Tips and advice

Measuring and calibrating a TV: Interpreting measurements and graphs

Measuring and calibrating a TV: This article will help you to interpret and understand the measurement and graphs we publish in our reviews
4.4/5 - (14 votes)

Measuring and calibrating a TV: In our TV reviews we provide you with a number of measurement results and graphs. They give a solid picture of what you can expect from television. But unless you are familiar with color theory, screen measurement and calibration, they are probably difficult to interpret. In this article, we’ll give you the necessary basic knowledge to solve that.

Measuring and calibrating a TV – How are our measurements taken?

All our measurements are made in the image mode closest to the relevant standards (Rec.709 for SDR and Rec.2020 for HDR). In almost all cases, that is the ‘Film’ or ‘Cinema’ image mode. We disable all image processing that interferes with the measurements. Typical examples are the light sensor, or local dimming. In this way we measure the performance and therefore measurement results that you as an end user can expect.

Measuring and calibrating a TV – Contrast

The contrast of a screen is the ratio between the brightest value the screen can display and the darkest value. For example, a screen showing white with a maximum luminance of 200 cd / m² and black with a minimum luminance of 0.1 cd / m² has a contrast of 200 / 0.1 = 2,000: 1.

Measurements for contrast are very much influenced by the way you measure. The ANSI contrast uses a chessboard pattern and is therefore a good indicator of the display’s own contrast. After all, black and white are displayed simultaneously. On / off contrast is often much more favorable because the measurement successively measures a white test image and a black test image. To prevent the screen from turning off completely, we perform that measurement with a white or black window that occupies 10% of the image on a light gray background (which varies to create an identical average brightness). In addition, you must take into account the accuracy of the meter. Our C6 HDR2000 has an accuracy of 4%. There is therefore an error of 8% on the contrast.

LCD screens typically achieve native ANSI contrast at around 1,000: 1 (IPS panel), and 3,000: 1 (VA panel). Local Dimming can greatly increase that number, to 10,000: 1 or more. OLED screens register on our meter as perfectly black, although a laboratory-class meter would still measure something there. For the living room, you can consider OLED black as perfect.

Measuring and calibrating a TV – Color temperature

White light is not a universal given: the ‘white’ light of an incandescent lamp looks very different from that of a typical tube lamp, and both are different from sunlight. Therefore, it is imperative that we define ‘white’.

The white of the HDTV standard is defined by D65, a standardized theoretical light source that corresponds to daylight. D65 has a color temperature of 6504 K (Kelvin, a temperature unit from thermodynamics). Higher color temperatures have a blue tone, lower color temperatures have an orange / red tone.

Because we usually associate light blue with winter and ice, we refer to ‘bluish’ images as cool. That is exactly the opposite of what pure science says: after all, we see blue shades at high color temperatures! Keep this in mind: when we speak of ‘cool’ images, the color temperature is high, and when we experience the images as ‘warm’, the color temperature is low. Not exactly intuitive, but it is no different.

On a television, the correct color temperature (6504 K) is almost always the ‘warmest’ setting. So choose ‘warm’, ‘warm2’ as the color temperature, which setting is almost always correct when you select the ‘film’ image mode.

Measuring and calibrating a TV – Grayscale

The gray scale shows the measurement result in 21 steps from black to white. All gray values ​​in between must have the same neutral tone, which corresponds to D65 white. The uniform rendering of the gray scale is extremely important for correct color reproduction of the image and has an impact on all images, not just gray tones.

SDR Grayscale – Color temperature

Top right: This graph shows the average color temperature. Ideally, we aim for 6504 K (yellow line). Higher color temperatures are ‘cooler’ and give the screen a blue tint, lower color temperatures are ‘warmer’ and give the screen a red tint.

SDR Grayscale – RGB balance

Bottom right: This graph explains how the color temperature deviates from the standard. The graph shows the relative balance of red, green and blue for each gray value. Ideally, those three lines coincide in measurements at 100. If they do not coincide, you still want the same horizontal course of each line. For example, if the red line runs consistently at 95, then there is a light shortage of red and the image has a slight cyan tint. The most visible deviation is an excess or shortage of greenery. If the graph is very erratic, this may be very visible because one shade of gray is, for example, somewhat red while another is somewhat green. You can see how visible these errors are from the following graph.

SDR Grayscale – Delta2000 errors

Top left: This bar chart shows the DeltaE2000 error for each gray value. Errors less than one are not visible to the naked eye (reference level). Errors below three are visible in some cases but are probably invisible in the living room and without a reference. Errors above three can have a visible effect. Below the graph we give the mean error. Of course we aim for an error of less than three in the measurements.

SDR Grayscale – Gamma value

Bottom left: Here you see the progression of the gamma value. The gamma value determines how the brightness of the grayscale changes as the signal evolves from dark to bright. This is not done linearly, but according to an exponential curve. The gamma value determines how the curve develops, which mainly affects the lower and middle gray values. Lower gamma values ​​make blacks and grays brighter. They provide more black detail, but also a somewhat paler image. Higher values ​​make black detail darker and the image therefore gets a more contrast-rich appearance.

The standard for the gamma curve is BT.1886. This uses a 2.4 gamma value but takes the true black value of the television into account, so that black detail remains more visible. Manufacturers sometimes also aim for a classic 2.4 gamma value, or the still popular 2.2. As long as the gamma value remains between 2.2 and 2.4, we are generally satisfied. In a dark viewing environment you can use 2.4, in a bright viewing environment 2.2 makes more sense. BT.1886 should be relatively universally applicable.

Measuring and calibrating a TV – SDR Color Checker

In this graph we measure the performance of the general color rendering. The test colors are based on the widely used GretagMacbeth color chart. On the right you see the representation of the colors in the CIE 1976 color diagram. The squares are the target, the round dots are the measured values.

On the left you see the bar graph with the deltaE2000 error. As always, error values ​​below one are not visible (reference). Errors below three are visible in some cases but are probably invisible in the living room and without a reference. Errors higher than three can often be visible. Skin tones are the first two colors in this graph next to the gray values, and in the diagram they are briefly circled in white for convenience.

SDR Color saturation

Color saturation determines how bright a color looks. At 0% saturation you have white, and as you have greater saturation values, you add more color.

Top left: This bar graph shows the DeltaE2000 error, for each color in five saturation steps. Error values ​​lower than one are invisible, error values ​​smaller than three are visible in some cases but are probably invisible in the living room without a reference. Errors higher than three can often be visible.

You will find the color diagram on the right side of the measurements. Squares are ideal values, circles are measuring points. For each color you start from the center of the diagram (white) and see the gradient to the full color in five steps. This screen has a relatively minor error in red and magenta. On the color diagram you can see that red is generally too saturated (measuring points run too quickly to the full color). Of course, this also affects magenta (a secondary color: blue + red), but magenta is clearly also drawn a little too much towards blue.

Measuring and calibrating a TV – HDR measurements

We also measure the performance in HDR. Of course different standards apply there, but the interpretation of the results is largely identical. In HDR we also use a different error formula, namely DeltaEICTP_240. All measurements are made for HDR10.

HDR Grayscale – RGB balance

Top left: This graph explains how the color temperature deviates from the standard. The graph shows the relative balance of red, green and blue for each gray value. You interpret this in the same way as with SDR.

HDR Grayscale – DeltaEICTP_240 errors

Bottom left: This bar chart shows the DeltaEICTP_240 error for each gray value. Errors less than one are not visible to the naked eye (reference level). Errors below three are visible in some cases but are probably invisible in the living room and without a reference. Errors above three can have a visible effect. Below the graph we give the mean error. Of course we aim for an error smaller than three.

We also provide two graphs. The top graph also shows the luminance error, the bottom graph only looks at the color deviation. If the error without luminance is very small, but the error with luminance is very large, then the gray scale is neutral and correct, but the gray values ​​are too dark or too bright.

HDR EOTF – Luminance

Top right: This is the progression of the EOTF (Electro-Optical Transfer Function). The EOTF determines how the electrical signal is converted into brightness. The gamma curve we use in SDR is also an EOTF. HDR10 uses a different standard, namely PQ (Perceptual Quantizer). The yellow line indicates the ideal curve, the black line is the measured result. If the result is above the yellow line, the image is too bright and vice versa. The EOTF ends in a plateau, depending on the maximum luminance of the screen. For this we look at the following graph.

Bottom right: The gamma curve for SDR is relative. The course has been determined, but the absolute values ​​depend on the measured maximum. However, the PQ EOTF is an absolute EOTF. This means that a certain signal value always corresponds to a certain brightness, regardless of the maximum brightness of the screen. In this graph you can see how the curve develops and to what maximum it ends. Where the curve approaches the maximum of the screen, it bends slowly or to hide no or minimal white detail. We call this operation tone mapping. There are no rules for how this is done. But if the screen flattens the curve too quickly, you can darken content where it is not necessary. If the screen flattens the curve at a right angle, you may lose white detail.

Our measurements are made with metadata that signals a maximum of 1000 cd / m². The curve must therefore eventually reach a plateau at value 75.

HDR Peak luminance vs Window size

In this graph you can see the maximum brightness that is achieved in a white window that occupies x% of the screen. The measurement on the far left is the 1% window, and can be seen as an indication of how bright the highlights can be in the picture. The measurement on the far right is a 100% white image. If it is much lower than the maximum, it means that the screen has certain limitations, usually related to power consumption. The example here is a typical OLED. On a completely white screen, they are considerably less bright than their maximum.

A maximum of 500 cd / m² is required for good HDR playback. If you really want to see the impact of HDR, 750-1,000 cd / m² is required.

HDR Color Checker

In this graph we measure the performance of the HDR color reproduction. The test colors are based on real HDR content. On the right you see the representation of the colors in the CIE 1976 color diagram. The squares are the target, the round dots are the measured values.

On the left you see the bar graph with the DeltaEICTP_240 error. As always, error values ​​below one are not visible (reference). Errors below three are visible in some cases but are probably invisible in the living room and without a reference. Errors higher than three can often be visible.

These measurements are very difficult for many screens, mainly because the colors are often very bright, and the errors with luminance are therefore very large.

HDR Color saturation

Color saturation determines how bright a color looks. At 0% saturation you have white, and as you have greater saturation values, you add more color. In HDR, we measure the saturation with respect to the UHDA-P3 color range as maximum, because no TV covers Rec.2020 completely.

On the right side you will find the color diagram. Squares are ideal values, circles are measuring points. For each color you start from the center of the diagram (white) and see the gradient to the full color in five steps. The interpretation is the same as for SDR.

Left: This bar chart shows the DeltaEICTP_240 error, for each color in five saturation steps. Error values ​​lower than one are invisible, error values ​​smaller than three are visible in some cases, but are probably invisible in the living room without a reference. Errors higher than three can often be visible. Again, the error with and the error without luminance, so that you can estimate whether the error is mainly due to color errors or from a too bright or too dark display.

HDR Color range

In addition to maximum brightness, the color range is also very important for HDR display. After all, the Rec.2020 standard covers a much wider color range than SDR Rec.709. Full coverage of Rec.2020 is currently beyond the capabilities of all TVs. We therefore aim more at UHDA-P3. In this graph you can see what percentage of Rec.2020 and UHDA-P3 the television can cover. You can also pretty well see in which color the TV falls short or not.

For good HDR images we aim for 90% UHDA-P3 (measured in the CIE 1931 xy model). We also include the numbers measured in (CIE 1976 uv), which are often used by manufacturers. However, the xy numbers are a better indication of performance.

More information

Do you want to know more about optimally setting or calibrating your television? Or are you looking for information about connecting your TV? Then read our tips and advice section .