Jump to content

Canon 1D C vs Sony A7S 4K - dynamic range - preview


Andrew Reid
 Share

Recommended Posts

EOSHD Pro Color 5 for Sony cameras EOSHD Z LOG for Nikon CamerasEOSHD C-LOG and Film Profiles for All Canon DSLRs

Challenge me with the facts and I will gladly accept it!!

Wait a second, is this a trap? How does that saying go?

"Never argue with an idiot. They'll drag you down to their level and beat you with experience".

http://www.goodreads.com/quotes/tag/idiots

It's going to be major turmoil around here if Canon releases a 5D4 with all the features you want! ;)

Link to comment
Share on other sites

  • Administrators

jcs I really don't understand your point. Maybe you are indeed arguing with an idiot :)

I can show you stuff... present my findings in black and white... you can deny the evidence and have it your way. It doesn't matter to me.

If Canon releases a 5D4 with all the features I want I will be very happy! why wouldn't I!? Turmoil on the forums? eh!? There would be a lot of excitement for it yes.

Link to comment
Share on other sites

Andrew- we presented evidence and test results from around the internet: it appears you didn't even look at the results and dismissed our posts out of hand. Even Adam Wilt only measured 11 usable stops for the 1DC. Most forums seem to agree that the A7S is really only 12 usable stops. Canon rates the 1DC at 12 stops. Sony rates the A7S at 15. Clearly Canon is being more honest and Sony is exaggerating, however in this case in the real world (vs just lab tests where the A7S does much better), the A7S is better than the 1DC in DR (mostly in the shadows).

While I am a Canon fan, I have nothing to prove. The A7S is a challenging camera, however it's a useful tool which can be improved with the right techniques. I worked with your examples to see if your claim that the 1DC provides better DR than the A7S out of curiosity. Debate, when kept civil, is fun and entertaining. If the other party isn't responding in a productive way, progress isn't made. If we post images and links to other sites corroborating our test results, and your reply simply implies that we're blind, there's nothing else to say (other than to stop participating or to join in the absurdity with jokes).

Why not pick up a Xyla 21 test chart and do these tests yourself if you don't agree with them- it's part of the scientific method:

http://***URL not allowed***/dynamic-range-sony-a7s-vs-arri-amira-canon-c300-5d-mark-iii-1dc-panasonic-gh4/

 

 

 

Link to comment
Share on other sites

  • Administrators

Those 14 stops are coming from the sensor, then they are compressed into the LOG image in 8bit. That loses you at least 2-3 stops of USABLE dynamic range OR fucks colour up, depending on how you grade it. To even SEE those 14 stops you have to leave the image flat and measure the chart off that like C5D have done. Who uses that as their final look?! SLOG is there to be graded and those blacks crushed back! If you don't you have horrible ugly contrast and colour and noise in the shadows, not worth it to say you've achieved 14 stops really is it!?

1D C is 12 stops and almost all of it is usable in your grade.

If A7S is 14.1 stops like an Arri Amira or Alexa like C5D claim with their flawed testing then I am santa claus.

You might be able to get 14 stops out of a test chart with it but 14 stops of usable colour and detail after grading it? NO WAY.

If anything it has been scientifically proven in other tests that 14 stops in 8bit LOG is impossible anyway!

The bottom and top of the file on the Amira will be golden, lovely colour & accuracy, zero noise. The bottom and top two stops on the A7S are a total mess by comparison but because of C5D's Sony bias they don't mention that.

It's the quality of the dynamic range that matters not the fact that SLOG has such a flat image it can see some grey shades in the shadows of a test chart!

I would measure the dynamic range of the FINAL graded image and then assess the way it looks artistically.

Any NX1 owners around? Try putting contrast -10. You get more dynamic range but it looks shit. Happy with that? No? Thought not :)

Link to comment
Share on other sites

The A7S's HDMI 4K doesn't appear to have the same processing as internal XAVC-S. Slog2 must be tweaked (but can be fixed apparently), and noise reduction must be applied in post (perhaps the A7S hardware can't handle NR at 4K). The 1DC isn't full frame in 4K- it's 1.3x crop- the A7S is full frame. The 1DC files are huge and apparently require transcoding before editing. The A7S files are tiny and don't need transcoding. Different cameras, different uses and price points.

Your readers might appreciate a Xyla-21 chart test, which would be useful for all camera tests moving forward.

Artistically, the difference in DR between the A7S and 1DC isn't why one would chose one camera over the other.

In my recent tests I have found the output of the latest version of mlrawviewer to ProRes444 10-bit to look amazingly good. It has a really nice browser and can process clips reasonably easily directly from the CF card after quickly setting WB and in/out points (if desired). Comparing 5D3 RAW from mlrawviewer ProRes444 10-bit to 1DC would be an interesting test in terms of usable DR and color fidelity (full frame, S35 mode, and 4K downscaled). Given that the 5D3 has about the same (better?) DR vs. the 1DC and has Canon colors at 10-bit 444 vs. the 1DC's MJPEG-compressed 422 8-bit, I would expect the 5D3 to perform better in post.

Link to comment
Share on other sites

  • Administrators

The A7S's HDMI 4K doesn't appear to have the same processing as internal XAVC-S.

Slog2 must be tweaked (but can be fixed apparently), and noise reduction must be applied in post (perhaps the A7S hardware can't handle NR at 4K).

Noise reduction in post doesn't give you extra dynamic range JCS or make colour any better, you still need to adjust the curve to put contrast back into the image, thus trading the dynamic range of the flat LOG profile for acceptable colour and contrast.

It makes ZERO sense to count the number of stops in a LOG file from a well lit chart. What matters is the quality of colour, tonality, how much detail is in there. When detail is smudged away by noise in the lows, noise reduction isn't going to bring it back.

The 1DC isn't full frame in 4K- it's 1.3x crop- the A7S is full frame.

I am aware of that :)

The 1DC files are huge and apparently require transcoding before editing. The A7S files are tiny and don't need transcoding.

The A7S 4K files aren't tiny at all. ProRes LT is around 350-400Mbit/s compared to 500Mbit/s on the 1D C.

The 1080p internal is 50Mbit/s so of course that is going to be smaller isn't it!?

You're readers might appreciate a Xyla-21 chart test, which would be useful for all camera tests moving forward.

Thanks for the tip assistant editor!

Artistically, the difference in DR between the A7S and 1DC isn't why one would chose one camera over the other.

Haven't you just counteracted your own argument with that then?!

Artistically the 1D C looks better and if you can't see that, then you've got, shall we say, "interesting" tastes!

In my recent tests I have found the output of the latest version of mlrawviewer to ProRes444 10-bit to look amazingly good. It has a really nice browser and can process clips reasonably easily directly from the CF card after quickly setting WB and in/out points (if desired). Comparing 5D3 RAW from mlrawviewer ProRes444 10-bit to 1DC would be an interesting test in terms of usable DR and color fidelity (full frame, S35 mode, and FF downscaled). Given that the 5D3 has about the same (better?) DR vs. the 1DC and has Canon colors at 10-bit 444 vs. the 1DC's MJPEG-compressed 422 8-bit, I would expect the 5D3 to perform better in post.

​Already compared 5D 3 raw dynamic range to 1D C briefly but since this was a year and a half ago, it's a bit hazy in my mind.

The 1D C won but can't remember by how much.

The 5D 3 shadows in raw get a bit noisy and the highlights clip quite suddenly.

They are not a million miles apart though.

The advantage of the 1D C is that it works properly, 100% of the time, with smaller file sizes and 4K resolution.

Link to comment
Share on other sites

The A7S files indeed aren't tiny at 4K, however they are smaller and don't require transcoding as with the 1DC. The HD files are tiny, and that's also where Cinema5D likely made their test. Have you compared internal Slog2 to external for DR for the A7S?

If my prior posts haven't been clear, I prefer Canon due to superior color and skin tones vs. Sony. However, this has not biased my opinion and in all fairness, in the tests presented so far and evidence from around the internet, the A7S has more usable DR than the 1DC. Happy to agree to disagree until there's test charts or visible evidence to the contrary.

When I tested 5D3 RAW DNGs in PPro against AE+ACR against mlrawviewer's ProRes444 10-bit files, I was surprised how clean the output was and how good the color looked with mlrawviewer vs. the other two methods.

If 4K is the goal, the GH4 is quite the deal with 100Mbit/s files, usable DR, usable lowlight, and more detail vs. the 1DC. If I had the choice for a production, I might chose the 1DC, however the GH4 is more relevant for most consumers.

Link to comment
Share on other sites

  • Administrators

GH4 more detailed than the 1D C? It depends where you look! Not in the blacks it isn't!

In the end, 4K is 4K if 1:1 sampled. The only thing that can change your perception of sharpness is the micro contrast and digital sharpening applied in-camera. The 1D C Canon LOG has very low micro contrast and virtually no sharpening. Apply that in post and it will match the GH4 for detail. Or you can just shoot with a normal Canon profile like Portrait with sharpness not turned all the way down.

Personally I prefer the softer look for 4K. It's enough res already.

Link to comment
Share on other sites

While I agree a softer 4K (to a point) can look more filmic, it's better to start with a detailed image, then have the option to run a filter or blur in post. The 1DC 4K footage looks a bit soft from what I've seen online so far. The GH4 looks more detailed than the 5K Red Epic in this test. The difference is so high that it doesn't look like any post sharpening will get it looking like the GH4. High amounts of sharpening will produce artifacts, hurting the filmic look. If you have examples showing the 1DC out-detailing the GH4, it would be insightful to see.

In your quick test it looks like you got the same results for DR on 5D3 RAW vs 1DC as others have noted (e.g. 12 vs 11, 1-stop more for the 5D3).

http://www.eoshd.com/2013/05/canon-1d-c-vs-5d-mark-iii-raw-and-c300-gh2-resolution-comparison/

This example is not a proper dynamic range test as we didn’t have much time for the actual shoot setup, but upon spending an hour experimenting and grading the DNG to match Canon LOG I felt dynamic range was about 1 stop more on the 5D Mark III.

Raw also has a much more film-like gradation in the shadows and with skin tones than 4K MJPEG on the 1D C, without the odd colours and banding the 1D C throws up.

At ISO 800 the blacks are nice and clean in Canon LOG on the 1D C, because the camera applies noise reduction to the raw sensor output.

I did the same to the 5D Mark III raw DNG to see how close I could get it to the 1D C and indeed the black levels can compete…

This is an extreme push of the blackest BMW blacks, bearing in mind the camera was exposed for the sky! Magic Lantern offers raw zebras and these were extremely accurate. There was more noise in the 5D Mark III blacks but the noise reduction got rid of it without reducing resolution.

Link to comment
Share on other sites

Noise reduction in post doesn't give you extra dynamic range JCS or make colour any better, you still need to adjust the curve to put contrast back into the image, thus trading the dynamic range of the flat LOG profile for acceptable colour and contrast.

To some extent this addresses the core issue that is missing in this discussion. The deep shadows in the 1DC images are featureless, lacks texture and details. These are all tell-tale signs of noise reduction. Most if not all cameras do some form of noise reduction when recording internally even if noise reduction is set to zero. I will be extremely surprised if the 1DC is an exception. On the other hand the Shogun records footage from the A7s without any internal noise reduction which means that noise reduction have to be done in pp for a proper comparison. Alternatively direct recording from the 1DC sensor is necessary if you want to omit any pp in the comparison.

Link to comment
Share on other sites

​It's the other way round on the clipping. Just look at the images I've put up from the original files.

Don't try to grade the web compressed JPEGs!

As an 8-bit originating source it doesn't make much of a difference in determining things as long as it hasn't been fiddled with prior, it would just introduce potential artefacts, not clip DR or fiddle too much, if you're that worried about it, put up uncompressed 8-bit png.

 

In the samples you have provided, in the log images, the one labelled as the Canon camera has the clipping, not the A7s. In the corner above the book in the log image sample, the Canon example is clipping, the A7s is not. The Canon sample appears brighter overall in the image, but this is just a case of the black point being essentially at a higher value - the blacks are simply lighter, the A7s sample does that too, just to a lower extent than the Canon sample, neither exhibit shadow clipping in the log sample image, but the Canon exhibits highlight clipping in the sample you have given.

 

In regards to "you dont get x stops after grading!" you can if you want, it's a decision to decide the look of the image, but not an automatic given, 14 stops is low contrast, you generally have a reasonably punchy image in most things, which require a black and white clipping point, unless you want funky curves, masking or other things like that etc.

 

Link to comment
Share on other sites

A lot of water under the bridge, so I have to summarise. I used to have a 5D mk3 and now own a 1Dc.

It makes ZERO sense to count the number of stops in a LOG file from a well lit chart. What matters is the quality of colour, tonality, how much detail is in there. When detail is smudged away by noise in the lows, noise reduction isn't going to bring it back.​

Actually, measuring a chart in log is the ONLY way it makes sense. I agree that the result will never be conclusive since DR stops might show up in the chart that eventually proves to be unusable: color pollution in the highlights and excessive noise in the shadows. There is no common standard and "image  overclockers" like Sony and RED have their issues. It's the price they pay by going for a marketable spec that looks good on paper. Some manufacturers lack the integrity and discretion, since their customer base don't ask for it.

But even worse than this is a single individual that subjectively grades the files until they fit the agenda of the day.

Comparing 5D3 RAW from mlrawviewer ProRes444 10-bit to 1DC would be an interesting test in terms of usable DR and color fidelity (full frame, S35 mode, and 4K downscaled). Given that the 5D3 has about the same (better?) DR vs. the 1DC and has Canon colors at 10-bit 444 vs. the 1DC's MJPEG-compressed 422 8-bit, I would expect the 5D3 to perform better in post.

I know for a fact that the 1Dx/1Dc (and probably 6D as well) has a higher dynamic range than the 5D mk3—and I do mean in raw DR, comparing photos in Lightroom. The 1Dx/1Dc have technical specs that support the idea (larger photo sites, more advanced processing), but the proof is really in the pudding. The 5D mk3 still suffers from noise and banding in the shadows, something that has been plaguing Canon for many years. The 1Dx/1Dc and 6D sensors (possibly the 7D mk2) are different. I can lift shadows on my 1Dc in Lightroom by 4 stops without encountering the famous Canon banding. It is the distinguishing feature difference on the stills side.

The 12 stops in Canon log (I think it's fair to say 12 if you are going to compare to other, significantly more optimistic numbers from other manufacturers) are what were left over after Canon discarded what they didn't feel lived up to their design standards. It isn't the highest number they could have come up with. The whole Canon Log image is usable without color artefacts.

 

If anything it has been scientifically proven in other tests that 14 stops in 8bit LOG is impossible anyway!

I agree that going after more than 12 stops in 8 bit seems like over reaching. 10 bits? Well, that's something else…

 

While I agree a softer 4K (to a point) can look more filmic, it's better to start with a detailed image, then have the option to run a filter or blur in post. The 1DC 4K footage looks a bit soft from what I've seen online so far.

​I am very happy with the level of detail I get out of the 1Dc. It's of little point discussing this, since I think it's largely up to individual taste. I would say this: as long as the image processing doesn't add artificial detail by haloing, and the detail is strictly optical, I'll take as much as possible. I think this 1Dc 4k screen grab has enough detail for my taste (click to open):

Bern-train_SH.jpg

 

Link to comment
Share on other sites

 

As an 8-bit originating source it doesn't make much of a difference in determining things as long as it hasn't been fiddled with prior, it would just introduce potential artefacts, not clip DR or fiddle too much, if you're that worried about it, put up uncompressed 8-bit png.

 

Well that depends on how the file was saved.  The gamma could of changed if a rec.709 (or Bt.601, same luma different chroma) image was saved in an Srgb color space (or vice versa). Now admittedly, it would be a slight change, but still noticeably visible. (test with a b/w gradient in rec.709 toggled between color spaces) Of course both the 1Dc and A7s files were probably saved using the same procedure, but the s and c  log would each respond differently to the gamma curve of the new color space.  If this were an issue,  one can still adjust each image to get an idea of dynamic range, but not one that's entirely accurate. 

The accuracy of one's display device and color management is also crucial when making an objective comparison, The difference between what Andrew and others are seeing could be somewhat drastic.   Based on the knowledge of Alexis Van Hurkman, Allan Teppur and Charles Poynton, I wouldn't use any monitor  without hardware calibration for objective comparison purposes.   A great article that briefly explains.  http://diglloyd.com/articles/Recommended/display-calibration.html  

 

 

Link to comment
Share on other sites

  • Administrators

Nobody complains about dynamic range or image quality off a 1D X JPEG. Certainly not the professional photojournalists, action photographers at the highest level of their sports and every other 1D X user for that matter.

Well Canon LOG does wonders for the highlights in comparison... same lens, same exposure (1/50, F2.8, ISO 1600)

1D C JPEG

1D C 4K video with Canon LOG

Link to comment
Share on other sites

Nobody complains about dynamic range or image quality off a 1D X JPEG.

​Lot's of photographers can make it work and still plenty of people do complain about it. Or rather, they even complain about the DR of 1Dx raw.

And there isn't such a thing as a "1Dx jpeg". If you compare a PP 'neutral' with contrast all the way down to a PP 'EOS standard' with default contrast, the difference might be even bigger than what you show here. What settings did you use for your jpeg?

What people need to understand is that there is a tug-of-war between DR and tonal reproduction. Especially in 8 bit. I can somewhat understand the greed for DR when you shoot raw. But when we shoot to a compressed video format we need to consider the "big picture".

Canon modeled Canon Log after Kodak's Cineon system from 1993. It's seems like a good compromise to me, and to base it off of ISO400 is reasonable. Putting base ISO at 2500-3200 seems like over reaching to me. Fans of Sony's implementation love to talk about 14 (I even hear 15) stops, but I rarely see any discussions about possible drawbacks.

Nice 4k frame grab by the way.

Link to comment
Share on other sites

Higher ISO = more noise. Is it possible the reason they are forcing higher ISO for Slog is to create noise for dithering to prevent banding? I've been noticing a lot of banding in my indoor low light tests with the A7S when not using Slog2- basically noise free and clear banding in mid-dark regions. Next tests will boost ISO to see if the noise will help reduce banding.

Agree re: tradeoffs for Slog- skin tone is much more important to me than highlight and shadow behavior (though I'll take it all when available ;) ).

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

  • EOSHD Pro Color 5 for All Sony cameras
    EOSHD C-LOG and Film Profiles for All Canon DSLRs
    EOSHD Dynamic Range Enhancer for H.264/H.265
×
×
  • Create New...