Smartphones v. Digital SLRs v. Film - Technical details

This page has all the technical details to complement the main article "Smartphones v. Digital SLRs v. Film - How far have we come?".  If you haven't read that article yet, it would make sense to head there first.

The bag packed ready for this test.

Without doubt, this was the strangest collection of cameras I've ever taken on a job.

7 SLRs (one film!), and 4 phones

Why did we resize all of the images to around 20 megapixels?

When cameras have similar resolutions, it’s easiest to compare them by looking closely at their pictures in ‘Actual pixels’ or 100% view, where one pixel from the camera is one pixel on your monitor. But these cameras vary more than 6-fold in resolution, and I found it too hard to interpret differences by using actual-pixels views. It was comparing a small 6-inch print from one camera with a 15-inch enlargement from another, and trying to decide which camera is better. I don’t care how good they look per pixel, I care how good they look per picture.

So we levelled the playing field by making them all the same size first. We blew them up or shrunk them down in Photoshop CC to about 20 megapixels - equivalent to an 18-inch wide print scrutinised really closely. But when you change the size of pictures, they end up fuzzier than they should be, so we used the closest thing to CSI’s mythical “Enhance that!” button - Photoshop’s Smart Sharpen - to show off the detail that each camera can give.

Resizing pictures adds loads of biases, especially in choosing the size to compare them. Make it too small, and the lower-resolution cameras look better than the others, because any sharpening that the camera adds is more appropriate to smaller photos. Make it too big, and the higher-resolution cameras gain an advantage that’s out of proportion to what you’d ever see outside of a lab.

We experimented, and found making all of the photos about 20 megapixels gave a reasonable compromise. It’s equivalent to full-quality 18 inch print at 300 dpi. It’s a good stretch for the low-resolution cameras, while the high-resolution cameras don’t lose all of their potential advantages. We resized each photo’s long edge to 5,400 pixels, giving 19.4MP for cameras with their 3:2 ratio of sides, and 22.5 MP for the cameras with 4:3 sides.

Why 20 megapixels? We thought at first we’d need more, but found that we didn't. The higher the resolution that we used to upress the files, the more we had to manually play with the images to keep them comparable. Upressed files, like film scans, are naturally blurry and it’s unfair to compare them directly with higher-res files without some sharpening. But as we upped the resolution higher and higher, we needed progressively stronger and wider-radius sharpening, and we ended up in the territory where several rounds of sharpening at different radii worked much better than one. But then noise went crazy, so Smart Sharpen’s noise reduction slider came in.

Eventually, we found we were choosing our own balance of contrast at various spacial frequencies, noise, and smeary noise reduction. We’d degenerated into comparing our skills at enhancing photos, rather than comparing camera quality.

So to keep us out of it a bit more, we kept the size of the final pictures down to something more reasonable and practical.

So the workflow was this:
Shoot with camera sharpening set to normal/standard (applies only to jpegs; raw has no sharpening). Pick the sharpest photo from each camera. Open in Photoshop with no capture sharpening applied to raw files. Resize the long side to 5,400 pixels in Photoshop CC using ‘bicubic preserve details’ for enlargement and ‘bicubic sharper’ for reduction. Then apply Smart Sharpen to taste, limiting myself to once only.

In a final round of experimenting, we tried letting Lightroom do the upressing directly from the raw files, with different amounts of capture sharpening applied. We saw slight differences versus leaving all the upressing to Photoshop (and liked the look with judicious masking applied in ACR/Lightroom to minimise artefacts). But what suited some cameras didn’t suit others - we were back to choosing the look ourselves.

We opened raw files in Photoshop without applying sharpening in Camera Raw, and did all the resizing and sharpening in Photoshop.

The films were scanned in a Durst Sigma scanner, and the scans resampled down to 5,400 pixels using 'bicubic sharper' before sharpening.

Please download and play with the original unretouched pictures here. You’ll certainly be able to make any of the pictures better with careful, multiple sharpening. Please let me know if this changes any of the conclusions.

Shooting the Sunlit Scene (EV 15)

We set all of the cameras to their best-quality settings, using base ISO where possible, and shot on a tripod, without zooming the phones. We focused on the building in the centre background in bright sunny light and took several shots at different exposures, choosing the sharpest. All of the photos from phones are jpegs; the DSLRs shot both in jpeg and raw, processed via Adobe Camera Raw 8.2 from Photoshop CC. Most lenses give softer pictures at the edges compared to the centre, so we’re showing both the centre and the edge of the image.

We roughly matched the zoom lenses to the iPhone's angle of view, at 29mm. This biased a little against the Nokia, with its wider 27mm lens.

Normally, a landscape like this would demand f/16 or f/22 of the DSLRs to keep everything sharp, but instead we chose the best aperture of the lens, f/8, to avoid diffraction. So in real-life shooting, the DSLRs will probably do worse that the images represented here.

Shooting the Dim Scene (EV 2.7)

We coerced all the devices into using a borderline hand-holdable 1/15th second for this shoot, and then we put them all on a tripod anyway, and aimed them at a stationary target. I wasn’t interested in measuring how stable I was or how much coffee a model had had, but how good the lens and sensor were. We left their flashes off, and fired away with the self-timer, mirror lockup and cable-release where available.

The Nokia still showed a high proportion of blurry pictures, even on the tripod. This might have been me wobbling only the Nokia phone when shooting, or possibly its image stabiliser was working against it. On DSLRs, it normally helps to turn off any image stabilisers when you’re on a tripod, but the Nokia won’t let you turn it off.

Notice that all of the devices - including the phones - can get perfect shots in this light if you let them shoot for as long as they want:

Nikon D800 ($3.5k in 2012) 1/15s, f/2.8, ISO 3200, professional lens, jpeg		Nokia Lumia 1020 1s, f/2.2, ISO 200, jpeg

This shows two things: the Nokia can get perfect images in cadlelight if you let it use a tripod. But it's the other thing that amazes me: the Nikon D800 can get almost as good without the tripod.

The Nokia used “Nokia Camera” version 4.3.0.9, running on Nokia "Amber".

iPhone 5, 1/120s, f/2.4, ISO 80

Capturing priceless moments is - to me - one of the big joys of smartphone shooting

Want to get the most out of your iPhone camera? Get all the video materials for our iPhone Photography course without leaving home. With over 90 minutes of video tutorials that stay bang up-to-date, you'll find out the best apps for shooting and editing, and how to use them. Just $40 plus postage. Click here for an example of the videos included. Or click here to get the full set now.

Or if you're near to Brisbane, book a place on the World's only jargon-free iPhone Photography course.

Apple, iPhone and iPad are registered trademarks of Apple Inc. Take Better Photos does not claim any endorsement by and receives no sponsorship from Apple Inc. We are Apple developers, but the agreement with Apple does not restrict us from being critical of  Apple hardware where warranted. We queued and paid normal full price for the iPhone 5s and the Nokia. Photoshop and Lightroom are registered trademarks of Adobe Systems.