The design process

_8B40789 copy
Not a camera, but a watch is as good an example as any – perhaps more so, especially when you’re producing just one and it has to satisfy the most demanding client: the designer.

Whenever a photographer ‘has some ideas about camera design’, they often forget they’re only seeing one small portion of the puzzle. Inevitably, there are significant other considerations beyond the obvious – sometimes to the point of being physically impossible or functionally incompatible with their own intended result. At this point, having significantly more involvement in the design process will allow me to clarify why some things are the way they are, why some things should or don’t change, and where manufacturers shouldn’t have any excuses. Think of it as a candid ‘message from the other side of the fence’.

Case in point, and one of the more frequent messages I seem to get these days: “why don’t you make a square 6×6 sensor back for the V system and sell it at $8000? You’ll sell a lot.”

We’re very aware there’s still great passion for the V system and demand for compatible backs, especially given that there are so many legacy solutions out there beyond the V system itself – technical cameras etc – but at this point, the last V cameras were made more than 10 years ago, and production costs today would be prohibitively high. Second hand equipment is very good value – I’ve seen basic camera-80mm kits as low as $700. This creates an economic mismatch: not everybody with a V system has the funds required to digitize it; on top of that, wear, age and unknown provenance may mean the cameras are out of tolerance (e.g. if a back flange is worn, it may not meet infinity focus) and would need to be recalibrated and adjusted before delivering the quality potential of the back. Clearly, this is not a trivial task – with accompanying labor and expertise costs.

Assume though there are still enough serviceable V bodies within tolerance – perfectly reasonable – the next limitation is one of sensor. Whilst it’s theoretically possible to produce silicon up to pretty much any size you want, yields drop exponentially as sensor size increases. Errors develop with imperfections in the material substrate, and even with extremely tight quality control – an imperfection-free wafer is impossible. Imperfections may mean dead pixels – which can be mapped out – or something more serious in the wrong component area, where the entire sensor may land up being unusable. Whilst a standard wafer may yield hundreds of smaller sensors, the count for that size sensor may be as low as four – yet the total number errors per unit area remains the same. I think it’s easy to see the costs per unit rise very quickly – and we haven’t even talked about design or layup costs for a custom sensor, which easily run into seven figures and beyond.

As with all technology, at some point the costs may well be low enough. Hopefully that also means the supporting circuitry will also be small enough – as things stand, even the 54x40mm chip (sensor plus are minimum circuitry that would be considered inclusive) won’t physically fit inside the V back mount standard – and for obvious reasons, we can’t exactly change that. It’s important to remember film has very little thickness and can be made to bend conveniently around corners and structural components!

Let’s assume again that eventually, all of these challenges are solved. We have a sensor that literally extends to the edge of the silicon and can physically fit and costs little enough that we can make the economics work over a sensible number of backs – remembering that increases in price translate to strange changes in demand – the next question becomes one of architecture. There are really only two choices: dedicated chips (ASICs) or general purpose processors (FPGAs) – there’s generally of course some programmability left in an ASIC solution to allow for custom functions and the like, but nowhere near as much as a FPGA, which is in effect a mini-computer. ASICs have all major instructions hardwired (e.g. Image processing) which leads to great power and speed, but for limited tasks. An FPGA solution has to run an OS of sorts, which affects startup times.

As you may probably have guessed, an ASIC solution is really the only option for speed. It’s also expensive because the chips have to be designed for that specific function; camera makers go for one approach or the other across the entire product line to amortize development costs and maximize production volumes (influencing bulk component costs etc). It also makes achieving consistency in color and tonal response easier. Development times are also significant, so platform architecture has to be somewhat modular if possible to handle higher data rates in future cameras. Note however that different functional priorities require very different engineering and electronic solutions; FPJs aren’t always the best solution for peak image quality as some algorithms can’t be completely hard coded (data stream dependent) – to date, this has been Hasselblad’s solution as the highest possible image quality is our priority.

Stills-focused DSLRs prioritize single frame responsiveness but allow for buffer-type processing with the sensor mostly off. There are limited electronic shutter requirements, but on the other hand, additional bandwidth is required for phase calculations and tracking AF algorithms across multiple points. A separate metering control unit is also required.

Video-centric solutions must focus on maintaining consistent data bandwidth and low rolling shutter (i.e. fast global sensor readout times) but the demands on peak image quality are somewhat lower. During output you never really spend that much time focused on an individual frame – very good video color is 10bit, and whilst there’s a significant difference over 8bit, I doubt the increase to 14 or 16 bit would be as noticeable as for sills only. However: the camera must continuously be reading, processing and writing data so long as the recording happens; there is no buffer per se as it would eventually be overwhelmed by the incoming data stream.

A live view mirrorless camera has the challenges of both types: you need a continuous video output stream for framing, at fairly high quality; but it also has to be able to deliver high peak image quality. If designed around video architecture, the stream has to be interrupted and sensor reset to deliver peak stills quality. It’s the switching of sensor and processor modes that often causes the irritating lags and delays in responsiveness that annoy the spontaneous stills shooter so much.

Even with the best electronic platform – assume you have the best of both worlds somehow – the next considerations must be around features, ease of use/ deployment and ergonomics – these are not independent. It doesn’t make sense if the overall camera is minimalist but feature rich and everything must be accessed through unintuitive menus; this doesn’t make sense for a general purpose camera, for instance – but it might be fine for technical or scientific use. The main problem with trying to accommodate features for every possibility is that inevitably they become simply too difficult to make easily accessible – it’s a bit like subject isolation in composition: you cannot prioritize everything because then nothing will be stand out, and the button you want will get lost when you’re in a hurry. Programmability and customizability can go some distance to alleviate this, but then you can’t avoid a lengthy menu process to get you set up and to access the remaining occasional features. It’s the same case with finder information: too much and it becomes impossible to compose anything at all because there are simply too many distractions. I’d be the first to argue that we were able to make strong images just fine on manual film cameras (and still can) without virtual horizons, focus peaking, three color histograms and knowing exactly how your picture profile has been set all the time.

The reality is that as much as we want sweep panorama modes, switchable front/rear electronic curtain and leaf shutters, zone spot metering modes – the majority of people only use the basic exposure controls most of the time. Knowing this allows for a much simpler camera that’s also easier to use, which in turn you’re more likely to want to shoot with. Our job as designers is to figure out where the priorities split – what really matters, and what doesn’t? Which of the new technologies is genuinely useful on a frequent basis, and how can it be implemented without distracting or detracting from the camera’s primary objective as a seeing tool? These are not easy questions to answer, and that’s before we’ve even started to take into account form factor preferences, different hand sizes and material choices – not to mention cost.

Over the last few years, I’ve talked a lot about what I consisted good, bad and critical design considerations; I’ve never attempted to define an ‘ideal’ camera for everybody because there simply isn’t one. There isn’t even an ideal single camera for myself, because different subjects and different creative states work better with different tools – variety is not a bad thing! But there are a few qualities in tool that I’ve always valued:

Responsiveness – a camera must take a picture when you intend it to; anything else is pointless. This isn’t necessarily write speed or high frame rates – it’s lag.
Sensible default choices or configurability – you want it to do what you intend it to do, and be sure it isn’t going to do something unexpected at a critical moment.
Reliability – I think this one is self-explanatory.
Tactile quality and ergonomics – when you have the camera in hand up to 12+ hours at a stretch, it’s important that the object quality is something that’s both pleasant and physically comfortable to hold. Anything else will start to annoy quickly, and distract from your main objective.
The final one is a bit harder to define: I think of it as ‘certainty’. You need to have a very high level of confidence you got the shot, or at least easily be able to confirm – qualities like exposure, critical focus etc.

These properties – at least the first four – are why the best cameras from the film era still feel perfectly fine to use today despite lacking just about everything else we’ve come to take for granted. It’s why a V camera with a film back still feels ‘sharp’, and attaching a CFV back doesn’t feel like a compromise (assuming you can find the right focal length). Yet – fundamentally, you’ve got only the basic exposure controls – focus, shutter speed, aperture, sensitivity and timing of release.

The bottom line is that we could build in every single function under the sun – and even dedicated buttons to run them all – but it’s important to realize this is likely to result in a camera that nobody will actually want to or enjoy using. In that sense – creating a tool that doesn’t feel compromised is really about finding the ideal balance of compromises. MT

__________________

Visit the Teaching Store to up your photographic game – including workshop videos, and the individual Email School of Photography. You can also support the site by purchasing from B&H and Amazon – thanks!

We are also on Facebook and there is a curated reader Flickr pool.

Images and content copyright Ming Thein | mingthein.com 2012 onwards. All rights reserved

Comments

  1. Hi. Thanks for that. Very good read.

  2. Hi Ming, really interesting article, that I have enjoyed reading and thinking about. Thank you. I think Albert Einsteins advice still resonates “make everything as simple as possible, but not simpler”. Not just for cameras either 🙂

  3. Ming,

    Thank you for this article. Its pleasant to actually LEARN something from the current photo blogosphere.

    Now that your are more familiar with the design side of things, do you believe that the little quibbles and big annoyances in many current cameras come from poor execution of legitimate technological limitations? Take the A7 series, for example. They have so much annoyances and weird qualities that it seems like any of us could have designed it better. Conversely, SRLs never seem to have the same annoyances no matter what the price point. Are their hands tied or is it just bad design?

    Perhaps this is just part of the design process itself: weighing the options of cutting-edge tech vs. the inherit drawbacks of adopting young technology. Canikon hasn’t introduced anything that new in a while, but their cameras are reliable and refined. Sony, on the other hand, have undoubtedly the most high tech cameras on the market; but they have so many ifs and buts and seemingly illogical quirks that they are unpleasant to use. The modern Leica is an other example of a brand that appears to struggle to find that balance between tech and user experience.

    I suppose, like so many things in design, its all a balancing act.

    • No problem.

      I’d say 99% of the annoyances are due to a lack of field testing and a disconnect between the design team and the end user – certain things like form factor and weight are restricted by engineering requirements, so we’ll never see a medium format iPhone that’s 10mm thick, but there’s no excuse for poor button placement or bad menus or an uncomfortable grip.

      Personally, the only sustainable and differentiable strategy is to pursue on thing relentlessly: reliability? Speed? Overall image quality? At least all other compromises can be defended and prioritised. We choose image quality.

      • Understood. Ultimately it seems like it really is the design process at work, with a clear design goal to guide your thinking and make your choices/compromises clear.

        Thats why I love my Coolpix A, its clear what their intention was: a pocketable, high IQ compact. It may not be the most feature packed camera out here, but at least all of the missing features are justified. Nikon’s goal was clear, and they succeeded at this without diluting the core concept.

        My old A7, on the other hand…

  4. Peter Bowyer says:

    > Programmability and customizability can go some distance to alleviate this, but then you can’t avoid a lengthy menu process to get you set up and to access the remaining occasional features.

    Coming at this as a software consultant I would like to see cameras coming with a companion mobile phone app (or website) where I can customise the menus and setup.

    Rather than a ‘custom menu’ (yawn), I can drag and drop items between menus and build my own, assign functions to buttons and create saved presets (not new by any means). Setups can be switched between from the app, when the camera is connected to WiFi. The entire creation can be shared like Lightroom presets, and in one swoop you let people have a camera that is closer to ideal in operation for their specific needs – and crowdsource what works (which frankly camera manufacturers need, given the current dire state).

    Now that cameras are getting WiFi this is more feasible. The elephant in the room is that electronics companies have never been good at software. Plus, at present I suspect what I’ve suggested above would currently need to be a custom-built firmware deployed each time to the camera, which would be terrible.

    • Peter Bowyer says:

      My point being that the experience of doing any of the above on the camera itself really sucks. Yet because camera companies are camera-focused, only rarely do they think beyond the camera.

      • …And if they really thought about what was required to begin with, menus for unnecessary functions would not be required in the first place… 🙂

    • I’m with you on this one, and it’s something I want to implement. We are looking into it though the challenge is supporting so many mobile platforms seamlessly…

      • Another tough one would be to do a perfect RAW ETTR so no individual channel would be clipped. Would improve the IQ even more.

        • The tricky part with that is to define where you want the cutoff: 1% area clipping might be okay (but clipped by how much – there’s no way to know) on paper, but that 1% could be a lightbulb or a face on a massively low key image. It’s not so simple…

          • The camera could show the overexposed RAW channel areas on the viewfinder as blinkies. Another nicety would be an uncorrected RAW histogram as well.

    • Peter Bowyer says:

      Speaking of which, I’ve just seen https://www.dpreview.com/news/3902928422/arsenal-is-an-intelligent-assistant-for-dslr-and-mirrorless-cameras. If (and it’s a big if) it lives up to its hype, it’s an interesting direction to head in.

      • I am not convinced. Algorithms are just going to average everything to the same point – making it even harder to make an exceptional image (‘good’ – sure) because the really outstanding shots are the ones that precisely don’t follow the rules…

        • Peter Bowyer says:

          I agree, however look at it as what a smart camera firmware could do to simplify menus and navigation. Consider it putting the options it thinks you will want to change for the current image on a first “Did you want to..” screen.

          That’s why I find this interesting. It’s adding smartness to complex devices which are currently very dumb, and has the potential to reduce the photographer’s cognitive load, allowing them to focus less on the tool and more on the scene.

          Plus the average camera buyer is going to struggle to reach ‘good’ on their own, let alone ‘exceptional’. If camera maker A produces a camera that gets a good image 80% of the time for little effort, that’s a market differentiator.

          To me it sounds like an extension of Nikon’s “Advanced Scene Recognition System” (http://www.nikonusa.com/en/learn-and-explore/a/products-and-innovation/scene-recognition-system-and-advanced-srs.html) applied in a different way.

          • I think it’s a dangerous path to take, because there really are no shortcuts – and you’re more likely to miss something if you’re second guessing what your camera is thinking. When Nikon switched to the second version of that metering system, I found it far less consistent, and requiring a second try + exposure compensation. This is not exactly a good thing…

            • Peter Bowyer says:

              We’ll have to agree to disagree. With the increased pervasiveness of smart assistants (Google Now, Siri, Alexa etc) and their improvements in the next 18-36 months, this type of enhancement will become expected. “OK Google, set my camera to capture my daughter’s football match” or “OK Google, set my camera to capture the birthday party tonight” is natural for someone who thinks of photography as recording (cf scheduling a PVR so you don’t miss something) rather than an artistic pursuit.

              Proposing this feature has nothing to do with whether it’s good for photography, and I take it as read that Pros won’t use it. But Pros don’t keep camera manufacturers in business. And if camera manufacturers don’t keep up, software companies will take over. With pretty sucky hardware.

              Pan out.

  5. “The main problem with trying to accommodate features for every possibility is that inevitably they become simply too difficult to make easily accessible ” Guess Sony went the other way huh?

    Great write up! It’s rare that someone can appreciate the difficulty of manufacturing or delivering a product that demands great hardware engineering and software design(interaction, coding, etc.). Even rarer that someone can have insight from both the end user and behind the scene perspective.

    Blazing a new trail for a new system is always a big risk: Canon EOS, Sony E, Leica’s disastrous R and S series all took time to mature/go bust. Even digitizing Leica M systems had been a great challenge for Leica that almost bankrupted them. People often don’t know about the work that goes into the image processors… it’s not as easy as slapping a sensor solution from Sony into a camera body and calling it a day. It often speaks volumes about a company’s legacy when they can’t nail the final image output despite great sensor capabilities. I highly doubt Sony’s habit of crippling themselves in RAW output (forced lossy compression, aggressive sharpening and “star-killer” noise reductions… the list goes on) are done consciously. More likely that they’re paying the entry price of making a professional system.

    I wish Hasselblad the best given how much potential their heritage has, and that they finally stopped making those funky ultra-luxury Sony rebrands…

    • Does everybody really use every feature in their Sony on a regular basis? I have yet to meet that person 🙂

      I too am glad we are no longer rebranding anything – that was fortunately way before my time…

  6. Hi Ming,

    Nice essay. It raises several interesting questions, both philosophical and practical.

    Whilst I agree with your end point that all you really need are “focus, shutter speed, aperture, sensitivity and timing of release”, I shoot all the time with a Fuji GX680III with a leaf AptusII back that has no interface with the camera save for a PC sync cable.
    The preceding arguments as to why there could not be an economical “V” back, that seems to be what everyone wants are open to query.

    The argument you seem to put forward makes it sound like the process is starting from scratch, and this is clearly not the case. I realise it is not a trivial task, but all people really want is an update to the CFV with the biggest possible chip in it. Another approach would be to utilise all of the work that has gone into the H6D/X1D platform and release a V mount back with these internals. I hear you about chip yields and size, and we may never get a 56x56mm chip, but surely the current 40x54mm crop could be made workable.

    Kind regards,
    Jason Berge

    • Thanks. Yes – no connection is necessary other than the sync cable; the CFV doesn’t even use one with the V camera. Instead, it is fired solely by a mechanical connection – the trigger pin that was previously used to advance the film counter; this actually moves before the shutter activates, which turns on the back, then the shutter fires, and when you release the button, the pin stops the exposure. In actuality, the exposure time on the sensor itself is much longer than the selected shutter speed – but the shutter determines exposure. A rather clever solution by the engineers of the first CFV, I think.

      Sorry, I think you missed the part where I said the physical sensor board for 54×40 doesn’t fit inside because of the back mount standard (which cannot be changed, since this would invalidate compatibility with all V cameras and other system accessories) – short of doing an entirely new sensor board, or throwing away the mount standard – it isn’t possible to do a larger sensor inside the CFV. The former is always possible, but it isn’t going to be cheap 😉

      • I am sure you know more about it than me, after all you work there.

        I just find it odd that the 54x40mm chip will fit in the H6D “back” which is physically smaller than a “V” back, but not into the CFV back apparently. Just saying.

        Also way back when I was a young fellow (1996) the Dicomed Bigshot managed to jam a 56x56mm chip into a back for the 553eld. Now it had all sorts of issues (mostly related to scsi) but it could be done. Also Leaf seemed to manage to fit a 54x40mm chip into the AptusII 12, and a 56x36mm into the AptusII 10 that rotated internally no less.

        Cheers,
        Jason.

        • That’s because the mount plate between back and body is different on the H compared to the V, and has a larger ‘well’ to accommodate the sensor board. We also need to accommodate the trigger pin, which isn’t all the way to one side (no such restrictions on film) and would cut through the board otherwise. The Sony CMOS boards are not the same dimensions as the Kodak or Dalsa boards that were used in the Aptus, and both of those need sync cables to trigger…

  7. Jaap Veldman says:

    Hi Ming,

    Every complex design is full of compromises and your conclusion is correct.
    That FF smaller version of the X1D still feels good 🙂

  8. Anders Vinberg says:

    Great paper, thanks.

    A square sensor on an X1D would have the additional benefit of eliminating the need for a portrait mode handle. Ok, the X1D doesn’t have one, but there is still portrait mode and the X1D isn’t convenient to hold. The Fuji GFX has a portrait mode handle attachment, and it makes the camera square but huge. Without a mirror, a 44×44 sensor would fit with the XCD lenses, and the camera would remain svelte.

    Ok, your other considerations remain valid…

    • Except, Sony doesn’t make a square sensor. The solutions would be a) leave it rectangular, like what people are already used to; b) use a larger standard sensor (i.e. the 54×40) but not the whole image area and implement crop modes – but then people would be frustrated they were paying for but couldn’t use the whole thing, or c) get a custom sensor, and build the cost in. Clearly, none of these are ideal…

  9. richard majchrzak says:

    very interesting and informative paper ,greatly written , thanks,

  10. Has anybody read that book called the ‘ Blind watchmaker ‘ by Richard Dawkins
    you can read an e-book for free https://docs.google.com/file/d/0BzjGpjXSw6aCY2FhNzQzZmQtMjQ1Yi00MGJiLTk4MTktOGZmMWM3NTdhN2Jm/edit

  11. Ming
    So nice to read about design without someone wanting to discuss ‘narrative’ (I previously worked with Architects, still in therapy).
    I guess that for my camera budget, I will always get lots of electronic features, simply because it cost not much more to design/program in for products with high turnover. Just wish there was more effort with interface though.
    Regards

    • Nope, no narrative: just objective 🙂

      Actually, programming stuff that works properly isn’t so easy – even harder when it has to be genuinely useful in the field and intuitive enough that you remember it’s there (and how to use it) under pressure…

Thoughts? Leave a comment here and I'll get back to you.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: