First Sony NEX5 Footage

Could just be the settings, but it looks like roughly around 6 stops of latitude, similar to a dvx, but at least the highlights roll off nicely.
 
CC might be a bit all over the place but for 600 bucks I think it's amazing. Plus there's already a thread on the way for hacking this thing for better compression and manual control. The rolling shutter is pretty much under control. The sensor is an optimized version of the DSLR-A550, which has a latitude of about 11.5. Sony is working on an actual video camera with this sensor and this mount. It will certainly have manual control and better codec.
 
The Sony 550 provides 8 to 10 stops depending on contrast setting, which is better than most SLRs. 0.5 to 1.0 more stops over mid grey compared to a top of the line Canon or Nikon.

The sensors can do 11 stops or more but the output will not look good at such settings. You need high contrast for good color and progressive highlight compression for good highlights. In cameras that record compressed 8 bit video, you have to get that in the camera, you can't do it in post because that will produce heavy quantization.
 
rawfa said:
The latitude is 11.8 (http://dxomark.com/index.php/eng/Image-Quality-Database/Sony/Alpha-550), but it's based on RAW, which obviously doesn't relate to video.
Click to expand...

Have you seen an image at the settings that are using to achieve the maximum dynamic range in the tests? It's unrealistically flat and unusable. The color errors are huge. Skin tones are dead.

With DXO, check the tonal range test, that represents actual pixel quality better. I have seen many ISO200 images from the 550, and the pixel quality is inferior to any Canon/Nikon SLR I have seen.

Dpreview measure dynamic range but they ignore pixel quality. It is still useful if you want to see how they map the original dynamic range compared to other cameras because they do this test for all cameras.

http://www.dpreview.com/reviews/sonyalphadslra550/page15.asp
 
Otis Grapsas said:
Have you seen an image at the settings that are using to achieve the maximum dynamic range in the tests? It's unrealistically flat and unusable. The color errors are huge. Skin tones are dead.
Click to expand...

Why does this matter? Have you seen the skin tones from the RAW D21? Or RED, or SI2K, or the especially the Viper?

Apply LUT, grade, move on.
 
PerroneFord said:
Why does this matter? Have you seen the skin tones from the RAW D21? Or RED, or SI2K, or the especially the Viper?

Apply LUT, grade, move on.
Click to expand...

You can't do that on GOP compressed 8bit material.

But my point is that the maximum dynamic range test has no relation to real world images.

We are becoming obsessed with measurements and the manufacturers know it. The processing in these cameras is designed to manipulate the "scientific tests". The Sony a550 lab tests produce wonderful results. Its ISO200 measures as good as a Canon 1d IV, the S/N results is over 41dBs, etc. It matches medium format CCD cameras in most tests!

But look at these ISO200 images, see the shadow detail and the texture in hair. The camera is using ultra heavy noise reduction even at its ISO200. What is a test good for if the images have the consumer look of a small sensor multimegapixel camera?

EDIT: expired links.

http://www.dpreview.com/galleries/reviewsamples/albums/sony-alpha-dslr-a550-review-samples#page=3

Sample 16,50,53, etc.
 
Last edited:
PerroneFord said:
Are you telling me you can't grade 8-bit GOP footage? Because if you are, you better call up Fox and George Lucas and tell them to hold their horses...
Click to expand...

You really believed that story?

Anyway, grading has nothing to do with it. Getting a usable image out of the high dynamic range images requires very heavy processing. You end up with 4bits if you are lucky.
 
Otis Grapsas said:
You really believed that story?

Anyway, grading has nothing to do with it. Getting a usable image out of the high dynamic range images requires very heavy processing. You end up with 4bits if you are lucky.
Click to expand...

Not sure which story you're referring to, but I have no reason to believe that FOX falsified the information about a Canon 5D shooting the House season Finale. And that was most certainly 8-bit long-GOP, that was undoubtedly heavily processed.

Otis, I don't know your background, but getting good images out of high dynamic range images has been pretty common for the film workflow for a good number of years. Considering that modern Kodak film is hitting around 14 stops right now, I'd say folks are doing a pretty darn good job with it.

As for 8-bit material, I don't know how others do it, but I generally take my 8-bit long-GOP and transcode to 10bit intra (DNxHD) or DPX, grade, then output to wherever I need it to go. I'm not sure how you do your work, so I cannot comment.

I will say that I agree that cameras that are recording 8-bit in-camera are not giving nearly as much as most of us would like, but the images are workable with some care.
 
PerroneFord said:
BTW, I believe the 5D will serve as cockpit cams, and I think that was the intent. Hard to fit an F35 with an Optimo Rouge in the cockpit of a WW2 fighter plane...
Click to expand...

Accordind to Phillip Bloom (I added the bold type):

" Much of this was going to be intimate in cockpit stuff and flying shots. They shot the movie on the Sony F35 but this is a beast and utterly impractical for the cockpit stuff. Hence the Canon DSLRs. They were also looking at new ways to shoot the much anticipated “Star Wars” live action TV series. So they wanted to see how well the 7D and 5D Mk. II stood up with this in mind. Not necessarily shooting on these cameras but perhaps the next evolution of them. If only we could connect that NanoFlash up to the 5dmkII or 7D…"

These companies have engineering support, they have alternatives. I can fit uncompressed 1080p on a plane cockpit or 720p on a very small bike. Small consumer camcorders are also better than the Canon for this use.
 
PerroneFord said:
Not sure which story you're referring to, but I have no reason to believe that FOX falsified the information about a Canon 5D shooting the House season Finale. And that was most certainly 8-bit long-GOP, that was undoubtedly heavily processed.

Otis, I don't know your background, but getting good images out of high dynamic range images has been pretty common for the film workflow for a good number of years. Considering that modern Kodak film is hitting around 14 stops right now, I'd say folks are doing a pretty darn good job with it.

As for 8-bit material, I don't know how others do it, but I generally take my 8-bit long-GOP and transcode to 10bit intra (DNxHD) or DPX, grade, then output to wherever I need it to go. I'm not sure how you do your work, so I cannot comment.

I will say that I agree that cameras that are recording 8-bit in-camera are not giving nearly as much as most of us would like, but the images are workable with some care.
Click to expand...

I haven't seen the House finale. It will take a while till it gets to Greece.

High dynamic range helps producing good images, a clean image is always an advantage. That's not the subject here. Film relies on high contrast to produce its saturation. High dynamic range stock combined with grading, or low dynamic range (10 stops in some stocks). The high contrast of the distribution print finish the job by increasing contrast further. But we are talking about 8bit GOP here, 8bit gop does not provide enough precision for good post. There is low spatial precision, the quantisation by the compression is very heavy so shadows are very bad, it uses YCbCr which cannot achieve 8bit RGB quality, and any motion makes things even worse. Your best bet is to get the image you need on camera, good exposure and white balance, and only do slight grading in post. An exposure mistake of 1 stop wastes at least 6dB of noise (from an original 30dB or less). A white balance error can be an extra 6dB for tungsten vs daylight. Combine the two and you have increased noise by a factor of about 4x, the equivalent of 12dB of gain, and that is digital gain with higher alias not analog in-camera gain which only increases noise.

The transcoding you use only preserves the quality you already got in multiple generations. If you have 5bit shadows due to compression quantisation and blocking, you are stuck with that. Try thinking in film terms, using the lowest contrast you can and compensating in post, it will be very inferior to the in-camera result.

It does take great care and it helps if you understand the limitations.
 
You must log in or register to reply here.
Back
Top