Last updated:- 19st April 2024
DxOMark has been in the image quality business since 2003. They empower photographers with good information and knowledge, helping them make informed and autonomous decisions. They translated this goal into a unique Image Quality Database, which many photographers use to learn more about their cameras. The test data also enable photographers to use their cameras more efficiently.
Did you know DxOMark does not support the "size and capture" theory? It's exciting to discover how they use a theoretical model to test, evaluate, and present their test data on camera sensors.
Please visit DxOMark.com and demonstrate your support by liking my EM1 II conversation with DxO. Add comments and ask DxO for updates on the EM1 III, EM1X, EP-7, EM10 IV, and OM-1.
Note:- The Olympus EP-3, EM5 II, and G11 images were done for this article...
With a small camera portfolio, DxOMark became one of only a few companies using a theoretical test procedure for Camera Sensors and IQ. With almost 400 tested cameras since 2012, they created one of the most reliable test results (databases) for camera sensors, lenses, and image quality.
This article reviews the test results DxOMark published over the past 11 years. These test results can be found in the sensor (image quality) database at DxOMark. I also included a short example of using these test results and how they benefit M43 photographers. Finally, I added an example of preparing any camera for "acceptable image quality" and what that looks like in real life.
See this DxO article discussing the challenges of designing mirrorless camera sensors...
Olympus EM1 II
How should we interpret the above DxO Scorecard? DxO publishes a scorecard on its website for each tested camera. The scorecard includes an Overall Test Score, Color Depth in Portrait mode, Dynamic Range in landscape mode, and Low Light ISO performance in Sport mode. The summary markers (below) apply to the overall score plus the 3 applications DxOMark reviews.
Also, study the summary notes from DxO in blue:-
Overall Score
- The DxO tests are done at the camera's base ISO and using a logarithmic scale.
- They measure sensor sensitivity, color depth, and the sensor's noise floor.
- The difference in sensor sensitivity is 0.3EV for every 5 points.
Portrait
- The higher the color sensitivity, the more color nuances can be distinguished.
- A 22 Color Depth means the sensor or camera is an excellent performer.
- Anything less than 1 point between 2 cameras is insignificant.
Landscape
- DxO regards a dynamic range of 12EV as sufficient (excellent).
- There is no significant DR variance between the EM1 II and 13.3EV.
- A 0.5EV difference between cameras can be regarded as insignificant.
Sports
- DxO measures the sensor's Saturation and SNR (logarithmic) at the camera's base ISO.
- The Low Light ISO rating is set at an SNR of 30db, a DR of 9EV, and an 18-color depth.
- With a linear progression on M43 sensors, an estimated OM-1 Low-Light ISO is +2000 *.
- Considering the GH5 & GH5 II, how much did the EM1 III and EM1X improve (+1600)?
- A Low Light ISO difference of 25% is the same as 0.3EV and only slightly visible.
* Why did I up my estimated OM-1 low-light ISO rating from 1867 (calc.) to +2000? The BSI sensor is more sensitive than Live MOS sensors, plus several TruePic X efficiency improvements. See this link for a more detailed description of the above DxO test procedures.
The DxOMark sensor database has several benefits for photographers. For example, the theoretical principles discussed in my articles reflect the theoretical model DxO uses to test camera sensors. A theoretical model says each sensor has a unique Noise Floor while highlighting the importance of Sensor Sensitivity, Saturation, and SNR as indicators to monitor sensor performance...
The illustration below explains image sensors, saturation, and SNR (noise).
What do we learn from the DxOMark database? Image sensors are better in 2023, and no 2 sensors are technically similar. The DxOMark database shows that the practical differences between sensors are tiny, and each image sensor is unique from a design or technical aspect. The DxOMark test results also showcase the improvements sensor manufacturers achieved in the past 20 years.
Study this link.
How to use DxOMark to Compare two cameras. The following example teaches us how to interpret the DxO test database or how we create a visual reference for Acceptable Image Quality. I used the Olympus EM1 II and Canon R because they represent Pro cameras from 2016 to 2018.
The Canon R and the Olympus EM1 II.
DxOMark uses two "Image Quality" benchmarks for digital cameras. The first is what they define as Good Image Quality, and the second is Acceptable Image Quality. Good image quality means a color depth of 22, a dynamic range of 12EV, and a saturated sensor (+30db SNR) at base ISO. Each of the three variables has to be within the specified range for a camera to have good image quality.
What do we learn from the Sports Low-Light ISO? Many "experts" mistakenly suggest this category determines the camera's ability to use high ISOs. This category only specifies Acceptable Image Quality. That means the ISO (amplification) and specific exposure at the sensor to saturate the sensor with an SNR of 30dB, a Color Depth of 18, and a DR of 9EV. Study Figure 1 and this article.
Acceptable image quality is helpful because it shows us more about the sensor's exposure behavior. Do the following test to get a feel for what acceptable image quality looks like with an EM1 II:-
- Find a similar scene to the above EP-3 photo and take an image in A-Mode at ISO400.
- Increase the ISO to 1250 (1312) and take another photo. Inspect your Raw File in WS.
Acceptable IQ does not imply we shouldn't use higher ISOs. This DxOMark rating says that any ISO value above 1312 (EM1 II) will underexpose the sensor, which reduces the color depth, DR, and SNR. We discussed sensor Saturation and SNR in these articles. Those studying my articles should be able to optimize any camera at higher ISOs. For example, we can increase (ETTR) the EM1 II/III exposure by up to 1EV. We can also explain why, plus manage visible image noise (Sony A7R4).
Is the visible noise the same for different cameras? An SNR of 30db does not mean the structure of the noise floor is identical for any two sensors. The visible part of the sensor's noise floor is different for the Olympus EM1 II and the Canon R. The reason is the sensor's noise floor is a function of sensor design and not sensor size. Take a few minutes to study the illustration in Fig 1.
How does one apply the Overall DxO Rating? One of the most critical IQ variables in sensor design is Sensor Sensitivity. The sensor's sensitivity depends on its Optical and Quantum efficiencies plus its Pixel Area. For example, Live MOS, BSI, and low pixel count sensors are Optically more efficient than higher MP or Standard CMOS sensors. See my articles for more info on sensor sensitivity.
Why is it incorrect to say the ISO adjusts sensor sensitivity? The native sensitivity of the sensor is measured and adjusted at the factory. This means the sensor is calibrated for digital cameras as it leaves the factory. Size and capture advocates like to provoke meaningless discussions about the ISO offsets in the DxOMark database. The ISO standard states that the image brightness should double or half at every full-stop ISO adjustment (amplification). Nothing more or less...
Conclusion.
- Basic sensor info - readout speed, AST, and DR.
- Info about the sensor's noise floor, like size and types.
- Create a test criteria to test & evaluate Image Processors.
- Test & present the main benefits of using different sensor types.
- The safety margin (ETTR gap) manufacturers apply at "100%" exposure.
See this article from DxOMark on sensor size and mobile phones - link.
All the best and God's Bless...
Siegfried