Chinese 2K RAW S35 Cinema Camera, 8000 dollars

Zephyrnoid said:
The pro video and movie industry can't let that happen to them and the key is to moderate the ability of amateurs to access the pro tools.
It's that simple.
So it's pointless anguishing over the obvious facts.
We do what we can with what we can afford and let the pros and big studios keep their jobs.
Click to expand...

I don't really agree. It takes a lot more to make a good movie than just the right tools. It takes a good script, quality acting, attention to detail, great audio/music and good looking special effects. Beyond the people and the time it takes, it's about attention to detail. So iIdon't buy into the conspiracy theory that major studios and manufacturers are trying to hold the little guy down. It doesn't benefit them to do that. If the little guy makes a great film the big guys will buy it and take their piece of the pie. Everyone wins. But if the quality of the little guys work is less than stellar because of the limitations of the equipment he had access to, it just makes it harder on the big guys who are trying to sell it.

The reason we end up with what we do on the low end of the video spectrum is because manufacturers are constantly trying to dumb down their video cameras for soccer moms. They aren't targetting no-budget filmmakers. They're trying to sell as many cameras as possible by making them easy to use and compatible with the TV's they sell. Like AVCHD, it's existence is to cram as much video as possible onto SD cards that are also compatible with panasonic TV's and blu-ray players. They have this visionary idea that every soccer mom is going to buy a panasonic GH2, panasonic flat panel and panasonic blu-ray player and move the card from one device to another, from shooting to viewing. Are they right or are they wrong is the question. That's why we don't get 12bit 4:4:4 with high bitrates in camera. They don't think their target demographic even cares.
 
Last edited:
As I view the S8 camera its a low cost replacement for 35mm movie film rather than Super8mm film, giving an end resolution on the 35mm anamorphic release print akin to blowing up 35mm Techniscope (tm) film, or perhaps a bit better when shooting in the 2.5K mode.

The advantage of the Super8 lenses is that they are more compact and lighter weight and also lower cost then big heavy 35mm movie camera lenses.
Click to expand...
Thanks for the detailed update. I did a little reading up over the weekend, on some of your posts and have a little more idea of what the camera is about. I'm looking forward to seeing the handheld footage.

One thought I've had. Would the S8 version be diffraction limited in any way? I'm presuming it would be, like smaller chip camcorders. Also, like Kurtf, I'm interested how the over crank modes would work. Is there any resolution loss?
 
diffraction

diffraction

AdrianF said:
Thanks for the detailed update. I did a little reading up over the weekend, on some of your posts and have a little more idea of what the camera is about. I'm looking forward to seeing the handheld footage. One thought I've had. Would the S8 version be diffraction limited in any way? I'm presuming it would be, like smaller chip camcorders. Also, like Kurtf, I'm interested how the over crank modes would work. Is there any resolution loss?
Click to expand...


I've been working on the hand held shots today, I've shot like 180GB so far, about 30 minutes. I'm shooting 1920x1440 mode to crop 1920x817 then letterbox that into 1920x1080 since the firmware does not support 1920x817 right now, I'm doing that because my experience with 1828x1556 mode requiring a small enlargement to 1920 width for Vimeo (not so for 35mm film transfer since 1828x1556 is the Cineon (tm) Panavision (tm) pixel size for so called "2K" filmout on a Celco (tm) film recorder) showed that I may need to look at my upscale code and how I am sharpening the images to minimize some de-Bayer or other artifacts. Anyway, the shots I have processed look good so far, I work-printed all footage shoot and viewed it playing 24fps in my NLE and also made an AVI of one hand held moving camera shot and looked at in my Brother's 8 core W7 64bit machine that plays smoother. I should note that you should have your 1920x1080 monitor set to 96fps or 120fps to avoid rolling issues from the monitor refresh that can look like camera issues even though they are from the monitor if your Hz is low or not a multiple of 24. I'm not sure if any LCD use global update, but CRT are rolling display, and if you use analog cable to the monitor the monitor's buffer will update as needed and may not be in sync with the graphics card, so you can get some jutter just from the monitor and graphics card. The current top overcrank mode is 220fps. The drop in resolution is from the smaller pixel size, and so you need a sharp lens that covers the smaller area, the camera has movie 8mm D-mount lens support, when I get time I'll try the D-mount Switar 5.5mm and 36mm to see how they work. The production camera may support a sensor about twice as fast as an option or something so that might allow for some faster fps. At top speed currently its in the SD/DVD range for pixels, there are several modes like 144fps and 220fps as well as 48fps and 72fps, 29.97fps, 59.94 etc.



If you were making a feature film you could rent a RED (tm) for a day if you need to shoot matching resolution in slow motion. With a compressed camera like RED (tm) the bandwidth is like constant, so as you increase the fps the compression rate should increase, but with a True RAW camera in uncompressed mode the image size needs to decrease in order to increase the fps because you have maxed out the record bandwidth. There are uncompressed high speed cameras made, you could also rent those, look on Vimeo for some demos of the Phantom, The S16 and S35 KineRAW (tm) are forecast to use Cineform (tm) so that can have its compression ratio adjusted to record overcrank within the sensor's limits without image size reduction to some degree maybe, it depends then when the sensor maxes out. I plan on shooting some overcrank tests when I have time, but I saw that mostly for internet use or for scientific instrumentation cameras uses perhaps, more than for cutting into a theatrical feature, maybe. It did not cost anything really to support over and under crank, so some modes were added for testing in the lens demo reels. == As to the question of diffraction on the S8p, it shows up as you stop down past f/6.3 (T/8) maybe. Dust also can be more of an issue at small stops so its better not to stop down past f/8 too much, although you can sharpen the images in the de-Bayer by adjusting the anti-OLPF blur correction. Some diffraction helps reduce aliasing as its not an on or off issue, the blue sensor resolves about twice what the red sensor does because blue wavelengths are about half the size, so you can still get some detail from the blue and UV light that is in focus. I have been trying to shoot between f/3.5 and f/8 to get best results from the lenses I have been using, but in the anamorphic reel much of the footage was shot at f/0.95. == No camera is without issues, but you use the one that gets the look you want or need for your project, and that does not mean that every shot is done with the same camera, most movies were shot with Mitchell NC cameras that top out at 32fps, when high speed shots were needed the Mitchell High Speed was used up to 144fps. Likewise with digital cinema, if you are shooting 95% of your footage at 24fps, do you need to purchase a high speed camera at a much higher price and NOT use the high speed modes all the time, or can you just rent or barrow another camera for the 5% of your footage that needs some special feature like High Speed or HDR that you don't use for 95% of you shots. One think I like about the CMOS sensor in the KineRAW-S8p (tm) is that it takes over bright highlights well, with no vertical streaks like CCD sensors can show.

The digital Bolex (tm) as three main issues as far as I can see:

1) The CCD is too large for 2/3" lenses to cover the 2048x1152 mode, they work file on the smaller KineRAW-S8p (tm) sensor. Soft corners may be an issue for using Bolex 16mm lenses as well wide open etc. The smaller sensor in Acam dII (tm) may show less soft corner issue.

2) The CCD may show vertical streaks on over bright highlights making a very un-movie film like artifact. Not the case with KineRAW-S8p (tm) to the same degree as far as I have seen. Acam dII (tm) also seems to show some vertical streaks in frames I have seen.

3) The quadrant readout of the former Kodak (tm) sensors may produce splits in the FPN making higher ISO shooting something of a problem with getting the black calibration very close enough so that you cannot see the splits after grading up quite a bit. The Acam dII (tm) also has this issue although they seem to be working on improving their black calibration. The KineRAW-S8p (tm) only has a single readout, so there is no splits issue, and my free de-Bayer program has FPN and bad pixel masking built in.

KineRAW (tm) asked me about the 1" CCD sensor and I mentioned that its sort of a "no mans land" for lenses, although there are 1" megapixel lenses now from companies like Kowa (tm). The price on the 2K 1" sensor may be lower than the Kodak 3K CCD sensor that works with 35mm movie lenses for wide screen. If you look at the time lapse sky shot made with the Digital Bolex on Vimeo the corners look soft maybe. Also check that other companies are showing 1920x1080 videos rather than all 1280x720 as its much harder to get 1920x1080 to look good with regard to de-Bayer artifacts and noise, going down to 1280x720 improves both from a sensor shooting 1920x1080 from oversample. The KineRAW-S8p (tm) has oversample to 2592x1104x12bit@23.976,24,25fps.

==

A digital cinema camera is not a "replacement" for a HDSLR, its a different sort of tool, so if you are not interested in Digital Cinema then you need to ask what you need a Digital Cinema camera for, perhaps?

If you find a better True RAW DNG recording 2.5K or higher Digital Cinema Camera selling for less than $1500-$2000 with 1280x720p HDMI monitoring and live 200% zoom, waveforms, and histograms and monitoring 'look' tables that works with 2/3", 1/2", 1/3", R16, S8 and R8 lenses let me know.
 
Last edited:
I dont think that 1" CCD sensor is bad idea ...u can always use crop part of sensor for 16mm lens and still have option to use s16(optimized for 12.85mm) or anamorphic lens ...and new 1" cctv lens are very good ..
just look kowa they go 120 lp/mm more than sensor can resolve....
 
Last edited:
So I see endless technological posts here, but where in the pipeline in this camera? When is it due to be released?
 
Kineraw S35 expected to go on sale in China in May according to someone that contacted the company:

http://vimeo.com/38907938

and elsewhere one month later according to various other forum posts.


... and as Red says "It's all subject to change" ...
 
In reply to some of the comments.

If people using the digital Bolex (tm) or other camera with a 1" sensor crop the image to like 1600x900 pixels so they avoid soft corners with regular 16mm lenses, then you get into the upscale problems that I had going from 1828 to 1920 in the C1 anamorphic test reel except that they would not have the advantage of the 1556 to 817 reduction in the vertical direction so the enlargement artifacts would be on both axis.

I'm processing some hand held KineRAW-S8p (tm) tests today and they are a 1:1 pixel crop 1920x817 from 1920x1440 shot. A Bayer sensor of only 1920x1080 is not really quite enough for theatrical projection using DCP, on a 35mm film print its probably usable but its at the edge of enough not so much because of resolution but because of aliasing and moire, if you under-sharpen then it may look a tad soft, and if you get the sharpness up you may see some aliasing. It has to do with how the de-Bayer is done and how that interpolation interacts with the sharpening and the OLPF filter and lens all together.

The larger sensor in the KineRAW-S35 (tm) works with lenses that have higher intrinsic resolution than smaller format lenses in relation to the image height, so the combination works out a bit differently.

Originally I thought I was going to be doing test demos on the production S8 camera, like a year a go, but things did not work out that way, so I've been doing the lens tests and using that footage to de-bug my de-Bayer and DI software and make improvements as needed so that come the day the cameras go on sale they will be supporting the revised meta-data and such hopefully.

The S35 demos linked to above were done before they got the 3D-LUT color matching working, the greenish faces and pink highlights are not the camera's fault, but the way those shots were color corrected without my input on the matter. I re-corrected some of those DNG and the look better processed through my programs. With the new color matching if I then export my color correction as a camera monitoring 'look' file the camera can turn that into 3D-LUT to match the color corrections downstream in the workflow, in that case I think it was Iridas Speed Grade (tm).

I'm waiting to get some trial DNG using the revised OLPF filters, they may need to modify the test cameras a bit to install the new filters for testing, but those tests may happen in a week or two.

While I'm waiting I keep grading and processing the hand held shots for my next lens test reel. In general the Kowa 3.5mm f/2.4 seems to be working, I'm applying lateral chromatic aberration correction to some shots because of a slight red fringe at the edges of the image.
 
Dan Hudgins said:
...

The S35 demos linked to above were done before they got the 3D-LUT color matching working, the greenish faces and pink highlights are not the camera's fault, but the way those shots were color corrected without my input on the matter. I re-corrected some of those DNG and the look better processed through my programs. With the new color matching if I then export my color correction as a camera monitoring 'look' file the camera can turn that into 3D-LUT to match the color corrections downstream in the workflow, in that case I think it was Iridas Speed Grade (tm).
...
Click to expand...

Hi Dan

Thankyou for you quick response. Just to clarify, were the poor colors totally the result of the person color correcting, and not the result of using Speedgrade to process the image from this sensor?

Cheers

Adam
 
adam777 said:
Hi Dan

Thankyou for you quick response. Just to clarify, were the poor colors totally the result of the person color correcting, and not the result of using Speedgrade to process the image from this sensor?

Cheers

Adam
Click to expand...

I think it was more complex than that. They did not tell me exactly what they did but from what I could make out the de-Bayer was done before loading into SpeedGrade (tm) for correction, but I'm not sure exactly what the steps were since I was not there.

If you matrix the color with the wrong clip points you can get that sort of result in any program, and you cannot correct it once the matrix has been applied I think even with a 3D-LUT because you get two colors represented by one color so its a non-reversible corruption as far as I can make out. Some of the early Acam dII (tm) samples processed in Adobe (tm) products from what I can make out showed the same sort of greenish skin tones and pink highlights. Its an easy problem to fix and not caused by the camera's sensor for the most part unless the ADC outputs have been processed in some way, then it may become imposable to correct without clipping off all the data above some level, maybe.

The reason I mentioned SpeedGrade (tm) was that they now have 3D-LUT support for that, and if the right steps are followed, then maybe the color matching would work out better now. As for the noise level and other de-Bayer issues that can vary from one program to another.

The point being that if you want good results you have to be careful about having the clip points right at each step, since if you matrix with the wrong clip points and compensation for non-linear response then it may be imposable to un-do such folding of the gambit. Imagine taking the CIE color chart and folding it into a paper airplane, then mix the overlaping colors, now tell me how you can reverse that and get the right colors back? So its not a matter of fault of the grader perhaps as asking someone to fix something messed up so bad its impossible to recover is not going to give good results no matter what software you use, maybe?

I posted the link to that video and some of my samples here before, but deleted them since as I understand it the current S35b uses a different configuration altogether, so those early results would not be relevant anyway?

I was just pointing out that it was not my software that processed that S35a test, that I was not involved in the color correction of that test, and that it is not relevant to the current workflow with SpeedGrade (tm) since the 3D-LUT taken from the camera monitoring tables was not in place when that video was done. So although it shows something was working, you should not base an opinion on the look of the S35b based on that test as far as I can see, probably.
 
Last edited:
Hmmm....just a comment. After today's announcements....I'm still very interested. Still no RAW 2K for the same price as these cams.....
 
camera du jour?

camera du jour?

HHL said:
Hmmm....just a comment. After today's announcements....I'm still very interested. Still no RAW 2K for the same price as these cams.....
Click to expand...

Which "After today's announcements..."?

I've been under my rock grading footage for the next reel today so have fallen behind the "camera du jour" reports?

EDIT: Thanks for the reply. 4K 8 bit?
 
Last edited:
The camera du jour is currently the Canon "4k" DSLR. That could well change by lunchtime though:)

Thanks again for the detailed updates. If they are going to launch this camera over the summer, then they really need a little more info on their website.
 
I saw a used SI-2K mini (tm) on ebay but it was not as low, do you have any leads to ones down in that range?

I have not seen what their sensor data looks like at -3 stops, the only RAW frame I have is a green screen shot with very flat bright light.

I have seen the SUMIX (tm) data (same sensor as SI-2K?) someone sent me and there was "some" moving pattern noise, I thought that maybe it needs some bypass caps if its external to the sensor, but if its internal to the sensor then it might be an issue for low light. If anyone has an SI-2K (tm) and can shoot some tests under my direction I can get a better idea of how things compare maybe. SUMIX (tm) is also odd because they have a median noise filter and gamma correction acting on their "raw" data before it is saved to DNG. It seems there is a way to turn that off, but its very peculiar to me to do that before the de-Bayer as you muck up the data for all time. I think they also do white balance clipping before saving the DNG, which is just strange to me, as it ignores the point of DNG which is to save linear sensor data as it comes out of the ADC for archiving what the sensor captured to fiddle with in post later. I have some support for SUMIX (tm). The guy that sent me the data said he could not get it to record above 960x480 without issues, so maybe you need a VERY fast computer to use it?

That's what I like about the KineRAW-S8p (tm) its one unit designed to work as one unit, no external recorders or cables needed other than the monitor to see what you are doing, but I think you can un-plug that and shoot with an optical finder once its setup(?). So it works like a movie camera. It beeps when the SSD is full so you could just put another SSD in when it beeps and not even need a monitor maybe? They are adding a beep-enable-disable to the S35b it seems, so its getting close to working all up.

I still don't have any data from Aaton (tm) as they have not sent me any of their DNG when I asked. I can process un-calibrated data so they don't need to worry about that part of the issue. Has anyone seen anyone shooting with an Aaton (tm) 4K camera on a real project?
 
vladszp said:
@Dan Hudgins
crop part of 1'' kodak sensor for 16mm is 1980 not 1600...
full size of that sensor is 2336 for super 16 ....
in 16:9 s16 mode image size iz 3mp
that sensor is perfect for their s16 camera ....
and it gives cleaner and more organic image than si 2k ...
Click to expand...

Lenses made for regular 16mm may only be sharp within the 16mm projection aperture, not the camera aperture which was never shown on the screen, so the image dimensions may be smaller than you think at the single pixel level. That is why I was responding to the person that posted about cropping the blurry part off, if you do that then you get a non-standard image size, so you need to enlarge it to 1920x1080 or 1998x1080 or 2048x858 which are standard sizes.

5.5micron is 0.0055mm

1920x0.0055 = 10.56mm

R-16mm camera width 10.26mm x 7.41mm

16mm camera is 0.402"x0.292"

16mm projection is 0.380"x0.284" = 9.652mm x 7.2136mm

TV field is 0.297"x0.242"

9.652mm / 0.0055mm = 1754 pixels width.

So 1600x900 should be within the astigmatic circle of regular 16mm lenses.

If you watch their video I think you will see them say that Super16 lenses are to be used with the 2K mode because its larger than regular 16mm lens coverage (implied). And where do they (d16) say that the 2336 width will be supported for recording?

KineRAW-S8p (tm) is 0.0022mm pixel size, so:

2592x0.0022 = 5.70mm

So that is well within the 7.54mm TV-16 projection aperture area near the sharp part at the center of the image.
 
Last edited:
i dont care for dbolex im talking about kineraw guys let them put 1'' kodak sensor and put option to use full size at 2336 ...
what u say can imply for all sensor with 16 mm size and regular 16mm or s16 film ..
by the way they supported 2336 recoding:
"5 We will allow full frame capture at 2336 x 1752 for people who want the maximum resolution the sensor can offer. This is a 4:3 format, but could also be used with Anamorphic lenses for a widescreen look."
but i dont have much hope for dbolex project
im not interested for s8 kineraw but i will buy their s16 if it comes with good spec and price...
and u can always use adapter and put something like SLRMagic 12mm f1.6 ...
 
Last edited:
HHL said:
Hmmm....just a comment. After today's announcements....I'm still very interested. Still no RAW 2K for the same price as these cams.....
Click to expand...

I'm not going to lie, but I'm also still keeping an eye on this - after all this rumpus, these guys could still have a large market (if they decide to push it properly outside china).
 
Dan Hudgins said:
Lenses made for regular 16mm may only be sharp within the 16mm projection aperture, not the camera aperture which was never shown on the screen, so the image dimensions may be smaller than you think at the single pixel level. That is why I was responding to the person that posted about cropping the blurry part off, if you do that then you get a non-standard image size, so you need to enlarge it to 1920x1080 or 1998x1080 or 2048x858 which are standard sizes.

5.5micron is 0.0055mm

1920x0.0055 = 10.56mm

R-16mm camera width 10.26mm x 7.41mm

16mm camera is 0.402"x0.292"

16mm projection is 0.380"x0.284" = 9.652mm x 7.2136mm

TV field is 0.297"x0.242"

9.652mm / 0.0055mm = 1754 pixels width.

So 1600x900 should be within the astigmatic circle of regular 16mm lenses.

If you watch their video I think you will see them say that Super16 lenses are to be used with the 2K mode because its larger than regular 16mm lens coverage (implied). And where do they (d16) say that the 2336 width will be supported for recording?

KineRAW-S8p (tm) is 0.0022mm pixel size, so:

2592x0.0022 = 5.70mm

So that is well within the 7.54mm TV-16 projection aperture area near the sharp part at the center of the image.
Click to expand...

From people's experience with 16mm lenses on 4/3 format cameras, the Kodak sensor at 2k just fine, many longer than 25mm focal length cover 4/3 (22mm image circle) with surprisingly little vignetting.
Reference optical circle for standard 16mm is 12.75mm diameter
For S16mm: 14.54mm diameter
for 2048 x 1152 the Kodak image circle diameter is 12.93mm approximately.
2048 x 1080p 2k digital widescreen format: 12.73mm
1920 x 1080p HD : 12.12mm
2336 x1752 full area: 16.06mm diameter
Standard 16mm lenses cover RED 2k 16:9 just fine. (12.69mm image circle)
 
You must log in or register to reply here.
Back
Top