For little more than a year now, I have been making video marketing productions using two Canon DSLRs as my weapon of choice: a 5D mark II (for its ability to do bokeh-porn) and the 7D (for those smooth Glidecam shots at 50 or 60 frames per second). Although both Panasonic (AF100) and Sony (F3) are introducing affordable ($5k and $15k respectively) new cameras that take in some of the best aspects of DSLRs, it's time to wonder what route Canon should take with its forthcoming DSLR successor to the Canon 5D mark II: the Canon 5D mark III.
Already, the Canon 5D mark III (or will it be called the 6D?) is on the India edition of PC World's most awaited cameras list. Since the 5D mark II was introduced two years ago, the new breed of DSLRs may very well be announced at NAB (in April), but surely we should expect to see them in the shops by IBC and PhotoKina (in September).
But what are the top features we would like to see in the next generation? Here it is: an ultimate wishlist of features for the 5D mark III.
I love DSLRs. No surprise there. I love them for the image quality that you can achieve with them, but I especially like the form factor. They are incredibly small, which allows me to travel with two or more bodies at the same time. Compare that to a 'proper' sized video camera with a humongous video lens attached to it. Not only are you carrying more weight, traditional video cameras are also much more mistrusted by the general audience. People behave differently when they know a camera crew is present. Not so for still photographers. Whenever I shoot in public places, almost everyone thinks I am shooting stills, not video. Often, people will pose for my camera, waiting for that click-noise from the shutter. I usually let them keep that pose for about 5-10 seconds and just when they start to wonder why they don't hear any click, I tell them that I am shooting video. At that point they relax, laugh and give you the best natural looking footage you can dream of (and use in your edit).
Altogether, Canon should definately NOT take the route Panasonic and Sony have taken. In other words: keep the form factor the way it is. Nice and small. Instead, Canon should allow for external add-ons to complement the camera with extra features that DSLR shooters require (more on that below). Much like the BG-E6 battery grip that adds extra power (and weight for stability), Canon could create similar add-ons for recording uncompressed video or allowing for XLR-inputs.
The small form factor also allows me to easily mount the camera to a Glidecam for ultra-smooth moving camera shots. This would not be as easy with a much heavier piece of gear. Many of the shots in the video below would not be as attainable:
Since the majority of these cameras are sold to a different market than DSLR shooters that use it for filmmaking (that is, the photographers that use it for stills), Canon will most likely continue to focus on development of features for this market. Righteously so, I like to add. The power of DSLRs already lies in the ability to interchange lenses as you please (and thus have a shooting range from 8 to 500mm). Being able to extend this with additional components allows for more wider use. Look at RED's building blocks and learn.
Maintain Full-frame sensors
The recent arrival of the Panasonic AF-100 has given rise to the notion that micro 4/3rd sensors are the way forward. I truly hope Canon will not follow that route. The bad thing about micro 4/3rd is that it gives a whopping 2.0 crop factor to existing 35mm glass. Also, Panasonic does not provide particularly fast glass itself, as Stu Maschwitz concluded earlier. And fast glass with at least a constant aperture is what we need to achieve beautiful shots. Sure, you can mount PL or compact primes to the Panny as well, but that won't solve the crop.
The video from Panasonic above compares the size of the sensor to 35mm film cameras versus 35mm still cameras. However, it does not discuss the crop factor. Yet, with a crop factor of 2, all of a sudden your nice-n-wide 24mm becomes a not-so-wide 48mm. Smaller sensors may be cheaper but the existing breed of cropped-sensor sizes of the 1D mark IV (1.3 crop factor) or 7D (1.6 crop factor) is enough for my taste. Combined with a 2x extender, I can turn my 70-200mm F2.8 into a 224-640mm F5.6 on my 7D (loosing two stops of light in the process), so there is no real need for further cropping to narrow my field-of-view.
Instead, the Canon 5D mark III should continue to have a full-frame sensor, meaning that your 24mm remains a 24mm. The real investment for any DSLR shooter is in the glass and one would expect this to last for the lifetime of multiple bodies.
Although full-frame sensors are harder to focus, they also provide a type of bokeh-porn from the shallow depth-of-field that you can't achieve with any other affordable camera. In the words of Stu Maschwitz:
By the way, being able to shoot shallow depth-of-field does not mean that you should. One of the most famous movies of all-time, Citizen Kane, was known for its use of deep focus. But, at least, with a full-frame sensor and fast 35mm glass, you can achieve that shallow DOF, whereas with a small-chip video camera, you can only shoot deep focus. It's all about having options and using the tools wisely and with proper effect.
Provide clean HDMI out
The current breed of DSLR cameras record footage in a highly compressed format, using the H.264 codec. This means that a lot of information gets thrown out and is therefore no longer available for post production image manipulation.
Being stills cameras, the DSLRs already do allow for RAW image recording, albeit for the stills function only. In terms of speed, they allow for between 3.9 (5D mark II) and 8 (7D) frames per second in consecutive RAW (CR2) images:
- Canon EOS 5D mark II at 21 Megapixels and a resolution of 5616x3744 pixels
- Canon EOS 7D at 18 Megapixels and a resolution of 5184x3456 pixels
If Canon would be able to handle the heat that these sensors generate they might be able to up that to 24-25 frames per second, thus giving you a camera that shoots RAW video. One thing is for sure - for video purposes they don't have to move the mirror 25 times a second.
However, the truth is that today's de facto video resolution is not 5k, 3k or even 2k - it is FullHD, or 1920x1080 pixels. Something you might call a 2k sub-standard, since true 2k is 2048x1080 pixels. Most digital theaters have 2k projectors, while the typical home cinema displays only support FullHD at best.
Image source: Wikipedia
Therefore, a better route than trying to shoot 24 or 25 consecutive RAW images per second would be for Canon to provide a clean HDMI out-signal and allow for external recorders to record the image. Again, these could be from external companies, or as a nice bolt-on package from Canon itself. Below are images of two notable products, the Ki Pro Mini and the Ninja from Atomos:
And by clean HDMI out signal, I mean clean. The image should not have any On Screen Display functions. More importantly, however, Canon should optimize this 5k-to-2k translation process and then allow for an external device (or: their own add-on) to record the image, instead of the camera's on-board recording function. The latter would cause the DLSR to overheat in a very short timespan, something an external device could handle better.
Recently, I read an excellent article on FreshDV from Matthew Jeppsen entitled "RAW video - be careful what you wish for", which discusses the pros and cons of adding RAW video to DSLRs. Another important issue with RAW video recording is the ability to process the images in the edit. Most current video editing systems are not fit for handling 5k images. Anyone who has ever tried to make a timelapse from RAW CR2 files knows how much of a strain this puts on your computer's resources (Adobe After Effects CS5 handles it remarkably well). Although RAW allows for tons of more options for image manipulation in post, it also comes at a cost. As Jeppsen correctly concludes:
Given a choice between H.264 and a substantially more space-hogging codec on a DSLR, I would probably still choose H.264 most of the time. Which is why the only "raw video" I want Canon to add in the next DSLR is a proper HDMI output. That would give me all the "RAW" options I'd need, for the handful of times I'd have to use it.
Aside from enabling a clean HDMI out, Canon should also allow for a variable setting of the bit-rate at which the H.264 codec operates. Already, the guys at Magic Lantern have shown that these cameras are able to go upto 70Mbps, thus giving you far less compression and more information to work with in post production.
Proper audio in- and outputs (and monitoring)
Much akin to traditional cinema, using a dual audio system is the safest route to good results with a DSLR. The on-board audio just cannot be trusted for professional audio quality. Sure, you can attach a Beachtek with XLR inputs (and an audio output), but there is no way for you to monitor the audio output on the camera itself. If you attach your headphones to the AV-out on the camera you will not get audio output and instead loose your video signal on the LCD screen. So there is no way for you to know if the audio is properly recorded on the DSLR. As the hackers from Magic Lantern have shown, Canon could instantly enable audio monitoring through a simple firmware upgrade.
Until the March 2010 firmware upgrade, which gave us 24/25p and manual audio, the Canon 5D mark II only had audio with a bismal Auto Gain Control (AGC), a feature the Canon 7D still has. After the upgrade, Canon did enable the setting of manual audio levels, but this feature is buried so deep inside the menu, that you cannot access it whilst recording, let alone change its settings easily.
The next breed of Canons should allow for on-board monitoring of audio via the AV-out (you will use the HDMI connector for video output anyway) and allow for an add-on box featuring two balanced XLR inputs on the L and R channels. Again, could be from Beachtek or as a battery-style grip from Canon itself.
Although the Canon L-Series glass is of phenomenally great quality and speed, it has two main drawbacks: the focus ring and the lack of manual aperture control. In short, these are stills lenses and they were made for — stills.
First, the focus ring was made for electronic, not manual control. In order for the brain inside the camera to get focus fast, the amount of travel on the focus ring is kept to a minimum distance. Although great when shooting stills in auto focus mode, this tight focus throw makes maintaining focus on moving subjects when shooting video more of a challenge. A larger lens throw would be wellcomed. Such larger focus travels can be found on the CP2 Compact Primes series from Zeiss, which feature a 330° throw.
Second, aperture - or F-Stop - can only be controlled from the camera body, not on a separate ring on the lens itself. Instead of allowing for the aperture blades to smoothly open or close down, this feature is controlled electronically with noticeable clicks: the shortest range possible is with a 1/3rd stop setting. Noticeable, since you will see the stopping down (from say F2.8 to F5.6) in your recorded image and there is little you can do to fix that to a smooth curve in post production. Older lenses from the seventies used to have a aperture ring on the lens, although I am not sure how smooth these could be controlled.
Manual control of aperture is a desired feature when shooting video. If you move the camera from a very dark to a very light room, you want the transition to be as smooth as possible. Also, it allows you to enlarge the focal plane, provided you have some control over the incoming light at the same time.
Another thing video lenses have as a benefit over stills lenses is the ability to remotely control zoom and focus. Especially when the camera is mounted on a crane or a steadicam, remote control is a key issue. Traditional video cameras have an on-board control mechanism for this - a LANC. DSLRs, on the other hand, require the addition of external motors to control these two lens rings.
If Canon were to allow for a series of lenses that feature a) a larger throw of the focus ring, b) smooth manual aperture control and c) a LANC-like control function, they would appeal to a new market of video enthusiasts and filmmakers alike. Canon could learn a lot from the popular Zeiss CP2 Compact Primes series which retail at premium prices. For me, this would be a reason to sell my current primes and upgrade my glass.
Fix Moire & Rolling Shutter
In order to go from an image captured by the sensor at full frame resolution (5616x3744 pixels) to a FullHD image (1920x1080 pixels), the Canon DSLRs use an algorithm that is not very precise. The current sensor processes the image in such a way that lines are skipped. This is causing moire. In other words, pure nastiness. Trying to shoot repetitive patterns, such as a brick house or a shirt with fine lines quickly introduces those funky, radiating patterns which cannot be easily fixed in post either. Unless you like frame-by-frame painting ;-). Canon should therefore improve this translation process.
One problem persist, though: the stills guys want Canon to up their game in the Megapixel race. Some rumors exists that the Canon 5D mark III will feature an upgrade from 21 to 28 Megapixel. If that translates into higher resolution, the translation to FullHD may even become a bigger challenge. Guess we'll have to wait and see how the new camera handles this.
Another issue has to do with the way CMOS sensors record the image. Unlike a CCD-chip, they start at the top and proceed towards the bottom. What happens when you record a fast-moving image (e.g. a bus passing by a steady camera or a whip-panning camera on a set of straight lines), is called Rolling Shutter.
Although there is a way to fix this with a plugin by The Foundry, Canon could fix this by allowing multiple CMOS chips to work on the image at the same time, hence reducing the time difference between the top and the bottom parts of the image. The Canon 5D mark III is rumored to have dual Digic-IV sensors (the 5D only has one), so that will hopefully help solve this problem once and for all.
Various other neat functions
Traditional video users are accustomed to various features that help when shooting, such as Zebra marks that indicate the extend to which your whites are clipping. Although I do not really miss such a feature, I can imagine this to add value for others. Again, Canon could learn a lot from the Magic Lantern firmware hack. Alternatively, the forthcoming electronic viewfinder, or EVF from companies like Zacuto and Red Rock Micro are likely to have such a feature integrated.
Aside from Zebras, a really neat feature would be the ability to digitally zoom in whilst recording. Currently, you can use the 5x and 10x zoom button on the back of the camera to check focus before hitting the record button. Use of this feature during a recording session would allow you to re-check focus. Of course, the zooming in would not need to be recorded on the flash card.
Another useful feature would be a short-cut button that allows you to see a short review (5 seconds) of the last recorded clip. Something pro-sumer level video camera's provide and can help you frame up your next shot.
Finally, an A-B setting mode to automatically rack focus between object A and object B would be a nice addition. This is already common in most (semi-)professional video cameras.
It will be interesting to see which features Canon is able to implement from the list above. One should not forget that Canon will continue to sell these as stills cameras to the photography community and that they have different demands than DSLR shooters. Some photographers I know rely on auto focus to a great extend, while I tend to operate these cameras almost exclusively in manual mode.
Also, Canon has to deal with the internal issue of having a professional video department that sells small-sensor-big-form-factor cameras to the prosumer and professional markets. These are the people that may consider using DSLRs, especially if they start to look and work more like video cameras. Given that you pay about $7000-8000 for a Canon 1/2inch chip video camera (standard video lens included) and only $3000 for a DSLR body and nice L-Series Zoom lens, it's not hard to do the math. On the other hand, if Canon were to take an approach of add-ons, as suggested in this article, the total spend on the camera could easily go up. Plus, people will be more likely to build-up a range of lenses with DSLRs than with video cameras, which is more profitable for Canon.
Also, the unexpected popularity of DSLRs over the past few years has given rise to a whole ecosystem of manufacturers that make complementary products. Canon should embrace this and nurtur it. If your product constitutes an ecosystem, you can be right at the center of it all.
Taken together, I do hope that Canon will listen to its users and give them what they want (and, perhaps, need). The Canon 5D mark II was the best thing that ever happened to Canon, so please let the mark III (or 6D?) be even better than that.