Open Letter to DxO — We Need a Universal Method for Lens Testing and Comparison

Bengt Nyman

First a thank you, DxO, for introducing the expression P-Mpix — or Perceptual Megapixel — for describing the sharpness of a lens. It elegantly translates the older one-dimensional line pairs per millimeter measurement into a two-dimensional resolution matrix directly comparable to the MP specification of a camera image sensor.

I believe that the P-Mpix concept also opens up the door to more effective testing and comparison of lenses.

DxO’s present testing method for camera image sensors is straightforward and easy to use for comparisons and decision making. However, this is not the case for DxO lens testing. The present method involves testing one and the same lens on a large number of cameras.

Not only does this create an unreasonable workload with never ending delays and customers waiting for test results, but worse, it obscures any meaningful comparison between lenses for different types of cameras.

In today’s market with mirrorless cameras using adapters to mount a range of old and new lenses we need a lens test method that is independent of the camera.

For example: how sharp is a Zeiss T* 35mm F2 lens?

We know that on a Sony RX1 it scores 37 in total and 20 P-Mpix in sharpness.

Since the camera used for testing is a 24MP camera I can guess that the lens is of decent quality since it scores 20MP out of 24MP possible. A theoretically perfect lens on a 24MP camera should obviously yield an image with 24MP clearly distinguishable pixels.

Testing a lens on a camera with a low quality image sensor is therefore only meaningful when comparing the result to a different lens on the same camera. Even a theoretically perfect lens will never make this camera see any better than what its image sensor is capable of.

Maximum resolution of a lens must therefore be measured with a technique where the image sensor is capable of resolving much finer patterns than what the lens is capable of.

The technique would involve using the highest resolution sensor available to measure all lenses using the same test pattern and the same super resolution sensor.

For the sake of argument let us do a thought experiment using a Nikon D800E 36MP sensor to test all lenses. We are lucky in that the Nikon consumer lens with the best sharpness — or highest resolution — is the Nikkor 200mm F2 with a sharpness score of 28. I say lucky because this score of 28 is sufficiently lower than the 36MP of the sensor to make meaningful comparisons to other lenses.

Now test another full-frame lens, for example the Zeiss T* 35mm F2 in the same setup. To estimate the outcome of this thought experiment we have to make some assumptions and calculations.

The Zeiss scores 20MP on the 24MP Sony RX1. This does not mean that it will score 20MP on the Nikon 36MP sensor. On the Sony RX1 this lens enables 20/24 = 83% of the sharpness potential of the sensor. I would think the Zeiss lens will score similarly or approximately 83% of 36 when tested on the Nikon 36MP sensor. In other words: it will score approximately 0.83×36 = 30MP.

Assuming that my rough approximation is acceptable, we can now compare the Nikon lens yielding a maximum resolution of 28MP to that of the Zeiss lens yielding an approximate resolution of 30MP.

DxO needs to come up with (or go back to) a universal method to test lenses independent of manufacturer specific cameras. This would give us users a much better way to compare and choose lenses.

This would also reduce or eliminate DxO’s need to test the same lens on a large number of cameras. Especially if the lens resolution was given as a maximum P-Mpix resolution.

The consumer can then quickly see how well the maximum resolution of a particular lens compares to the resolution of a particular camera, and approximately how well a particular lens will enable or compromise a particular camera.

Thank you, DxO.