Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

2K Scaled Is Vital

David Mullen ASC said:
Regarding the whole resolution of projection issue, I think we need new ways of thinking about sharpness because generally 720P and 1080P photography look softer, less detailed, compared to 35mm on the big theater screen, despite what the numbers tell you. Maybe it's the whole "oversampling" notion that it's always better to start out with more resolution even if the final display format has less. For example, I've seen 65mm footage on an HD monitor that looks amazingly detailed, yet you'd think that on an HD monitor, anything from HD to IMAX would end up looking about the same. Or consider 35mm on an NTSC monitor, which still looks better than 16mm or even some HD, despite the low resolution of NTSC.
Click to expand...

SO TRUE.

I dont know if it makes any sense, but i always feel that with oversampling,
each pixel of the destination screen looks sharper AND smoother at the same time, like if each pixel was representing its own value plus the value of surrounding pixels from the source. ( I know...my english is bad...)
 
Full frame high speed is basically the only topic I bother to post on. Why? Because the camera is kicking ass on all the other fronts. Full Frame high speed is essential for commercial and music video work. Especially when shooting 4:3 (still a common format for commercial and mv) and we have to crop inside the 16:9 3 perf rather than shoot 4-perf 35mm Big TV.

Has the RED team ever said that 60fps 4k REDCODE is impossible? CF probably can't handle it.
 
Another epic 2K scaled thread! hahahaa!
 
Dominic Jones said:
That's just the point though - 2k scaled will not be the same datarate as 4k - it will be 1/4 of that datarate for the same framerate, hence why 2k scaled would allow for higher fps shooting (both in terms of the encoding FPGA bottleneck and recording media bottleneck, which should be ever growing in any case).

That's assuming we're talking about 2k scaled RAW, not RGB (which would only yield a ~25% decrease in datarate over 4k RAW). But Red have certainly in the past mooted the possibility of 2k scaled RAW - whether or not this is still on the cards I do not know. The proposed max fps was the full 60fps of the sensor (at 35mm size), although I would imagine that without real-world testing that's a theoretical limit that may not be possible, in practice, to reach.

Any and all 2k scaled overcranking options would certainly be very welcome though, and I will certainly add my vote to those calling for a 2k scaled RAW mode...
Click to expand...

Hi Dominic,
It still seems to be on the cards (no pun intended) but on the back burner.
You would tend to leave the camera on 2K SCALED RAW for
most Hi Def television work, potentially making Red the first HD Digital Acquisition
Camera (for sale) with 35mm DOF/FOV and framerates up to 60fps.
That's worth celebrating I think, new features over and above what the
HD competition have to offer.
2K scaled and a few other minor (Redcine) things are the only reasons I would
seriously hold off considering Red for purchase in my small rental business.
I think Red is hobbled with 25fps max framerate in 4k RAW 35mmDOF/FOV
mode and other options need to be considered. Add my vote as well.
Mezmo
 
Yes the datarate of 2K is 1/4 of 4K so the datarate of 2K 60FPS is 1/2 of 4K at 30FPS (respectively 50FPS an 25FPS). Hence the CF card is not the bottleneck.
BUT.
1) You have to handle twice more data from the CMOS in given time,
2) at the same time COMPUTE them to create scaled 2K
(It's not the same as create 2K reference movie - it doesn't affect the 4K RAW data.)

Simply: it may be (maybe not - we don't know yet) bit too much load for in-camera computer.
 
You'd think though that the faster frame rate happens first before the scaling processing step happens -- i.e. it would have to take 4K at 60 fps and then scale it to 2K for recording. If that's true, then basically it is still limited to the processing speed of 4K at 60 fps.

I don't think the scaling can happen first, before the 4K sensor is firing 60X a second, so that the camera is only dealing with 2K at 60 fps in the processor.
 
Hmmm... Interesting point David.

You seem to be pretty up on the technical side of stuff, so I'm loath to disagree, but I'm sure I remember posts from the Red team to the effect that 2k scaled could potentially achieve higher framerates than 4k - perhaps it was the recording bottleneck that they were discussing and I've confused the two...

Or, perhaps (you may know better than me and this may not be plausible, I'm really not too confident on the nitty-gritty tech) it's possible that there is a less FPGA intensive way to scale which then allows the encoding FPGA cycles to achieve a faster speed. Or maybe we're just hoping!

In some ways I think that the Red team are victims of their own success - they've delivered so much that seemed so highly implausible at the start of the project that now we all seem to believe they can achieve anything!!!

Still, it would be great to have higher fps shooting from the full chip. That said, if circa 75fps in windowed mode is the best it gets, that's still pretty damn good compared to what else I can afford.

Out of curiosity, how much emphasis (disregarding the equipment required to achieve it) do you put on the importance of off-speed shooting for dramatic effect? I see it as a very important tool for a cinematographer, but my background is very indie, and so I don't usually get the chance to play with framerates as much as I'd like, and perhaps I'm over-emphasising the importance of the technique simply because, for me, it's a relatively rare tool to have access to - care to share any thoughts on that?
 
I think you're remembering part of it...

I think you're remembering part of it...

Dominic Jones said:
...but I'm sure I remember posts from the Red team to the effect that 2k scaled could potentially achieve higher framerates than 4k - perhaps it was the recording bottleneck that they were discussing and I've confused the two...
Click to expand...

What I think you are remembering was that 4K Redcode was limited to 30fps (currently 25) but if they scaled 4K they could potentially get faster framerates with a hard cap at 60fps as that is as fast as the full sensor could handle.

The only way to get faster than 60 was to window the sensor.
 
David Mullen ASC said:
You'd think though that the faster frame rate happens first before the scaling processing step happens -- i.e. it would have to take 4K at 60 fps and then scale it to 2K for recording. If that's true, then basically it is still limited to the processing speed of 4K at 60 fps.

I don't think the scaling can happen first, before the 4K sensor is firing 60X a second, so that the camera is only dealing with 2K at 60 fps in the processor.
Click to expand...

The RAW port is supposed to be able to output 2540p at 60fps, but that may be "true" raw sensor data (not REDCODE) with no compression or processing done in-camera whatsoever.

That being said, I vaguely remember seeing somewhere on these forums that the camera can support 4K@60fps internally, and that the inability to record 60fps 4K is a flash memory/HDD transfer rate issue. If that's true then 2K scaling is just a question of programming and could probably be added in a firmware upgrade.
 
Thom said:
What I think you are remembering was that 4K Redcode was limited to 30fps (currently 25) but if they scaled 4K they could potentially get faster framerates with a hard cap at 60fps as that is as fast as the full sensor could handle.

The only way to get faster than 60 was to window the sensor.
Click to expand...
That's certainly correct as far as my memory goes Thom - thanks for helping me regain clarity!!

What I imagine is the question is whether (a) the scaled mode is still on the cards and (b) what it's practical top fps would be, i.e. whether the bottleneck would end up being the sensor clock speed or the FPGA processing/data throughput...

I guess we'll find out in time!

BTW, your sig is great!
 
Hey noclip,

You're certainly correct that 2540p @ 60fps (uncompressed - i.e. no RedCode) is (or rather, will be) possible through the RAW PORT.

I'm pretty sure though, that even if recording options were available, > 30 (currently 25) fps RedCode RAW is not possible due to the processing overheads - I think those FPGA's have got their work pretty well cut out encoding 4k @ 24 fps!!
 
Dominic Jones said:
Out of curiosity, how much emphasis (disregarding the equipment required to achieve it) do you put on the importance of off-speed shooting for dramatic effect? I see it as a very important tool for a cinematographer, but my background is very indie, and so I don't usually get the chance to play with framerates as much as I'd like, and perhaps I'm over-emphasising the importance of the technique simply because, for me, it's a relatively rare tool to have access to - care to share any thoughts on that?
Click to expand...

Well, I haven't done too much of it in my narrative features, but that's partly due to the limitations of some sync-sound 35mm cameras, plus a time issue - you mostly shoot sound scenes in a narrative feature, since, unfortunately, most screenplays are loaded with dialogue. So on a short schedule, it doesn't leave much time to shoot off-speed versions of set-ups except for obvious stuff, like an action scene with explosions or gunfire, etc. which is when you'd bring in an MOS high-speed camera.

However, when I've worked with sync-sound 35mm cameras capable of 50 to 60 fps, I've used the effect more liberally as a result. I don't find much reason to run faster than 60 fps though. In fact, my favorite slow-motion rate is 48 fps.
 
David Mullen ASC said:
You'd think though that the faster frame rate happens first before the scaling processing step happens -- i.e. it would have to take 4K at 60 fps and then scale it to 2K for recording. If that's true, then basically it is still limited to the processing speed of 4K at 60 fps.

I don't think the scaling can happen first, before the 4K sensor is firing 60X a second, so that the camera is only dealing with 2K at 60 fps in the processor.
Click to expand...

You mean something like the sensor itself would send just quarter of the pixels out?

I doubt it. First it suppose some kind of intelligent on-chip controller that would be able to omit 3/4 of data, and secondly there is a Bayer mask, so it's not just like omitting 3 pixels, sending the fourth.

I'm afraid it's work for controller (or computer - whatever it's called) after the sensor outputs all the data .... just at twice the speed.
 
Just that 48 fps is not so slow that it takes up a lot of screen time, yet it is visibly slowed-down action, and it's a nice and neat one-stop exposure correction.

Most features spend a lot of time in the editing room trying to tighten the pacing, which gets hard if the director has fallen in love with shooting everything at 60 fps...
 
I think the sensor could send all 4K pixels out, and scale before doing the compression (getting a speed advantage so it is compressing 2K instead of trying to compress 4K) It may not reach the full 60fps but maybe David's magic 48.

I played around a little in photoshop, trying to think this through. Maybe I'm oversimplifying it (I'm sure I am) but sometimes the simple solution actually works.

I"m sure Graeme already tried this and has a better looking, more complex solution, in the works--but it is fun to speculate.

If I had more time, I would poke around in a Canon Raw file and see if I could see how it would scale. Scaling the R3D file would be a bit unfair as it is already compressed and I think it would be better to do the reduction prior to compression.

Its pretty obvious, so who knows who has the patent on it, but it may work.

bayerScaling.png
 
Brook Willard said:
I wonder how the moiré characteristics of such a scale would be...
Click to expand...

I don't know. It's essentially a direct 1/4 scale in place, ending up with the same ratio of RGB, so it may be pretty good.

Of course, there is really only one way to find out... :)

But As I said, I'm sure Graeme has already got something cooking...
 
Yes that's it. It's simple on the picture, but quite complicated task for sensor itself. I'm afraid it wouldn't be possible to achieve such an effect just by addressing the rows and columns of pixels. (In principle CMOS sensor is just CMOS memory array exposed to the light).
Note that on the picture above it's operation over the complete 4K data. So I still think, the sensor has to deliver all the 4K data to the RED's on-board computer that is due to execute them to receive 2K data and, of course, compress them at the end.
 
I'm wondering how such a scale would actually work out too. I don't see any other way that it would work. Thom's example is pretty much it, I suppose the algorithm could expand and sample overlapping sections for more detailed averaging and weighted results.

There's always a moire pattern when resolving an image to a grid of pixels or photosites and that has to be dealt with in the processing / demosaic. But I'm curious too if this will amplify such an effect, scaling like this or what sort of effect this may have on edge details or fine details in general. Obviously, it isn't simple or we'd have it already. :)
 
You must log in or register to reply here.
Back
Top