EOS DSLR Review
For over year and a half now, public interest in the new field of the so called “HD-DSLR video” has been certainly overwhelming. And a little bit over the top most of the time, if I may say so.
Canon wasn’t the first brand to present a DSLR camera with HD video functionalities, nor the most imaginative once they did, but one has to agree that it’s been the company with a better overall vision of this niche market so far.
To this day, five Canon DSLR cameras (out of a total range of seven inside the EOS sub-brand) are capable of shooting HD video (the 500D, 550D, 7D, 5D Mark II and 1D Mark IV) and it’s not unreasonable to think the rest will be updated including HD video capabilities at some point during 2010.
This review aims to help users squeeze all the video functions available within their Canon DSLRs, bearing in mind this could be a very powerful video tool but -at the same time- a limited one in many ways.
Not even Canon was aware of the potential hidden under the case of their first HD-DSLR product (the EOS 5D Mark II) when it was unveiled, back in September 2008. Developed in order to respond to major agencies and newspapers demands for a new tool allowing their photojournalists to shoot good pictures and some additional HD video coverage for their media websites –hence those damn 30fps-, no one could expect the excitement and the huge level of attention low-budget filmmakers, DOPs and little production companies all around the world would devote to it. And that was only the beginning.
Don’t get me wrong. Canon’s marketing departments did count on some minor agitation in the video world and their engineers were already prepared to develop firmware updates allowing -or to be more accurate “unlocking”- manual control and 24/25p shooting modes from day one. But it was no big deal.
Typical clean and boring demo footage from Canon Japan was all you could see for the first few days after the launching. There were no ads focused on videographers at all. No one was paying too much attention to it. After all this was (and still is) a photo camera.
And then, out of the blue, came Vincent Laforet and “Reverie”.
Laforet, a well-known french-american photographer borrowed a pre-production camera from Canon USA to test it and decided to spend his own money in the making of a video using a bunch of lenses, a couple of models, a fancy car, a little classic lighting, some aerial shooting of NYC by night and lots of little gadgets. He edited those clips adding a kinda confused plot, a glossy grading and some timeworn Moby’s musical background and the result was an “in your face” HD video which shocked half the internet users in less than a week.
Canon USA got the short movie quickly transferred to the company’s own site and it didn’t take much longer before it actually collapsed due to so many downloads. The wave kept rising and “I didn’t know a DSLR could do THAT!” became the commonplace sentence of the quarter.
While we could argue for ages about its doubtful taste, “Reverie” turned out being the best (and probably the cheapest) commercial Canon could ever dream of, so Mr. Laforet clearly deserves every penny he earns as advertising photographer (just one of his many successful facets).
From then on, the japanese manufacturer has tried to be a step ahead the rest. First releasing an entry level DSLR with video features (EOS 500D), then unlocking manual control on the EOS 5D Mark II, later on releasing three new HD-DSLR cameras (with some noticeable improvements as 24/25p and 50/60p (720) in the EOS 7D, 1D Mark IV and 550D).
Eventually Canon launched -in mid march 2010- a firmware update for the EOS 5D Mark II which unlocked 24/25p features. It also included new histogram display for shooting movies in manual exposure, the addition of shutter-priority (Tv) and aperture-priority (Av) in movie mode and improved audio functionality that allow users to set sound record levels manually using a sound-level meter displayed on the LCD screen. The audio sampling frequency was also increased from 44.1KHz to 48KHz, providing the audio signal typically required for professional or broadcast material.
On the other hand, Canon always kept this advances and updates to the minimum in order to optimize mass production. This “little by little” strategy rises the sales and the life of a camera like the EOS 5D Mark II and helps setting differences between the EOS ranges (“Full HD” 1080/20p in the -already dated- EOS 500D being the most ridiculous example of that policy).
Shallow depth of field is, obviously, the main reason behind that “film look” everybody loves about HD-DSLR video. The size of an APS-C sensor like the one used on the EOS 7D and 550D (with a crop factor of 1,6x) is pretty close to a Super 35mm frame or a S35 sensor. On the other hand the EOS 5D Mark II’s full frame sensor is equivalent to a Vistavision Frame. The APS-H sensor of the EOS 1D Mark IV (with a crop factor of 1,3x) stands in the middle of the other two (with no equivalence in film). Such sizes allow a richer tonal range, a better bokeh and a much better behaviour with high ISO speeds than videocameras with sensors of 1/3rd, 1/2nd or 2/3rd sizes.
Electronic noise, however, does appear sooner or later with all digital cameras. Specially in the darker areas of the image. There are several reasons why this noise might be noticeable:
High ISOs: We can expect clean images in most situations up until 800 ISO in the EOS 7D, 1600 ISO in the EOS 5D Mark II and 3200 ISO in the EOS 1D Mark IV. But with good lighting conditions you can reach even higher ISOs without notable flaws.
Overheating can also cause thermal noise and a general diminishing of the image. And that’s something you won’t be able to distinguish in the camera LCD without an external viewfinder. Therefore it is advisable not to shoot continuously over long periods of time. Besides, the EOS 7D and 1D Mark IV will automatically turn off the video shooting in such cases no matter how important and unique those images might be so… watch out!
Canon includes some preloaded gamma curves (called “picture styles”) on each one of its DSLRs so the user has -to a certain extent- the final say on the eventual look of the in-camera images before shooting. Chances are not nearly as many nor as amazing as in their -already dated- range of XL/XH prosumer camcorders (which will be replaced this year), but then again these happen to be photo cameras. Good news are each style allows little changes in sharpness, contrast, saturation and color tone so you don’t have to be stuck in their basic curves for long. Those preloaded styles are:
Standard: According to Canon this style is “set to produce the vivid colors and contrast level that people tend to prefer for general photographic subjects. It provides the optimal sharpness for printing image without post-processing and makes it possible to produce beautifully finished prints for a broad array of subjects, from snapshots to sports shooting, with no retouch”. In the real world that means high contrast levels and pretty saturated images (specially in the reds). While it’s true that most photographs will benefit from this look (which is the default one on any EOS camera) whenever you’re shooting video this might be your worst nightmare. Such contrast produces stacatto and aliasing pretty often. Besides, moiré and even maze artifacts are inevitable on any detailed texture from a colourful subject.
Portrait: It does adjust the color tone magenta-to-yellow close to red range and adds brightness. Canon claims it’s better to reproduce the skin tone of women and children. In my experience, this style is something to be avoid at all times (at least when shooting caucasian faces). Otherwise, every skin will look like pig skin. Period.
Landscape: Basically rises saturation and changes brightness on blue and green colors independently, turning them deeper (blues), brighter (greens) and more vivid. As moiré in Canon DSLRs is specially critical within reds we might get away with this one, but being extremely careful on avoiding too much detail within our landscapes. This style also uses a stronger sharpness than any other, so please turn it down or you’ll host an aliasing feast in your videos.
Neutral: By far the most appropriate style most of the time. Contrast and saturation settings are moderated, so there is less risk of overexposure and color saturation compared to other styles. Manually reducing contrast down to “-4” and saturation down to “-2” you’ll be the closest you can to a general all-purpose style without creating curves on your own. Richer detail is retained as data, so corrections can easily be rendered later on post.
Faithful: Pretty close to the standard style, only less contrasty and maintaining the color taste near the actual subject. Can be quite useful whenever the original colors of a scene must be accurately expressed.
Monochrome: Obviously, the ideal one for black and white images. Sharpness is set relatively strongly (beware). Sepia, blue or other toning effect is applicable to create a monochrome image in that color. Also, red, green or other filter effect can be applied to control the color. Using them in a proper way (and avoiding too much sharpness) this might be another good style.
In addition to the six preset styles, three more Picture Styles can be registered through connection to a personal computer. And some more can be applied in post (with such uninspired names as “studio portrait”, “snapshot portrait”, “nostalgia”, “clear”, “twilight”, “emerald” and “autumn hues”).
Or you might create your own curve with the Picture Style Editor software, which allows you to control color freely using numeric values. Needless to say, some know-how and some dexterity are required as well as an appropriate color management throughout the process. There’s a growing online library of custom styles available wherever there’s a HDSLR forum so you might want to test some of them too.
Debate still rages about the exact nature of the motion perception process, but these days it is generally agreed that what makes us perceive motion in place of a rapid succession of still images is a psychological effect, known as the phi phenomenon, an optical illusion in which the rapid appearance and disappearance of two stationary objects such as flashing lights are perceived as the movement back and forth of a single object.
At low spatial frequencies, the human visual system’s temporal contrast sensitivity function (CSF) is roughly bandpass, whereas at high spatial frequencies it is low pass [Robson 1966; Koenderink and van Doorn 1966]. This implies that at low spatial frequencies the visual system is more sensitive to high image speeds, and at high spatial frequencies it is more sensitive to low image speeds. For example, the visual system can detect much larger frame-to-frame displacements in images that are dominated by low spatial frequencies than it can for images dominated by higher spatial frequencies.
The perception limit of our human visual system is delimited by a maximum amount of 43 to 45 images per second. Film cameras (rolling at 24 fps) depend on the shutter angle to get as close as possible to our human vision. A 180º angle is the average choice in most cases (however, this is never a strict rule to follow, for there are many other considerations a DOP has to bear in mind in order to obtain the kind of image and motion required by the director).
We must “translate” these shutter angles to shutter speed values and that can be done with the following simple formula:
Frames per second X 360º / desired shutter angle = shutter speed needed.
For an instance, if we are shooting at 24fps we have to calculate 24 X 360 / 180 = 48 (meaning the shutter speed will be 1/48).
Therefore, a shutter speed of 1/50 (there’s no 1/48 available) is the closest one to our vision allowed by a DSLR (whenever we shoot at 24/25 fps, 30 fps will require 1/60 shutter speed instead).
With such sensor sizes, any faster shutter speed will produce stacatto while slower speeds will increase motion blur significantly. So, unless we want our video to look like the first 20 minutes of “Saving private Ryan” or -on the other side- like a cheap phantom’s movie we must stick around the 1/30 and 1/125 values.
There’s yet another reason to be careful about shutter speeds. The frequency of the electrical system varies by country but most electric power is generated at either 50 or 60 Hz. So if you’re not working with professional lighting equipment the only way to avoid flickering is by shooting at 1/50 or 1/60 shutter speed (depending on the country you’re shooting at).
Some Canon EF lenses incorporate an image stabiliser to prevent camera shake from spoiling the shot. Once its turned on the IS will keep stabilising during the whole video shoot. But just as well as with ENG video lenses, the IS must be disconnected when attaching the camera to a tripod in order to avoid jumping images while panning.
As with most CMOS sensors, those of Canon DSLRs do record each individual movie frame not as a single snapshot of a point in time but by scanning across the frame either vertically or horizontally. As a result not all parts of the image are recorded at exactly the same time though the frame is stored as a single still on the CF card. This produces predictable distortions of fast moving objects or when the sensor captures rapid flashes of light.
Partial exposure seems to be the biggest issue with the rolling shutter of this cameras so whenever there are flashes or ER vehicles around be aware those lights might lit up part a third of your frame while the rest remains in the dark.
As for the jello effect it is noticeable in pretty fast pans but it’s no big deal generally speaking. So skew can be avoided and i’ve found no wobble so far.
In case you can’t go for another take try correcting this effects with the “Rolling shutter” (by The Foundry) plug-in for After Effects applying a correcting factor of 0.56. That’ll more or less do the trick.
Heads and Tripods
Most people are looking for a good priced fluid head to use with this kind of cameras. Manfrotto HDV Pro Fluid Video series (which are friction plate designs instead of true fluid heads) have become very popular and have earned a reputation of very cheap and quite reliable heads. I’ve briefly tested the 701HDV, 501HDV and the 503HDV models so far.
I was quickly done with the first two. The 701HDV is not strong enough to operate with heavy lenses and/or rigs. And there’s no way to pan smoothly with the 501HDV. Not at all. Besides, tilt, lock and drag functions are uncomfortable and uneasy to manage fast and properly. And then there’s that terrible spring-back (enough to ruin any decent pan). To be honest, it got on my nerves.
The 503HDV is quite a different beast. First of all, it’s wider and sturdier and therefore has a much better mount plate for a DSLR body. The tilt, lock and drag functions are much easier to manage (situated on the left side of the head) and fast enough to operate with.
Unfortunately, the pan drag control sits under the camera support plate and it can be a bit tough to reach and a little nightmare to adjust, specially once the camera is mounted.
One of the best features on this head is the four-step counterbalance spring for camera balance. Bearing in mind there are (significant) weight differences within the EF lenses range (conformed by more than 60 lenses) this counterbalance spring is extremely useful.
Another useful feature is the leveling bubble which includes a LED light powered by a disc battery. I can’t even recall how many times did i thank Manfrotto for having that little light in dark enviroments.
The telescopic arm is strong enough and easy to attach and remove (you can move it to the left side or even add an extra arm for LANC remotes purposes). But even being a sturdy head, there’s a bit of spring-back on the 503HDV too, only you can easily deal with it.
The only way to avoid spring-back completely is using a multiple tube tripod. Manfrotto offers a kit with the 525PK tripod, which is ok. It’s major disadvantage is the lack of height once extended. On the other hand, those thick legs are light and strong at the same time thanks to their combination of magnesium and carbon fiber which made them a real “built to last” product. And –believe me- your back will appreciate their lightness. So i definitely think if your looking for the best quality/price ratio on a video tripod this product is a something to be considered.
Now, if you can spend your money on a true fluid head go for the Sachtler FSB-8. You won’t be disappointed. It’s extra weight capacity (extending the payload range to 19.8 lbs (9 kg) is perfect to use with all kinds of gadgets attached to your camera allowing you to adjust the center of gravity with ease. It provides a 10-stage counterbalance as well as five steps of drag (plus 0) for both tilt and pan. Speedbalance and Sideload technology allows for quick mounting and un mounting and combined with any of the tripods offered with it by Sachtler it becomes a bliss to use right after the unpacking.
Zacuto, Arri, Red Rock, Chrosziel, Cavision and Genus (among others) manufacture rigs for HDSLRs. All those solutions are mostly ok, although quite expensive for what they’re worth. The basic parts of those rigs are usually:
Matte Boxes: Not essential for every shoot, but indispensable to avoid IR pollution with ND filters. The cheapest ones are dangerous for filters, which might get several scratches during use. Among the reliable ones (such as the ones from the already mentioned brands) the main issue is their huge size (too big for most EF lenses). Genus and Red-Rock are the main exceptions in this case, and their DSLR models suit perfectly Canon lenses.
Follow-Focus: By far the most important part of a rig for DSLRs if you want your focus puller to survive the shooting without killing himself or anyone else. It has to be used gently not to force the lenses’ torque too much. Arri has the most unbelievable ones, well build and very accurate –hence their also unbelievable prices-. On the other hand, the Red-Rock follow-focus is not trusty enough. IDC’s one is a plain and simple disaster.
Finder: Must provide a good magnification and has to be easy to adjust properly in order to see the whole LCD frame. Here Zacuto’s Z-finder stands out as the best one. It has a 3x magnification and a fine diopter adjustment wheel with a pretty wide range.
A portable video camera dolly offers silky smooth tracking camera shots without any of the hassles and constraints of traditional video camera tracking and dolly systems. Glidetrack offers the wider range of portable rails for that purpose, the Glidetrack HD being ideal to use with DSLR cameras. It’s available in compact 0.5m length, standard 1m length and extra 2m. lenght. It can be used as a somewhat strange rig too.
Kessler Crane’s CineSlider is the best portable dolly around. Much more versatile than other models in the market though a bit more expensive too.