2K scaling...

[David Mullen ASC]

Of course I could shoot 2K windowed, but why could'nt you have a shallow DOF on docs ? More than once I've shot docs with mini-35 + HVX, and it looked fine. P2 was a pain in the ass, but 35mm DOF looked great and focus was ok. Why should docs always have a crappy video look ?

For those saying 2K scaled is not doable in camera, then how comes it is able to debayer and downconvert to 1080P and 720P RGB multiple times in camera ? Most high end computers struggle doing that, but the camera is able to do it just fine. Why wouldn't that apply to scaling RAW ?

There is some range of DOF between 1/3" consumer cameras and 35mm -- Super-16 and 2/3" HD cameras for example, which is what 2K windowed would be like. Ken Burn's documentaries are all shot in Super-16 / 16mm and I'd hardly call that a "crappy video look" because of the DOF.

In 2K windowed mode, you'd get the equivalent of 2-stops more depth of field than in 35mm mode. So if you're shooting 35mm at f/5.6 just to make focusing easier, wouldn't you like to be able to shoot at f/2.8 and get the same depth of field as f/5.6? 2 stops is not the difference between "crappy" and attractive depth of field. The five stops extra depth of field of 1/3" cameras, now that's a different story.

Certainly 35mm depth of field is not a problem for static subjects or when you can stop down the lens. But some documentaries are shot in more uncontrollable situations than that. I don't think people understand how hard focusing 35mm at f/2.8 on a medium to long lens is, or how bad the mistakes look on the big screen.

Converting RAW to RGB is simpler than converting 4K RAW to 2K RAW if you just stop and think about what a Bayer pattern is! Plus the camera would still have to convert that to RGB for monitoring. I'm not saying it's impossible, just very, very hard and processor intensive and considering how picky people are about artifacts, it may not be a perfect conversion. I believe when a DSLR shoots in lower resolutions, it always ends up as a JPEG or other file where the color is converted. I don't think think they store smaller files in RAW mode but I could be wrong.

I'm sure Graeme is working very hard on the problem and if anyone can do it, he can.

 

[Emmanuel Cambier]

Actually there is an option for smaller Raw on some DSLR.

Nice to have you moderating around here David.

Emmanuel

 

[Nils Ruinet]

You're right David, S-16 DOF won't give you a "crappy video look", I didn't mean that, this was a bit exaggerated.
What I meant is that even in documentary there are situations where you want a small DOF, and where you are able to keep the subject in focus even with 35mm lenses. Your subject doesnt always move like crazy, and you don't always use very long lenses...
Of course, if you're shooting wildlife animals with long lenses, you'll want to use the windowed mode to make focusing easier, otherwise it will be very hard.

But having the choice is always a good thing, isn't it ?

Also, what I really appreciate with 35mm DOF is that even with shorter lenses (in the 25 - 50mm range) you are able to achieve some nice blur in the background if your subject is not too far away. Such a thing is very hard to get even with S16 or 2/3 lenses... And in this particular case, I think the 2 stops make a difference.
When using longer lenses, you'll get a shallow DOF anyway, even with 2/3 lenses...

Regarding feasibility, I know it's probably very difficult to do, even if processing power wasn't an issue. In fact, a few month ago we all thought it was impossible, until Jim and Graeme mentionned it may be an option.
I'd be very curious to see a quality comparison between 2K windowed and 2K scaled. Usually oversampling results in better image quality so 2K scaled should be better, but as you mentionned there may be artifacts due to the RAW conversion.

Nils.

 

[Leif Thomas]

In 2K windowed mode, you'd get the equivalent of 2-stops more depth of field than in 35mm mode. So if you're shooting 35mm at f/5.6 just to make focusing easier, wouldn't you like to be able to shoot at f/2.8 and get the same depth of field as f/5.6?

That's actually a myth:
Between 35mm and 16mm are 4 stops worth of dof. You can check your Guild Kelly, DOFMaster or other focal-charts for this:
It's the same dof with 50mm f5.6 than with 25mm f1.4. for example.

You can also get it with a little math from the base-formular to calculate dof
(H = hyperfocal distance, f = focal length, u = coc, k = f-stop):

H = f² / (u * k * 1000)

If you double the focal length, you need to multiply the f-stop by 4 to get the same hyperfocal distance (and with that - other related focus-values).

 

[Stephen Williams]

That's actually a myth:
Between 35mm and 16mm are 4 stops worth of dof. .

Hi,

If you project the images the same size you will see David is correct as usual on this matter. Your mistake is to use the same coc

Stephen

 

[David Mullen ASC]

Yes, if you don't change CoC then it's 4-stops, but you should be changing the CoC by half when cutting the size of the original in half (and thus enlarging it by twice to match the larger format), so if your figure for 35mm/4K is .001", then for Super-16/2K, it would be .0005", so the difference is 2-stops.

---

Maybe the easier solution would be to build and sell a separate 2K Bayer-filtered RED camera with a 35mm sized sensor, then it's simply a 2K RAW image naturally without processing (which is the whole point of RAW.) There may be some benefits from that, fewer photosites in a large area, like more sensitivity or less noise.

 

[Eddie]

So if you're shooting 35mm at f/5.6 just to make focusing easier, wouldn't you like to be able to shoot at f/2.8 and get the same depth of field as f/5.6?

Maybe the easier solution would be to build and sell a separate 2K Bayer-filtered RED camera with a 35mm sized sensor

I don´t really get these arguments David. You seem to be arguing against scaled 2k by exagerating some very construed situations where shooting windowed mode has benefits?!?
And the idea that it might be easier to construct a new chip and market a wholly different camera, can´t be your honest opinion?!

The EFP shooters don´t demand much, but 2k scaled would truely be a nice christmas present... As TJ said, -we would be using it all the time.

Again, as I said it is about working hours and storage... Why are people all of a sudden being condencending about the great possibility of not having to bring 50 harddrives in a back-pack, and not having to wait long hours while in the field?!? I mean REDCODE RAW is already a "storage/quality" - compromise, exactly as scaled raw would be. The only difference is that for efp shooters having an less storage-demanding full chip shooting mode would be a HUGE advantage.

Hoping for santa...

All the best eddie

 

[David Mullen ASC]

I'm not against a 2K scaled option at all, I'm just pointing out how difficult it will be to achieve and the basic reason why people are asking for it at all, which is that the size of 4K files makes recording times shorter and processing times longer, so if that problem could be solved, people probably would not be as desparate for 2K scaled.

Also, philosophically, the whole point of a RAW camera is somewhat confused if you have to start processing the image a lot in camera. Just think about it -- RAW is a natural form of 3:1 data compression. But scaling 4K RAW to 2K RAW is not natural at all, it requires a lot of processing.

If what people really want is a 1080P camera with a 35mm sized-sensor, or a 2K RAW camera with a 35mm sized-sensor, it might be simpler to actually build such a camera optimized for that. Maybe not, maybe in-camera 4K RAW to 2K is simpler to do ultimately.

But I'm not against more options in the camera if they are possible. I'm sure 2K scaling, if it works well, would be highly useful.

I mean, what if one of the artifacts of 4K RAW to 2K RAW, and then converting to HD for onset monitoring, is a two-second delay, which would require audio delaying (in a buffer) in camera or shifting the audio by two seconds in post? Would you still use it if the director is screaming at your operators because he can't get used to a 2 second delay?

 

[Harry Clark]

But the camera ALREADY does a downconvert live for monitoring. I think what some of us are asking for is a 1080p or 2K camera with 35mm DOF. I KNOW that the specs are "subject to change, etc., etc." but it seems as if a lot of the promised flexibility of the camera has been tossed. I also wonder if the folks at RED understand real-world production timelines and requirements. If I'm offering up my Red One for a TV commercial, and the agency just wants a 1080p HDCam tape at the end of the day, I'll probably be stuck with a Genesis (which I suppose is not the worst thing in the world) Or I can offer up my HDX-900 with Pro-35 and Cookes (again, not terrible)
And I get it; the camera is optimized for 4k, especially the OLPF. Can we have a way to get the client footage at least as fast as a downconvert on tape or film dailies? I can shoot 35mm, wrap at 10:00 pm, and see HD dailies at 6:00 the next morning. As far as I can tell, I can do that with Red with the computers available 3 years from now. Not today. Not tomorrow.
Again, a dedicated processor box might be a nice solution for professionals with less time and more money. For the indie-film set, then yes, a render farm in Mom's basement will do...
And remember, lots of us are DPs; not editors and not post companies and not production companes. Our involvement after wrap is typically minimal. That's why the notion of "just set it to render back at you office" doesn't work. Please. Small Red Box. Makes other formats fast. Don't care how much $$$.
Cheers,
Harry

 

[David Mullen ASC]

Well, a 2K RAW recording doesn't really solve that problem -- it still needs to be rendered, just that the rendering will take less time (unless by shooting in 2K scaled, you shot twice as much footage... )

Trouble with uncompressed 4:2:2 and 4:4:4 1080P are the data rates for recording are quite high. So would you choose that option or 2K RAW and having to render it? I mean, if the client wanted an HD tape at the end of the day, wouldn't the footage have to be processed into an HD codec like HDCAM or DVCPRO HD or transferred from some sort of hard drive to HD tapes?

Don't all of those steps (4K RAW, 2K RAW, uncompressed HD to hard drives, etc.) take some sort of time after the shoot is over to create a standard HD tape, short of sending a signal from the RED directly into an HD tape deck on set? Except for that scenario, it seems to me you are stuck doing some sort of conversion and transfer work after the filming is over, if the client really wants an HDCAM or DVCPRO HD tape to take back.

Again, this is not an argument against 2K scaled, like I said, I know there is a use for such a feature.

But short of putting dual-link HDSDI outputs on the camera sending 1080P out to HD decks, I don't see a way around having to render the RED footage in post. Even if you recorded 1080P internally to CF cards or something, compressed using REDCODE maybe, wouldn't it still have to be converted to a standard codec for HD tape, if that's your delivery requirement?

 

[Jeff Kilgroe]

But the camera ALREADY does a downconvert live for monitoring.

But that's not a RAW conversion. That's a conversion within RGB space... 4K scaled to 1080p RGB was in the specs up until a few months ago. There were some heated debates here on the forum and talks of possibly (some day) having multiple firmware versions to allow for RGB modes like 1080p from the full sensor.

I also wonder if the folks at RED understand real-world production timelines and requirements. If I'm offering up my Red One for a TV commercial, and the agency just wants a 1080p HDCam tape at the end of the day, I'll probably be stuck with a Genesis (which I suppose is not the worst thing in the world) Or I can offer up my HDX-900 with Pro-35 and Cookes (again, not terrible)

Oh, they're aware of such production considerations. The camera is still in development, and it's still gaining acceptance. Many of these concerns presented themselves with the HVX200 and recording to P2. Short record times initially on 4GB cards, consolidating and copying media then was a lot slower than it is with RED's current workflow. People were worried if producers would accept it, etc.. They did.

As RED becomes more established and accepted, I don't see where there will be too many problems. Clients who want to shoot with RED will just have to accept the workflow restrictions (or slower turn-around time vs. shooting most HD cameras). It's still faster than developing film. If time is that much more important to them than quality and they choose Varicam or F900 over RED because they need that tape as soon as the director yells 'cut', then I guess that's what they want. Eventually, I don't see where RED will be any different than HDCAM. The software tools are free, no reason that at the end of the day you can't just hand over a RED DRIVE or a box of CF cards for the producer to forward to his post house or editor.

And I get it; the camera is optimized for 4k, especially the OLPF. Can we have a way to get the client footage at least as fast as a downconvert on tape or film dailies? I can shoot 35mm, wrap at 10:00 pm, and see HD dailies at 6:00 the next morning. As far as I can tell, I can do that with Red with the computers available 3 years from now.

For dailies or rushes, you can just use the QT wrappers. I can take 2 hours of RED footage and burn it via the QT wrappers for viewing on a DVD just as fast as I can with DV or HDV footage. Or I can play them instantly to an HD monitor right from the CF card they were shot to. You can take the contents of an 8GB CF card and have it viewable in any standard DVD player in a matter of minutes. You don't have to process through the 4K footage until you need full resolution.

In the near future we will have the ability to create color grade LUTs that can be loaded onto the camera to apply to monitor output in place of the standard REC 709 curve. These LUTs will also transfer to the QT wrappers for immediate playback or drop-in to NLEs or other software.

 

[Larry McKee]

In the near future we will have the ability to create color grade LUTs that can be loaded onto the camera to apply to monitor output in place of the standard REC 709 curve. These LUTs will also transfer to the QT wrappers for immediate playback or drop-in to NLEs or other software.

That will make my clients very happy. Some of their clients won't understand why the picture doesn't snap. Having the LUTs will make things go a lot smoother on set.

 

[Alan Skinner]

Adapt and Conquer

Adapt and Conquer

External realtime converter, with many settings possible, to be connected to the camera, or not.........

Everyone gets what they want. I get 4K 60 fps in a fairly compact package that (by the the time I receive RED #2973 at least) RED drives and CF card sizes make practical.

Folks in the field without access to render farm get their output to hand off and others who are expecting to take some time with the big, nasty RAW files back at the studio.

I know it makes for one more thing to hang on the camera or off of your belt but I second the idea and would pay $$ for it (when/if I needed it)

I don't want the camera to be any bigger than necessary.

My .02 FWIW..................

 

[Gavin Greenwalt]

http://digitalproducer.digitalmedianet.com/articles/viewarticle.jsp?id=183311

 

[Eddie]

Dear Graeme,

do you have any idea of the improvements in conversion times with 2kscaled compared to 4k...
On another thread 0.9 secs per frame was mentioned as realistic with a decent setup. That is around a 24hours for an hour worth of material. How would the numbers crunch for a 2k scaled, all things being equal?

This is of major importance for doc shooters. if you want to convert before editing... nobody likes to sit around and babysit a computer for a month...
If you can pull a rabbit out of the hat.. (that is, if there is anyone left) I believe it would make a hugh difference for a lot of shooters out there.
Don´t listen to the high-end boys with regard to this feature... they are not the ones who babysit their material themselves...

 

[Graeme Nattress]

I can't answer such a hypothetical question.

Graeme

 

[I Bloom]

do you have any idea of the improvements in conversion times with 2kscaled compared to 4k...

It seems like you can test the speed of conversion using 2K windowed footage.

 

[Axel Mertes]

It seems like you can test the speed of conversion using 2K windowed footage.

I suggest to make that speed comparison!

Is anyone able to volounteer/supply the same kind of footage (e.g. a fixed still image) in both 2k windowed and 4k as .R3D files, same length in time?

I suspect the speed difference in decoding should be minimal depending on the footage.

Anyhow, if at some point we can use other apps for decoding as well (such as Digital Fusion) we can easily make use of our Render Farm for bulk conversion. Maybe RED introduces such functionality at some point on its own. It simply a difference if you process on one CPU or hundrets at the same time.

Cheers,
Axel

 

[Eddie]

Another interesting speed test would be cropping a 2k to 1080p.
This would give an idea of the speed improvements for 2k pre-scaled.
If i remember it correctly graeme mentioned considerable benefits from cropping compared to scaling in redcine...

If this holds true, 2k scaled in-camera would benefit everybody with a pressing deadline, and limited processing powers... doc-shooters, indies, hell even the big boys are sometimes in a hurry.


all the best

 

[Jeff Kilgroe]

Eddie,

The biggest advantage of cropping 2K to 1080 rather than scaling is the scaling operation can have a softening effect on the image. I don't see a big difference in speed one way or another (crop vs. scale from 2K to 1080p).

As for post-processing 2K scaled, it should theoretically be the same as processing 2K windowed footage. It will be a 2K frame of wavelet-compressed bayer data, regardless of whether a windowed section of the sensor was captured or the 4K frame was scaled-down.

In terms of post-processing speed, I think the biggest thing coming down the pipe is going to be the command-line render tools. These will allow REDCINE frame processing to be delegated out to a render farm, or even to make better use of multiple CPUs/cores on multiprocessor machines. I think we'll also get speedier REDCINE performance in the near future... It currently doesn't make use of the full system resources on any system I've tried. Most of the time, the CPUs it does utilize are loaded to about 40%, so somewhere there's a bottleneck in the pipeline or the operations aren't parallel enough and we have CPUs spending more time waiting on other CPUs rather than executing the next operation.