(commentary)
Quite a few readers posed the same question to me in recent months as I’ve written about the woes of the camera industry and just a few of the reasons why I think we’re experiencing a decline in camera sales. That question was basically: what would you design?
In all likelihood, my answer would probably surprise many of you, despite my having written plenty of bits and pieces of the answer over the years, starting with my original CPM idea (communicating, programmable, modular).
Let me try to bring my personal ideas up to date:
- The original UI was fine. There’s nothing wrong with aperture rings on lenses. Indeed, controls that impact the lens ought to be on the lens, don’t you think? Likewise, shutter speed, ISO, and exposure control dials aren’t problematic, either, as long as they don’t lie to you. One of the reasons why I like the recent Panasonic LX-100 is that it is mostly faithful to the idea that the original camera UI was perfectly acceptable (unfortunately, someone with a modern bent managed to muddy up the waters with a few non-programmable, not-needed-for-photography buttons).
Photographers want to control the basics, and they want to do so directly. Having an “A” (automatic) mark on any of those old controls is fine, too, giving us automation when and where we want it (ISO but no aperture and shutter speed, aperture, but not ISO and shutter speed, etc.).
So my camera of the future will go back to the past: direct rings and dials for the primary decisions photographers make (aperture, shutter speed, ISO, and exposure compensation). It would also tend to put further controls on the thing you’re controlling ;~). For example, one common question I get from people is this: “I can’t seem to set autofocus any more, it’s grayed out.”
Well, that’s because the camera maker put focus controls in two places: one via a push/pull focus ring on the lens (where it belongs), another often buried in the menu system (but maybe assignable to a button). Guess what, they can be set differently, and contradict one another, which the camera programmer discovers and takes the easy way out: just gray out the option. Therefore, a critical UI concept is that things only get set in one, consistent place.
Update: for those of you complaining that dials can be accidentally moved. Sure they can. That’s why you can virtually lock them (as you’ll see, they aren’t permanently marked as to position, which means we can add a LOCK parameter to any dial). - Sensor size is 24mp APS/DX. I’m willing to give up a stop even for a camera I’m targeting at serious shooters. Why? Because you make compromises for volume and for price. The more specialized and capable you try to target, the fewer customers you’re going to get, and the more it will cost. If you haven’t shot with one of the state-of-the-art 24mp crop sensors, you should. They’re quite good. And with a little work, we can make the results from them even better (I’m looking at you Sony, and your 11-bit compressed raw files ;~).
Products over time go in one of two directions: bigger or smaller. Something like a refrigerator, which you don’t move around and which you’ll eventually fill, get replaced by bigger ones. You’ve got plenty of wall space, so your TV will get bigger now that it’s flat. But other items get smaller over time, and if you hadn’t noticed, most of the consumer electronics products do, especially those that you carry around. Some of that has to do with electronics, but things you carry all the time, or even part of the time, will have a tendency to downsize generation to generation. (This, by the way, is written by one of the people responsible for the first popular portable computer, which was 28 pounds at birth. ;~)
Cameras should get smaller and lighter, within some boundaries. Too small and light and they can’t be handled well and we have to stick lesser components into them (e.g. even smaller sensor). Likewise, trying to go the opposite direction—bigger—is fighting the trend line. That’s especially true given that the primary age of the serious photography practitioner is going up as the Boomers age.
As for 24mp, that’s about as far as I want to go with APS/DX. Diffraction and lens designs start to be problematic if we try to squeeze too far in pixel count, and frankly, the number of folk needing more pixels than that is small (and they ought to probably be using MF gear ;~). That said, it wouldn’t stop me from putting more pixels on a follow-up product, as long as I could insure that the image quality notably improved in the process. - The camera is not a DSLR. Simply put, we can’t carry the mechanical/alignment issues of the DSLR very much further into the future. Especially if we want to price reasonably and still make a profit. We need the built-in benefits that semiconductors and electronic components give us when produced in quantity, so I’d keep my design mostly non-mechanical. There really are only two primary issues with electronic viewfinders (EVFs) that the DSLR crowd tends to complain about anyway: visual lag and focus performance. The first can be directly dealt with, as Samsung recently showed with the NX1. The second is going to be a tougher problem to solve, but keep reading, as there is more than one way to skin a cat.
So mirrorless. Do we need a new mount? Probably. I’ve got some communication and control additions I want to make between camera and lens. Perhaps we can repurpose a mount. Frankly, I think mount legacy is the least of our worries. If we build a sufficiently better camera, we’re going to need new lenses and I don’t think the customer will complain when it all comes together into something that works better than what they’ve got. - Six lenses to start. 16mm f/2, 24mm f/2, 35mm f/2, 58mm f/2, 16-50mm f/4, 50-135mm f/4. We must cover the basics (covering the 24-200mm equivalent in primes and zooms). Note that I’m keeping apertures modest. Why? Remember that smaller/lighter thing? I want you to carry this camera around as often as possible. I don’t want you leaving it home because it’s too bulky to deal with.
- Everything revolves around NewEXIF. Welcome to the future. Every parameter in the camera is settable via NewEXIF tools (more on that in a bit), and recorded in the NewEXIF data fields of the resulting files. Every NewEXIF field is completely and publicly documented. NewEXIF header files are available to programmers.
So how is NewEXIF different than EXIF? In many ways, it is the same. Everything that’s currently in the EXIF definitions (including what would be proprietary Maker Tags) lives on in some way in NewEXIF. But there are some interesting new additions. Instead of recording “focal length,” would you believe that we record “focal length at beginning of shot” and “focal length at end of shot”? Ditto for focus, aperture, and many other things that we might want to vary. We also have “Part of Sequence” fields where we have a sequence number and the position of the image in the sequence. (This part is a little tricky; I’m not going to go into the full set of APIs/tags we need here, but suffice it to say they’re be there when the product ships ;~).
And yes, users can add “tags” to NewEXIF. Indeed, that is one of the differences between old and new: users not only have access to NewEXIF, but they are the ones that create new tags, not camera makers in secret chambers hidden deep beneath the surface of the Earth and guarded by dragons. If I want to put “Galen spot” as a tag into a UserTag field and have it follow the image around, I can. If I want to later start a Lightroom Collection of just images with the “Galen spot” tag, I can. In fact, if Adobe is paying attention, they’d automatically pick that up when the image imports.
User tags can be entered when you program the camera (program tool), or on the spot when you’re shooting (via built-in camera “tools”), or after the fact (with a NewEXIF editing tool). That last might be a bit of a surprise to you, but what if I wanted to repeat everything in a shot except one parameter? Why can’t I just edit the existing NewEXIF data and reapply it to the camera?
Why do I want NewEXIF to be so extensive and so user accessible? Because the camera shoots from what the current state of NewEXIF says. The simplest program thus becomes:
LOAD NEWEXIF(“File”)
SHOOT
Want to perform a 10-shot focus bracket?
LOAD NEWEXIF(“FocusBracket10.NEX”)
WHILE TagSequenceNumber < TagTotalShotsinSequence
SHOOT
ENDWHILE
Yeah, I think you can program this camera. Given the right set of tools. But remember, the camera itself shoots from what’s set in NewEXIF any time you press the shutter release, too. Thus, we can have menu settings that program NewEXIF variables that allow you to do the same thing in real time (e.g. SEQUENCE/SETSEQUENCECOUNT, SEQUENCE/SETSEQUENCETOTAL, SEQUENCE/SETSEQUENCECHANGE, SEQUENCE/EXECUTE).
Our new camera doesn’t care whether the NewEXIF fields are getting loaded externally (from a smartphone, laptop, desktop, whatever), whether from a file you previously saved and loaded from the card, or via some menu choices.
While I’m trying to make this simple, behind the scenes it isn’t quite so simple. There are a lot of dependencies (you can’t start executing a sequence before some variables have been set up, for instance). I’ll bet I’d fill a large room’s worth of white boards with all the bits and pieces of NewEXIF and how it out interconnects and relates. I tried doing just one piece and ran out of white board in my office.
Also there’s a contradiction in NewEXIF versus one of my previous points. Did you catch it? Yep, dials. Nikon just punted when they hit that problem with the Df design. I refuse to punt. The solution is actually simple: dials aren’t preprinted. The current setting is actually shown via a small monochrome LCD. And yes, as you turn the dial, the setting “scrolls” just like a real dial.
But the takeaway is this: my camera is programmable. By users.
Note: While I call this NewEXIF, it’s actually likely that we need to move away from the TIFF-defined tag world into something more solid, such as XMP or a real database. I used NewEXIF here for a reason: it shows what could be done with the current approach camera makers are using. - The camera is modular. Yeah, I knew you were waiting for that. Ricoh almost got this right with the GXR. Indeed, you could say they did get it right with the Leica M-mount module. Having the sensor/ASIC/lens mount all in one unit means that you can guarantee alignment in the one thing that still needs critical alignment in a mirrorless camera. The rest of the camera is just a battery holder, LCD display, EVF, memory card system, WiFi connection, knob/UI holder, and grip. I’d consider putting 16GB into the camera itself for temporary image storage (remember, the camera communicates, so a card could just be “backup”).
It’s a little premature to say exactly how exposure metering and autofocus work, as both those things are rapidly changing in technology, but suffice it to say that we’d seek the highest possible performance in both.
My vision is this: you buy two camera bodies (one to shoot, one as a backup). You then buy sensor modules as needed (24mp Bayer, 24mp Monochrome, 24mp IR to start, but demand probably would allow for eventual 12mp Hi ISO capable, non-Bayer three layer, and more). Technically, if you were clever, you could even design this new camera so that a full frame sensor/mount module could be attached ;~).
Note that by having a lens mount on every sensor module, you can also set up quick lens changes without exposing the sensor (just have two sensor modules, one with each lens). I suspect there’s a weather advantage there. I certainly can imagine that—should the long lenses become available—this would solve my dust problem in Africa that keeps me from changing lenses. Yes, an expensive way to do it, but also a protective way to do it, too: no more fiddling with rear lens caps, either, so I don’t have to worry quite so much where I’m seeing the lens when I’m changing it quickly. Still, this is a small side benefit of this design, and only for those who are buying multiple sensors, which means “rich users” ;~).
You’ll note that wireless—the communicating part of the camera—wasn’t modular. While wireless technologies can and do change, you’ll just have to buy a new body if they do ;~). Hey, we have to generate replacement customers somehow. Seriously, I don’t take this as a big problem. If you start with state-of-the-art LCD, EVF, WiFi, Bluetooth, Card Write mechanisms in the “body,” that should last users a long time. If there’s demand for newer, better, faster, or even previously unknown things in the camera body, that’s obviously the point where the company offers an updated body to the market. - Everything is documented. There are no secrets in the Thom camera. We don’t have proprietary fields in the NewEXIF. We document all connections (including lens mount), not just pin outs but protocols, as well. We document spectral and other information about the sensor itself, and we document where we put black in the data structure and why. As best we can, we document how the exposure meter and focus system work. We even publish our repair manual for any and all to look at.
So, if you have US$250m to spare and want me to start a camera company that “gets it,” that’s an approximation of where I’d start. I say approximation because, as a manager of some very exceptional engineers over the years I know to let those folks create the actual designs. The real trick is to have a clear goal and not let the team deviate from the primary vision into nuance that’s not important and doesn’t serve the user’s purpose. My goal would be to produce a camera that’s as capable as about any on the market today in terms of image quality, has photography-friendly controls, includes user programmability (and in multiple ways), is modular in ways that allow the user to update the primary image quality producers in the camera, and that connects and integrates with the other tools we use once we have our images captured.
How’s that Nikon J4 looking when compared to that? ;~)
One word of warning about my ideas. They are for the camera I’d want to deliver today. The minute we start talking about the future, all bets are off. The further into the future we talk about, the more has to change. Thus, if a camera maker decided that I was right and tried to make the camera I define above by starting an engineering project today, it means the camera would be out in two or three years and if it didn’t take into account how things will change in that time, it could look just as impractical as today’s DSLRs.
That’s one of the real problems the camera makers have, actually. They’re still iterating 2007 state-of-the-art for the most part (mirrorless is a bit further along, probably iterating 2010 state-of-the-art). By letting themselves fall behind and not being aggressive about what could (and should) be done in their products moving forward, they’re falling progressively further behind, and making the big leap to “modern” is going to get more and more difficult.
Apple is often criticized for moving before it’s necessary (dropping floppy drives, dropping CD drives, moving to Firewire or Thunderbolt, using SSDs, etc.), but I think they’re actually doing the right thing for the most part. Interestingly, we don’t really have any camera company trying to be as forward-thinking as Apple is (Nikon’s use of the XQD card in exactly one camera notwithstanding ;~).
