Is the 2k raw in 7Q going to be as bad as 2k in sonys raw recorder?
- Thread starter Sami2
- Start date
SergeSmArt
Well-known member
What about this way to solve the problem? 
-->
If we use this "Crop" 17 X7.23mm in special "Cropped 2.35:1 Mode" and DO NOT RECORD anyhting else... I see some positive moment in it:
1. We save space on expencive media for filming in 2.35:1 Aspect Ratio ( instead to "paint it black" in post )
2. Rolling Shutter effect will be less visible as vertical way of strokes is shorter
3. This size is still big enouth to use footage for "image stabilisation" in post.
4. We will be able to shoot Slow Motion with speeds like 120 and 240 fps without loosing some quality ..even in buffered way
I don't know - what resolution will be on this 17 X 7.23mm part of FS-700 sensor.. so I hope someone can calculate it even in average ...
maybe it's a crazy idea... but you know - I like 2.35:1 Aspect Ratio so much!! ...
-->
If we use this "Crop" 17 X7.23mm in special "Cropped 2.35:1 Mode" and DO NOT RECORD anyhting else... I see some positive moment in it:
1. We save space on expencive media for filming in 2.35:1 Aspect Ratio ( instead to "paint it black" in post )
2. Rolling Shutter effect will be less visible as vertical way of strokes is shorter
3. This size is still big enouth to use footage for "image stabilisation" in post.
4. We will be able to shoot Slow Motion with speeds like 120 and 240 fps without loosing some quality ..even in buffered way
I don't know - what resolution will be on this 17 X 7.23mm part of FS-700 sensor.. so I hope someone can calculate it even in average ...
maybe it's a crazy idea... but you know - I like 2.35:1 Aspect Ratio so much!! ...
Last edited:
Barry_Green
Moderator
Taking two lines, and discarding two lines, as harddrive discussed, would technically work -- and it would result in horrible video. The aliasing would be terrible. I hope that's not what Sony's doing.
Taking the Canon route, of combining each 2x2 RGBG block into one RGB pixel, isn't "binning". That's just using an alternative demosaic method, it's a very simplistic demosaic, and works quite well. But it's not raw, so that's likely not what Sony's doing either.
The only way to get a true, artifact-free, quality raw 2K off of a 4K sensor, is to do it the way Red does, and the way SergeSmArt is proposing, and the way the GH2 gets a true 1080p out of its 4K sensor when in ETC mode -- extract a window of the sensor at full resolution. That would give you a completely pure 2K raw, but obviously the field of view is tighter.
I don't know how Sony proposes to deliver a 2K raw signal from a 4K sensor, but if they're trying to retain the entire field of view then I'm not surprised at all that there's disappointment in the results.
If you want high-quality raw from a 4K sensor, the way to get that is to stick with 4K. If you need 2K afterwards, then demosaic your 4K, do all your processing on it at 4K, and then scale it down to 2K. But getting a true 2K raw from a 4K sensor is a problematic issue no matter how you slice it. Something's gotta give. In my opinion, the thing that should give is field of view, but apparently that is not the way they've chosen to handle it.
Taking the Canon route, of combining each 2x2 RGBG block into one RGB pixel, isn't "binning". That's just using an alternative demosaic method, it's a very simplistic demosaic, and works quite well. But it's not raw, so that's likely not what Sony's doing either.
The only way to get a true, artifact-free, quality raw 2K off of a 4K sensor, is to do it the way Red does, and the way SergeSmArt is proposing, and the way the GH2 gets a true 1080p out of its 4K sensor when in ETC mode -- extract a window of the sensor at full resolution. That would give you a completely pure 2K raw, but obviously the field of view is tighter.
I don't know how Sony proposes to deliver a 2K raw signal from a 4K sensor, but if they're trying to retain the entire field of view then I'm not surprised at all that there's disappointment in the results.
If you want high-quality raw from a 4K sensor, the way to get that is to stick with 4K. If you need 2K afterwards, then demosaic your 4K, do all your processing on it at 4K, and then scale it down to 2K. But getting a true 2K raw from a 4K sensor is a problematic issue no matter how you slice it. Something's gotta give. In my opinion, the thing that should give is field of view, but apparently that is not the way they've chosen to handle it.
SergeSmArt
Well-known member
But I hope my way gonna be better than RED Scarlet 2K as it use bigger area ->
..so, it's looks more as "2.6K" or even "3K" ...and make less problems for FOV for wide angle lenses .. but still give us some level of fast frame rates...
The more I think about this, the more one point comes home to me. We're used when evaluating camera performance to looking at a final image - the result of lens, sensor, read process *and processing of that data in camera*. With a move to RAW, the last element is taken out of the equation, it becomes a post process, no longer intrinsic to the basic camera. Hence chart performance becomes dependent on factors other than those intrinsic to the camera itself, demosaicing being the most relevant.
In the context of this thread, is what we see (and are disappointed with) really fundamental to the FS700 and it's 2K RAW mode? Or does it have a lot to do with the 2K demosaicing not being done as well as it should be.....? I'm beginning to wonder if the latter may not be a more likely scenario, but without further information it's impossible to say.
"Horrible" video? "Terrible" aliasing? Everything is relative. It won't look good on charts, but on pictures then most of the time it will be more than acceptable enough. It's better than earlier DSLR video with poorer methods, and is much used on current DSLRs for video. (See the references to the GH2 on charts below.)
It's also the technique that seems to be used in the FS700 in high speed (greater than 60fps) non-RAW mode. Read at 240fps and it's 4x the normal 60fps rate. Is it really a surprise that you can then only read a quarter of the total photosites? It seems to be the case that at normal speed, the FS700 does something similar to the Canon C300 - derive an R,G,B pixel from each 2x2 block - so 1920x1080 resolution from a 3840x2160 chip. So at 240fps, if it only reads one block in four in the same basic way, what could we expect? Half resolution (540lpph), bags of aliasing, but importantly, no colour aliasing. And look at Adam Wilts charts from last year and that's exactly what you get. http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/ - and scroll down to the second zone plate. Exactly what you'd expect. Resolution halved, bags of aliasing - but none of that coloured.
Yet many users, looking at real pictures, are very happy with the 240fps FS700 slo-mo images. Quality is not as good as at 60fps - the charts make that very clear - but the consensus seems to be far from "horrible video".
It's also worth looking at http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/P2 , to see what it looks like on the DSC charts. 240fps is the second one down. Now just look at what the GH2 and GH3 do on the same chart.... http://provideocoalition.com/awilt/...mix_dmc-gh3_dslm_hybrid_still_video_camera/P2 . Whoa!! Very, very similar looking patterns, the big difference being that the null zone on the H,V trumpets (where first order aliases give way to second order) seems to centred around 1350 for the GH2,3, and 1080 for the FS700 in 240fps. Which corresponds to unaliased resolutions of a bit below 700 lpph for the GH, 540 for a high speed FS700. That's to be expected because the GH2 sensor has a far higher photosite count than the FS700 - it's why it needs to pisel-skip in normal mode. This model would predict it to be about 2x1350 for height (2700). In practice, it's windowed to 4976 x 2800 in 16:9 stills mode - which bears the theory out well.
And whilst the charts make clear GH2(3) video is far from perfect, it's good enough for many users.
Coming back to the real point, if that's what the FS700 does in 240fps mode, it's reasonable to assume that it's using the same principle to derive 2K RAW, at least as far as sensor reading. That theory is backed up by it being able to output 240fps 2K RAW indefinitely - no buffering.
In the context of this thread, is what we see (and are disappointed with) really fundamental to the FS700 and it's 2K RAW mode? Or does it have a lot to do with the 2K demosaicing not being done as well as it should be.....? I'm beginning to wonder if the latter may not be a more likely scenario, but without further information it's impossible to say.
In principle, it (or something similar) is a widely used technique, even if manufacturers don't shout about it. If you can't read all the photosites of the entire frame at the speed required, you have to miss out a percentage of the photosites - end of story. The first attempts were clumsy line skipping, the idea being to simply get a signal for "live view" viewfinding - quality was not important. Then someone hit on the idea that live view could give a video mode on a stills camera, and the rest is history. Next question was how to make it better - and the system I mentioned (discard every 3rd and 4th line and column, and make the one in four 2x2 block left into a single RGB pixel) is probably as good it gets - if you can't read the full chip at the necessary speed.
"Horrible" video? "Terrible" aliasing? Everything is relative. It won't look good on charts, but on pictures then most of the time it will be more than acceptable enough. It's better than earlier DSLR video with poorer methods, and is much used on current DSLRs for video. (See the references to the GH2 on charts below.)
It's also the technique that seems to be used in the FS700 in high speed (greater than 60fps) non-RAW mode. Read at 240fps and it's 4x the normal 60fps rate. Is it really a surprise that you can then only read a quarter of the total photosites? It seems to be the case that at normal speed, the FS700 does something similar to the Canon C300 - derive an R,G,B pixel from each 2x2 block - so 1920x1080 resolution from a 3840x2160 chip. So at 240fps, if it only reads one block in four in the same basic way, what could we expect? Half resolution (540lpph), bags of aliasing, but importantly, no colour aliasing. And look at Adam Wilts charts from last year and that's exactly what you get. http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/ - and scroll down to the second zone plate. Exactly what you'd expect. Resolution halved, bags of aliasing - but none of that coloured.
Yet many users, looking at real pictures, are very happy with the 240fps FS700 slo-mo images. Quality is not as good as at 60fps - the charts make that very clear - but the consensus seems to be far from "horrible video".
It's also worth looking at http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/P2 , to see what it looks like on the DSC charts. 240fps is the second one down. Now just look at what the GH2 and GH3 do on the same chart.... http://provideocoalition.com/awilt/...mix_dmc-gh3_dslm_hybrid_still_video_camera/P2 . Whoa!! Very, very similar looking patterns, the big difference being that the null zone on the H,V trumpets (where first order aliases give way to second order) seems to centred around 1350 for the GH2,3, and 1080 for the FS700 in 240fps. Which corresponds to unaliased resolutions of a bit below 700 lpph for the GH, 540 for a high speed FS700. That's to be expected because the GH2 sensor has a far higher photosite count than the FS700 - it's why it needs to pisel-skip in normal mode. This model would predict it to be about 2x1350 for height (2700). In practice, it's windowed to 4976 x 2800 in 16:9 stills mode - which bears the theory out well.
And whilst the charts make clear GH2(3) video is far from perfect, it's good enough for many users.
Coming back to the real point, if that's what the FS700 does in 240fps mode, it's reasonable to assume that it's using the same principle to derive 2K RAW, at least as far as sensor reading. That theory is backed up by it being able to output 240fps 2K RAW indefinitely - no buffering.
Absolutely, 100% agreed. Treated as a 4K RAW camera, full marks to the FS700. Same with it's conventional mode and the slo-mo options. But I'm very suspicious of the 2K RAW mode. Heck, even if it was a normal 1920x1080 Bayer sensor, and was deBayered as well as possible, the eventual true resolution will only be about 80% of 1920x1080.
Again, fully agreed.
SergeSmArt
Well-known member
So..? Let's asking Sony to give us this " 2.35:1 Cropped Mode" as additional, separate option...! I don't know if it's simple or hard ( technically ) for Sony to do.. but someone here probably know it better ..
NextWaveG
Well-known member
Here's something to consider...
CD just announced some more specs on recording options and times, and for those who don't feel like doing the math:
2K RAW: 277GB/hr
4K Compressed: 296GB/hr
1080p Compressed: 74GB/hr
Now one thing that caught my eye was this on CD's website:
I know we're talking about RAW here but how often will you NEED RAW? I see myself shooting compressed HD and 4K more often than RAW. Depending on how good CD is at super-sampling the 4K image, 1080p compressed may be sharper than the 2K RAW stream.
CD just announced some more specs on recording options and times, and for those who don't feel like doing the math:
2K RAW: 277GB/hr
4K Compressed: 296GB/hr
1080p Compressed: 74GB/hr
Now one thing that caught my eye was this on CD's website:
I know we're talking about RAW here but how often will you NEED RAW? I see myself shooting compressed HD and 4K more often than RAW. Depending on how good CD is at super-sampling the 4K image, 1080p compressed may be sharper than the 2K RAW stream.
SergeSmArt
Well-known member
I don't think we really need RAW for final production... ( except if we get an order to shoot for Hollywood ...
rogar-BigGrin(DBG) ) But I don't see any option to shoot 240 fps in "1080p compressed" - maximum they have is 1080p YCC 422 10-bit Compressed 120fpspaulcurtis
Veteran
+1
The only reason to have RAW is to be able to choose to debayer it in a different way, but in reality, beyond people testing i really don't think most people agonise over a debayer method (not that we actually have many choices either).
But if you have 4K 12bit files then frankly that's more than enough to work with. If CD can downsample both 1080 and 2K from those 4K source files that would be perfect (for me).
If that handles burst 120fps even better.
You know there are so many parts to the image chain that affect quality that obsessing over one part is always detrimental. Speaking as someone who's just found a weird lens aberration at a certain aperture screwing with some green screen... All that 4K RAW would have gotten me is *more* aberration
cheers
Paul
I don't think we really need RAW for final production... ( except if we get an order to shoot for Hollywood ...
Maybe you and you all don't but some of us do need it.
Who cares about the bitrates CD anounces if they can not produce even a single sample clip out of their magic machine. Have they told what kind of a prosessor is in the machine and how much it will actually consume power driven at full speed or is that something that is subjected to change once again in next weeks. If they actually managed to capture images from fs700 for the first time in the beginning of June on their prototype recorder they must have a brilliant production pipeline hence they can turn that prorotype to a working device in a ew months. For some reason something does not ad up...
"The Odyssey7 and 7Q have now unfortunately been delayed by Convergent and is now expected approximately August/September time."
WOW, what a suprise.
paulcurtis
Veteran
Unsure whether to reply to something that looks rather troll like...
There're no public samples from Sony either. And as soon as any are released they're going to be picked apart and aspects of the shooting will be blamed on hardware.
The CPU (actually a FPA) has been discussed along with the tiny power requirements. What does the CPU matter anyway?
They captured images from a very beta camera. I think perhaps even Sony wouldn't want images from a beta hardware test camera out there. To the best of my knowledge there are no upgrades done yet (mines in the queue at sony anyway).
The 'delay' was from quite a while ago, no surprises there.
So unless you're from a competing company why are you throwing all your toys out of your pram?
cheers
Paul
powervideo
Well-known member
All this Sami2 guy does in his limited number of posts is bag-out, criticise, tear-down and offer sarcasm and unproven opinions. What is your real name, who are you and where are you?
SergeSmArt
Well-known member
My cameras updated at Tuesday
... only a problem to try 4K RAW is - I cannot get Sony HXR IFR5 and AXS-R5 in New York.. even those who have it, don't want me to try it... probably don't want me to publish what I able to get from RAW....
rogar-BigGrin(DBG)There are a few raw samples out of sonys recorder hooked up to fs700 online.
Well if you want to say that bad 2k raw is an unproven opinion look at the res charts in the video.
i did place an order for 7q but cancelled it after seeing how low res the 2k is and seeing the first samples of 4k raw with epic resolutions out of sonys recorder. Also i did the cancellation since there are no evidence supporting the arrival of 7q this year. Of course the company will say that they will get it working but bare in mind that they finally managed to get the 7q (or that is what cd says, who knows what kind of machine they actually used for the first recorded 2k file) to work with fs700 a month ago.
we have too many similar examples with high promises that turn into year long delays, red one, blc camcorder,etc.
lets hope that sony fixes the 2k algorithms to get more res ( hopefully not in reds way with crop) and then cd should atleast have a brilliant starting point if they can just get the wonder monitor really working.
cpu and tech specs matter if there is a lot of calculations done with it itmwill drain more power and generate heat
Well if you want to say that bad 2k raw is an unproven opinion look at the res charts in the video.
i did place an order for 7q but cancelled it after seeing how low res the 2k is and seeing the first samples of 4k raw with epic resolutions out of sonys recorder. Also i did the cancellation since there are no evidence supporting the arrival of 7q this year. Of course the company will say that they will get it working but bare in mind that they finally managed to get the 7q (or that is what cd says, who knows what kind of machine they actually used for the first recorded 2k file) to work with fs700 a month ago.
we have too many similar examples with high promises that turn into year long delays, red one, blc camcorder,etc.
lets hope that sony fixes the 2k algorithms to get more res ( hopefully not in reds way with crop) and then cd should atleast have a brilliant starting point if they can just get the wonder monitor really working.
cpu and tech specs matter if there is a lot of calculations done with it itmwill drain more power and generate heat
paolo.negro
Member
FS700 1080p formats are super-sampled HD, derived from 4K RAW stream to improve the image quality (10-Bit Bit Depth).
This is the news I was waiting for!
I see all the CD efforts to fulfill our requests, but I think now we need to see some footage as promised time ago (some people is becoming anxious and skeptical)
Dan, where are you?
This is the news I was waiting for!
I see all the CD efforts to fulfill our requests, but I think now we need to see some footage as promised time ago (some people is becoming anxious and skeptical)
Dan, where are you?
powervideo
Well-known member
The thing is much can change by the time this gets to market. All the manufacturers (except maybe Apple in the PC world) per-announce too early these days. It creates a buzz and causes buyers to reconsider purchases with other brands they may have had in mind. The downside is the long waits are destabilising and someone else will inevitably announce a competing product, which will also go into long-term development hell!
I agree it's a bit worrying about the 2K but I try to ignore per-production sample footage and hearsay and wait for the real deal. After NAB I was very tempted to order the 7Q but I'd rather wait until production models are shipping before deciding what to do. That said..... I did pre-order a BMC Pocket Camera which I'll purchase whatever happens.
"FS700 1080p formats are super-sampled HD, derived from 4K RAW stream to improve the image quality (10-Bit Bit Depth)."
just more promises once again, how about some prores 444 samples on this?
also placed an order for bmc, even though i wish they had also made it with a mount that has a tiny flange for example c-mount.
just more promises once again, how about some prores 444 samples on this?
also placed an order for bmc, even though i wish they had also made it with a mount that has a tiny flange for example c-mount.
I can understand where your coming from but I would not compare the first black magic camera to cd.. From what i know it was their first camera and maybe had less experience, and from the look of it it seems they learned from past misstakes. Even if black magic wont be perfect this time I still think things will be better this time.
When it comes to CD, this is not their first monitor/recorder so I think it is at least safe to assume they know what they are doing. It might get delayed yes but thats not really that uncommen is it? Things get delayed all the time.. And as far as the samples goes, sony just realesed their final upgrade. Sure CD had a prototype before in the lab but honestly a lot is at stake and a lot can happen. I sure as hell would not post a lot of samples unless I knew it would be perfected and final. As far as we know things might have changed from the upgrade they had to the one sony is doing now on the fs700. Plus cd seems to be working really hard to travel and go to different expos.. They have been to two big ones in the states, 1 in asia & the upcomming one in australia. This within a few months. If they can deliver what to me is most important (1080p from 4k raw) in order to preserve best image and 10-12 bit thats sooo much more mobile than the bulky sony solution, then im willing to wait a little longer.
Also it is a little unfair to expect sample from a codec they just recently (from what i know) just licensed to use. This is a big thing their doing, it is not to be handles lightly from their part.
But thats me
When it comes to CD, this is not their first monitor/recorder so I think it is at least safe to assume they know what they are doing. It might get delayed yes but thats not really that uncommen is it? Things get delayed all the time.. And as far as the samples goes, sony just realesed their final upgrade. Sure CD had a prototype before in the lab but honestly a lot is at stake and a lot can happen. I sure as hell would not post a lot of samples unless I knew it would be perfected and final. As far as we know things might have changed from the upgrade they had to the one sony is doing now on the fs700. Plus cd seems to be working really hard to travel and go to different expos.. They have been to two big ones in the states, 1 in asia & the upcomming one in australia. This within a few months. If they can deliver what to me is most important (1080p from 4k raw) in order to preserve best image and 10-12 bit thats sooo much more mobile than the bulky sony solution, then im willing to wait a little longer.
Also it is a little unfair to expect sample from a codec they just recently (from what i know) just licensed to use. This is a big thing their doing, it is not to be handles lightly from their part.
But thats me
Last edited: