Canon 1Dc review/ Contact
let's talk about goodness
If you have questions, or need anything - including source files, frame grabs, advice, or you would like to buy me a 1Dc, don't hesitate!
All rights reserved.
Initially I wanted to call this article “Canon 1Dc – the king of HDSLRs”, but then I thought: what is the first thing that come to mind, when people mention the word HDSLR today, in 2013?
I mean in the beginning when the 5DmkII hit the market no one complained about its shortcomings, because it made something available to the masses, that wasn’t possible before. But in the following years, as these HDSLRs got their piece of cake in bigger productions, more and more people started pointing out the possible problems with them, and some went all the way to simply complaining instead of trying to get the best possible image with the given gear. Some say the sensor in the 5D is too big, creating too shallow depth of field for cinematic work, some say it’s hard to color correct these footages including the C300 footage, some say that moiré makes their life harder, some miss the dynamic range in movie mode compared to still mode, some find it not sharp enough at all, and some say you should be embarrassed, and Canon should be ashamed because of the line skipping, and the 1080p resolution…
And none of this true for the 1Dc.
It doesn’t line-skip, so it doesn’t have moiré or any other artifact, it’s not recording the full frame sensor area, it has a C-log gamma curve maximising dynamic range, and has a hefty compression – all of this in 4K resolution. You can’t compare it to previous, or current HDSLRs, but if you do it will be a clear winner. Instead, you can compare it to current, large sensor cameras. This makes it even a more interesting animal, because under the hood it is also the current top-of-the-line Canon still camera, the 1D X. The price of the camera tells us that Canon will have to find the user base who are top photographers, and top cinematographers at the same time which is going to be very difficult
Luckily we had more time to test the camera than we had with the C300, so the idea was this: put the 1Dc in real-life situations where we can push it’s sensor and codec to the limits, and try to find the difference between the 1Dc, HDSLRs. This wasn’t enough, we had to put it up against the C300 as well, because it performed so well. As you will see, this adventure ended up bringing the 1Dc to a studio, outdoors, shooting with it at night, bringing it underground, and into a hangar. No resolution charts, no dynamic range tests, I’ll leave those to others. Also I won’t write much about ergonomics, or usability, I’m only curious about the image quality. We will use our most precious tool, our eye to judge this camera. I’m sure if you watched enough source footage from different cameras, you will be able to judge resolution, dynamic range, and noise after this review.
So thanks for reading the intro, scroll down, and enjoy!
Note: all images will open in a preview window. These are resized to your display, so if you’d like to take a look at the full size image, there’s a Download link at the bottom. Videos are uploaded to vimeo, so they are compressed – these won’t really show the real quality of the footage. Upon request I would be more than happy to share Prores files – as I did in the Footage section.
The 1Dc has a full-frame sensor, although you can only fully use it in still, and in 1080p movie mode. When using the 4K recording mode it crops into the sensor area, to create an image without line-skipping. This ensures that the image will be artifact free, and sharp as can be.
Compared to the full frame video recording, this is how much you will crop into the sensor in 4K recording mode, in other words, this is how much field of view you will lose:
The color part of the image is the 4K recording, B&W is full-frame
Canon has used H264 compression in previous HDSLRs, and MPEG compression in the XF cameras, also in the C300. Now they introduced a new compression in their camera-line, available only in this camera. It is a Motion-JPEG compression, packed in a MOV container. Bit rate is about 500-530Mbits/s, which is about a 1:7 compression – if I calculated it right. This is not a bad ratio considering that for example the Red One had a 1:7.5 ratio using the highest RedCode42. Of course you will find the limits of this codec if you push the image very-very hard. In the test I’ve included some uncompressed TIFF files, and at the end, there’s a 2K Prores file, so you will be able to test the codec yourself. It is still 8bit as the C300 which is strange for me, as 10bit wouldn’t require much more data throughput. I don’t know what’s the limitation of using 10bit in this camera, but I would be glad to see 10bit bit depth in future Canon cameras.
If you’ve read some headlines from the CES 2013, I’m sure you already have 4K-poisoning, but bear with me. The 1Dc is capable of recording a 4096 pixels wide picture, and it does it by recording the file internally to CF cards. This puts it in a higher league – a lot of the cameras only 4K capable using an external recorder. The image size difference between 1080p and 4K is quite big:
For a better example, I’ve grabbed two frames from a 1080p, and a 4K file, and composited the statues together, both in their original size:
It is also worth mentioning that the 4K mode has a different, wider aspect ratio, not 16:9.
This is one of the main features that differentiates it from previous DSLRs. This is the first time that a camera manufacturer creates a log(ish) gamma curve for it’s video cameras – the C300 was the first in the line by Canon- and includes it in a still camera body. Yes, you can still load Technicolor’s Cinestlye on the 5DmkIII for getting a flatter image, but there are two things that makes the 1Dc/C-log pair better: Cinestyle was made by Technicolor, a great company, but still Canon knows their own sensors, cameras the best. The second thing is that what you need for a good log image is a good compression, and the reason for this is simple: after recording the picture with a log gamma, you will push the original flat image very hard to get a more pleasing image at the end. During this operation things can fall apart with a bad compression. For me it sounds crazy that Canon included C-log in the 1Dc, but am I happy they did? Absoloutely. Check the sample image, and the waveform beneath it:
Without the C-log gamma, you would lose all the highlight info that’s there in the C-log picture. Of course judging exposure viewing a log picture would be hard, so there’s a View Assist function works the same way as in the C300.
The OMG Visuals team has managed to get the location and the car for their own shooting, but I was able to pop by, and bring the 1Dc with me. It was a leap of faith from them to use the 1Dc without any previous tests, but I hope they won’t be let down when they start editing the footage.
The hard thing about these kind of places is that it’s almost unavoidable to have crushed blacks, and blown out highlights in the picture, no matter how you expose. But what differentiates cameras is that how much they can handle both ends, and when they reach their limit, will they introduce some artifact, or will they crush/blow smoothly. I was lucky enough to have a film camera with me at the shooting, so for a reference, we can checkout this picture:
Where film gets ahead of digital is that when you overexpose it – let’s take a look at the lamps – light bounces between among the emulsion layers, creating a blooming, glowing effect. The other thing that film has in advantage is the dynamic range, with a logarithmic senisitivity – this means that it gets less sensitive as the light gets brighter and brighter compressing the highlights.
Now let’s take a look at a framegrab from the 1D c:
First thing that pops up is sharpness. I’ve uploaded the image as a 2K picture, so it’s not even 4K, but I’m sure if you watched enough footages, you’ll notice the difference – it’s as sharp as a 2K picture can be. And this image wasn’t even shot with a prime lens…
The next thing is the dynamic range we mentioned. The Canon log gamma will help you a lot to use the sensor’s full dynamic range. Instead of recording a nice contrasty image that is pleasing to the eye, it’ll record a flat, log-like picture as you can see on the left side. Without this, you would lose all information in the car’s headlights for example. Canon included the View Assist feature, which means that even though under the hood the camera will record a flat picture, you can watch – and send to the clean HDMI feed – a nice contrasty, REC709-like picture. You will be able to judge your lighting, exposure with this tool better. One thing to note is that in playback mode, and also when you zoom in to the Live View picture, you will see the Canon Log picture even if you have the View Assist on – I wouldn’t mind if this would be a $3000 HDSLR, but if it sells for $12.000, and we call it a Cinema camera, these things should be customisable.
After we shot some exteriors of the new Golf VII, the guy who’s responsible for the indoor pictures grabbed the camera, and went ahead. Because there was no additional lighting, and the interior of the car was all black, he asked how much can he push the sensitivity without seeing any noise. I’ve said ISO 1600, and it turned out quite well:
When using high ISOs there’s one other thing that’s even more interesting than noise: how much detail we lose at higher sensitivity. Maybe because of the 4K frame size, but it seems to me that the 1Dc won’t lose image detail, even if you push the ISO above 1600. I’ll show you more example of high sensitivity sharpness in the Subway test.
For testing the codec, I grabbed a frame from the opening shot, and started pushing it as hard as I can. The first image shows the original C-log footage. The middle picture shows a contrast adjusted image – not like the REC709 LUT, but using the Levels tool. The third one has the final grade. This included even some HDR-like tone mapping (mostly in the shadows, mid tones), so during a “normal” color grading you won’t even push the footage that hard.
Passed? I think so.
This was one of the most interesting locations we brought the 1Dc. Even though we had limited time we tried to shoot sample footages as much as we can, so you would be able to download them.
The first thing that popped out at this location is that no matter where you pointed the camera, there were super-bright highlights in the frame - I would love to see how other cameras perform in this environment. And again, here comes the C-log gamma, and the dynamic range of the sensor, and together they solve the problem:
In the log image you can see that all the information is there without any clipping, this means you can use it later in post production. Without the C-log gamma you would record an image more like the middle image where I applied the C log-REC709 LUT. You would get brighter – in this case surely clipped – highlights, and lower blacks. The third image shows that even if you apply contrast on the image, you will be able to bring back image data from the first image selectively – check the lamps at the ceiling.
For a comparison I put together a few shots, so you could compare how much information is lost without the C-log gamma:
This location didn’t allow us to shoot at low ISOs, even with fast lenses (f1.4, f2.0) we needed to shoot between ISO 1600 and ISO 3200. As we raise the sensitivity to this range we should lose a lot of detail because of the noise.
When we had a bit of time left, I shot a noise test:
The first thing I loved when I put this one together is that there’s absolutely no color shift. We cover a huge sensitivity range by going from ISO 160 through ISO 25600, but color doesn’t seem to suffer at all.
Yes, there will be more noise as you raise the sensitivity, but the image is still very clean even at ISO 1600-ISO 6400
Check it in motion too, although we will se it through vimeo’s nasty compression:
When we had some time between the scheduled tests, we went out, and started shooting in the city. These weren’t meant to be scientific tests, but I wanted to shoot some test footages for you to be able to downlad them.
Here’s a noise test I did in daylight. My opinion is that two types of noise tests can be interesting when testing a camera: one of them is when there isn’t enough light to shoot at low ISO (at night, indoors, low lit locations), and the other one is when you have enough light to shoot at base ISO, but you would like to stop down. The latter can be found here, I’ve shot seven videos with different sensitivity, and composited the videos together:
If you would like to see the noise in movement, check the video below:
What needs to be recognised is that – with this sensor, and dual Digic 5+ processors – even using very high ISOs sharpness, and color won’t be affected, I’ll say this it a lot, but sometimes this is more important than how much noise you have.
Next I wanted to compare the difference between using the 4K Motion JPEG recording, and the 1080p ALL-I recording. Because the two recording options uses different sensor sizes I compensated by changing the focal length on the Canon 70-200mm. The best way to evaluate these comparisons is to hit the download link in the pop-up window.
Again, the first difference is sharpness. Even though both images had been resized to FullHD for this comparison, the 4K file still looks sharper, more detailed.
You could say that you always apply a little sharpening on DSLR videos, why not solve the softness in post? You could do that, but if both sides have been applied the same amount of sharpening, the right image looks even more detailed, almost too real, while the left side looks just ‘normal’
For checking the compression difference, I’ve created a green channel only,
and a blue channel only comparison. H264 compression tends to soften details, and blocking larger gradients.
This is not the case with the Motion JPEG compression, check the dark part in the bottom right corner.
And you can download all these comparisons together by clicking on this image, and hitting the download link
This frame has been uploaded in the original resolution, so you can download it, and pixel peep it. I’ve applied some contrast to the original footage, but otherwise it’s untouched.
I didn’t have the 5D Mark II, and the 5DmkIII available to test moiré side-by-side, so I grabbed the test files from last years quick test:
Be sure to check out the full-size image to be able to judge the moiré, for me the 1Dc is absolutely moiré-free.
For people who come from the still world and started shooting videos with their camera, one of the difference could be the different sharpness they saw on footages.
So to test how the 1Dc behaves we did a comparison between a still frame (shot to JPEG), and a frame grab from a movie file :
What struck me is that this camera won’t lose sharpness at all, if you use it in Movie mode. Of course even though the compression format is the same – JPEG -, the still picture is much less compressed. We can spot out the difference in the color picture too, but in the blue channel only picture, it’s very easy to see that the Movie frame is more compressed.
Obviously this isn’t a fair test because of the different processing done to the two files. The still JPEG was made out of the raw data in-camera to this nice contrasty picture, so no post adjustment was needed.
As for the movie still, I had to push the C-log very hard to get to the point, where the contrast was the same. Still I found it nice to compare the two, and see how the 4K Motion JPEG holds up.
Of course the outdoor tests included some time spent at night shooting gorgeously lit buildings. This meant that the 1Dc had to fight with the low light conditions – ranging from ISO 640 to 25600 -, and the high contrast scenes.
Here’s a frame grab of one of the first tests during the test days. Notice the highlight falloff, even when it’s close to be blown out, looks fairly natural to me.
You can download this frame in 2K resolution as a TIFF file here so you can grade it in Photoshop and see how it holds up.
And of course another few noise tests was done at night: recorded in C-log, but I applied a LUT on all of them.
And here’s another frame grab, where everything is given to see some bad things – given the high contrast, and the ISO 2500 sensitivity
What I really like in this image is detail. The Canon 135mm is a sharp lens, but I’ve never seen it performing this sharp on cameras.
And again, if you simply open the image in your browser, you will see a downscaled, 2048 pixels wide image. So yes, there is a point shooting at 4K even if the final resolution is 1080p, or 2K.
For those who want to examine this image, I uploaded it as a 4K uncompressed TIFF file, which you can download here
I was lucky enough to know Attila, who offered his Canon C300, and 5DmkIII to test them against the 1Dc.
There are two things that are interesting and should be mentioned before talking about the images. One of them is that using the C-log gamma on the 1Dc results a flatter image than using the same (should be the same) C-log gamma on the C300 . I don’t know the reason for it, but it happened all the time when we compared the two cameras. So because of that, for the following tests I applied a little contrast for the 1Dc footage – except the first, color image.
The other strangeness/mistake was that somehow the Zeiss lens on the C300 was a bit darker at f2.8 than the Canon lens on the 1Dc. They are equally bright at f4.0, and f5.6. I didn’t catch this at the shooting, sorry for that.
Here’s the first comparison, where you can see that the 1Dc is somehow more flat..
Here’s the same setup, but with the little adjustment on the 1Dc footage that I mentioned above, and displaying the blue channel only. This is the weakest color channel of the three, so we will be able to spot the noise, and compression artifacts better.
Same setup at ISO 3200
And pushing it to ISO 6400
As much as I love the compression of the C300, the 1Dc’s compression looks better for me – just look at the gradients on the door. Of course we need to mention that the picture doesn’t look ten times better, but the bit rate – and storage needed for recording, archiving – is indeed ten times more which is a huge sacrifice for anyone using the 1Dc. Of course the noise is different too but I’ll leave you to judge that one.
The other little test was done to compare how each camera can handle the gradient in the background – which doesn’t hold too much detail, so compression algorithms will compress it more than the rest of the image. Of course it isn’t a fair test, because the C300 has a base sensitivity of ISO 850 – so noise level will be higher on the HDSLRs – but it’s still interesting.
So we have the 5DmkII at the left, the C300 in the middle, and the 1Dc at the right.
Again the same thing can be seen, even though both the 1Dc, and the C300 were set to record with the C-log gamma, the 1Dc has a flatter picture. This is of course better: more information in the shadows and in the highlights.
If we look at the blue channel only picture, we can spot the different compression schemes easily. The h264 compression tends to smoothen these gradients by creating larger pixel blocks. The mpeg-2 compression tries to maintain sharpness, but still looks for places where it can simplify the image. The Motion JPEG compression looks very clean to me, even though the ISO was set to 800 on the DSLRs so we should see some noise along the compression artifacts. What needed to be mentioned is that all three cameras are 8 bit, and I’m sure if we would pair the 1Dc’s compression with a 10bit bit depth, this gradient would be even smoother.
So what’s the conclusion, or as Ali G would ask ”Is it good, or is it whack?”
I didn’t wanted to emphasise this before, but the 1Dc has a steep price, currently it sells for $12.000 at B&H. The 1D X sells for $6800. The difference between these bodies is that the 1Dc has over clocked processors, because of this it needed larger heat sinks, but these wouldn’t cost $5200. So in my opinion, the reason for the price difference is that little red “C” logo at the front. For the “Cinema” features Canon charges as much as they would like to, and given this logic, the camera could have been priced at $16.000 as the C300. But there’s one understandable reason behind the price, and it’s the ability to record 4K onboard to CF cards in a small body like this.
Of course drawback of this price is that if Canon would have set it lower, it would be easier to overlook the shortcomings of this body. With this price we expect the same features as other cameras have in this price range, working the same as they do, creating the same images they do, being just as flexible as they are.
I’m sure it needs a few years before 4K will be the new 1080p, but it’s clearly coming. If you ask yourself: ”I work, and finish in 1080p, would I benefit from 4K?”, the answer is yes. Recording 4K, and then downscaling to 1080p results a more detailed 1080p image, and an image with less noise. Of course there are downsides of this frame size, one of them is the storage space needed. Let’s say you shoot 1TB worth of footage with the C300. You would need 10TB to store the same amount of footage if you use the 1Dc. And of course it will make your post production a bit more pain. On a dual-processor Mac Pro I wasn’t able to playback the footage coming out of this camera, so I had to always render them before watching them real-time – of course if software companies would accelerate the decoding of this type of compression, everything would be a tad better. Also you will need pricier CF cards than what you had for your 5D: for being able to record 4K continuously, here’s the approved CF card list.
One thing I would add is the possibility to record a QuadHD image size. This would be a 3840×2160 sized picture, which would make life easier for people who would use the footage in 1080p at the end of the pipeline. It has the same aspect ratio (16:9) as 1080p, and because it is exactly twice as high, and twice as wide, downscaling would be faster, and better.
Joking aside, I praised the noise level of the C300, but this camera is…I can’t say the word “better”….but I see less noise. And a sharper image at high ISO. Without color shifts above ISO 5000. We should have done more comparisons to be able to state something clearly, maybe next time…
When I first put up the image on a plasma screen, the first thing that came to my mind is that this image is clean. Don’t get me wrong, but today we watch more and more videos on vimeo, youtube, and this we watch a lot of compressed formats – including DSLR footage. Our eyes adjust to these, and after a few videos we won’t even notice the artifacts, the blocking, the noise, we just ignore them. But when you open a better footage – Alexa, Epic, of film footage – the first thing that’s different for me is that their picture is clean. Clean from blocks jumping around, clean from broken gradients, it is simply a picture. The 1Dc has this same cleanness which made me happier than anything else about this camera. I’m still hoping that someday Canon will license the Prores codec for it’s Cinema cameras, but I don’t think that’ll happen soon. Until that, for me this Motion-JPEG compression seems to be the best compression format in their camera-line currently.
One question remains unanswered: if they put this new compression in the 1Dc, why didn’t they use the same codec for the 1080p modes?
Things I miss:
- the already mentioned QuadHD resolution with Motion JPEG compression
- look around/surround view: given that the sensor is a bit bigger than what it uses in 4K recording, it would be possible to have a look around, just like the Alexa, and REDs have this option
- fully customisable display options: For example to be able to separately turn on/off the View Assist function on the display, and the HDMI out feed. Also, the zoom-in function in the Live View shouldn’t turn off the View Assist function. The other request would be to have a display mode, where no info is displayed (no focus box, no settings), but the red dot is still visible – even on the external display.
- Sensor crop mode: to have smaller resolution recording options without line skipping, for example 3K, 2K, 1080p all with the same, Motion-JPEG compression. This would crop in the sensor, turning lenses into tele lenses.
- Bit depth: 10bit recording, if possible.
You can download this whole video in 2K Prores here
All footage were shot with the C-log gamma, and left untouched
You can watch the same video graded:
Koscsó Ferenc – PennaMédia
Lató Péter – Canon Hungária
Sztarenki Dóra, Dékány Barna, Dévény Zoli, Nagy Marcell, Damokos Attila
Marosi Gábor, Bíró Attila, Bíró András
Phil Harvey for the exiftool
LUTBuddy for applying the Clog-REC709 LUT
Clog-REC709 LUTs converted by Abelcine
contact me at hello at carousel.hu or the via the form below: