Is light falloff a problem with a full-frame sensor?

I think so. Measurements of light falloff were made on my Canon EOS 5D camera using the Canon EF24-70mm f2.8L USM lens. The graph shows the light falloff in f-stops relative to the centre, as a function of aperture. The lens was tested at 24, 35, 50 and 70mm focal lengths. For each focal length, the solid line gives the mean falloff at the corners of the image (="vignetting"), and the dotted line gives the mean falloff at the centre of the left and right sides (landscape orientation).

Graph of 24 to 70 mm zoom light falloff

It seems to me that light falloff is noticeable if it is half an f-stop or more. On that basis the following conclusions may be drawn:

Of course, one can always correct falloff in Photoshop or DxO Optics Pro.


It is important to remember that, in so far as light falloff is caused by the lens, it will be no worse on a digital full-frame camera than it is on a 35 mm film camera. The one exception to this is where falloff is due to light striking the sensor at an oblique angle (a problem not suffered by film). Modern telecentric lens designs try to minimise this, but I don't know if that applies to the Canon lens tested.

I have only tested one sample of the EF24-70 f2.8L lens. Mine might be below par.

Technical details

The target was the external wall of my house on a dull, overcast day. Spot readings showed it was evenly illuminated. The wall is smooth rendered and painted in a Magnolia colour; however, the camera's automatic white balance converts that to grey. The camera was my EOS 5D set at ISO 100, picture style "standard", large fine jpeg output. The lens was focussed at infinity but the actual target distance was 0.7 m so it was deliberately out of focus. A cable release was used to allow me to keep out of the way to minimise any shading effect.

Exposures were made at about EV 17.6, corresponding at ISO 100 speed to, for example, f8 and 1/100 sec. This exposure put the peak corresponding to the target about two thirds of the way across the 5D's histogram.

The resulting images were analysed using Norman Koren's IMATEST program. The IMATEST light falloff analyser needs to be told the gamma of the camera so that it can convert pixel values into f-stops. The gamma value I used was 0.69 since this is the mean of the values I have obtained for this camera when producing jpegs in standard picture style.

Removing the Canon standard lens hood did not improve the falloff figures.

IMATEST output

Canon EF24-70mm f2.8L USM at 24 mm

Imatest figure 2 for 24 mm at f2.8
Download the JPG file (2.3 MB)
Imatest figure 2 for 24 mm at f11

Canon EF24-70mm f2.8L USM at 50 mm

Imatest figure 2 for 50 mm at f2.8
Download the JPG file (2.3 MB)
Imatest figure 2 for 50 mm at f5.6

Something wrong with this analysis? Let me know!

Peter Facey, Winchester, England