmetry was often larger than the difference between two competing models. That makes it a more useful discriminator than the headline number, and it is almost never published.

How does overdrive setting change the chart?

Dramatically. The same panel produces a slow curve with no overshoot at the minimum setting and a fast curve with a large overshoot at the maximum, and both are the same hardware. Any response figure quoted without stating the overdrive setting is describing an unknown condition.

This is why independent measurements usually test several settings. The useful output is not one number but the relationship between speed and artefact across the range, which lets you pick the setting rather than guess it.

Does brightness change the measurement?

Yes, because the crystal’s behaviour depends on the voltage range being used, and the brightness control changes where in that range the transitions happen. We measured the same panel at two brightness levels and got different curves, with more overshoot at the lower setting.

That means a figure measured at a default brightness does not necessarily describe how the panel behaves at the brightness you will actually use. It is another unstated condition that makes published numbers hard to compare.

What about the panel’s age and temperature?

Both move the numbers. Crystals respond slightly differently as they warm up, and a panel measured cold behaves differently from one that has been running for an hour. Age produces a slower, gentler drift in the same direction.

These effects are small compared with the measurement-condition differences, but they matter for anyone trying to reproduce a published figure. A result taken in a cold room on a new panel is not the result you will get in normal use.

How should you read a chart that includes the full curve?

Look first at whether the curve crosses its target and by how much, because that tells you about artefacts. Then look at how long the curve takes to settle, rather than to reach a threshold, because that is what the eye integrates. Finally compare the slowest transitions rather than the fastest.

Doing this in that order reverses most conclusions drawn from headline numbers. Panels that look ordinary in a specification table often have the cleanest curves, and panels with impressive figures often have visible overshoot.

Is there a trustworthy single number?

Not really, but a settle time measured across a defined set of transitions, with overshoot reported separately, comes close. It is more work to produce and harder to market, which is exactly why it is rare.

In the absence of that, the most useful single figure is the worst-case transition time at a stated overdrive setting. It is pessimistic, comparable, and much closer to what you will see.

What should a buyer do with all of this?

Look for independent measurements that publish curves rather than figures, and compare the worst cases. If only the marketing number is available, treat it as a rough ordering rather than a description, and weigh refresh rate and panel type more heavily.

Then test the panel you receive with a moving pattern and your own eyes, at your own brightness. That test takes a minute and it beats any number in a listing.

What is worth measuring after the response curve?

Does the measurement depend on where on the screen you sample?

It can. Panels are not perfectly uniform, and transition behaviour varies slightly across the surface, particularly near the edges. Measuring at the centre and near a corner on the same panel produced different curves in our logs.

That is worth knowing before comparing two published figures from different sources. If one sampled centrally and the other did not, part of the difference belongs to the sampling position rather than to the panel.

Why do some charts show a matrix of transitions?

Because a single transition cannot describe a panel. A matrix covering several value pairs shows where the panel is slow and where it is fast, and the spread across that matrix is more informative than any average taken from it.

Reading a matrix is easy once you know to look for the outliers rather than the typical values. The slow entries are the ones that produce visible artefacts, and they are usually clustered in a predictable part of the range.

Is there a relationship between quoted figures and price?

Weak. We measured expensive panels with mediocre curves and cheaper ones with clean ones, because the quoted figure is a marketing choice as much as a measurement. Price tracks other qualities, such as uniformity and build, far better than it tracks response behaviour.

That makes independent curves the only reliable input for this specific decision. Everything else about the panel can be judged from the listing, and this one cannot.

Does the backlight affect the measurement?

It can, particularly on panels that dim the backlight rapidly. A fluctuating light source changes what the sensor sees at the moment of sampling, which adds noise to a response measurement. That is a genuine experimental complication.

Measuring with the backlight behaviour known, and preferably stable, removes it. Where a panel uses a dimming scheme, noting that alongside the figure is good practice.

Is a published curve trustworthy?

More than a published figure, because it contains more information and is harder to spin. You can see the overshoot, the settle time and the shape, and those cannot be hidden by choosing a favourable threshold.

Look for the conditions alongside it, particularly the overdrive setting and brightness. A curve without those is still useful and still incomplete.

The next figure is how the panel’s behaviour interacts with frame timing, because a fast panel fed unevenly still looks poor. Logging frame intervals alongside the response curve is the natural next step, and it usually explains the remaining complaints.