PDA

View Full Version : thought bubble: S16 PL lens + EOS adapter + RAW cropped to 1080....



imdjay
10-12-2012, 11:59 PM
maybe someones tried, but with so few cameras out there, perhaps not...

the thought bubble is, what if you took a PL to EOS mount adapter, which i know does not work with many lenses, but stick with me here...

instead of s35 glass, what about s16 glass, and shoot raw to then do a post crop down to 1920x1080 to get rid of the vignette.

thats the thought. curious if it's been tried, or if theres solid reasons why it wouldn't work in the first place. the reason of course is for some nice vintage glass, and to solve the wide+fast lens conundrum with the EF BMC (take a guess what i've got reordered)

and just as a visual for instance. attach this:

http://www.ebay.com/itm/7mm-T1-3-Panther-Cinar-Super-16-lens-Arri-PL-mount-rare-Red-One-Epic-S16-EXC-/280971457232?pt=UK_Lenses_Filters_Lenses&hash=item416b3432d0

using this

http://www.adorama.com/CZPLEOS.html

shoot raw 2.5K. see the vignette, crop down to 1920x1080, hope vignette goes away and left with a wide fast solution. and this would TECHNICALLY be like having a 1920x1080 s16 camera with no crop factor

thoughts?

Nomad
10-13-2012, 05:08 AM
Well, this is no different from using the S-16 glass on µFT version, see the discussion on Zeiss Ultras.

Please remember: 2.5 K photocells from a Bayer sensor doesn't mean it resolves 2.5 K. With a very good de-bayering algorithm you'll get about 70 to 80 % of that. So you'll more or less use full HDTV resolution to it's limit when using all the RAW information.

If you crop, you won't outresolve HDTV any more, so it'll get a bit softer. If you still will be able to see the excellent resolution of a Zeiss (or the like) should be tested by someone who already holds such a camera and the lenses. I have the lenses, but no camera :-(

But I tested them on a Scarlet at 3K and that clearly indicates what I wrote here.

Amr Rahmy
10-13-2012, 01:45 PM
cmos resolution resolving problem aside, s16 will cover 1920(you still have 0.1mm difference), and you could also have some additional headroom before vignetting. or you could just imagine it's a s16mm camera with a native 1920x1080 cmos sensor. wide problem solved.

but is a vintage s16mm really sharp, or would you have to go to a zeiss ultra prime for this camera, is it a wise investment, are there other affordable solutions, a lot of cameras with s35mm and full frame sensors are coming out nowadays. and it's possible that the next bmc will have a bigger sensor.

razz16mm
10-13-2012, 02:21 PM
S16mm lenses are typically significantly sharper than most 35mm still lenses. They have to be to resolve the smaller format with high MTF.
The only reason I can see to do this is to use a fast compact wide range zoom, like a Canon 8-64 or 11.5-138 for instance, as a single lens solution. But you are looking at $3k-$8k or more for a good used zoom.
Cost vs benefit doesn't really work out for S16mm lenses on this camera. The money would be better spent on premium 35mm still glass the camera was designed for.

John Brawley
10-13-2012, 04:46 PM
My experience, coming from RED, is that windowing or cropping down the sensor usually leads to resolution disappointment. Just because you've got a 1920 file won't mean it will have 1920 resolution.

The 2.5k sensor size on this camera was chosen for rendering a really nice 1920 image. This is the way CMOS works. You want a bit extra so that the delivery resolution looks great.

It's also why a RED EPIC is shooting for 4K delivery with a 5K sensor.

Cropping down to 1920 will give you LESS than a 1920 *feel* and *look*.

The question being, does using a lens like the Super 16 ones mentioned here give you enough of an upside over the resolution hit of a cropped image ?

Maybe in some circumstances.....

jb

Amr Rahmy
10-13-2012, 05:22 PM
everything is a compromise of course, but i don't think file size is more important than having a better lens range, or better lens to choose from. a sharper lens will have more impact on perceived resolution. much better than having the entire image soft.

some of the s16mm other than the ultra will actually cover the sensor at any aspect ratio wider than 16:9. and the ultras still cover a tiny bit more than 1920, especially at wider aspect ratios. and red sensors are notorious for being soft and having limited color gamut, so it's not a good reference point.

and we are talking about wide lens for wide shots, what other options are there? not having the lens in that range at all. it is too expensive as an investment with all the s35 cameras released or coming in the next few years, but it's a viable option.

Amr Rahmy
10-13-2012, 05:27 PM
i also think, good films transferred to vhs look like film, and crappy tv shows captued in 1080p look like crap. that's resolution independent.

Nomad
10-15-2012, 06:23 AM
@ Amr:

You wrote: "and red sensors are notorious for being soft and having limited color gamut, so it's not a good reference point."

While color is always debatable, they are soft because they don't use any artificial sharpening like video cameras, and the same will apply to BMCC's DNG. So, the resolution issue is similar, or why does Arri use more photocells than 1080 for the Alexa? Or why is Canon using even twice the number in the C300 (without conventional de-bayering) ?

There is also the issue of very conservative OLPFs in RED and Arri vs. no OLPF at all in the BMCC, so it'll be aliasing vs. sharpness…

But for the true impression of perceived resolution, we'll need tests with the final combination. If you take Otto Schade's research of resolution perceived by humans vs. MTF into account ( see http://en.wikipedia.org/wiki/Otto_H._Schade ), the excellent optical properties of a Distagon with good micro-contarst might compensate for lower MTF.

It's quite a complex balance of factors in the end!

razz16mm
10-15-2012, 08:29 AM
@ Amr:

You wrote: "and red sensors are notorious for being soft and having limited color gamut, so it's not a good reference point."

While color is always debatable, they are soft because they don't use any artificial sharpening like video cameras, and the same will apply to BMCC's DNG. So, the resolution issue is similar, or why does Arri use more photocells than 1080 for the Alexa? Or why is Canon using even twice the number in the C300 (without conventional de-bayering) ?

There is also the issue of very conservative OLPFs in RED and Arri vs. no OLPF at all in the BMCC, so it'll be aliasing vs. sharpness…

But for the true impression of perceived resolution, we'll need tests with the final combination. If you take Otto Schade's research of resolution perceived by humans vs. MTF into account ( see http://en.wikipedia.org/wiki/Otto_H._Schade ), the excellent optical properties of a Distagon with good micro-contarst might compensate for lower MTF.

It's quite a complex balance of factors in the end!

Good post.

Another wiki article that illustrates the issues involved: http://en.wikipedia.org/wiki/Optical_transfer_function

imdjay
10-15-2012, 10:50 AM
nice direction the thread is going, but i'll clarify a bit on my OP: im specifically referring to solving the wide/fast option, not going entirely s16 glass for every situation. so basically just have one wide/fast s16 lens with an adapter always on it, and treat it just like the other canon glass on the shoot.

now of course the question with that situation is will it cut well with the other glass, i may be a nitpicker but i think i'd be more than happy with a slightly differing image to cut in if that means i get the fast/wide option.

Amr Rahmy
10-15-2012, 10:58 AM
there is sharpening in the processing of the red file at the processing stage before the files are ready, so it's there for all cameras(other cameras have the option to not use sharpening in camera). and the images are softer than most/all other cameras, hundreds of people and tests stated/proved such over the years for all the sensors red put out. it's not an opinion, or something up for debate. alexa and c300 produces sharp 1080p images because they are not pretending that their these cameras are 4k or 5k cameras. and the images are much cleaner which makes the red images all the more soft. i thought the first red one sensor was capable of producing 720p 6bit extremely noisy images with a limited color gamut and with limited exposure compensation. the newer sensors are better but still inferior to everything else remotely close to there price range.

but the matter at hand,

the bmc sensor from 2230-2096(the ultras and most s16mm might fall closer to 1920 depending on the aspect ratio and each individual lens) pixels will not resolve 1080p 4:4:4 image, or 1080p resolution, it will make a 2096 or 1920 file through assumptions/calculations made by the algorithms used in the software. so the color space and sharpness will definitely suffer(but at least the more expensive ultras and sk4 are sharper than inexpensive full frame still lens), it's a compromise. but it can be cleaner than noisy red one files and it can have a good color gamut reduced in a smaller color space. if you start with a good color rendition 12bit 4:4:4 worth of color information and go to 12bit 4:2:0 color information in a 4:4:4 file container, it's not that big of a compromise. and it's only for the really wide shots, so it would cut nicely with other shots. it's not for every shot, just a solution to the wide problem. do you have other rectilinear solutions.

imdjay
10-17-2012, 04:03 PM
nice thoughts. have there been any shots done yet with s16 glass with any of the BMCs in the wild?