What I Want To See In My Next DSLR

A recent satirical article on The Online Photographer, criticising the Canon 5DMkII for its lack of a phone, and a rather more serious essay on the State of Camera Design 2010 by Thom Hogan got me thinking. What's missing from the typical DSLR, anno 2010? What could be improved, using simple established technologies, to make the DSLR a better picture-taking device? And why don't the major DSLR manufacturers do some of these things, which might help sales?

Now I quite like the basic concept and form of the DSLR. This has evolved from 80 years of small camera development, and basically works. Almost any current DSLR is capable of making great photographs, and I wouldn't want to detract from that, by either distracting engineering (as mentioned in Thom's essay) or unnecessary "convergence". I don't want a camera cum phone cum coffee maker. I'm not too fussed about video. And I'm happy for the manufacturers to continue improving the underlying sensor, optical and processing technology as they will, although for my purposes they are already well past "good enough".

No, I'm talking about what are, for the most part, simple software or mechanical refinements to ease the process of taking and processing photos. At the risk of giving away some good ideas, here's my round dozen ways in which DSLR manufacturers could easily improve the product, and, who knows, maybe boost sales by doing so.

Please note: my main reference point is mid-range Canon DSLRs, and they provide most of the examples in this article. However no specific criticism of Canon (by inclusion) or other manufacturers (by omission) is implied, as I believe most of this applies equally to the equivalent models from most manufacturers. If your vendor of choice has already addressed some of these points, you're winning!

Michael Reichmann and others suggest that to make best use of a digital camera's dynamic range, one should "expose to the right", creating a RAW file in which the histogram is skewed as far right as possible, and then adjust the exposure as required in the RAW processor. While this is a great idea, in practice it's very difficult to do manually without blowing the highlights, and I find that in practice if anything I tend to slight underexposure.

It should, however, be pretty simple to program a DSLR to do this automatically, controlled via a simple "expose to the right"/"expose for 18% grey" switch in the setup menu. Even better, the camera could easily calculate the difference between the two exposure values, record this in the EXIF data for use by the RAW processor, and apply the required correction to the creation of the in-camera JPEG so that it doesn't look over-exposed.

While we're on this subject, when will the camera manufacturers realise that the majority of serious photographers shoot RAW, and optimise higher-end cameras for RAW rather than JPEG capture. One simple example: the histogram on most cameras shows the exposure of the processed JPEG (even if it's only retained as a thumbnail in the RAW file), not the RAW file itself. A "RAW Histogram" option definitely goes on the list.

Surely it would also be easier in high frame-rate situations to just capture and write the RAW data, and then (optionally) create the in-camera JPEGs (which can be useful as a rating and mnemonic tool) when the camera and memory are idle?

While manufacturers strive to improve the dynamic range of their sensors, it's still often exceeded by real subjects. "Manual" HDR techniques address this for static subjects, but have limitations which could be addressed by in-camera support. The ideal would be to deliver the required information from one exposure. Here's one way it could be done...

Let's suppose the camera meters the scene, and a "normal" exposure requires a shutter speed of 1/200s. Two stops greater exposure is 1/50s at the same aperture, so the camera sets that speed, and indicates it to the user. The user releases the shutter. 1/800s into the exposure, the camera scans the sensor and saves the data as the "2 stops underexposed" RAW file. 1/200s into the exposure the sensor is scanned again and data saved as the normal exposure file. Finally, at 1/50s the shutter is closed and the data saved for the final time. In post-processing you then blend the three shots using your favourite HDR program.

This technique has a multitude of benefits. In particular it would work for subjects with some movement: as long as the indicated shutter speed is fast enough to capture the subject, then the amount of movement between frames should be minimal. Also the user only has to hold the camera still for one shot (and knows precisely that shot's ultimate shutter speed), and if stability is adequate, there should be very little mis-alignment between the frames. If the indicated shutter speed is too slow the user is free to adjust the aperture or ISO to compensate.

Having established the underlying technique, there are several possible variants on the algorithm. The main alternative is to take the matrix metering data and calculate the "normal exposure" for the darkest shadows, set the shutter based on this, take the first "sensor snapshot" based on a normal exposure of the brightest highlights, then continue at two stop (or user selected) intervals until the shutter closes.

I'm not sure I want camera manufacturers to go down the road suggested by some (see, for example Thom Hogan's article referenced above), and make the camera the "hub" of picture processing. To my mind that invites too many compromises on usability, and you'd end up with something which is neither a good computer nor a good camera.

However, it is absolutely right that the camera needs to play better in the connectivity world, and interact more easily with our other devices. The obvious omission from most DSLRs is BlueTooth connectivity. This would be a natural (and probably relatively easy) complement to existing USB connectivity and enable the camera's use with other personal devices such as mobile phones and PDAs. There are many possible uses, including:

• Enriching the image with subject, exposure or processing notes (see next section),
• Geotagging, using either a separate standard BlueTooth GPS, or one built into the phone/PDA,
• Using the phone/PDA as a remote release/timer. This would have a number of advantages over dedicated wireless solutions, not least a "sensible" bit of convergence - currently I have to carry four dedicated "gadgets" to provide usable remote operation for both my 350D and 40D, which is at least three too many. A PDA solution would also be much more flexible, and easily able to manage multiple cameras,
• Wireless "teathered" operation and image review without carrying a large laptop, ideal for working in awkward spaces and able to work with more than one camera at a time,
• Uploading pictures from the field, via the phone or laptop.

It's important that manufacturers "get" the need for a standard solution here, not a proprietary one. We want proper BlueTooth support, with its broad device compatibility and design for use in complex multiple device environments, and the more the command set can exploit existing standards and profiles, the better.

Currently, if you want to capture model, subject, exposure or processing notes, you have to record these separately and marry them up with the images at a later stage. Some high-end DSLRs allow the capture of a voice note, but those need to be transcribed later. Similarly, if your review in the field identifies good or challenging images this has to be noted elsewhere and recovered manually in the rating/tagging stage of your workflow process.

There's no good reason why all this information cannot be captured by the camera and stored with the images to which it relates, stored directly in the image files as EXIF/ICMP data or alongside them as XMP or similar files.

The problem is that the camera isn't a good text input device, and I don't want to break my rule against inappropriate convergence. The best solution is probably BlueTooth support, and the ability to manage notes and related files via a phone or PDA. As well as simple note capture, this could allow more sophisticated options like attaching electronic business cards to relevant images.

That said, it might be useful to add a few simple tagging/rating options on the review screen, which would be a pretty simple software enhancement.

One specific piece of data I'd like against all my images is the geographical location at which they were captured, ideally embedded in the EXIF data. GPS would also ensure that the time set on the camera was always accurate, and allow easy prediction of sunrise and sunset times and positions via a screen on the camera. Separate geotagging accessories already exist, and I've suggested above the possible use of an external BlueTooth GPS receiver, but actually this is one piece of electronics which really should just be added in to the camera itself. High-end cameraphones have this capability, so why not DSLRs?

The only possible issue I can see is battery life, but given the way most photographers work a simple solution would be to take a fix whenever the camera is powered up, or on manual demand, and turn the GPS off as soon as you have a location. This should keep the impact manageable.

Another useful piece of information, which probably already exists in the camera, is the focal distance. Wouldn't it be useful in a near/far situation to focus on the nearest object, read the focal distance, repeat for the furthest object, and then be able to confirm the depth of field and hyperfocal point with some accuracy?

If I dive into the EXIF data from my cameras, it looks like they have all the requisite information. Oddly the Canon 40D actually reports near and far focus distances, but not the current focus, although that's easily estimated from the other two values. However, all this is only available on the PC after the event, and I want it at, or immediately after, shooting.

Useful as this would be, it might be even more useful to have...

As a landscape photographer, my most regular focusing challenge is to keep everything from a foreground interest point to infinity in focus. This can be a particular challenge with smaller-sensor cameras, as diffraction effects kick in at relatively large apertures. Why can't I simply instruct the camera that this is my intention, and let it do the work?

I'm imagining a mode in which I control the aperture (rather than the camera taking control as in Canon's A-DEP mode), but the focusing mode is such that the camera scans the focal points for the nearest subject, and then if possible sets the focus at a point which covers from that nearest object to infinity. If that's not possible I get a "stop down" warning (such as flashing autofocus points), but it's my choice how I proceed.

A natural extension of this approach would be automated focus bracketing: the camera focuses on the nearest autofocus point, and then takes multiple frames in quick succession gradually moving the focus from that distance to infinity, making sure the depth of field bands for successive frames overlap slightly. I can then use "focus blending" software such as Helicon Focus on the image set.

In current SLRs, the camera is unaware and unsupportive of multi-shot techniques, such as panoramic photography, HDR and focus blending. The only real support is for exposure bracketing, but even then I just get three sequentially-numbered files (e.g. IMG_1234/5/6), and have to recognise them manually at the workflow stage. If I want to do a complicated panorama with exposure brackets and/or focus brackets at each position, then I have two challenges, keeping discipline and correctly shooting the different variants in a controlled sequence, and then "decoding" this structure from a large linear file set delivered to my computer.

It would be much easier if related shots had a structured numbering scheme. For example, a simple exposure bracketed sequence (or an automated HDR set) could produce images IMG_1234_E0, IMG_1234_E2 and IMG_1234_E4 (where the "E" stands for exposure, and the numbers in this case are relative exposures). Focus bracketed shoots might get an F1/F2/F3 suffix, panoramas a P1/P2/P3 suffix and so on. For panoramas it might be even better to allow for two dimensions, with "row" and "column" numbering. It's important to make sure that combinations of techniques are possible, so for example the first shot in a linear panorama with exposure bracketing at each position would have a P1E0 suffix.

Aside from instantly identifying related shots to the user at workflow stage, such a scheme could be very easily exploited for automated image processing, removing an enormous amount of tedious work post-shoot.

Making the camera "panorama aware" should be trivial, as better compacts and cameraphones already have this capability, and most DSLRs now have a live view mode or electronic viewfinder to support it. Once the camera is in "panorama mode", the live view provides a guide to the shooting sequence (by showing a "ghost" of the previous image suggesting a sensible overlap), and the intelligent numbering scheme is applied.

The refinement is that on a DSLR I should be able to "program" the camera for a complex multishot sequence, (for example a panorama of two rows of four shots, with HDR exposures and focus bracketing at each point), and then all I have to do is move the camera between successive positions and trip the shutter. The camera automates the rest of the multishot process, using the techniques outlined in previous sections.

I was going to be boring and ask for a dedicated mirror lock-up button, but with my shooting style I'd probably rarely use it. What would be much more useful is to make the camera "tripod aware". When mounted on the tripod and fired by remote or timer, I'd like the camera to automatically lock the mirror before the exposure (at a point which the good engineers have determined is the optimum), and also turn image stabilisation off. I'm always forgetting to do the latter, and this has a much greater impact on the image than mirror vibration as the stabilisation "fights" with the tripod.

Detecting tripod mode is easy. An algorithm which applies when the shutter speed is longer than say 1/4s and the triggering mode is via timer or remote would get it right most of the time, but I'm sure the camera designers could do something even cleverer with optical sensors or microswitches in the camera base or hand-grip. However, like flash mode this needs to be under user control, so you can leave this on auto, or force it on/off.

I like shooting with the camera in portrait orientation, and it's no different when it's on the tripod. However, whenever I do so it's impossible to aim accurately as there's a continuing problem with the camera simply rotating against the camera plate, under the weight of a typical modern zoom lens. Gitzo may coat their plates with "a special techno polymer that maximises locking torque and equipment stabilisation ... to prevent the camera from moving or slipping", but it just doesn't work!

I don't really want the extra complication of something like an L bracket on the camera, and while I can get a better result by mechanically tightening the tripod screw, that 's not ideal, either, especially if I want to fix the plate in the field.

Fortunately Victorian engineering comes to the rescue, and many tripod plates feature a little spring loaded pin to slot into the camera to lock it in place.

Well... it would if the camera had the requisite hole. None of my Canons have it. My 1970s Russian TTL had one, but not my 40D. Come on Canon, this is just ridiculous.

The late, great automotive engineer Colin Chapman had a simple recipe for building better sports cars: "Just add lightness". His lesson would be well observed by camera manufacturers. Current higher-end DSLRs are simply too heavy.

This was really brought home to me when I was on a photo shoot in Venice last year, and several members of our party were taking long range shots using our telephoto zooms. Mine, on my 40D* with the Canon EF 70-300mm f4-5.6 IS USM lens were razor sharp, but other members of the party, using the supposedly "superior" combination of the 5DMkII and Canon EF 70-300mm f4.5-5.6 DO IS USM lens were complaining bitterly that theirs weren't. We resolved this by the simple expedient of me picking up one of their cameras - the combination weighed almost 200g more than mine, and was simply too heavy for some of its users, especially the older ladies. I managed to get some good shots with it, but then my sport is weight-lifting!

A lot of people say they want lighter and smaller cameras, but I'm not sure that's quite right. The size of a DSLR is to some extent dictated by ergonomics, and the need to balance lenses whose size is to some extent dictated by fixed laws of optics. That said, I'd be very happy to drop one size from my current Canon 40D, if that didn't mean (as it currently does) compromising on features, as the camera manufacturers currently seem to think that to charge more money, they have to provide a "bigger" camera.

Here's an idea. Canon like creating new niches, so how about a "9D", with exactly the same feature set and control options as the 7D but in a XXXD body? This would have, for example, the same rear dial, multiple custom modes, RGB histogram, and high frame rate as the larger cameras, but not the rugged chassis and big battery. If Canon need to signal its position in the range they could just put a red ring around the body (like an "L" lens), rather than making it bigger. I'd be prepared to pay somewhere between XXD and 7D money for such a camera. There would still be a place for the XXD/7D, but it would be a conscious choice for those who need the ruggedness, or have big hands, or are the sort of people who would choose a big, solid car when I'd spend the same money on a smaller, faster one.

I'd settle for the 7DMkII being the same size as the current camera, but the same weight as my 350D - this must be possible now the technology is maturing.  If all else fails drilling a couple of anti-rotation holes would save 2g!

* I've used Canon model numbers here, because that's what I and many readers understand best, but I believe that equivalent arguments apply to most of the major manufacturers.

The modern DSLR is a great image making machine, and I don't want to change it's core paradigm. But as I use my cameras in different ways, I recognise a number of lost opportunities and frustrations which have yet to be addressed. Now is the time to divert a small fraction of that enormous engineering effort from the hunt for ever more megapixels to the development of software and mechanical features which will make the DSLR a more flexible and better-integrated photographer's tool.

[…] jagged fireworks pictures, and others, over the years. Regular readers will recall my suggestion in What I Want In My Next DSLR that it would be easy for camera design to include automatic detection of “tripod mode”, and […]

Three years and several camera purchases later, how are we doing against these targets. Read my progress report:
What I Want In My Next DSLR – Progress Report