Sign in to follow this  
Followers 0

Low light test of 5D Mark III raw vs H.264

74 posts in this topic

Posted · Report post

[media]http://vimeo.com/66206596[/media]

The 5D Mark III is already a very capable low light camera. In its factory guise whilst not quite as clean as the Canon C300 or Sony FS100, it is the best DSLR for low light shooting (though the Nikon D5200 puts up a good fight).

But that was before the latest developments with ML Raw.

Has low light improved even further?

[url=http://www.eoshd.com/content/10338/low-light-test-of-5d-mark-iii-raw-vs-h-264]Read the full article here[/url]
Julian, tomekk and matt2491 like this

Share this post


Link to post
Share on other sites

Posted · Report post

I am new to video and would like to ask some questions over AVCHD and H.264
 
I have an impression that AVCHD is a "compressed TIF file similar to the still image while H.264 is just a non-compressed JPG. Is AVCHD good enough as a compromise comparing with RAW footage? Which means in quality and the tolerance of post processing, the order should be RAW > AVCHD > H.264. Am I right in this point of view?
 
It makes me curious about the footage of RAW comparing with Sony which can apply Cine Style for FS series while Canon provides Canon Log Gamma. I am looking forward that there will be a comparison in the near future.  :)

Share this post


Link to post
Share on other sites

Posted · Report post

I am not sure where you got that impression from, but it is broken get a refund :)

 

The way to think of it is:

 

Raw is top of the pyramid.

 

From that there are ways to compress and mangle the image and none of them are as good quality as raw.

 

JPEG and TIFF have nothing to do with AVCHD or H.264.

 

In the post world, TIFF is for sequences of very high quality stills.

 

AVCHD is a consumer video format with H.264 inside, at 24Mbit for 1080/24p.

 

Cine Style, log, etc. is just a way to get a bit more dynamic range out of a crappy compressed image, with raw you don't need log, it is N/A and if you just want a flat low contrast look then you apply it to the raw image in post.

Share this post


Link to post
Share on other sites

Posted · Report post

I'm sorry.  I didn't see it anywhere but maybe I missed it.  Are the shots with the lantern, lit by only the candle or is there another light source in the room at all?

Share this post


Link to post
Share on other sites

Posted · Report post

I'm sorry.  I didn't see it anywhere but maybe I missed it.  Are the shots with the lantern, lit by only the candle or is there another light source in the room at all?

 

One more dim light source for fill lighting around 2m away from the candle.

Share this post


Link to post
Share on other sites

Posted · Report post

Can you add one more benefit to your list at the bottom?

-Many of us, if not most, of us already own the 5D Mark iii and the entire system surrounding it (lenses, rigs, etc.)

 

A big thank you to the Magic Lantern team. 

Share this post


Link to post
Share on other sites

Posted · Report post

What's up with the magenta tint to all this RAW footage? The H264 versions seems to filter it out?

Share this post


Link to post
Share on other sites

Posted · Report post

Andrew, simple yet incredibly relevant point:

 

What picture style and associated settings are you using for the H.264 footage?

 

Because Cinestyle is a FREE download. And friends don't let friends shoot 5D3 internal with anything worse when DR matters. No "ProLost flat" or "FLAAAAAAAT" or any nonsense. If DR matters, then you use a high DR picture style and then grade in post with a LUT.

 

That looks like Neutral or Faithful with reduced contrast and saturation. Why are you crowing about the dynamic range difference, when you chose to burn it out with those settings?

 

Also, why aren't you using the pull-down ISO's (2500, 5000, 10,000) as we've all learned to by studying the 5D3 noise floor?

 

Show us that it beats the best we can do otherwise, fair and square, and that will be plenty sensational enough.

Share this post


Link to post
Share on other sites

Posted · Report post

The article says that 5d mark III can make 60 fps raw, is that for sure already? In the tests it says 24fps continous?

If it's 720p raw as it says it's still very nice. And what about 5d mark II? Will it be possible to boost fps up to 60 fps even though the native camera can't do that?

Share this post


Link to post
Share on other sites

Posted · Report post

does anybody know if it´s possible to build a connector from cf card slot to sata. i think about running a cable out of the cf slot from the mk3 to a ssd drive. then we could use ssd´s direct with the 5d which would be awesome. there are adaptors from cf card to data but i could not find the other way around.
maybe the powering of the ssd would be a problem. could also be that the power from the cf slot is enough. the power of an usb is enough...
 

Share this post


Link to post
Share on other sites

Posted · Report post

Andrew, simple yet incredibly relevant point:

 

What picture style and associated settings are you using for the H.264 footage?

 

Because Cinestyle is a FREE download. And friends don't let friends shoot 5D3 internal with anything worse when DR matters. No "ProLost flat" or "FLAAAAAAAT" or any nonsense. If DR matters, then you use a high DR picture style and then grade in post with a LUT.

 

That looks like Neutral or Faithful with reduced contrast and saturation. Why are you crowing about the dynamic range difference, when you chose to burn it out with those settings?

 

Also, why aren't you using the pull-down ISO's (2500, 5000, 10,000) as we've all learned to by studying the 5D3 noise floor?

 

Show us that it beats the best we can do otherwise, fair and square, and that will be plenty sensational enough.

1. There is no such thing as pull down ISOs once you are that high. Above ISO3200 everything is simply 1/3 more and more and more digitally pushed.

 

2. Cinestyle can be dicey to work with, that flat and 8 bits doesn't always grade so easily. Some of the others have even a touch more DR but keep some better pop in the mids.

 

3. The RAW gives truly amazing rich colors and all the advanced tools of Adobe ACR processing with ACR pseudo-HDR sliders, raw pre-sharpening, advanced NR, etc.

 

4. Maybe this is not quite scientifically carried out, but believe me Cinestyle doesn't process the way these RAWs do and you get somewhat less DR and a lot less beautiful tonality (and of course the detail is simply much less).

Share this post


Link to post
Share on other sites

Posted · Report post

What's up with the magenta tint to all this RAW footage? The H264 versions seems to filter it out?

A sensor issue. The image processing engine filters it out and counter acts it normally.

Share this post


Link to post
Share on other sites

Posted · Report post

1. There is no such thing as pull down ISOs once you are that high. Above ISO3200 everything is simply 1/3 more and more and more digitally pushed.

 

2. Cinestyle can be dicey to work with, that flat and 8 bits doesn't always grade so easily. Some of the others have even a touch more DR but keep some better pop in the mids.

 

3. The RAW gives truly amazing rich colors and all the advanced tools of Adobe ACR processing with ACR pseudo-HDR sliders, raw pre-sharpening, advanced NR, etc.

 

4. Maybe this is not quite scientifically carried out, but believe me Cinestyle doesn't process the way these RAWs do and you get somewhat less DR and a lot less beautiful tonality (and of course the detail is simply much less).

 

I have no problem working with Cinestyle on the 5D3, and with just a touch of green/magenta correction I can get it lined up fairly well with Canon Log (Cinelock) C100/Ninja 2 footage. ISO 10000 on the internal codec is significantly cleaner than 12800 in the same way 2500 and 5000 are...it may be in-camera NR that you can't turn off, or some other mechanism, but the pattern is consistent and I shoot that way accordingly. The 5D3 clean HDMI out to the Ninja 2 does not display the pull-down ISO pattern as strongly so you might be right in that case.

 

The A/B we are all waiting for is the absolute *best* you can do "officially" vs. the best you can do with the ML hack. I will test that once all the bugs are shaken out of the hack and we know what works best. The best I can get out of a 5D3 at this time officially is clean HDMI out to the Ninja 2 using ProRes 422 HQ on the Ninja, Cinestyle with default settings on the 5D3, and then applying NR (I use Neat Video) and a LUT (I use Pomfort) in post. I don't see banding after grading the Cinestyle in typical use. Even so, that image is blown away by the C100 in Cinelock to the Ninja 2 in every capacity (other than full-frame DoF and angle of view of course).

 

I have no delusions that the C100/Ninja 2 will be displaced by the 5D3 RAW output in absolute IQ. That sensor and system is just purpose-built and far superior for filmmaking. I think a lot of people overestimate the value of RAW, perhaps egged on by Jannard. For stills of course I only shoot RAW but that's because the only other choice (JPEG) sucks and RAW stills aren't onerous. Direct-to-ProRes HQ with a log gamma is an absolutely credible option and anyone skillful enough to get decent exposures from pre-Log DSLRs will have no trouble at all with that codec. The hassle with RAW is so intense in comparison to just capturing hours of ProRes on an SSD and loading it directly into your NLE ready to cut that it better completely blow it away even if the best possible settings and workflow is used with the compression.

 

The main losses for the HDMI 8 bit 422 are going to be chroma detail (422 and 444 are VERY similar to the eye, even for keying work..it's the 420 used in the internal codecs that truly sucks) and maybe a touch of DR. 8 bits with the Cinestyle gamma is not far from representing with high fidelity the full DR of that sensor on most ISOs in practice...see DxoMark's reports on all the read noise of the 5D3 sensor in stills mode. The main hit resolutionwise is going to be done at the downsampler in-camera taking the 22MP to 2MP. If you use a crop 1:1 pixel mode, you will get a ton of sensor noise and probably some aliasing/stair-stepping the OLPF doesn't fully blur out. We aren't recording 24 RAW 22MP stills in a second on this camera...there is damage done in both pathways.

 

What I would like as a working professional for whom any gains from RAW are likely overkill for a 5D3, is a nice internal codec from ML. If MJPEG is the best they can do, OK, make it a good one with 422 and an easy workflow in post. If that, plus the CF cards needed, is overall cheaper and better in practice than getting a second Ninja 2 I'm happy. Will be more compact to just use the Zacuto loupe and hopefully not require too much card swapping while shooting.

Share this post


Link to post
Share on other sites

Posted · Report post

I have no problem working with Cinestyle on the 5D3, and with just a touch of green/magenta correction I can get it lined up fairly well with Canon Log (Cinelock) C100/Ninja 2 footage. ISO 10000 on the internal codec is significantly cleaner than 12800 in the same way 2500 and 5000 are...it may be in-camera NR that you can't turn off, or some other mechanism, but the pattern is consistent and I shoot that way accordingly. The 5D3 clean HDMI out to the Ninja 2 does not display the pull-down ISO pattern as strongly so you might be right in that case.

 

The A/B we are all waiting for is the absolute *best* you can do "officially" vs. the best you can do with the ML hack. I will test that once all the bugs are shaken out of the hack and we know what works best. The best I can get out of a 5D3 at this time officially is clean HDMI out to the Ninja 2 using ProRes 422 HQ on the Ninja, Cinestyle with default settings on the 5D3, and then applying NR (I use Neat Video) and a LUT (I use Pomfort) in post. I don't see banding after grading the Cinestyle in typical use. Even so, that image is blown away by the C100 in Cinelock to the Ninja 2 in every capacity (other than full-frame DoF and angle of view of course).

 

I have no delusions that the C100/Ninja 2 will be displaced by the 5D3 RAW output in absolute IQ. That sensor and system is just purpose-built and far superior for filmmaking. I think a lot of people overestimate the value of RAW, perhaps egged on by Jannard. For stills of course I only shoot RAW but that's because the only other choice (JPEG) sucks and RAW stills aren't onerous. Direct-to-ProRes HQ with a log gamma is an absolutely credible option and anyone skillful enough to get decent exposures from pre-Log DSLRs will have no trouble at all with that codec. The hassle with RAW is so intense in comparison to just capturing hours of ProRes on an SSD and loading it directly into your NLE ready to cut that it better completely blow it away even if the best possible settings and workflow is used with the compression.

 

The main losses for the HDMI 8 bit 422 are going to be chroma detail (422 and 444 are VERY similar to the eye, even for keying work..it's the 420 used in the internal codecs that truly sucks) and maybe a touch of DR. 8 bits with the Cinestyle gamma is not far from representing with high fidelity the full DR of that sensor on most ISOs in practice...see DxoMark's reports on all the read noise of the 5D3 sensor in stills mode. The main hit resolutionwise is going to be done at the downsampler in-camera taking the 22MP to 2MP. If you use a crop 1:1 pixel mode, you will get a ton of sensor noise and probably some aliasing/stair-stepping the OLPF doesn't fully blur out. We aren't recording 24 RAW 22MP stills in a second on this camera...there is damage done in both pathways.

 

What I would like as a working professional for whom any gains from RAW are likely overkill for a 5D3, is a nice internal codec from ML. If MJPEG is the best they can do, OK, make it a good one with 422 and an easy workflow in post. If that, plus the CF cards needed, is overall cheaper and better in practice than getting a second Ninja 2 I'm happy. Will be more compact to just use the Zacuto loupe and hopefully not require too much card swapping while shooting.

i really don´t know what you are talking about. ML is able to record RAW to CF cards. they are not able to process something else to the card. i would also like to see a mk3 doing really good 422 proress on a cf, for better workflow. and i would also like the sun to shine the whole year, but this will not happen. what has happened is that ML IS ABLE TO RECORD RAW WITH THE 5D MK3

Share this post


Link to post
Share on other sites

Posted · Report post

I have no problem working with Cinestyle on the 5D3

 

 

Well I have. These are my problems:

 

Cinestyle raises the blacks from 0 to 16. This is to match Cineon film scanners. It decreases dynamic range, mostly from the midtones as the highlights and shadows are compressed. Good highlight rolloff is gone. Cinestyle clips them ugly. You have to create them back with a gamma curve. It increases the noise in the blacks and when you pull those blacks down in post, the noise is there. And it doesn't really help in grading at all plus way more difficult to assess exposure. If you want milky blacks, just raise them afterwards.

 

For example the Flaat10p profile is lots better and also Prolost Neutral.

 

You personally don't see banding but that's because you use Neatvideo and it creates very good 16-bit files out of 8-bits. Neatvideo actually fills in the gaps. It's actually quite remarkable how much banding it can take away. But it will do that to any profile, not just cinestyle.

 

Those intermediate ISOs are cleaner in h.264 mode because they are digitally pulled down. This pull down only decreases DR in RAW files but because h.264 loses so many stops anyway, it doesn't really matter there. So basically, when shooting stills, try to go with the original ISOs and when doing h.264, drop down once to get the shadow noise gone and more for the highlights.

Share this post


Link to post
Share on other sites

Posted · Report post

I have no problem working with Cinestyle on the 5D3, and with just a touch of green/magenta correction I can get it lined up fairly well with Canon Log (Cinelock) C100/Ninja 2 footage. ISO 10000 on the internal codec is significantly cleaner than 12800 in the same way 2500 and 5000 are...it may be in-camera NR that you can't turn off, or some other mechanism, but the pattern is consistent and I shoot that way accordingly. The 5D3 clean HDMI out to the Ninja 2 does not display the pull-down ISO pattern as strongly so you might be right in that case.

 

The A/B we are all waiting for is the absolute *best* you can do "officially" vs. the best you can do with the ML hack. I will test that once all the bugs are shaken out of the hack and we know what works best. The best I can get out of a 5D3 at this time officially is clean HDMI out to the Ninja 2 using ProRes 422 HQ on the Ninja, Cinestyle with default settings on the 5D3, and then applying NR (I use Neat Video) and a LUT (I use Pomfort) in post. I don't see banding after grading the Cinestyle in typical use. Even so, that image is blown away by the C100 in Cinelock to the Ninja 2 in every capacity (other than full-frame DoF and angle of view of course).

 

I have no delusions that the C100/Ninja 2 will be displaced by the 5D3 RAW output in absolute IQ. That sensor and system is just purpose-built and far superior for filmmaking. I think a lot of people overestimate the value of RAW, perhaps egged on by Jannard. For stills of course I only shoot RAW but that's because the only other choice (JPEG) sucks and RAW stills aren't onerous. Direct-to-ProRes HQ with a log gamma is an absolutely credible option and anyone skillful enough to get decent exposures from pre-Log DSLRs will have no trouble at all with that codec. The hassle with RAW is so intense in comparison to just capturing hours of ProRes on an SSD and loading it directly into your NLE ready to cut that it better completely blow it away even if the best possible settings and workflow is used with the compression.

 

The main losses for the HDMI 8 bit 422 are going to be chroma detail (422 and 444 are VERY similar to the eye, even for keying work..it's the 420 used in the internal codecs that truly sucks) and maybe a touch of DR. 8 bits with the Cinestyle gamma is not far from representing with high fidelity the full DR of that sensor on most ISOs in practice...see DxoMark's reports on all the read noise of the 5D3 sensor in stills mode. The main hit resolutionwise is going to be done at the downsampler in-camera taking the 22MP to 2MP. If you use a crop 1:1 pixel mode, you will get a ton of sensor noise and probably some aliasing/stair-stepping the OLPF doesn't fully blur out. We aren't recording 24 RAW 22MP stills in a second on this camera...there is damage done in both pathways.

 

What I would like as a working professional for whom any gains from RAW are likely overkill for a 5D3, is a nice internal codec from ML. If MJPEG is the best they can do, OK, make it a good one with 422 and an easy workflow in post. If that, plus the CF cards needed, is overall cheaper and better in practice than getting a second Ninja 2 I'm happy. Will be more compact to just use the Zacuto loupe and hopefully not require too much card swapping while shooting.

All I can say is try the hack and you will see. It's a night and day difference for dynamic range, detail, color tonality, ease of color and tone processing (just think how power ACR is for stills and imagine using those sliders and all that power for video correction; just think that you can adjust WB in post without losing anything). At times it is almost laughably better.

 

The workflow is a tedious slow mess to get stuff converted true and it burns up CPU cycles like mad (although color correction and processing the steps where you do stuff are actually much easier and quicker thanks to the power of ACR) and yeah less than a minute and it cuts off at the 4GB limit. But man the quality!

 

I'm telling you if you think internal vs Ninja 2 is much of a difference then this ML RAW will blow your mind. The difference is like 1000x, literally, more noticeable.

Share this post


Link to post
Share on other sites

Posted · Report post

We're not even through the middle of 2013 and the camera scene has been turned upside down twice already this year.

2013 - the year of RAW wars!

Julian likes this

Share this post


Link to post
Share on other sites

Posted · Report post

co-sign with peederj.

 

first of all, I have no doubt that RAW is always going to look better than any other format, but the stills from H264 footage posted here look awful and are not an honest comparison.

 

one rule of thumb for shooting with DSLR's (and digital in general) is to expose for the highlights. 
all highlights are severly blown out in these stills, so this makes no sense at all.

 

furthermore, if you're going to compare a very flat still (which RAW is), you should compare with a flat video still. 

using a flat picture style such as Visioncolor, flaat, Cinestyle or even neutral with contrast and sharpness turned down.

 

even stills from a T2i which were exposed for the highlights and shot with a flat picture style will look better than what we got here.

Share this post


Link to post
Share on other sites

Posted · Report post

Well I have. These are my problems:

 

Cinestyle raises the blacks from 0 to 16. This is to match Cineon film scanners. It decreases dynamic range, mostly from the midtones as the highlights and shadows are compressed. Good highlight rolloff is gone. Cinestyle clips them ugly. You have to create them back with a gamma curve. It increases the noise in the blacks and when you pull those blacks down in post, the noise is there. And it doesn't really help in grading at all plus way more difficult to assess exposure. If you want milky blacks, just raise them afterwards.

 

For example the Flaat10p profile is lots better and also Prolost Neutral.

 

You personally don't see banding but that's because you use Neatvideo and it creates very good 16-bit files out of 8-bits. Neatvideo actually fills in the gaps. It's actually quite remarkable how much banding it can take away. But it will do that to any profile, not just cinestyle.

 

Those intermediate ISOs are cleaner in h.264 mode because they are digitally pulled down. This pull down only decreases DR in RAW files but because h.264 loses so many stops anyway, it doesn't really matter there. So basically, when shooting stills, try to go with the original ISOs and when doing h.264, drop down once to get the shadow noise gone and more for the highlights.

 

Do you use the LUT or are you yanking curves around yourself? It's supposed to be used with LUT. Prolost Nuetral was a joke of Stu's, it's just the settings people were already using before Cinestyle existed.

 

Almost every other flat profile is a poor thing made by playing with the packaged editor. some of them are horribly inverted and have no LUT to correct to a good starting point for post.

 

Neat Video solves the problem yes, which is why we use it. Any picture profile is 8-bit and will have banding, so I also solve it in post with Neat Video, it's smart to use the best tool for a job!

 

The only exceptions I've found are VisionColor, they're nice and have LUT built into FilmConvert. Technicolor spent about a year or so working with Canon to make Cinestyle according the PR, I'm not willing to laugh that off for these oddly named experiments.

 

I am willing to shrug it off for Magic Lantern Raw though. ;)

Share this post


Link to post
Share on other sites

Posted · Report post

I agree, ideally any tests should use the very best settings for the h264

Share this post


Link to post
Share on other sites

Posted · Report post

Could you inform us about the lens you´ve used?

Share this post


Link to post
Share on other sites

Posted · Report post

@EOSHD Andrew I'm about to read the rest of the article but the line " In its factory guise whilst not quite as clean as the Canon C300 or Sony FS100, it is the best DSLR for low light shooting (though the Nikon D5200 puts up a good fight)." needs to be updated, since it is factually inaccurate if not qualified.
 
The 1DX, 1DC , D4 and D3s all offer superior high ISO performance and the D4 and D3s stretch their max ISOs further than the 1DX in video mode. Now the 5DMkIII may be the "best DSLR for low-light shooting under x number of dollars" or the best combination of certain features in combination with lowlight shooting, but by every metric (including the posted comparisons of video clips, the scores for the sensors on DXOMark and the RAW stills posted on DPReview for the stills side) the sensors used in the cameras I mentioned outperform the 5DMkIII for lowlight/high ISO, both in stills and video in their compressed modes.
 
Looking forward to reading clarified version of your original remark on your blog and thanks for the early RAW footage and testing. :)

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0