Max Field Posted July 5, 2018 Share Posted July 5, 2018 (edited) I've come to understand the Sony F35 is a 1080p camera, however was recently reading it had a 5K sensor?? After reading after that reading, it had something to do with the theory of doubling the photosites to ensure optimal quality, but is there any way to rig it so it records at a higher resolution? I saw some guy on eBay saying he was recording at 4K with no quality loss, but could this be the chroma subsampling taking a hit for luma resolution? Thank you! Edited July 5, 2018 by Macks Fiiod Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted July 5, 2018 Premium Member Share Posted July 5, 2018 Its not a Bayer CFA sensor, it has RGB stripes, an equal number of total photosites for each color unlike with a Bayer. So to create 1920 x 1080 RGB, it has about 2MP per color so 6MP total. But because its not a mosaic color array pattern, you cant intelligently reconstruct a full color image at a higher resolution than 1920 x 1080 per color. Plus there is some greater tendency towards moire with this striped system. But since a Bayer CFA only allots 25% of the photosites for red and blue information, Id suspect one could uprez the F35 signal fairly well at least in terms of detail in the red and blue channels. Link to comment Share on other sites More sharing options...
Max Field Posted July 7, 2018 Author Share Posted July 7, 2018 Its not a Bayer CFA sensor, it has RGB stripes, an equal number of total photosites for each color unlike with a Bayer. So to create 1920 x 1080 RGB, it has about 2MP per color so 6MP total. But because its not a mosaic color array pattern, you cant intelligently reconstruct a full color image at a higher resolution than 1920 x 1080 per color. Plus there is some greater tendency towards moire with this striped system. Just to make clear, are you saying that the F35 has moire issues? Or only if I try upscaling a sensor of its nature? Link to comment Share on other sites More sharing options...
Premium Member Phil Rhodes Posted July 7, 2018 Premium Member Share Posted July 7, 2018 People have said it has moire issues. I think the reality may be that it has moire issues that are easier to see because they tend to form unbroken lines as opposed to Bayer moire which is, in theory, just as present but harder to see. If a tree falls in the woods and nobody is there to hear it... Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted July 7, 2018 Premium Member Share Posted July 7, 2018 I shot on the Panavision Genesis (more or less the same camera as the F35, same sensor) for over three years on a series and some pilots and it had moire issues — you had to test wardrobe to avoid it. I don’t know if it was more due to the striped sensor design or a weak OLPF in the pursuit of better sharpness. Shooting for a shallow depth of field helped. 1 Link to comment Share on other sites More sharing options...
Premium Member Mark Kenfield Posted July 7, 2018 Premium Member Share Posted July 7, 2018 Interesting. I wonder if the Genesis and F35 had different OLPFs? I shot a feature and a half dozen other projects on the Sony version of the camera and never encountered any moire. Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted July 7, 2018 Premium Member Share Posted July 7, 2018 Sure, it’s possible, if not probable, that Sony refined the OLPF after the two years or so that the Genesis had the market to themselves. But I think the physical separation between a color stripe and the next time it appears in a row would make color moire almost inevitable. I only did a few things on the F35 compared to the hundreds of hours I did on the Genesis but I don’t recall much difference in the image. Link to comment Share on other sites More sharing options...
Jack Jin Posted May 11, 2020 Share Posted May 11, 2020 On 7/7/2018 at 1:51 PM, David Mullen ASC said: Sure, it’s possible, if not probable, that Sony refined the OLPF after the two years or so that the Genesis had the market to themselves. But I think the physical separation between a color stripe and the next time it appears in a row would make color moire almost inevitable. I only did a few things on the F35 compared to the hundreds of hours I did on the Genesis but I don’t recall much difference in the image. Hi, sorry for replying to such an old thread, but isn't the genesis pixel binning the 5k sensor, and the f35 is doing a full sensor readout? Thus the moire could just have been caused by the pixel binning of the genesis. Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted May 11, 2020 Premium Member Share Posted May 11, 2020 No, the Genesis did not pixel bin, it was the same sensor as the F35, approx. 2K per color, 6K total, so it could simply read out each color stripe and create 1920 x 1080 per color, no demosaicing. There weren't extra pixels to bin from anyway, there were just enough pixels for each color in HD plus some unrecorded lookaround I think (that part I don't remember.) I guess you could consider that to create red, for example, you extracted every third stripe from the sensor, maybe you'd call that "binning" but it's not the definition normally used. It's the nature of the color stripe design that color moire is sort of inevitable unless you really want to soften a lot of fine detail to get rid of it. 1 1 Link to comment Share on other sites More sharing options...
Jack Jin Posted May 11, 2020 Share Posted May 11, 2020 (edited) 1 hour ago, David Mullen ASC said: No, the Genesis did not pixel bin, it was the same sensor as the F35, approx. 2K per color, 6K total, so it could simply read out each color stripe and create 1920 x 1080 per color, no demosaicing. There weren't extra pixels to bin from anyway, there were just enough pixels for each color in HD plus some unrecorded lookaround I think (that part I don't remember.) I guess you could consider that to create red, for example, you extracted every third stripe from the sensor, maybe you'd call that "binning" but it's not the definition normally used. It's the nature of the color stripe design that color moire is sort of inevitable unless you really want to soften a lot of fine detail to get rid of it. Oh I see, I saw the wiki for the genesis that said it cutted the vertical resolution by half through binning ? https://en.wikipedia.org/wiki/Genesis_(camera) So I presume that this moire issue is the reason why so many manufactures decided to use a 3ccd design instead? Or is it to just maintain compatibility with the b4 lenses? Edited May 11, 2020 by Jack Jin added text Link to comment Share on other sites More sharing options...
Premium Member Phil Rhodes Posted May 11, 2020 Premium Member Share Posted May 11, 2020 Some camera designs do various types of what could be called binning; Viper did similar things so as to be able to achieve either widescreen-ish or cinemascope-ish aspect ratios without overall loss of resolution. If it's built into the camera at that level, though, it's not as if there's anything the user can do with that information. Especially on a CCD camera like Genesis, getting at the individual photosite information would require a large electronic engineering effort that would likely be more work than simply building a camera around a more modern sensor that you could buy off the shelf with full support and documentation (which would still be a lot of work, as various wannabe camera companies have found.) 3CCD (or more often these days 3 CMOS) designs improve lens compatibility and sensitivity/noisefloor/dynamic range in some combination. Three times the sensor area per pixel helps. Link to comment Share on other sites More sharing options...
Mike Krumlauf Posted July 12, 2020 Share Posted July 12, 2020 On 7/6/2018 at 10:21 PM, Mark Kenfield said: Interesting. I wonder if the Genesis and F35 had different OLPFs? I shot a feature and a half dozen other projects on the Sony version of the camera and never encountered any moire. I can attest after having used both the Genesis their OLPFs are different, just as the F900 Stock version and Panavision modified HD900F had different OLPFs. Link to comment Share on other sites More sharing options...
Carl Nenzen Loven Posted January 2, 2021 Share Posted January 2, 2021 On 7/11/2020 at 10:23 PM, Mike Krumlauf said: I can attest after having used both the Genesis their OLPFs are different, just as the F900 Stock version and Panavision modified HD900F had different OLPFs. Did you do any moiré tests? C Link to comment Share on other sites More sharing options...
Raymond Zananiri Posted January 3, 2021 Share Posted January 3, 2021 I have an F35 but I never experienced any moire/aliasing of any kind. But then, maybe the conditions were always as such that they did not appear. I can run some tests if that is critical for someone. In 1080p the image looks almost as sharp as 4k for other cameras. Link to comment Share on other sites More sharing options...
Premium Member Aapo Lettinen Posted January 4, 2021 Premium Member Share Posted January 4, 2021 On 5/11/2020 at 12:46 PM, Phil Rhodes said: Some camera designs do various types of what could be called binning; Viper did similar things so as to be able to achieve either widescreen-ish or cinemascope-ish aspect ratios without overall loss of resolution. If it's built into the camera at that level, though, it's not as if there's anything the user can do with that information. Especially on a CCD camera like Genesis, getting at the individual photosite information would require a large electronic engineering effort that would likely be more work than simply building a camera around a more modern sensor that you could buy off the shelf with full support and documentation (which would still be a lot of work, as various wannabe camera companies have found.) 3CCD (or more often these days 3 CMOS) designs improve lens compatibility and sensitivity/noisefloor/dynamic range in some combination. Three times the sensor area per pixel helps. 3ccd tends to have different, more efficient way to handle colour separation (colour-reflecting dicroic coatings instead of traditional substractive filters) so if done well it should be even more efficient than the traditional single-chip system using "normal" filters. I have understood that the A/D conversion is the most challenging part on the CCD cameras. Is that F35 sensor handling anything on chip level or is the conversion external like in the traditional CCD? it should be theoretically possible to hack a sensor if having the documentation for it but it is extremely demanding and time consuming so probably not worth it for that camera, better to purchase a used Alexa instead ( I would assume it would be a 2 year job for a person to do something like this to the F35). If you can get the sensor data efficiently from analog stream to digital format then you can probably do a custom system using powerful fpga's for signal processing. CMOS cameras are easier in this regard because you already have a digital output on the image sensor level so you "only" have to handle the timing control etc. stuff and the very high bitrate data streams coming from the sensor which is a perfect job for a powerful fpga's to do if one has the expertise to custom program them for the job Link to comment Share on other sites More sharing options...
Premium Member Aapo Lettinen Posted January 4, 2021 Premium Member Share Posted January 4, 2021 2 minutes ago, aapo lettinen said: I have understood that the A/D conversion is the most challenging part on the CCD cameras. Is that F35 sensor handling anything on chip level or is the conversion external like in the traditional CCD? it should be theoretically possible to hack a sensor if having the documentation for it but it is extremely demanding and time consuming so probably not worth it for that camera, better to purchase a used Alexa instead ( I would assume it would be a 2 year job for a person to do something like this to the F35). If you can get the sensor data efficiently from analog stream to digital format then you can probably do a custom system using powerful fpga's for signal processing. CMOS cameras are easier in this regard because you already have a digital output on the image sensor level so you "only" have to handle the timing control etc. stuff and the very high bitrate data streams coming from the sensor which is a perfect job for a powerful fpga's to do if one has the expertise to custom program them for the job the described job is basically taking the sensor, the physical body and the optical parts from the original F35 and making most if not all of the camera electronics by yourself from ground up. Similarly demanding than purchasing a plain image sensor from Mouser and building your own digital cinema camera from it doing all the control and signal processing electronics and custom software by yourself. This is probably not a good solution for any cinematography related use... just wanted to say that it is possible and how it can roughly be done if one is experienced enough and has tons of spare time for the project. it does not make any financial sense compared to purchasing a used cinema camera or a new mid range movie camera because the amount of work is huge. maybe a youtuber or similar person could make a project out of it... they have lots of spare time and funding available to do things which do not make any financial sense :P if not having the sensor documentation one could maybe probe the sensor outputs with oscilloscope to find out how the timing control and signal outputs of it work. without that information one can't even start working with it. Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted January 5, 2021 Premium Member Share Posted January 5, 2021 On 5/10/2020 at 11:04 PM, Jack Jin said: So I presume that this moire issue is the reason why so many manufactures decided to use a 3ccd design instead? Or is it to just maintain compatibility with the b4 lenses? 2/3" 3-sensor prism block cameras were around for decades, the single 35mm sensor shows up in the 2000's, so why it is odd that the first HD cameras were 2/3" 3-sensor designs? I also suspect that there is less computer processing needed when the camera naturally delivers three color signals. Link to comment Share on other sites More sharing options...
Mike Krumlauf Posted February 28, 2021 Share Posted February 28, 2021 On 1/2/2021 at 11:16 AM, Carl Nenzen Loven said: Did you do any moiré tests? C Just seeing this, sorry, have not been on this website in awhile. Yes I did and they reacted near the same. No noticeable difference between the 2 cameras. The genesis is a relic and a rare gem of a camera if you can find one but the F35 is by far better than the Genesis all around. Link to comment Share on other sites More sharing options...
John Shell Posted September 14, 2022 Share Posted September 14, 2022 On 7/5/2018 at 1:03 AM, David Mullen ASC said: Its not a Bayer CFA sensor, it has RGB stripes, an equal number of total photosites for each color unlike with a Bayer. So to create 1920 x 1080 RGB, it has about 2MP per color so 6MP total. But because its not a mosaic color array pattern, you cant intelligently reconstruct a full color image at a higher resolution than 1920 x 1080 per color. Plus there is some greater tendency towards moire with this striped system. But since a Bayer CFA only allots 25% of the photosites for red and blue information, Id suspect one could uprez the F35 signal fairly well at least in terms of detail in the red and blue channels. Do you think a 4K or 8K camera with this RGB stripe sensor technique would be feasible ?? The photosite density would be very high of course but I'm just wondering Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted September 14, 2022 Premium Member Share Posted September 14, 2022 1 hour ago, John Shell said: Do you think a 4K or 8K camera with this RGB stripe sensor technique would be feasible ?? The photosite density would be very high of course but I'm just wondering Sure, I think John Galt at Panavision might have once been working on a 4K version of the Genesis, maybe with a full-frame sensor -- I think the original Primo 70 lenses were developed for that. Just a rumor though. Link to comment Share on other sites More sharing options...
Premium Member Phil Rhodes Posted September 14, 2022 Premium Member Share Posted September 14, 2022 This is a complex one but suffice to say that a lot of different sensor layouts and arrangements have been tried over the years and the upshot has been a resounding meh. Sensitivity and colour performance are to some degree a zero some game at any point. I would say we now have more sensor performance than we really need in order to make it clear that this doesn't really matter very much anymore. Link to comment Share on other sites More sharing options...
John Shell Posted September 16, 2022 Share Posted September 16, 2022 On 9/14/2022 at 4:58 PM, David Mullen ASC said: Sure, I think John Galt at Panavision might have once been working on a 4K version of the Genesis, maybe with a full-frame sensor -- I think the original Primo 70 lenses were developed for that. Just a rumor though. Interesting. Another thing, what does RGB stripes or striped sensor mean exactly ?? Is it just a Colour Filter Array with RGB lines stripes in that order or does the sensor itself have stripes or something?? I tried googling to see what it looks like or an illustration but I couldn't find anything. Link to comment Share on other sites More sharing options...
Premium Member David Mullen ASC Posted September 16, 2022 Premium Member Share Posted September 16, 2022 Red, green, and blue stripes rather than a Bayer mosaic pattern. The main advantage is that all three colors are equally sampled. It's also a "simple" conversion -- in the case of the Genesis camera, you had full HD resolution total for each color, i.e. 2K per color, a 6K sensor total. The main disadvantage is color moire, plus if you want a 4K image, you'd have to have a 12K sensor. 1 Link to comment Share on other sites More sharing options...
Robin Phillips Posted September 16, 2022 Share Posted September 16, 2022 On 9/14/2022 at 8:39 AM, John Shell said: Do you think a 4K or 8K camera with this RGB stripe sensor technique would be feasible ?? The photosite density would be very high of course but I'm just wondering feasible sure. but the CCD on the F35 / Genesis was so god damned expensive (and Im not sure if the manufacturing yields were that good, though thats just a recollection) that it probably wont happen. Remember that the F35 was a $250k USD camera back in the day. CCDs have always been very expensive, hence why when CMOS sensors came to market they swooped in and dominated, they were that much cheaper. That being said 1 boy do I love the F35s CCD imager at 444 and 2 if you were to upscale to 4k, at least you'd be doing it with a true 1080 444 image. With VFX companies still upscaling 2k vfx shots to 4k for delivery (though this is changing), its not the end of the world to upscale, especially if its just going somewhere like youtube. Link to comment Share on other sites More sharing options...
John Shell Posted June 15, 2023 Share Posted June 15, 2023 Forgive me but I’m still confused about this camera’s sensor. How exactly does this camera not perform any demosaicing when each photosite only gets 1 color just like a Bayer sensor ?? Link to comment Share on other sites More sharing options...
Recommended Posts
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 accountSign in
Already have an account? Sign in here.
Sign In Now