First straight out of camera Canon EOS 5D Mark III footage *UPDATE – pulled

by planetmitch34 Comments

UPDATE: The video has been taken down with no explanation.


I first saw this Canon EOS 5D Mark III straight out of camera footage on NoFilmSchool and am sharing a bit of their post:

It seems we’ve now been treated to the first Canon 5D Mark III raw footage directly from the camera. EOSHD was the first to find this footage posted by Tõnis Liivamägi. Keep in mind that this footage is from a pre-production camera and is not necessarily indicative of final results, but at least we can start making some real conclusions about what this camera can and cannot do over the 5D Mark II.

Note, I'm opting to not use the phrase “raw footage” just to make sure people don't get confused. This is straight out of the camera and “raw” means just that… not in the sense of stills RAW.

RAW 5D MKIII low light footage for download (iso 3600)



Footage description

Make sure you view this with QuickTime player to witness the actual quality.
This is as random of a low-light, jello, moire and aliasing test as I could set up in few hours, texture wise at least :)..
Pure raw touch-and-go footage directly from a $30 16G low end CF card.
… hopefully no-one will be in such of a(n un)real ISO3600 situation with all those patterns around.
Notice how 3600 ISO kind of “bands” in the direction of camera movement when the worst of moire would have occured on the stapled grid screen.
This was shot at @ 50th of a second under pulsing fluorescent lights.
The in camera settings were set to “factory defaults” so there might be room for improvements.
Remember: this was shot hand held with 85mm lens to pronounce camera movement.

NofilmSchool also taught me something I wasn't aware of – you can see the details about the video file and even the bitrate via this tool…

I loaded this footage into MediaInfo and was given stats about the raw file

you can see the results on the noFilmSchool page

So, what do you think?

Are you seeing what you wanted to see? The footage isn't very “exciting” but it does give us a chance to see some early details (and remember this is pre-release) and can evaluate moire and rolling shutter.

(cover photo credit: snap from the video)

Comments

  1. This is probably the most underwhelming fist footage I’ve ever seen. Can the people holding the camera actually use one?

  2. Think you are missing the point with this vid—
    It is clearly defined as a test for the typical problems that plagued the 5d II – moire and aliasing.

    Not saying there couldn’t have been a better test–but for me it is what I’m looking for — and I’ll be looking for more tests regarding moire, aliasing and the jello effect–so please keep them coming!

    It’s that reduction that would make me want to buy the III–but still on the fence until I see more reviews and more results which I have no doubt will start flowing!!

    And it definitely seems to show a great reduction in moire and aliasing from what I saw.IMHO

  3. moire is well controlled but jello is still there
    I got headache anyway

  4. Yeah, I have to agree, this is very underwhelming.

    The moire and rolling shutter being reduced helps. But, 1080p should look like actual 1080p already and not a soft 720p. Seriously, why is the GH2 still resolving more lines of resolution than this camera at a fraction of the cost?

    It’s about that time now and I can’t be a Canon fanboy in favor of this either. There’s no reason whatsoever why it still looks outdated and at $3500 for what smaller HD cams do for a fraction of the cost. Better dynamic range (still 9-10stops?) or not.

    I hope this isn’t the final reality of the camera, personally.

    1. It all comes down to the sensor size. Here me out, I’m not gonna argue asthetics of full frame vs 4/3’s, ect.

      Let’s just talk scan lines. The 5dmkII/III is 864mm2 vs the GH2 residing at 199.5 mm2, or 4.33 times larger.

      Now, roughly speaking, you need a processor 4.33 times as fast since you have 4.33 times the area to decode. It comes down to how fast the Digic 5 really is.

      Let’s be honest hear: full s35 sensor read outs like with RED require HUGE data processing and heat sinks. I just don’t think the Digic 5 is capable of a full sensor scan line output for video.

      1. In terms of how much processing power is needed, it’s not the sensor size (which determines how much light can be captured), but rather the number of pixels.

        The GH2’s sensor is 18 MP while the 5D’s is 22 MP. True, the 5d’s gotta push some more pixels, but I doubt the extra 22%’s the killer issue.

        I’m guessing the bigger issue affecting quality is the algorithm’s they’re using to resize the image down to 1080 (or 720) plus any algorithms they’re using to reduce noise, reduce jello, etc. It’s certainly possible Canon’s algorithm’s just aren’t as good at maintaining a sharp image as the GH2’s.

        Again though, would really need to see a side by side comparison & know what the settings (such as “Sharpness”) are in the cameras for those tests.

  5. I’m with Ivan on that… are you sure we haven’t just witnessed someone finding the Mark-iii, not quite sure what it is and accidentally switching it on while discussing last night’s football game…?

  6. Certainly not the best test footage, but the moire improvement is huge.

    And when he does have it in focus (or if you manage to find the areas of the picture that are in focus), it seems relatively sharp.

    Coupled with 2 stops of light sensitivity improvement & the noise reduction, and the Mark III seems like a very significant overall improvement to me.

    That said, products usually are improved over time & remain at a similar price point. The extra $800 (comparing new-original release prices) to $1500 (compared to current 5d price) is a tough pill to swallow though…

  7. Yeah, clear evidence that moire has been kicked to the curb, and we’ve got dinks on here saying they are underwhelmed. Boo hoo. Grow up.

    Mostly due to Canon’s retarded marketing department, the 5DIII announcement follow-up is quickly producing a culture of chronic complainers, like Canon owes them something.

    Go to vimeo and see the radball vitriol. It’s so bad that Laforet got on there to tell people to stop whining.

    Hey soon-to-be D800 converts: Consider how happy you’ll NOT be when your line-skipping “clean hdmi out” footage still has moire (which is a FACT–see making of Joy Ride) and the Canon doesn’t.

    Canon has fixed the biggest issue with DSLR video. Is that not even worth a breath of praise? Yes, they are marketing/PR morons, but the 5D3 is going to be a winner.

    Cheer up.
    Grow up.
    Be patient.

    1. If I plan on spending another several thousand on a new DSLR, I’d most certainly be expecting it to out-resolve the sensor capabilities of a sub $1000 camera (GH2).

      Yes, moire is gone and there’s a new codec. Now is that worth investing $3500? Seriously? Be truthful about it.

      Everyone on here knows that I’m a huge Canon supporter with my body of work. But, it’s time for a REAL upgrade beyond that of 2008. Eliminating moire should be the tip of the iceberg, not the full gamut of its improvements.

      Again, a “1080p” claim when the sensor is only resolving much less is fraudulent and insulting when you consider the money needed to buy the hardware. And even worse when considering that it’s a 5k capable sensor that is CLAIMING to prevent binning this time around.

      Yet low and behold, other than moire (as rolling shutter seems to be just as bad), that’s about as far as the upgrades go for the video capabilities. Oh, and extreme low light ISO that would normally not be used to the extreme degree as advertised beyond a 10% time frame.

      So again, why upgrade from a ML modded 5D2?
      Headphone jack out of the box? Are you kidding me?

      1. Seriously? No actual tests have been done with this camera & you’re calling them “fraudulent” claims and saying it’s resolving “much less” than 1080p? Exactly what is this claim based on?
        Let’s take a breath & find out what the camera’s actually capable of before making such slanderous statements.

        The reality is that the moire elimination is a HUGE plus for people using the 5d.
        And for most people (at least us on almost all our shoots) the extra 2 stops of light sensitivity is HUGE. Almost all our shoots are indoor, and I can’t remember a shoot we’ve done where having that extra light sensitivity (or lower ISO, or smaller aperture, or faster shutter speed) wouldn’t have been a big benefit.
        We’re already ordering ours.
        Seriously.

        1. Updating you on the latest memo:
          vimeo.com/groups/eos/videos/38087210

          So yes, there HAVE been several tests already.
          This includes streams, “raw” files and downloads.

          Just so you know, the conclusions were drawn from tangible evidence. Not hearsay or “guesstimations”

          1. Where is there a test showing how many lines of resolving power the 5d Mark III has in its video?
            There have been a handful of video samples that people have shot, but that’s nothing like an actual scientific test showing resolving power.

  8. Additionally, the “2 stops” of light people are referring to appears to be confusion with actual stops of light per ISO and lowlight ISO boosts.

    For instance, claiming that ISO boosting of 10,000+ISO is an extra stop of some sort isn’t the right terminology for it. It seems people are using “2 stops” in conjunction with “high lowlight ISO” as one in the same.

    The Arri Alexa boasts 14stops for its sensor
    The Red Epic 13.5 stops
    The Canon C300 12 stops
    And the DSLRs (both Digic 4 and Digic 5) appear to remain in the 9-10 stop range.

    Hopefully I’m wrong about this specific point. But, I haven’t seen any conclusive tests that have measured how many stops the sensor can see in a single ISO. That includes highs, mids and lows, not just lows via 20,000 ISO.

    1. Yes, there’s been nothing showing how much “range” or stops can be displayed at a single setting, so I’m not sure if that’s been improved, but what I’m referring to is the amount of light it picks up at a specific setting.

      For instance, shots that would previously need to be shot at ISO 800 can now be shot at ISO 200. Or as mentioned with the iris closed down more, or using a 2x faster shutter speed.

      I do wish it had more “range” as well, but I haven’t heard anything about that either.

      1. So what you’re saying is that you want numerical charts that display how many lines the MKIII shows, but have zero of similar value for measuring DR stops on your end?

        Alright.

        While there aren’t NUMERICAL measurements [yet], there are in fact comparisons in lines resolved. Currently the GH2 is the king of the hill with resolution, as it’s the only DSLR that actually resolves 1000 lines (or at least close). The MKIII has been compared to it side by side. It doesn’t hold up, and looks almost exactly like the MKII did (sans moire).

        So now we have to wait and see how many stops the Digic-5 sensor actually sees. In a single ISO, not by boosting for the lows.

        1. Nowhere did I say they shouldn’t also be measuring the dynamic range. They definitely should.

          But what I’m saying is that your value for measuring the Mark III is completely flawed. You’re basing it on a handful of sample videos (using random assorted lenses, with random settings/focus, random/assorted materials, who knows how it was pre-processed/edited/exported/compressed/etc). There haven’t been any proper tests or comparisons.

          Have you see the Zacuto shootouts? Here’s the link for the 2011 one:
          www.zacuto.com/the-great-camera-shootout-2011

          THAT’s a proper test/comparison. Or even just having an actual resolution chart & showing footage of it to see how many lines of resolution are visible as well as dynamic range stops, etc. When something like those are done for the Mark III, we’ll have some proper answers. But basing it on a few sample videos is way premature.

          1. Agreed on that. We DO need a shootout test. But, it doesn’t even need to be as high budget as a Zacuto one as of yet. I’m waiting for Bloom’s where he compares each camera side by side. You know, the way EOSHD did w/ the MKIII footage and the GH2.

            He did it for the F3, C300, FS100, GH2 and MKII respectively. All of which revealed awesome results, even without the complete accurate science behind every nuance of each test.

            While the latter won’t have numbers, they’ll have the evidence right smack dab in your face. The Zacuto one will just add confirmed numerical factors for what you can already see with your own eyes.

  9. And the hair splitting over basically nothing continued into the night.

  10. I downloaded Tonis’ footage and added a sharpen of ~30 in PPro CS5 (please contact Adobe and ask them to add support for the 5DM3 for CS5.5.x: currently requires renaming files from MOV to MPG to work around a file loading bug/ CS5 does not have this problem).

    When sharpened, this footage looks excellent. That said, would like to see Canon’s published horizontal resolution spec (“lines of resolution”). The 5D2 was reported to have 700 lines (if done correctly, that’s horizontal pixels, not related to 1080p or 720p, but rather 1920 horizontal pixels). Decent cameras can resolve around 1/2 the pixel resolution. For good 1920 output cameras, they are rated at 1000+ lines of resolution (C300, F3, etc.). I would expect the 5D3 to be able to do the same with optimal filtering (current examples have sharpness turned down too far). Again, this footage looks pretty good once sharpened.

    1. I don’t mean to rain on your parade but, almost any footage sharpened can look good. I tend to do the same with my current footage out of the Canons now, so I understand where you’re coming from.

      The negative part about this is that we shouldn’t HAVE to sharpen the footage in order to realize proper resolution. It’s nice to do it, don’t get me wrong. But, it shouldn’t be necessary.

      At 1080p frame in 2012 should actually resolve 1000 lines by now. Especially if it runs $3500.

      1. From my analysis, it appears an optical low-pass filter is throwing away frequencies above 1920*3/2 (sensor native resolution) to capture an initial anti-aliased image (the /2 is required as that’s the maximum frequency possible per the Nyquist-Shannon sampling theorem). The target resolution is 1920, so any frequencies above 1920/2=960 must be removed before downsampling: the Digic 5 performs some form of low-pass filter before decimating down to 1920. When we speak of “lines of resolution”, that’s alternating black and white vertical lines: it means horizontal resolution, as opposed to vertical resolution, and is not related to 1080p or 720p, etc. The maximum alias-free resolution provided by 1920 pixels is 960 pixels, per Nyquist. Consider how many black lines we can have if we need to be able to count alternating black-white horizontal pixels. It’s 960. Camera manufacturers appear to count white lines as well, so under perfect conditions, they have 960+960 = 1920 lines. The worst case condition occurs when the pixel grid is shifted 90 degrees- after low pass filtering the sensor would see no lines at all: gray pixels. The C300 and F3 are rated at “1000+” lines of resolution (optical filters aren’t perfect): again, that’s horizontal pixel resolution.The GH2 appears to be only 700 lines of horizontal pixel resolution. The 5DM3 appears to be closer to 1000, as I can see clean horizontal anti-aliased pixel detail in Tonis’ footage. A proper test with a line chart will let us know the actual number: it will be much higher than the GH2. The GH2 footage may look sharper at this point, but that is only because it is aliased, which causes the perception of higher resolution.See the Cinema5D thread for a great deal more information about resolution and anti-aliasing.

  11. Do we have a bottom line on the 5D mIII yet, or is it still too early? Are there subtantial improvements in this camera to justify the cost? Improved low light performance? Reduced “jello” (rolling shutter) like the c300? Better autofocus? IS THIS CAMERA WORTH IT? I’m getting the impression this camera is like when Canon introduced the 30D–it was basically the 20D with a few minor improvements–

    1. As a still camera, I definitely think it’s justified the cost. 2-4 stops of low light improvement, a massively improved autofocus system (moving from something like 9 AF points to 60 something, with the majority of them cross-points), improved speed, dual memory slots, new options for bracketing/hdr, wireless flash sync improvements. Lots of stuff.

      For video, I personally think it’s worth it due to the moire/aliasing reduction and the improved low light performance. From the MKii, we’ve had shots that were completely unusable due to the moire, so for us, the MKiii is absolutely going to be worth it to ensure we don’t miss shots.
      Others though had hoped for more, don’t think the handful of sample videos out there so far are impressive (find them too soft), and felt the price shouldn’t have been increased.

      We ordered ours.

      1. Still photographers are just as upset as the video camp because the claimed 2 stops of low light is only achieved by noise reduction in their JPEG engine. The raw files show less than a stop of improvement. Overall, it seems that everyone is unhappy with the camera.

          1. Agreed, that was a generalization. There is, however, strong factions of both video and still users that are disappointed with the release as evidenced by numerous forums.

            Personally, for me, I’m eagerly awaiting the arrival of my Mark III as well. It will be my first Canon FF and I’ve been waiting 1.5 years for this announcement. For my use (enthusiast), I’ve always wanted the child of a 5D and 7D. The only source of my disappointment is the cost as I would have hoped for it to be lower.

            Apple has me trained to expect new models of products with new features to remain at a relatively constant price point and since the IQ appears to have remained mostly constant (from a still RAW point of view). I was expecting the cost of a FF sensor to have dropped in the last 3.5 years.

  12. Reading over comments again—

    Nothing has changed my thoughts—- as previously mentioned that if all the III did was to reduce aliasing and moire it would be enough—which was to me the biggest upset factor with the 5d II. (However–Really like the II for still photography–full frame sensor makes huge difference with wide angle shots)

    If I were going to go past that price range I think I would have to make the jump all the way up to the C300 or Scarlet–for another 16,000 to see a noticeable difference.

    And even then different tools for different applications!!

Leave a Comment