A closer look at the DXOMARK Display protocol

Earlier we presented you with the key points of what we test and score in the DXOMARK Display protocol. In this article, we’ll provide a closer look at our process of testing smartphone displays. We will look at the tools and methods that we use to scientifically evaluate display quality attributes, which are based on specific use cases that reflect the ways in which people use their phones: web browsing, night reading, in-car navigation, taking photos, viewing photos, gaming, and watching movies, including how smoothly and efficiently a display’s auto-brightness function responds to changing light conditions.

Before we head into the main topic, it’s important to remember that smartphone display performance isn’t just about display panel quality. Smartphones embed software with dedicted algorithms to control many display functions, and manufacturers choose which settings to use with those algorithms (a process known as “tuning”). Of course, some algorithms are more efficient than others, and the way an algorithm is implemented on a smartphone can make a big difference in performance, as in these examples:

  • Software determines how smartphones balance the trade-off between frame rate and battery usage; depending on the apps used, some phones automatically adjust the frame rate to extend a battery charge (and thus autonomy). What this means is that a smartphone with a refresh rate of 120 Hz does not always refresh the screen at 120 Hz (for example).
  • Many smartphones include an ambient light sensor, a photodetector that gauges surrounding lighting conditions; tuning determines how quickly and appropriately the auto-brightness feature responds to the input from the light sensor, in addition to how well the display adapts to the content being viewed.
  • When people watch videos on their phones, motion interpolation algorithms generate frames in between “real” (existing) frames with the aim of making animations or moving actions appear smoother, and again, the battery vs. frame rate trade-off can have an impact here. (We will visit algorithms again in some of our articles about specific display attributes. Read about the pivotal role software tuning plays in display performance here. )

DXOMARK conducts tests under many different (and sometimes changing) light conditions so as to recreate as closely as possible the real-world experiences of smartphone users, rather than just simply pitting display performance against “ideal” viewing conditions as defined in standards/norms.

Finally, as we head into our toolbox, just a short reminder that first, we test each and every display under the exact same conditions so as to ensure that our results are fair, scientifically rigorous, and repeatable. Second, apart from certain well-defined exceptions, such as color accuracy,  we test devices using their default settings. And third, DXOMARK measurements differ from those of other sites in that we not only include lab-based objective measurements, but perceptual measurements as well.

Objective testing tools

The images below show the array of tools our evaluation experts use when testing display light spectrum, color, luminance (brightness), contrast, uniformity, frame drops, and judder:

Testing devices that DXOMARK uses to measure display quality, from left to right, spectroradiometer, video colorimeter, video colorimeter with conoscope, and compact camera.

We use these tools to measure reflectance, gloss, flicker, and lighting conditions:

Other testing devices, from left to right: spectrophotometer, glossmeter, flickermeter, lux meter, and colorimeter.

We use the tools below to measure touch responsiveness, accuracy, and smoothness:

Ultra-high-speed camera and robot for measuring display touch attributes

We conduct many of our objective tests within the DXOMARK Display Bench, which is a special testing chamber that facilitates testing automation and ensures that our engineers test all devices under the exact same conditions. It includes mounts for devices being tested and for testing tools (principally a spectroradiometer and a video colorimeter), computer-controlled LED lighting arrays to imitate all kinds of lighting types and brightness levels, and lux meters.

A device under test and video colorimeter inside the DXOMARK Display Bench
A device under test and spectroradiometer inside the DXOMARK Display Bench

In both photos showing the inside of the DXOMARK Display Bench above, you can see a device under test (DUT) mounted on the left, with the testing instrument on the right mounted on a rail; testing engineers use computer-controlled servo motors to move the instrument to various distances from the DUT. During testing, the Bench is first sealed against any external light sources, and an engineer controls the tests via computer.


In addition to the Display Bench, we have developed a fully computer-controlled Dome System, which serves to reproduce more intense outdoor lighting conditions of up to 50,000 lux. The shape of the dome allows the very intense light to be diffused so that it hits the smartphone’s screen from all directions, in the same way that we experience lighting conditions outdoors. But the dome’s ability to reach extreme levels of brightness permits us to really challenge the limits of a device’s screen capabilities.

The Dome System in display testing

In the photo above, the DUT is attached to a rail within a chamber, with the screen facing the testing instrument. A lux meter sensor, which monitors the intensity of the light, is next to the DUT. The testing instrument, the Radiant imaging colorimeter, which is mounted on an external rail on the other side of the dome (not pictured), acquires contrast and brightness measurements through a hole at the top of the dome as the DUT’s screen displays testing patterns for measurement.


Every element of the system–the DUT, the motors (or modules) controlling the light levels, and the instrument– is controlled by a computer.

In the objective part of our testing, we measure color, contrast, luminance, reflectance, EOTF curve, etc. by displaying calibrated video and photo patterns on the device’s screen. The perceptual part of our testing involves playing a set of videos that we have produced and mastered, enabling a repeatable evaluation using real content in SDR and HDR formats.

A small sample of DXOMARK’s charts, patterns, and visuals that are used in display testing.

Perceptual testing tools

One of the most important tools DXOMARK relies on for its perceptual testing is the human eye. Our perceptual tests confirm and complement our objective tests, in that we want to be sure that we can see in real life what the objective measurements are telling us. Further, objective tests measure only what they are strictly aiming to measure. Given the complexity of the software driving the display as well as the complexity of the human visual system, perceptual tests are an essential ingredient in evaluating display quality.

Our Display protocol engineers receive careful and extensive training before conducting any perceptual tests, some of which involve closely evaluating multiple devices (a DUT and two or three comparison devices) against reference images displayed on a professional monitor. The color and brightness values of each carefully chosen image on the pro display have been precisely calibrated and measured. When making comparisons, engineers follow a very strict and scientifically sound protocol that requires conducting the test multiple times using different engineers each time to avoid any bias.

In addition to our most important perceptual tool (the human eye) our display engineers use a specially-designed smartphone apparatus for holding several smartphones at once, pro-level monitors, and lux meters.

Our engineers perform all perceptual evaluations by looking directly at the device’s display. We take photos only to use as illustrations, but never use them as a basis for any kind of test or evaluation.

Display protocol tests

The tables in each sub-section below cover all of the attributes that the DXOMARK Display protocol currently tests, and include information about the equipment we use, some of the testing conditions, and some of result parameters and definitions.


In our reviews, we regularly remind people that the most important consideration for end-users is how easily and comfortably they can read the display under different real-life conditions. DXOMARK uses its Display Bench and its dome system to recreate ambient light conditions ranging from total darkness to bright daylight (0, 25, 250, 830, 20 000, 50 000 lux).

Objective measurements done in the lab are always completed with perceptual evaluations, allowing us to assess the device’s performance in real-life situations.

Below is a sample graph of comparison data showing brightness/contrast measurements for three devices:

Luminance under various lighting conditions
This graph shows the screen luminance in environments that range from total darkness to outdoor conditions. In our labs, the indoor environment (250 lux to 830 lux) simulates the artificial and natural lighting conditions commonly seen in homes (with medium diffusion); the outdoor environment (from 20,000 lux) replicates a situation with highly diffused light.
Contrast under various lighting conditions
This graph shows the screen’s contrast levels in lighting environments that range from total darkness to outdoor conditions. In our labs, the indoor environment (250 lux to 830 lux) simulates the artificial and natural lighting conditions commonly seen in homes (with medium diffusion); the outdoor environment (from 20,000 lux) replicates a situation with highly diffused light.

In the example above, you can see how the measured contrast in daylight conditions does not live up to the claimed contrast values of 1:1,000,000 (or infinite), which are based on measurements taken in dark conditions (< 0.01 lux). Our measurements show what users experience — that is, it hard to read our screens in sunlight.

Another test of display readability measures the homogeneity or uniformity of brightness output, as shown in the illustrative image below:

Brightness uniformity test, left; False-color luminance map measurements, right.
Photos for illustration only

Our readability tests also include looking for things that affect readability and the user experience, such as artifacts, mainly flicker and reflectance.

Flicker is a phenomenon associated with the temporal modulation of light, the quick oscillation of light output between on and off on a screen. All screens have temporal light modulation to some degree because of the interaction between the screen’s refresh rate (whether 60 Hz, 90 Hz or 120 Hz) and Pulse Width Modulation (the power that turns the light on and off for a certain duration.) Flicker relates to temporal modulation frequencies inferior to 90 Hz .

Flicker is known to create unease, eye fatigue, or in the most extreme cases, seizures. Additionally, its impact varies considerably among individuals; some people are even able to perceive the modulation. The effect of flicker tends to be stronger under dim environments as the screen and our eyes adapt to the darker light.

That’s why this measurement is important for assessing display comfort.

DXOMARK measures the behavior of a smartphone’s flicker in order to assess flicker perception. For our Eye Comfort label, the detection probability of flicker should be below 50%, in default mode or with the anti-flicker mode activated (if available).

For example, flicker tests reveal that slow pulse-width modulation (PWM) can have an impact on visual comfort even for devices with a high refresh rate. (In the graph below, the first spike corresponds to the refresh rate, and the highest spike corresponds to the PWM.)

Temporal Light Modulation
This graph represents the frequencies of lighting variation; the highest peak gives the most important modulation. The combination of a low frequency and a high peak is susceptible to inducing eye fatigue.

Reflectance is an artifact that affects readability. Smartphone screens are reflective by nature. But the degree to which they reflect light affects the user experience. Reflectance impacts the contrast of the content being displayed.

To give you an example, the reflection from a simple glass sheet is around 4%, while it reaches about 6% for a plastic sheet. Although smartphones’ first surface is made of glass (or plastic for foldables), their total reflection (without coating) is usually around 5% (or higher) due to multiple reflections created by the complex optical stack that is sometimes coated with an anti-reflection layer.

To determine the device’s reflectance, we measure the reflected light intensity as a function of wavelength over the visible spectrum (400 nm to 700 nm). We use a spectrophotometer in SCI (Specular Component Included) mode to perform reflectance level measurements on smartphone displays when turned off.  The SCI mode measures both the diffuse reflection and the specular reflection.

We then calculate the average based on the measurements within the visible color spectrum.

We also use our spectrophotometer in SCI (Specular Component Included) mode to perform reflectance level measurements on smartphone displays that are turned off. Below are measurements that show the reflectance level for each 10 nm-spaced wavelength within the visible spectrum range (400 nm to 700 nm).

Average Reflectance (SCI) iPhone 15 Pro Max
4.8 %
Apple iPhone 15 Pro Max
Google Pixel 8 Pro
Samsung Galaxy S24 Ultra
SCI stands for Specular Component Included, which measures both the diffuse reflection and the specular reflection. Reflection from a simple glass sheet is around 4%, while it reaches about 6% for a plastic sheet. Although smartphones’ first surface is made of glass, their total reflection (without coating) is usually around 5% due to multiple reflections created by the complex optical stack.
Average reflectance is computed based on the spectral reflectance in the visible spectrum range (see graph below) and human spectral sensitivity.
Reflectance (SCI)
Wavelength (horizontal axis) defines light color, but also our capacity to see it; for example, UV is a very low wavelength that the human eye cannot see; Infrared is a high wavelength that the human eye also cannot see). White light is composed of all wavelengths between 400 nm and 700 nm, i.e. the range the human eye can see. Measurements above show the reflection of the devices within the visible spectrum range (400 nm to 700 nm).

Unless specified otherwise, all tests are conducted at light levels ranging from 0 to 50,000 lux and using white color temperature/spectrum tungsten, white LED, D65 color illuminants, etc.

Sub-attribute Equipment Remarks
Vs. ambient lighting Bench + spectroradiometer (brightness, given as cd/m2) + video colorimeter (contrast, given as :1) Light levels: 0,25,250,830 lux

Dome +
video colorimeter (brightness, contrast, given as :1) Light levels: 20,000 lux and 50,000 lux

Brightness should adapt to viewing conditions; screen content should be readable in any condition and be as close as possible to original intent.
Vs. average pixel level Bench + spectroradiometer (brightness) + video colorimeter (contrast) at 20,000 lux

Dome + spectroradiometer (brightness) + video colorimeter (contrast) at 20,000 lux and 50,000 lux

Neither brightness nor contrast should change with APL.
Brightness vs. time Light booth with changing lights and brightness levels. Light levels: 0 and 830 lux We investigate reaction time, smoothness, transition time
EOTF* Bench + spectroradiometer Tested under various light conditions (0, 830, 20,000 lux) at 20% APL; the closer to the target value of gamma, the better.
Uniformity Video colorimeter + standard lens Tested at 0 lux; results are given as a percentage (the higher, the better).
Vs. angle Video colorimeter + conoscope Tested at 0 lux; the lower the loss of brightness, the better.
Screen reflectance Spectrophotometer (+glossmeter with display off) A reflectance result of under 4% is considered good.
Flicker Flickermeter Flicker frequency corresponds to the highest peak on the graph. The higher the frequency, the better.

*EOTF stands for Electro-Optical Transfer Function, which converts an electronic signal into a particular level of brightness on a display.

All objective tests done in the laboratories are then complemented by a series of perceptual tests.



From the end-user’s point of view, color fidelity — that is, having the display faithfully reproduce the exact same hues and shades that they see with their eyes — is second in importance only to readability.

We use a conoscope in the setup below to evaluate how color shifts when users view display content on axis versus when they look at content on a screen held off axis (tilted up to 70°).

Setup of conoscope-equipped video colorimeter (for illustrative purposes only; actual testing takes place at 0 lux).

We perform color fidelity measurements for different lighting conditions to see how well the device can handle color management under different ambient lighting conditions. Below is just one of our color fidelity results taken under a D65 illuminant at 830 lux.

Color fidelity