Almost always the lens rather than the camera, and usually at its widest aperture. Stopping down a stop or two is the test that costs nothing, and a filter or step-up ring in front is the second thing to remove.
Optical performance falls away from the centre of an image circle. That is a property of lenses, and it is loudest wide open.
The order worth testing
- Stop down one or two stops from the maximum aperture and take the same frame again. On a variable zoom, remember the maximum differs along the range — f/3.5 at one end, f/6.3 at the other.
- Remove any filter or step ring, then repeat. Four titles in this catalogue print a filter thread, in three different diameters; a ring pushes the glass forward and the effect shows first at the corners.
- Check the image circle. A lens built for a smaller format on a larger body dims and softens the outer frame by design, and a crop mode is the maker’s own answer to it.
What it is not
Not the sensor, which samples uniformly across the frame. Not the pixel count. And not stabilisation, which addresses movement and would soften the whole frame rather than a ring of it.
What is never published
How sharp the corners are at each aperture. It is the figure this question most needs and the one no listing states, which is why the two tests above are worth more than any specification sheet.
What people ask before they commit
Will stopping down always fix it?
It improves most lenses at the edges, and there is a limit past which diffraction softens the whole frame instead. Between those two, every lens has a range where it performs best — and no maker publishes that range in a form two lenses can be compared on, which is why the test is done rather than looked up.
Could a filter be causing it?
A thick mount can, especially on a wide-angle lens, and a step-up ring pushes the glass further forward still. Makers publish ring diameters; none publishes what appears in the corners as a result. Removing it for one frame settles the question in seconds.
Last reviewed 12 September 2026