Jump to content

David Stump's tests


Stephen Williams

Recommended Posts

  • Premium Member

Can't wait to see the images.

 

David posted some very meticulus notes on his methods. Right down to what lightmeters he used for which situations, thier serial numbers, when and by whom they were callibrated.

 

I'm a little disappointed though, he failed to mention what he had for breakfast before he shot the tests.

 

Very detailed. Very impressive.

Link to comment
Share on other sites

  • Premium Member

EDITED

 

I wonder if I can say what I think from these grabs, since it looks like anytime someone critiques the RED, on these forums, he gets offended by a team of so-called producers and DOPs who never came here but for promoting this camera... Maybe it's part of a commercial policy to be so agressive I don't know, but, hey, I'll try and prepare for the hard time to come...

 

- There is no pure black in the image, even though the exposition is low enough so that the whites are not white (at F 19 for instance)...

 

- The image doesn't look very contrast, in any situation.

 

- Pity he doesn't go further on the overexposition...

 

- The colors don't look very saturated, but maybe this is to match the Varicam's color rendering, that is loved by many DOPs, I don't know, but there is a need for color timing after that, just as Varicam needs it, I think.

 

- Pity there is no definition chart... The Siemens Star is not enough, to me.

 

Ok, now, RED team, pull the gun... :unsure:

Link to comment
Share on other sites

  • Premium Member

The most interesting part for me was the C-Stand on the left. There is a catch on it that is very telling in how the sensor deals with highlights. Can't say I like it too much, it looks clipped. I love the look of blow-out highlights and obvioulsy that is a problem with digital cameras.

Link to comment
Share on other sites

I'm a bit puzzled.

On the CML site they say: "As usual click on a frame for the full res version, they're 74Mb each!!"

 

But, a 4K 16 x 9 sensor would have about 9 megapixels. If it had 16-bit analog to digital conversion that would mean a raw data rate of 18 megabytes per frame with no compression at all.

 

Even if it had separate red green and blue pickups for each of the 4K pixels, (which it doesn't) that would still only a mean a file size of 54 megabytes for 16 bit RGB uncompressed.

 

So where do they get 74 megabytes from, and what possible use is such a massive file size anyway? (No I don't feel particularly inclined to download the file to find out!)

Link to comment
Share on other sites

  • Premium Member
So where do they get 74 megabytes from, and what possible use is such a massive file size anyway? (No I don't feel particularly inclined to download the file to find out!)

 

I dont know as much about digital interfaces for cinematography as I know about computer file formats, but Im guessing that the file is 24bit since that would be 16bit size * 1.5 which would put it at around the size mentioned.

Link to comment
Share on other sites

I'm a bit puzzled.

On the CML site they say: "As usual click on a frame for the full res version, they're 74Mb each!!"

 

But, a 4K 16 x 9 sensor would have about 9 megapixels. If it had 16-bit analog to digital conversion that would mean a raw data rate of 18 megabytes per frame with no compression at all.

The sensor is 12 Megapixel - not 9. 30% higher than you assume.

resolution is 4900x2580. on S35 AR it will be ~4520x2540, a bit higher than 4K.

 

Even if it had separate red green and blue pickups for each of the 4K pixels, (which it doesn't) that would still only a mean a file size of 54 megabytes for 16 bit RGB uncompressed.

it has 12 millions readout pixels - which are converted to RGB. add the 30% higher resolution compared to your assumed 9MP and there you go.

Also it is not only 16 uncompressed to store, some dozens bytes to add, which go into the LUT :)

 

So where do they get 74 megabytes from, and what possible use is such a massive file size anyway? (No I don't feel particularly inclined to download the file to find out!)

a) the oversampling of the sensor makes sense in order to compensate the given resolution reduction when debayering a CMOS-basing photographed image.

B) the files are not huge. that is 4k biz as usual - pretty common in top-end cinematography since some years. i know most of 35mm and digital releases are done 2k today, however: even if you dont use 4K for your release, but go for a typical theatrical 2K release, it gives you many important advantages aquiring and postproducing in 4k. Here a some.

 

- pan & scan: with 1080p & 2K, be it 35mm or cinealta or whatever, you dont want to enlarge or pan your image.

- noise : you can push the borders in DI & colortiming/grading quite a bit further in 4k - when noise begin to show up slightly in 4k, your downrezzed 2K will not show it, or at least much weaker.

- postproduction & VFX: for pulling keys, tracking/stabilizing etc the higher resolution is a blessing. Also, VFX assistants can be a little bit less tight when masking, rotoscoping etc.

 

- sidenotes: print: yes, i know this might sound bizarre, but certain dops will use RED as digital still SLR camera as well, and 4K is useful for printing larger sizes.

 

furthermore, with the optional redcode wavelet datareduction, you only need ~30MBytes/sec for 4k datareduced, and the higher resolution makes the datareduction, if one wants to use it, quite useable for cinematic release. however, we will have to push the footage through the DI workflow first in order to understand how much the datareduction will reduce colorgrading etc.

 

On cinealta we typically shoot a mix - uncompressed 10bit to disk for rather comples postprodution shots, to tape for "usual" shots.

 

 

 

 

I dont know as much about digital interfaces for cinematography as I know about computer file formats, but Im guessing that the file is 24bit since that would be 16bit size * 1.5 which would put it at around the size mentioned.

 

24bit is = 8bit/channel*3, the images from david stump are 16bit/channel + lut. the final will "only" be 12, not 16 bit. the 16bit a there as REDCODE isn´t available yet, and 16bit tif is common.

also, the images are not only 4K, but 2540+. 4900x2540 from the red frankie protoype. 4540x2540 will be the 4K in S35 from the serial red one.

 

the images are coming from the uncharacterized sensor, no sharpening at all, no colorcompensation etc. also the charts contain syncing problems the sensor was misaligned often etc. red however is ok with having the prototype shots published.

Link to comment
Share on other sites

Hi,

 

A very heavy & expensive SLR IMHO.

 

Stephen

 

even if we certainly won´t replace our making-of Canon 5D by the red - i overheard several talks of RED-buyers, especially indy-dops, who intend to do so at IBC 2006.

 

1) to buy 2 cams is more expensive

2) to carry 2 cams is heavier than one - and the 3.1 kg red has to be brought on set anyhow.

 

also, taking stills through cookes, zeiss, angenieux isn´t to bad either.

Link to comment
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
×
×
  • Create New...