Gear

What resolution costs you underwater

Fifty megapixels in Bunaken and Lembeh. Diffraction arrived at f/11 instead of f/25, and every high-resolution body since has made the same trade.

Gear · Tested 2015 in North Sulawesi · Revisited 2026

In 2015 I took a Canon 5DsR to Bunaken and Lembeh to find out what fifty megapixels does underwater. The answer was not what I expected, and the lesson has outlived the camera by a decade — because every high-resolution body since has made the same trade.

Diffraction arrives much earlier

This is the finding that matters. Macro work pushes you to small apertures because that is how you hold a whole subject in focus at life size. On the 5D Mark III I could generally get to f/25 before softness became intolerable. On the 5DsR the images started going soft at f/11.

That is not a fault in the camera. The diffraction-limited aperture is a property of pixel pitch: 6.7 on the 5DsR against 10.1 on the Mark III. Pack more photosites into the same full-frame sensor and the point at which diffraction becomes visible at the pixel level moves towards wider apertures. More resolution, less usable aperture range. You cannot have both.

Four macro frames at different apertures with matching one-to-one crops
Four frames at different apertures, each with a 1:1 crop. Deliberately not exposed to flatter — the point is where the detail goes, not how the picture looks. Shot full manual with a port focusing knob.
You do not buy resolution. You trade aperture for it, and macro is where the bill arrives.

What that means in the water

You stop reaching for the aperture ring and start solving focus a different way. Get closer. Put a wet dioptre or a doubler in front of the port instead of stopping down another two stops. Light harder so you can afford to open up. Compose so the plane of focus does the work rather than depth of field rescuing it.

It also punishes technique. A number of frames I was certain I had nailed turned out to be fractionally off once I saw them at full size — the sensor resolves your focus errors as readily as it resolves the subject. When you do nail it, the file takes a heavy crop and still holds, which is the compensation.

The smaller findings, which still hold

Low light
Two-thirds to a full stop behind the 5D Mark III, with more noise in the shadows. Underwater this matters less than it sounds — you are bringing the light with you.
Auto white balance
Less consistent than the Mark II, running cooler and warmer between frames of the same scene. Nothing that a pass in Lightroom did not settle, but a reason to set it rather than trust it.
Files
51 to 56 MB each. In 2015 that meant a year-old MacBook Pro with 16 GB of RAM struggled to render previews even with 40 GB of cache allocated. The files have not shrunk since.
Video tracking
Better than expected on small moving subjects with the 100mm, though 100mm is not the lens for macro video. A 60mm, two decent video lights and something to rest on is the honest setup for that.

What I tested with

Aquatica housing, twin Sea & Sea strobes, two focus lights, carbon-fibre float arms, a doubler, and the Canon 100mm macro. Full manual throughout, focusing with a port knob rather than letting the camera hunt.

A frogfish photographed in Lembeh Strait at f/13
Lembeh, f/13, ISO 100. The full-resolution file, so you can judge the rendering rather than take my word for it.

Why a decade-old test still applies

I shoot a Canon R5 Mark II now. It is a different camera in almost every respect, and the trade at the centre of this piece has not moved: high pixel density buys you detail and cropping room, and charges you for it in usable aperture and in tolerance for sloppy focus. If you are choosing a body to put in a housing, that is the question to ask about it — not how many megapixels, but where the softness starts.

Prototype — not live