Jump to content

Super8 Digital Cinema using Beaulieu lenses?


Dan Hudgins

Recommended Posts

I know so little about the Ikonoskopes. What did they do wrong?

 

They use a proprietary memory card format and at present only have a USB 2.0 card reader for it. (The cards are 80gb or 160gb which would take minutes of recording and hours of downloading! Not to mention the cost vs SSD.)

The built in viewfinder is VGA resolution only.

And the sensor is "only" HD (bayer).

Despite all this the pictures from the Ikonoskop look beautiful, really "film" like texture to my eye at least.

It just seems like their design wasn't updated as technology changed (drop in price and increase in speed of memory).

 

So far kineraw 16 looks like the perfect camera for me.

Records to SSD, Raw, maybe more resolution for a true HD result after debayering, and hopefully cheap.

Link to comment
Share on other sites

The 8 and 16 models have me and a few others curious. $8 grand for the s35 model would be very interesting if it delivers. But if the 16 is 4 grand and 8 is 2 grand, well, then that's beyond interesting. They have my attention, but the complete lack of video footage and information makes me wonder if it will end up being vaporware.

Link to comment
Share on other sites

The 8 and 16 models have me and a few others curious. $8 grand for the s35 model would be very interesting if it delivers. But if the 16 is 4 grand and 8 is 2 grand, well, then that's beyond interesting. They have my attention, but the complete lack of video footage and information makes me wonder if it will end up being vaporware.

 

I thought that was the point of this thread!

 

If Dan has the S8 version and is shooting tests with it already then it seems like it's pretty close to real already.

 

(I would doubt that will be the pricing structure and would guess as with the Scarlet the release price and the proposed price will be different. I'm hoping not quite as different.)

Link to comment
Share on other sites

Yeah, clearly Dan has been using it, but we need much more footage. I cannot find anything from the 16 sized camera. For the 8, there is very little.

 

The original price for the super35 sized camera was 8 grand. Competitive if it does what they say it will do. The super8 is supposed to be around 2 grand but the lenses are the issue on that one. What to use and where to find? the super16 version will have far greater lens choices.

Link to comment
Share on other sites

  • 3 weeks later...

I have uploaded some more footage to Vimeo, here is the link to KineRAW-S8p lens test reel #B1 (B1B),

 

 

The shot of the bird with the Scnheider Optivaron 6-66mm f/1.8 macro zoom is not such a great example because that was shot at the lower end of the exposure range before sunrise in very bad sky light under trees. I plan on doing some more tests with the Super8 lenses later. The Super8 Fujica 5.5mm f/1.8 EBC was more than a bit soft in the corners for the 2592x1104 resolution mode, although it may cover the 2048x1152 or 1920x1080 sensor sizes, I'll need to do more testing.

 

The MSRP on the S8 was said to be about 1/4 the S35 as the parts are smaller and there is less processing in the S8 to make it more compact and lower power etc. Someone on DVXuser posted that the S35 is projected to be priced at $6000 to $8000 so that would make the S8 about $1500 to $2000 (FOB China I would think as the airmail is an additional cost.) At least those seem to be the targets so far.

 

I may have some uncompressed frames from this second reel to email back to interested people, there are some losses and differences in banding and noise perhaps on the Vimeo version so its best to look at the uncompressed BMP frames used to make the video before passing too much judgment about the resolution and image quality. For making a 35mm film print or DCP for projection in digital movie theaters the extra quality from the 100% uncompressed CinemaDNG may be visible and more film-like.

 

The camera is now shooting to a KineMAG-60 2.5" high speed SSD, that uses standard SATA II interface so I can plug it directly into my computer's SATA or ESATA port for file transfer, or download to my 500GB 2.5" notebook disk in the camera. The camera supports BOTH NFS+ and FAT32 for the HDD to dump shots to and lets you review the shot and just dump one at a time, or dump all of them and re-use the SSD for making more shots, WITHOUT needing a notebook computer or anything outside the camera because it has TWO drive slots in the back. The camera can re-format the SSD itself, although that is not needed to clear the disk and start shooting more shots, it deletes the 60GB in a short time. 60GB is about 9 minutes of non-stop record time at the 2.5K resolution mode, longer for the lower resolutions. You can also hot-swap the SSD for quick change if you have more than one of them, or you can put two in and switch the recording from one to the other from the menu (automatic switching may come later with a firmware update).

 

They added 200% zoom so that is great for getting the lens in focus for 2.5K resolution, so when you press the zoom button you get full view with guide lines, 100% zoom with area of interest border showing as a small green outline moving in a larger green outline for the full frame, and 200% also with the area of interest being able to be moved around. The zooms work with the camera in setup, while shooting, and while in playback so you can check the focus of shots made to see if they were in focus.

Link to comment
Share on other sites

I'm working on the anamorphic lens test reel today, we have shot some footage using an Elmo Scope-II adapter made for Super8 cameras with various backup lenses like an Angenieux 26mm f/0.95 and Kern Switar 50mm f/1.4. I may have that up on Vimeo in several weeks. The camera shoots anamorphic mode at 1828x1556 which is the exact resolution used to scan 35mm movie film negatives for "Panavision " format on the Kodak Cineon film scanner, and as used for output on the Celco type film recorders, many of the wide screen movies you have might have seen were processed through a DI using that pixel resolution. So far we have been getting some great anamorphic flare and that unique look that only anamorphic lenses can give on the 2.35:1 "Cinemascope " format. For output to a 2.39:1 DCP for projection in a digital movie theater the height can be cropped a bit, that also gives a bit of room to crop in case the mic boom dips into frame, and shooting 2.35:1 gives more image area on the DVD and SD versions.

Link to comment
Share on other sites

I'm all aboard this thing. Maybe. If it has rolling shutter issues and so far it does, as discussed, I will have to pay attention to price point. If they go with a mechanical shutter,, price isn't as important. What I can say is this... the images you are getting look very good. I really like the look.

Link to comment
Share on other sites

If it has rolling shutter issues and so far it does, as discussed, I will have to pay attention to price point. If they go with a mechanical shutter,, price isn't as important. What I can say is this... the images you are getting look very good. I really like the look.

 

Yes, I like the look also, with time and using lights and actors I should be able to get better looking results.

 

As for the rolling shutter, there may be a solution to that as there are other sensors that can use, and also the mechanical shutter can be added, the combination of both those changes could result in a camera option with ZERO rolling shutter and no vertical streaks like CCD have, very much like using a Super8 movie camera.

 

If you are interested you should talk to them about your reservations for purchase because of the needed changes to the design, so they understand its impact on the world wide filmmaking community. My voice alone may not be enough to get that into the design.

 

If they still refuse to make the needed changes, like having interchangeable sensor modules so you can pick the look and image area you want for a particular shot, I think the camera is usable if you limit the shooting to more static shots since the temporal noise reduction works best on static camera shots anyway. For most of my projects moving camera is not absolutely needed in 'every' shot.

Link to comment
Share on other sites

Unless one is going to project Super8 film from the camera original or print, it gets viewed in digital form.

 

So in that digital form, the result is a digital image.

 

That digital image is a conversion of the light reflected from the subject into various color values in the end use pixels.

 

What advantages there are to capture on film vs. sensor depend on two factors:

 

1) The handling of the camera when shooting.

 

2) The appearance of the end results on the movie screen which will become increasingly from a DCP rather than a 35mm print.

 

Both issues can be brought closer to the original film camera experience if both the digital camera and image processing are planed to emulate both experiences. With True RAW full bandwidth recording you are closer to the uncompressed nature of shooting motion picture film.

 

I plan on doing more testing with the Super8 C-mount lenses. Right now I'm working on an anamorphic reel shot using a Super8 format anamorphic adapter the ElmoScope-II. The results with processing in my de-Bayer software have some corrector of the look of shooting with film due to the ability to apply crossover and heavy S-curve to the 12bit data. I may have that posted on Vimeo in several weeks, some additional shots are needed for the edited portion.

 

==

 

I don't know if they would get that far, but possible additions would be a mechanical shutter maybe spinning or reciprocating mirror so that an optical reflex could be added in addition to the HDMI video monitoring option. People interested in such additions should contact them to show their interest.

Edited by Dan Hudgins
Link to comment
Share on other sites

Is the point of this camera to emulate S8 or is it a digital camera which just happens to have S8 sized sensor?

 

Emulating S8 (to me at least) is not that interesting. It can already be done – rather poorly - by any number of cheap HD cameras and post processing or by ACTUALLY SHOOTING Super 8 :D

 

A high resolution digital cinema camera with a s8 or s16 sized sensor at a low cost is interesting, particularly now that the Scarlet and most of the other digital cinema cameras have 35mm sized sensors.

Link to comment
Share on other sites

It looks like they are trying to get maximum picture quality on the small sized censor and to me, they may be on the right track. The footage looks gorgeous. I am wanting one of these.... but only if there is no rolling shutter. That's the problem. Rolling shutter = kiss of death. I'd gladly pay an extra 500 to a grand if they had a mechanical shutter instead of the usual DSLR-like crap. Enough already.

Link to comment
Share on other sites

From my perspective is to achieve results on par with 35mm Techniscope like was used on American Graffiti and THX1138 as well as many other successful movies, from a compact camera system.

 

The small size of the Super8 Zoom lenses relative to ones of the same speed and zoom range makes them easier to be mobile with, and they have the quality of cinema glass since they were made by cinema lens makers.

 

The cost of the S8 sensor is much lower than for S16 and S35 so that keeps the camera cost down.

 

It would be possible to make a buffered USB camera for maybe $250 to $499 that would shoot the same image quality as the KineRAW-S8p but with the shot length limited to the buffer size. The advantage of the KineRAW-S8p is the self contained recording to standard SSD so you don't need a notebook computer to use it in the field.

 

There are other low cost sensor options and other ways to make the camera smaller. The goal of Kinefinity seems to have been to make a camera with an interface similar to other Digital Cinema Camera makers with on-screen waveforms, histograms, zoom and such, as well as HDMI monitoring. Those extras make the camera much more usable for filmmaking, but add some cost to the design and increase the camera size. None the less, there are no similar cameras you can purchase the record to CinemaDNG frames 12bit uncompressed data, Acan dII is only 1920x1092 pixels, and the 2.5K oversample gets a bit closer to Techniscope goal as it can reduce some of the de-Bayer artifacts.

 

The current rolling shutter may be an issue for some kinds of filmmaking, and if Kinefinity wanted to they could add additional improvements such as a changeable sensor module with some having a mechanical shutter. The S35 model does not have a mechanical shutter as far as I know, so its actually fewer pixels resolution and 4x the cost, but does take high cost and larger PL lenses.

 

If you are out in the field, packing around a S35 outfit and many large and heavy lenses, even DSLR lenses, you total weight for the outfit are probably going to be 2x to 10x what a S8 outfit would weigh. And after three for four mouths of packing the equipment around by yourself, the importance of that may take on a greater meaning of 'Super8' Digital Cinema Camera.

 

I still need to do some more tests on the Super8 C-mount Zoom lenses, I have a Schneider 6-70 f/1.4 I have not tried yet.

 

I processing frames from the Anamorphic lens tests today, there is some more to shoot with that before I can post it on Vimeo, it is planed to have sync sound and a mixed sound track.

Link to comment
Share on other sites

If you work out the pixel sizes, the Acam dII crops to 1920x1080, so

 

1920x1080 = 2073600

 

The KineRAW-S8p shoots 2592x1104 so,

 

2592x1104 = 2861568

 

So that seems,

 

2861568/2073600 = 138% as many pixels

 

If you have thoughts about the various camera options you should share those with Kinefinity so they can get some idea what people think.

Link to comment
Share on other sites

  • 2 weeks later...

If you work out the pixel sizes, the Acam dII crops to 1920x1080, so

 

1920x1080 = 2073600

 

The KineRAW-S8p shoots 2592x1104 so,

 

2592x1104 = 2861568

 

So that seems,

 

2861568/2073600 = 138% as many pixels

 

If you have thoughts about the various camera options you should share those with Kinefinity so they can get some idea what people think.

 

This looks like pretty much the same idea approximately Super 8 sized chip in the current camera module. C-mount lenses. (Another camera module coming soon that is 4k and frame rates up to 300fps, global shutter too.)

Link to comment
Share on other sites

  • 3 weeks later...

 

Is the Apertus camera still planning or using its 8bit JP4 block compressed file format?

 

Apertus seems to have started out its design as a compressed IP camera, not a from the ground up Digital Cinema Camera circuit design.

 

8bit data cannot be graded hard without introducing Histogram gaps developing, that is a problem shared with HDSLR. Adding noise to 8bit compressed data is not the same as recording actual sensor data as it comes out of the sensor boards ADC at full bit range.

 

The KineRAW cameras record 100% uncompressed True RAW to DNG frame files, so you get all the sensor data without block artifacts in the recorded data.

 

You can read more about progress on the KineRAW-S35 camera here,

 

http://nofilmschool.com/2012/02/latest-chinese-kineraw-digital-cinema-camera/comment-page-1/#comment-141285

 

http://nofilmschool.com/2012/02/latest-chinese-kineraw-digital-cinema-camera/

 

There are some renders.

 

As I understand it they have two working prototypes of the S35 back from the factory since the Chinese new year, and have been able to shoot both CinemaDNG [uncompressed] and Cineform [wavelet compressed like RED's REDCODE ]. Test shooting is planed for the near future.

Edited by Dan Hudgins
Link to comment
Share on other sites

  • 2 weeks later...

First, I'm surprised at the hostility shown to Dan. In a supposedly "pro" group calling someone a "shill" is out of bounds.

 

Secondly, I'm equally surprised that so many don't "grock' what these cameras are about. Of course, one can buy a consumer camcorder or a horrible-handling DSLR that have better specs and provide a "clearer" image. But, if you want true ring driven focus that holds and ring-driven aperture, plus a smooth manual zoom -- for little cost -- you are out of luck. (I've had a Beaulieu and Bolex H8 and H16 -- and today's camera are simply not not FUN to shoot with!

 

Of course, one could shoot film. process it, and then scan it to HD video. But, that seems pointless if one can get a digital FullHD file by using one's computer.

 

 

The why seems obvious to me. Had the Scarlet arrived at $3000 the S8 and S16 would have no chance. But, at the $3000 to $6000 price range there are no 4:4:4 RAW RGB cameras that record 12-bits. One has to bump to an F3, add S-log firmware, and buy a recorder. Not an economic proposition for MANY folks. (If it is for you, I'm wondering what you are doing reading and replying to posts about Super8!)

 

So assuming current readers really ARE interested in DIGITAL S8 and/or DIGITAL S16 -- here are some questions for Dan.

 

1) The chip aspect-ratio is about 2.35:1 and Dan says is slightly wider than S8 film. What are its size and diameter? And, how does it work with S8 lenses that were built for 1.33:1?

 

2) Does it record audio? Details?

 

3) What is the recorded data-rate of CinemaDNG with audio? How long does it record? Hopefully, at least 15 minutes to a disk!

 

4) Does it record TIFF images into CinemaDNG?

 

5) Can you describe the recording system. Is it built in? What drives are required?

 

6) For Mac users, your software may not run well under BOOTCAMP with Windows XP. But, Adobe has a CinemaDNG plug-in for Premiere. Does it deBayer? If not, any idea how we'll get deBayering and Image adjustments in Premiere?

 

7) Assuming a software 75% deBayer efficiency, 2592 yields about 1944 LUMA PIXELS. An image scale of 0.74 from 2592 to 1920 means the 1944 becomes 1440. The question I have is HOW will the 1105 be mapped to 1080. My guess is that it will be letterboxed info FullHD. If so, then 1440 will be divided, I think, by 2.35 which reduces H. resolution to 612 TVl/ph. Before folks scream that's not "HD" many big chip cameras at the same price point only offer about 750TVl/ph.

 

8) I swear I saw a pix of the S8, but can't find it now. Can you snap some pix of all sides of yours and post.

 

Thank you.

 

Steve

Link to comment
Share on other sites

  • 3 weeks later...

I posted the third test reel for the KineRAW-S8p last night.

 

It stutters a bit from the Vimeo playback on my Brother's computer, and there are some spots left behind in the scrolling titles from the compression it seems, but you get the general idea of what shooting at f/0.95 with a 2:1 squeeze lens gets you,

 

 

Hi Steve,

 

I will try to answer your questions,

 

 

Quote: [1) The chip aspect-ratio is about 2.35:1 and Dan says is slightly wider than S8 film. What are its size and diameter? And, how does it work with S8 lenses that were built for 1.33:1?]

 

The KineRAW-S8p supports many resolutions, aspect ratios, and frame rates.

 

I was shooting 2592x1104x12bit @24fps (it does 23.976 and 25 currently at that resolution also) because that gives the largest reduction ratio and because I am interested in shooting 2.35:1 and 2.39:1 for projection wide screen in movie theaters for my own projects, for DCP than means 2048x858 reduction and for 35mm filmout 1828x1556.

 

Quote: [2) Does it record audio? Details?]

 

There are two channels (stereo) 48000 samples per second 16bit PCM WAV uncompressed audio, the input can use used for mic level with plug-in-power for Sony type mics that have a built in J-FET preamp, or you can use an isolation cable and plug in line level signals, the viewfinder has two VU meters bar type with green, yellow and red colors, and the setup menu has three gain controls for the input to adjust the gain over a wide range and add mic boost for extra gain. You can hear some audio recorded in the camera, although there is some SPL issue with the mic used on some shots it seems. Some of that audio was recorded using an external mini-disk and synced sing the camera auto-slate beep from the internal Piezo, the internal tracks also get a wired beep on one channel, and there is an external line level beep on the front multi pin connector so you can wire the auto-slate beep to an external recorder or mic mixer, the head slate beep is 1000Hz and the tail slate Beep is 2000Hz so you can tell them apart by ear or by looking at the waveform when editing the sound. The camera also have SMPTE time code for using a time code slate, and internal slate data that gets saved in the DNG meta-data my program can burn that into the workprint frames as is shown in the above video for doing an off-line edit, and then later conform in a DI system, in the case of the above video my freeish DI system was used.

 

Quote: [3) What is the recorded data-rate of CinemaDNG with audio? How long does it record? Hopefully, at least 15 minutes to a disk!]

 

At the 2592x1104 resolution and 48000 s/s stereo 16bit audio, I get a bit over 9 minutes on a 60GB SSD non-stop recording. The camera has two slots so I have a 500GB 2.5" notebook drive in the second slot to dump shots to when I'm setting up the next setup, so that way you can never fill the SSD, the SSD quick hot-swap, so you can change fast, and they are standard SATA interface so you can download fast on a portable computer if you need to shoot sort of non-stop all day. With a larger SSD the non-stop single take should be able to be longer, in a few years 1TB SSD should be common maybe. The KineRAW-S16 and KineRAW-S35 are to have Cineform recording option that may work direct to a 1TB HDD in the camera and that should give very long non-stop single takes depending on the compression ratio used.

 

Quote: [4) Does it record TIFF images into CinemaDNG?]

 

The camera records DNG frames with 12bit Sensor data in them, my de-Bayer program DANCINEC.EXE can convert those color corrected to TIF 48bpp, DPX 48bpp or 30bpp, CIN 30bpp, and BMP 24bpp for color and black and white and 8bpp for monochrome.

 

Many other editing and DI systems support DNG conversion, and Adobe has a free program called DNG_validate.exe that converts DNG to TIF frames.

 

Quote: [5) Can you describe the recording system. Is it built in? What drives are required?]

 

The KineRAW-S8p is totally self contained (other than the battery) and can record to one of its two SSD SATA slots, the camera can record to one slot while the disk in the other slot is used to dump shots to, I have been using a 60GB SSD (KineMAG-60 in the left slot, and a 500GB HDD notebook drive in the right slot. I have a spare 60GB SSD also to swap for fast change so that gives me 18 to 19 minutes of shooting before I need to dump anything, I'm testing two sample drives for them to see if there are any issues, so far the 60GB drives are recording glitch free non-stop without issues.

 

 

Quote: [6) For Mac users, your software may not run well under BOOTCAMP with Windows XP. But, Adobe has a CinemaDNG plug-in for Premiere. Does it deBayer? If not, any idea how we'll get deBayering and Image adjustments in Premiere?]

 

People tell me DANCINEC.EXE runs under bootcamp, it has been tested under Windows ME through Windows 7 64bit Pro.

 

You can see an Acam dII frame converted by someone here,

 

---------------------------

 

40200001.BMP is on page,

 

http://www.ikonoskop.com/forum/?action=listPosts&forumID=8&parentID=1322

 

at link,

 

http://www.ikonoskop.com/begood/forum_attach.php?14159711654f0f0ac899cbb8.52359071

 

------------------------------

 

Since the KineRAW-S8p has 16:9 oversample more pixels can mean less de-Bayer artifacts in the results, maybe.

 

Quote: [7) Assuming a software 75% deBayer efficiency, 2592 yields about 1944 LUMA PIXELS. An image scale of 0.74 from 2592 to 1920 means the 1944 becomes 1440. The question I have is HOW will the 1105 be mapped to 1080. My guess is that it will be letterboxed info FullHD. If so, then 1440 will be divided, I think, by 2.35 which reduces H. resolution to 612 TVl/ph. Before folks scream that's not "HD" many big chip cameras at the same price point only offer about 750TVl/ph.]

 

The letter boxed image are 2592x1104 reduced to 1920x817. Its done with floating point math so the ratio does not need to be exact like some hardware scalers.

 

Other resolutions like 2400x1200 and such scale by various ratios in real numbers.

 

Please look at the three videos on my vimeo page to get an idea, you can email me to get some uncompressed BMP frames used to make those videos since some loss of fine detail can happen with compression for showing on Vimeo.

 

Email me at tempnulbox (at) yahoo (dot) com and put "cinematography.com user wants uncompressed frames" so I know your email is not spam, thanks.

 

Quote [8) I swear I saw a pix of the S8, but can't find it now. Can you snap some pix of all sides of yours and post.]

 

I'll see if I can post one here, its of the prototype, the production model may look different. The camera is narrow so that two can be in sync for shooting 3D stereoscopic movies and TV shows etc.

post-47872-0-08194000-1332372892.jpg

post-47872-0-12152500-1332372906.jpg

Link to comment
Share on other sites

About the image size of Super8 lenses, the KineRAW-S8p has lens mount rings for C-mount, CS-mount, and D mount movie and megapixel lenses.

 

The image diagonal is the important dimension and the aspect ratio does not matter, lenses are rated by their image circle not any rectangle shape, so as long as the image area on the sensor is smaller in diagonal than the image circle and the lens is corrected out to the limit of its image circle then you get a sharp image.

 

I did notice that the Fujinon 5.5mm f/1.8 Super8 lens is not very sharp in the corners although it covers the image area with light. The Angenieux 6-80mm f/1.2 seems to cover more than the needed image area, although I need to do more testing, I also have Schneider Super8 zoom lenses 6-66mm f/1.8 with macro-zoom at all focal lengths, and 6-70mm f/1.4 with just macro at wide end to test more, there is a shot made with the Optivaron 6-66mm f/1.8 in the B1 video, but the light was very low so I under compensated for the OLPF filter blur to not pick up too much grain, that lens is quite sharp in the center of the image. The camera can shoot 4:3 using 1920x1440 that can be pillar boxed into 1920x1080 to get more sharpness, that 4:3 ratio would be more like shooting with a Super8 film cameras, perhaps.

 

8mm film (from web), 4.8mmx3.5mm, 5.5mm diagonal

 

1/3" sensor (from web), 4.8mmx3.6mm, 6mm diagonal

 

Super8 film (from web), 5.30mmx4.0132mm, 6.647 diagonal

 

1/2.5" (from pdf), 5.7mmx4.28mm, 7.124 diagonal

 

1/2.5" 2592x1102, 5.7mmx2.426, 6.19mm diagonal (within C mount Super8 lens area)

 

1/2.5" 2048x1152, 4.5mmx2.536mm, 5.165mm diagonal (within D mount 8mm lens area and CS mount 1/3" lens)

 

2/3" lens (from web) 8.8mmx6.6mm, 11mm diagonal (for 4:3 ratio)

 

16mm (from web)=10.4mmx7.5mm, 12.82mm diagonal

 

Super16 (from web)=12.4mmx7.5mm, 14.49 diagonal (<seems right)

 

ARRI SR3 Super16 image (measured off negative) about 0.486"x.295" or 12.34mmx7.493mm, 14.44mm diagonal (close to values from web of 12.4mmx7.5mm, ARRI SR3 has thick frame line and clear band on outside edge of sound area on film).

 

KAI-02150 sensor (from pdf) 10.56mmx5.94mm, 12.12mm digonal

 

Scarlet 2/3" sensor (from RED.COM) 10.1mmx5.35mm, 11.43 diagonal

 

1" sensor (from web) 12.8mmx9.6mm, 16mm diagonal

 

===

 

This link gives,

 

8mm film as 4.9 mmx3.7mm, 6.11 mm diagonal

Super8 as 6.2 mmx4.2mm, 7.5 mm diagonal

 

http://en.wikipedia.org/wiki/Normal_lens

 

==

 

For high speed shooting like at 220fps, Regular 8mm D-mount lenses may give better sharpness on the smaller sensor area.

Link to comment
Share on other sites

 

8mm film (from web), 4.8mmx3.5mm, 5.5mm diagonal

 

1/3" sensor (from web), 4.8mmx3.6mm, 6mm diagonal

 

Super8 film (from web), 5.30mmx4.0132mm, 6.647 diagonal

 

1/2.5" (from pdf), 5.7mmx4.28mm, 7.124 diagonal

 

1/2.5" 2592x1102, 5.7mmx2.426, 6.19mm diagonal (within C mount Super8 lens area)

 

1/2.5" 2048x1152, 4.5mmx2.536mm, 5.165mm diagonal (within D mount 8mm lens area and CS mount 1/3" lens)

 

The site destroyed my reply -- so once again:

 

1) The S8's sensor is 1/2.5"?

 

2) Is the sensor size 5.7mmx2.426, 6.19mm diagonal?

 

3) A Super8 lens throws a 5.30mmx4.0132mm, 6.647 diagonal image on the sensor. Which means some of the image is lost going from 6.65 to 6.19? Thus, there is a crop factor. Correct? What is the factor?

 

4) The S8's sensor is 2592x1102? Or, 2592x1152? I ask because how can a 2592x1102 sensor shoot 2048x1152?

 

5) When shooting 2048x1102 -- your numbers show the capture area to be much smaller at only 4.5mmx2.536mm, 5.165mm diagonal. But, if the S8 is oversampling 2592 to 2048 then the actual capture size remains the same. Correct?

 

6) Likewise, when shooting 1920x1080 -- if the S8 is oversampling 2592 to 1920 and 1152/1102 to 1080 -- then the actual capture size remains the same. Correct?

 

 

We need a segment of cars moving at various speeds past the S8 to see the amount of rolling shutter. Right now the $3200 DigitalBolex with it's CCD chip is going to have a big advantage by offering what the S16 provides at half the price. Which may say the S8 needs to be much cheaper. :)

 

PS: will KineRAW be at NAB?

Edited by Steve Mullen
Link to comment
Share on other sites

Hi Steve,

 

I'll try to answer your questions:

 

Quote: [1) The S8's sensor is 1/2.5"?]

 

The readout time for full sensor is limited to less than 24fps, so only part of the sensor are is used for normal shooting, there are two modes 1fps and 6fps currently for full sensor, but others could be added as the sensor is programmable.

 

Quote: [2) Is the sensor size 5.7mmx2.426, 6.19mm diagonal?]

 

The sensor size for 2592x1104 is full width and part of the height, that gives CinemaScope Panavision Cineon pixel area 2.35:1 that can be cropped and reduced to 2048x858 for making DCP for projection in digital movie theatres, or letter boxed 1920x817 in 1920x1080 for Blu-ray use.

 

Quote: [3) A Super8 lens throws a 5.30mmx4.0132mm, 6.647 diagonal image on the sensor. Which means some of the image is lost going from 6.65 to 6.19? Thus, there is a crop factor. Correct? What is the factor?]

 

The horizontal angle of view is a bit WIDER than on Super8 when using 2592 sensor width, you could crop it to match Super8 width if you wanted to. Since you are inside the astigmatic circle the lens makes, in theory at least the image should be usable, but some lenses may have better periphery resolution than others. Camera lens makers may limit the real astigmatic circle to the Super8 PROJECTION aperture rather than the CAMERA aperture, although there is light in the corners of the frame, the resolution in lp/mm may be lower in the corners of the frame. I have a special sharpen filter in my de-Bayer program that can compensate for darkened and soft corners by lightening them and increasing the sharpens progressively going from the image center to the corner in anticipation of that issue.

 

The 1920x1440 shooting mode is closer to Super8 projection aperture plus screen over-projection and may show better over all resolution across the frame with the normal 4:3 Super8 aspect ratio. There are also over-sample 16:9 modes, and 2:1 mode each with slightly different image diagonals. 1280x720 is also supported at 59.94 frame rate, which may work with both Super8 and regular 8mm D-mount lenses.

 

You can work out the crop factor in the vertical, but you need to take into account the projector cut off, since even video transfer does not show the whole camera aperture perhaps, with the digital camera you get ALL the captured pixels on the computer screen without projector cut-off. So an exact crop factor cannot be found, since Super8 projectors had their reduced aperture and were over-projected on the screen so you would not see the fuzzy edge of the projector gate.

 

In this case the height is less and the width is more, but since the projector height is less than the camera height, the loss may not be noticeable and in fact you may end up with more in some cases.

 

Quote: [4) The S8's sensor is 2592x1102? Or, 2592x1152? I ask because how can a 2592x1102 sensor shoot 2048x1152?]

 

CMOS can window to any size that is an even number on the Bayer sensor, it has to be an even number because of the 4 pixel group that is two greens one red and one blue, other than that the sensor used can window to any size or aspect ration to maximize bandwidth, lots of modes are supported from full sensor down to smaller ones for 220fps. The zoom on the viewfinder does pan and tilt as well to look around the frame at 100% or 200% zoom to check for focus and other issues, both in setup and record and playback.

 

Quote: [5) When shooting 2048x1102 -- your numbers show the capture area to be much smaller at only 4.5mmx2.536mm, 5.165mm diagonal. But, if the S8 is oversampling 2592 to 2048 then the actual capture size remains the same. Correct?]

 

None of the current shooting modes use binning because that causes problems with the OLPF and image quality, so the pixels are 0.0022mm square and you can work out the sensor area from that.

 

The sensor area is in ratio to the pixel resolution, but the diagonal is in ratio to the aspect ratio and the total width.

 

Quote: [6) Likewise, when shooting 1920x1080 -- if the S8 is oversampling 2592 to 1920 and 1152/1102 to 1080 -- then the actual capture size remains the same. Correct?]

 

If you are shooting 1920x1080 then its 1:1 pixel with no oversample, likewise for 1280x720 to 1280x720 those are 1:1 pixel ratio modes.

 

Shooting 2048x1080 for DCP use gives you two crop choices at 1:1 pixel 2048x858 for 2.39:1 aspect ratio and 1998x1080 for 1.85:1 aspect ratio, both without oversample.

 

If you shoot 2048x1152 and reduce that to 1920x1080 then you have oversample, of if you shoot 2400x1200 and letter box that to 1920x1080 then you have oversample.

 

There is a higher than 2048x1152 16:9 oversample ratio, but I forget its exact dimensions right now, its between 2048 and 2400.

 

2048x1152 can be used 1:1 pixel for transfer to 35mm movie film for 1.85:1 masking with a bit more height to prevent frame lines from showing as the projector bobs the image up and down. 2K 16:9 is common for that use.

 

If you shoot 2592x1104 and reduce it to 2048x858 for DCP projection in digital move theaters than you have oversample.

 

 

Quote: [We need a segment of cars moving at various speeds past the S8 to see the amount of rolling shutter. Right now the $3200 DigitalBolex with it's CCD chip is going to have a big advantage by offering what the S16 provides at half the price. Which may say the S8 needs to be much cheaper. :)]

 

I have a shot of cars moving in the B1 video on my Vimeo page.

 

I just purchased a 3.5mm f/2.4 Kowa megapixel lens to shoot some hand held tests with.

 

There is a way for Kinefinity to make a Super8 camera without any rolling shutter problems, and I have talked to them about having that as an option. You can email them and tell them you are interested in such an option.

 

The prototype KineRAW-S8p has had its shutter values adjusted to reduce the rolling shutter to some extent and supports 2.35:1 at 48fps and 72fps, shooting at 72fps the hand held and pan shots would show less rolling shutter and using a wide angle lens is a help in that regard as well. I have support to skip frame these modes to get 24fps with normal motion, they can also be used for slow motion cinematography, or used as over-crank to do speed ramps in software via variable interpolation.

 

The target price for the S8 was originally lower, but the costs vs. sales volume issues make some factors that drive their guess of the MSRP higher, if the camera sales volume works out better than expected, then maybe they can get a volume discount. None the less, for a few hundred more live waveforms and SMPTE time code and 1280x720 HDMI monitoring make it a nice camera to work with.

 

As for the digital Bolex it is not without its own issues:

 

1) The dynamic range of the Kodak sensor seems to be about 1 T/ stop less than the S8p's sensor by looking at the mfg spec sheets, that may not seem like much, but as I understand it the Kodak sensor uses four readouts and can show splits seams on underexposed shots, that FPN is quite visible on Acam dII test DNG I have processed and lifted the shadows to increase the ISO. The KineRAW-S8p sensor uses only ONE readout had I have been successful in doing FPN subtraction on that to give a seamless background noise field, that works well for boosting the ISO to grade up shots that are under-exposed.

 

2) CCD sensors can show strong vertical streaks on highlights that CMOS sensors are totally free of in normal use. Those vertical streaks can destroy the suspension of disbelief while watching a movie as they scream "diagonal camera" rather than looking like movie film as the CMOS camera does.

 

3) The larger sensor pixels may not be such a good idea as they may cause soft corners in using some regular 16mm movie camera lenses, and may prevent the use of common 2/3" video lenses. More so at the digital Bolex 2048x1152 resolution. Its a case of more may not be better. The 1" sensor size is a kind of 'no-man's-land" for movie camera lenses, and the choice of 1" video lenses may not give the choice you want. I have been using regular 16mm movie lenses on the KineRAW-S8p with success in particular the Kern Switar and Schneider NON-RX lenses made for 16mm movie cameras and for 2/3" video format cameras. Also 1/2" video camera lenses like the 3.5mm f/2.4 Kowa mega-pixel lens would show less soft corners because the sensor area and image diagonal are a bit smaller than what the lens can cover, so the used image area is within the lens's astigmatic circle.

 

Quote: [PS: will KineRAW be at NAB?]

 

I asked them but did not get a clear answer as far as I can remember. You can ask them.

 

Tanks for your questions, hopefully there will be some improvements to the KineRAW-S8 to address the points you have made before they go in sale, but that is up to them I can only offer suggestions.

 

You can check my Vimeo page in maybe 6 weeks to see if I have added some more videos.

Link to comment
Share on other sites

Hi Dan,

 

I'll post my reply for others to read.

 

Quote: [1) The S8's sensor is 1/2.5"?]

 

The readout time for full sensor is limited to less than 24fps, so only part of the sensor are is used for normal shooting.

 

Quote: [2) Is the sensor size 5.7mmx2.426, 6.19mm diagonal?]

 

The sensor size for 2592x1104 is full width and part of the height, that gives CinemaScope Panavision Cineon pixel area 2.35:1 that can be cropped and reduced to 2048x858 for making DCP for projection in digital movie theatres, or letter boxed 1920x817 in 1920x1080 for Blu-ray use.

 

Reading your comments, my sense is that aiming for 2:35:1 (2592x1104) pushes the sensor and lens requirements/questions as well as needing cropping in post — something video folks don't want to bother about . Thus, it would seem easier to use one of common 2K resolutions: 2048x1556, (4:3) 2048x1152 (16:9), 2048x1080 (1.90:1), 2048x1024 (2:1).

 

When windowing the sensor for 2K, I'm wondering if the 1556 and/or 1152 fall outside the typical Super8 lenses angle of view?

 

If so -- then the obvious EZ capture option is 2048x1080 (1.90:1) because:

It should safely fall within a typical Super8 lens H. and V. angle of view — meaning less worry about corner resolution.

The 1.90:1 aspect-ratio matches the common Super35 sensor aspect-ratio.

A 1.90:1 aspect-ratio is slightly letterboxed on 16:9 devices -- giving a sense of widescreen "film" without the extreme bars of 2:35:1.

I think I'm correct that for a 1920x1080 frame, the vertical dimension must be downscaled by 0.067 to about 1012.

"Shooting 2048x1080 gives you two crop choices; 2048x858 for 2.39:1 aspect ratio and 1998x1080 for 1.85:1 aspect ratio."

 

"The sensor area is in ratio to the pixel resolution, but the diagonal is in ratio to the aspect ratio and the total width."

 

I'm lost.

 

None of the current shooting modes use binning because that causes problems with the OLPF and image quality, so the pixels are 0.0022mm square and you can work out the sensor area from that.

 

When I said "oversample" I didn't mean binning. But, your comments lead to two questions:

When you use the word "oversample" do you mean in-camera interpolation from the maximum capture window-size to the recording frame-ize OR do mean the scaling in post?

It sound like you mean a post operation. So am I correct that each capture size must create different angle-of view when shooting?

 

The prototype KineRAW-S8p has had its shutter values adjusted to reduce the rolling shutter to some extent.

 

How can shutter values be adjusted and yet deliver the expected motion blur?

 

The KineRAW-S8p sensor uses only ONE readout had I have been successful in doing FPN subtraction on that to give a seamless background noise field, that works well for boosting the ISO to grade up shots that are under-exposed.

 

Why do you need to do this if there is only one read-out? (Is the read-out analog or digital?)

 

The larger sensor pixels may not be such a good idea as they may cause soft corners in using some regular 16mm movie camera lenses, and may prevent the use of common 2/3" video lenses. More so at the digital Bolex 2048x1152 resolution. I have been using regular 16mm movie lenses on the KineRAW-S8p with success in particular the Kern Switar and Schneider NON-RX lenses made for 16mm movie cameras and for 2/3" video format cameras.

 

I agree that DB is going to have to deal with many questions about WHICH lenses can be used. Which makes me think that using regular 16mm C-mount lenses on the S8 will throw a much larger than image on the sensor making 2:35:1 (2592x1104) very capture EZ. Correct?

 

 

Quote: [PS: will KineRAW be at NAB?]

 

I'm going to contact them this week. I live in Vegas. Are you coming?

 

Despite the hype about the DB — I have my doubts they can deliver their fall units at the prices they think. Frankly, I would think it would be foolish to do so.

 

Thank you.

 

Steve

Link to comment
Share on other sites

Hi Steve.

 

The final specs of the resolutions and frame rates are not set yet, what I have been talking about are some that the prototye KineRAW-S8p supports.

 

 

Quote: [Reading your comments, my sense is that aiming for 2:35:1 (2592x1104) pushes the sensor and lens requirements/questions as well as needing cropping in post — something video folks don't want to bother about . Thus, it would seem easier to use one of common 2K resolutions: 2048x1556, (4:3) 2048x1152 (16:9), 2048x1080 (1.90:1), 2048x1024 (2:1).

 

When windowing the sensor for 2K, I'm wondering if the 1556 and/or 1152 fall outside the typical Super8 lenses angle of view?

 

If so -- then the obvious EZ capture option is 2048x1080 (1.90:1) because:

It should safely fall within a typical Super8 lens H. and V. angle of view — meaning less worry about corner resolution.

The 1.90:1 aspect-ratio matches the common Super35 sensor aspect-ratio.

A 1.90:1 aspect-ratio is slightly letterboxed on 16:9 devices -- giving a sense of widescreen "film" without the extreme bars of 2:35:1.

I think I'm correct that for a 1920x1080 frame, the vertical dimension must be downscaled by 0.067 to about 1012.

"Shooting 2048x1080 gives you two crop choices; 2048x858 for 2.39:1 aspect ratio and 1998x1080 for 1.85:1 aspect ratio."]

 

I turned the camera on and you get this progression of resolutions at sound speeds:

 

2592x1104 for down sample to 2048x858 for DCP or 1920x817 for letterbox HD 2.35:1 or 2.39:1 aspect ratio.

 

2400x1200 for down sample to letter boxed HD aspect ratio 2:1 for 1920x960 in 1920x1080

 

2228x1252 for down sample to full screen HD aspect ratio 16:9 or down sample and crop to 1998x1080 for DCP 1.85:1

 

2048x1152 for use 1:1 for 2K 35mm film prints 16:9 for projection mask 1.85:1

 

2048x1080 for use 1:1 DCP use crop 2048x858 and 1998x1080

 

1920x1440 for 4:3 down sample to 1828x1371 for 35mm prints or to pillar box 1440x1080 in 1920x1080

 

1920x1080 for 1:1 HD use, or DCP use within 1998x1080 pillar boxed as is done because of over projection anyway.

 

1828x1556 for 1:1 use for 2.35:1 35mm film prints Anamorphic mode

 

For full screen 16:9 the best mode would be 2228x1252 which as an area of:

 

2228x0.0022 = 4.90mm

1252x0.0022 = 2.75mm

 

That falls within the projection aperture of Super8 camera lenses so the corners of the frame should be less soft focus. But it gives some over-sample to help reduce de-Bayer artifacts and OLPF blur as well as reduce noise and increase tonal detail.

 

As I mentioned normal Super8 projection cut-off included over projection of the movie screen in addition to the reduced gate in the projector, that is also true of 35mm projection, just to maybe a lesser degree.

 

If you want 4:3 aspect ratio, then using 1920x1440 reduced to 1440x1080 pillar boxed in HD 1920x1080 would be the best option since there is reduction on both axis by a notable amount to improve the image quality.

 

1920x0.0022 = 4.22mm

1440x0.0022 = 3.17mm

 

As noted since ALL of the digital image shows on the computer screen or digital monitor, you don't really want to use the maximum area of the Super8 lens's coverage since the corners will show some rainbow and blur, its better to crop the lens's coverage down to get the sharper part that would be what was left on the movie screen after projection cut-off.

 

For use with very good regular 16mm movie lenses and 2/3" lenses of high quality, these image areas fall within the 0.707 zone where the astigmatism is best corrected, perhaps, so very soft corners would not be an issue even at full lens iris opening. That is important since 16mm lenses are not really meant for 2.5K resolution in their image extremities in the first place so using the center of the image gives more uniform results at resolutions from HD to 2.5K.

 

The Angenieux 6-80mm f/1.2 seems to cover all of the supported resolutions with its image circle varying in size outside that area depending on the focal length and focus distance selected. I need to shoot some tests and process them to check for what the down-sampled HD results look like, since the camera's video out is 1280x720 and everything looks better at that resolution on a 7" LCD than at higher resolutions on a larger monitor.

 

 

 

Quote: ["The sensor area is in ratio to the pixel resolution, but the diagonal is in ratio to the aspect ratio and the total width."

 

I'm lost.]

 

When checking to see if a lens will cover, its the image diagonal and not its width or height alone that matter.

 

The lens forms a circle not a rectangle. Within that circle is the astigmatic circle of sharp resolution that can be smaller than the circle of illumination that is light but without sharp detail. In other words, you can get light or dark corners, and you can also get sharp light corners of fuzzy light corners, depending on the lens used and the relationship of the image rectangles diagonal to the lens image circle diameter.

 

Wide screen ratios can have the same or smaller image diagonal, but be wider in width, so give a wider angle of view but less image diagonal using the same lens, so you can get sharper corners and wider angles in a wide screen image than a square image or a 4:3 image perhaps, as its the image diagonal that matters if the lens is centered on the sensor image area.

 

 

Quote: [When I said "oversample" I didn't mean binning. But, your comments lead to two questions:

When you use the word "oversample" do you mean in-camera interpolation from the maximum capture window-size to the recording frame-ize OR do mean the scaling in post?

It sound like you mean a post operation. So am I correct that each capture size must create different angle-of view when shooting?]

 

The capture size does impact the angle of view horizontally and vertically, and the anamorphic modes depend on what the horizontal expansion factor is of the anamorphic lens,

 

you can use 1920x1440 to shoot 16:9 using a 1.33x anamorphic lens

 

you can use 2228x1252 to shoot 2.35:1 using a 1.33x anamorphic lens

 

The reduction to final use size, which can be DVD size or HD or DCP etc. is done in post not the camera, the camera records True RAW sensor data and does not process it.

 

My de-Bayer program includes re-sample and boxing options in the one-step processing that also includes grading to final grade and use size. Other work-flows have their own options.

 

 

 

Quote: [How can shutter values be adjusted and yet deliver the expected motion blur?]

 

The readout time and the light storage time are not related, there is a dwell time after the readout is done, in some cameras like Alexa and RED ONE the readout time is very quick and in other cameras like HDSLR the readout time is slower and takes most of the frame rate cycle time.

 

If you readout in half the frame rate cycle time, then you get a rolling shutter like shooting 48fps when you are shooting 24fps, and so the jell-o and skew are half. The exact details of the rolling shutter ratios are not final in the KineRAW-S8p and I hope to test the 48fps and 72fps shooting modes and to post some videos when I get time. The final camera may be quite different from the prototype in regard to these issues, for some kinds of film-making the image quality of shooting 2.5K may be more important. A Digital Cinema Camera is mostly for shooting 24.000fps for going to DCP or 35mm film print, for other kinds of production like making TV shows they may have some uses, but there is no one size fits all best solution in an under $2000 digital cinema camera yet, so is for people that have some special use for a True RAW DNG recording camera I would think. If you need to shoot on a po-go stick for one or two shots, you can use the higher fps modes, or another camera maybe, as there is no reason not to use more than one camera for a feature film where effects shots are required for a single days shooting on a long production schedule, just as Mitchell High Speed cameras were used to shoot shots of cars going off cliffs in slow motion or for stop motion work.

 

 

 

Quote: [Why do you need to do this if there is only one read-out? (Is the read-out analog or digital?)]

 

All sensors have fixed pattern noise, but with one readout there is not a drift issue in which you are trying to match quadrants of the image that may drift in different directions on their black level and gain as the sensor heats or cools.

 

With all of the sensor drifting in the same direction you don't see splits and seams in the image maybe as much and its easier to null the FPN automatically over a days shooting where the camera is turned on and off.

 

 

Quote: [i agree that DB is going to have to deal with many questions about WHICH lenses can be used. Which makes me think that using regular 16mm C-mount lenses on the S8 will throw a much larger than image on the sensor making 2:35:1 (2592x1104) very capture EZ. Correct?]

 

You can see the results on my videos on Vimeo and in the uncompressed BMP frames like the ones shot with the 17mm f/0.95 Schneider 2/3" lens and the 10mm f/1.8 Schneider Cinegon R16 lens. No strain as the sensor is like 50% to 60% of the lens image circle and well within the 0.707 or 70.7% zone of astigmatism that is the "sharpest" part of the lens coverage, also the curvature of field at the 70.7% zone may be positive and that complements the slight negative curvature of field that the thickness of the OLPF may cause, so not using the maximum of the lens coverage can also improve the corner of image resolution that way.

 

Some CS mount 1/3" video lenses may also be sharp, as needing to cover less large of an image circle they can be sharper in the small area that they cover in an absolute sense rather than by lines per lens coverage diagonal, say one lens covers a diagonal of 1" with 1000 lines over that distance, bit another covers 1/3" with 500 lines over that distance, then the spot size of the 1/3" lens is smaller when used on a small sensor area, even though the total resolution in lines over its maximum coverage is lower.

 

 

Quote: [i'm going to contact them this week. I live in Vegas. Are you coming?]

 

I offered to show their cameras for them if they could not make it, maybe take a Casino bus tour for a day, but got no response on that issue as far as I can recall.

 

Quote: [Despite the hype about the DB — I have my doubts they can deliver their fall units at the prices they think. Frankly, I would think it would be foolish to do so.]

 

I wish digital Bolex success, but their concept is not without faults of its own. Having talked with Kinefinity.com and OptimaCine.com for like four years now, I know its much harder to get things like working without glitches that one might think, the fact that both companies have demonstrated cameras but are not selling any yet is a bit frustrating to me since I spent four years working night and day to get my software support working well enough for testing and hopefully use.

 

I was hoping for an under $500 solution and thought of doing my own design that would be passed into the public domain, but as developing the software for True RAW camera development is a time consuming task I thought experienced camera makers would have an easy time making something working quickly. I hope digital Bolex does not fall into the same sort of delays with the companies they will be working with, as they have taken peoples money and promised shipment dates.

 

For now, its good to keep up to date on what progress any of the camera makers have made, but as specs may change its too early to make any final decisions based on minutia I would think.

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...