why would staying in 3K not be as sharp as 3K to 2K

[Nicholas Natteau]

I didn't mean to start a new thread here but I'm not able to reply to Barry Green's post in the original thread i started "how many resolutions can Scarlet shoot in". It seems I don't have sufficient access as a member to post a reply. Can anyone tell me how to upgrade my membership so I can post a reply?

Any way, Barry if you're reading this: can you please elaborate a little more on that point you made in my original thread regarding Scarlet and resolutions. You wrote:

>>>>
With Red, you WANT to do conversions. The single Bayer sensor won't deliver as much detail as the nominal system says. For example, in a Red One, "2K" mode doesn't deliver anywhere near 2K of resolution, it's more like a little better than 720P. But if you shoot in 3K and then downconvert to 2K, you get astonishingly sharp, perfectly rich, detailed thick and lush 2K footage.

You can't make 4K of detail from a 4K chip, but you can make a great 3K's worth of detail from a 4K chip. Red's Scarlet is probably 3K specifically because a 3K chip will make extremely good 2K final images.
>>>>>>>>

Reading this I'm a little worried: "2K mode doesn't deliver anywhere near 2K of resolution, it's more like a little better than 720P"

I think I'm missing something here. On Red's brochure from NAB 2008, on the last page they show the different resolutions: 3K looks much bigger than 2K and 2K looks much bigger than 720p.

 

[Cüneyt Kaya]

red shoots in raw, if you decode this to a rgb image you will loose some resolution....

reds 4k raw is around 3,2 k rgb.....technically seen....but a lot of red footage was shown at 4k and no visible difference...

you are able to decode a 4k image out of a 4k raw image, but then you would have heavy aliasing effects....hence graeme did some magic trick and the sweet spot between aliasing and resolutio is technically seen 3,2k for a 4k raw image.
read graeme`s why bayer thread.

and dont be too scared as long as you dont have to shoot star wars.

----i am not a tech wizzard, if you need more in deph details, talk to red directly
----you should be able to answer posts----ask a moderator, send him a pm

 

[Lawrence Bansbach]

you are able to decode a 4k image out of a 4k raw image, but then you would have heavy aliasing effects....hence graeme did some magic trick and the sweet spot between aliasing and resolutio is technically seen 3,2k for a 4k raw image.
read graeme`s why bayer thread.

and dont be too scared as long as you dont have to shoot star wars.

If what you say is true and the arithmetic holds, why would shooting Star Wars pose a problem for Scarlet? A de-Bayered Scarlet image would be 2.4K, much higher-res than the 1080p of the last two films.

 

[Graeme Nattress]

The big issue is not that after OLPF (the optical filter that stops aliasing occuring) and the bayer sensor you "only" get around 78% of the linear rated pixel resolution, but that most cameras "allow" significant levels of aliasing so that they can meet their "rated" pixel resolution and "ruin" their image quality in the process. I think one of the reasons why people like the feel of the RED images so much is that we like to avoid the extremes of aliasing wherever possible....

FYI, 2k would measure about 1600 lines, which is significantly more than 720p and more than 1080p HDCAM.

Graeme

Graeme

 

[number6]

The big issue is not that after OLPF (the optical filter that stops aliasing occuring) and the bayer sensor you "only" get around 78% of the linear rated pixel resolution, but that most cameras "allow" significant levels of aliasing so that they can meet their "rated" pixel resolution and "ruin" their image quality in the process. I think one of the reasons why people like the feel of the RED images so much is that we like to avoid the extremes of aliasing wherever possible....

FYI, 2k would measure about 1600 lines, which is significantly more than 720p and more than 1080p HDCAM.

Graeme

Graeme

Guess this is more of a post question, but if you dounsample 3k to 2k in post, do you still end up with 1600 lines ersatz 2k? I would think not but would like to hear definitively.

 

[Graeme Nattress]

If you downsample 3k to 2k, you can get the resolution arbitrarily as close to 2k as you want, balancing the effects of the the downsample filter (ringing, edge sharpness, and aliasing). However, you should be able to get a much better image in all those above qualities than any camera shooting at or around 2k.

Graeme

 

[John Caballero]

Graeme, I am still learning is this area and my question is: when you downsample from 3K to 2K do you loose pixels, I guess it is the term, in the process or is it the same amount of them reduced to smaller image size? Sorry if I sound dumb but I want to find out if that is why the image would look better by doing the downsample.

 

[Graeme Nattress]

The pixels go into a downsample filter, which extracts all that is "resolution" and "detail" up to the amount that can fit in the lower resolution, so we extract all that is 2k say, and no more, from the 3k and then place that 2k detail into a 2k file. That's how the downsample works. It's essentially a low pass filter, set to only allow frequencies below 2k through.

Graeme

 

[John Caballero]

Thank you Graeme.

 

[Dj Joofa]

the eternal question, isn't it, my guess is you loose them, but I am not sure where they go

EDIT: Message remove since my example of square wave was contrived and confusing as pointed out by Graeme.

 

[Graeme Nattress]

Joofa, you're using square waves which you know are not bandwidth limited.

Graeme

 

[Dj Joofa]

Joofa, you're using square waves which you know are not bandwidth limited.

Graeme

You are right. But the point of the exercise is just an illustration into effect of photosite size.

 

[Dj Joofa]

Joofa, you're using square waves which you know are not bandwidth limited.

Graeme

You are right. But the point of the exercise is just an illustration into effect of photosite size.

The more I thought about it it does appear that Graeme is right and my example for square ware is a bit contrived and confusing. In theory when an analog signal is digitized it creates multiple copies of the same base band frequency spectrum at intervals determined by the sampling rate. That is why a low pass filter can use a sinc function to recreate original signal. Now ignoring the effect of integration time of the sensor (considering the signal is not changing) and assuming photosites have some width (not point sampling) then it means that those copies of the base band translated spectrum will have their amplitude adjusted by (a different) sinc function resulting from the fact that it is not point sampling. The original signal may still be reconstructed.

The bottom line even if the photosites have some width (which of course they do), since the width is less than the sampling interval, and the maximum bandwidth considered has been fixed by that sampling interval things may be fine.

 

[Barry Green]

, Barry if you're reading this: can you please elaborate a little more on that point you made in my original thread regarding Scarlet and resolutions.

What I meant was, if you want the best 2K you can get, you don't want to shoot in 2K mode. You'd want to shoot in the 3K mode and downconvert. Graeme's figures for Red One show that you get about 78% of the stated mode's resolution, so if you shot 2K mode his figures show you'd get about 1.6K of res. But if you shot in 3K mode and downconverted, you'd get a full 2K of res.

Now, that's math. Viewing may result in different interpretations. Having tested my Red in 2K vs. a 720P camera, I can say that the Red in 2K mode was a little tiny bit sharper, but not night and day. DuArt had a test of a Red in 2K mode vs. a Sony EX1 in 1080P mode, and (surprise, surprise) the EX1 looked sharper. I'm not surprised. Tying both hands and one leg behind the Red's back will give you results like that. But the Red in 3K would have spanked the Sony.

The main benefit to shooting in a lower res on a Red camera isn't because that's what your delivery method will be (i.e., if you're doing a 2K finish, you don't really want to be shooting in 2K mode). The benefit is that you can scan the chip in faster frame rates in the lower-res modes. So 120 fps in 2K, vs. 60fps in 3K, vs. 30fps in 4K... there's a very real benefit there, even if the image isn't as spatially sharp.

Reading this I'm a little worried: "2K mode doesn't deliver anywhere near 2K of resolution, it's more like a little better than 720P"

That was what I observed. It was better, but not by much, it looked like good 720P. Good 720P is good, by the way, but it's not 1080P and it's not 2K.

I think I'm missing something here. On Red's brochure from NAB 2008, on the last page they show the different resolutions: 3K looks much bigger than 2K and 2K looks much bigger than 720p.

Yes, but that's frame size, not resolved detail. A world of difference there.

The sharpest mode on the Scarlet will undoubtedly be its 3K mode. Undoubtedly. If you're aiming for theatrical projection or 3K deliverables, shoot 3K, no question.

The question comes about when you're aiming for delivering in a different mode, such as a 1080P final. If you want the sharpest, clearest, crispest 1080P you can get (for blu-ray or whatever), then you don't want to shoot in a native 1080P mode on a single-chip Bayer sensor system. You'd want to shoot in 3K to deliver super-ultimate-mega-crisp 1080P.

If you're aiming for standard-def DVD, you could use a 720P mode or higher (2K, 3K, 1080P) and still get a great crisp sharp DVD. Depth of field will be shallower when you use the larger frame size, of course, because the wider field of view means you have to get closer.

Basically, after a few months with the Red One, I predict that Scarlet users will put the camera in 3K mode and never deviate. The only reason to come out of 3K mode (that I can think of) would be if you needed faster frame rates (the smaller frame sizes should allow for faster scanning rates).

 

[S. Um]

The big issue is not that after OLPF (the optical filter that stops aliasing occuring) and the bayer sensor you "only" get around 78% of the linear rated pixel resolution, but that most cameras "allow" significant levels of aliasing so that they can meet their "rated" pixel resolution and "ruin" their image quality in the process. I think one of the reasons why people like the feel of the RED images so much is that we like to avoid the extremes of aliasing wherever possible....

FYI, 2k would measure about 1600 lines, which is significantly more than 720p and more than 1080p HDCAM.

Graeme

Graeme

Graeme,

I was under the impression that 2k measures the number of horizontal pixels, not vertical pixels. So for a 16:9 aspect ratio, 2k would be 2048 x 1152 pixels. If de-bayering yields a 78% resolution, then that would leave you with about 1600 x 900. This is somewhere between 1080p (1920x1080) and 720p (1280x720), and it supports the observations that Barry Green made.

 

[Graeme Nattress]

Well, it's often hard to see between 720p and 1080p.....

One of the points I'm making though, is that, for instance the EX1 from the charts I've seen allows through a very significant amount of aliasing. And one of the reasons people really feel comfortable with the look and feel of RED footage is that we don't (allow through significant aliasing).

You can't have it both ways. We're taking the aesthetics + engineering correct approach. Other designers are free to take other approaches, but until recently, a lot of forum readers didn't really know about this side of things, and the detrimental effects on an image caused by aliasing.

Graeme

 

[Lawrence Bansbach]

Basically, after a few months with the Red One, I predict that Scarlet users will put the camera in 3K mode and never deviate. The only reason to come out of 3K mode (that I can think of) would be if you needed faster frame rates (the smaller frame sizes should allow for faster scanning rates).

Scarlet, at 3K, will be able to record at a sustained 120 fps and burst 180 fps. It isn't certain that going down to 2K or 1080p, even if possible with Scarlet, would necessarily enable higher sustained frame rates. Although it might be technically possible, Red may not implement it. If that's the case, then there would virtually no reason to shoot in a sub-3K mode.

 

[Barry Green]

Scarlet, at 3K, will be able to record at a sustained 120 fps and burst 180 fps.

At 3K size? Has that been confirmed? Or do you have to drop to 2K for 120fps?

If that's the case, then there would virtually no reason to shoot in a sub-3K mode.

From my experience with the Red One, I'm fairly certain I'm gonna glue the Scarlet in 3K mode and never deviate.

 

[Lawrence Bansbach]

At 3K size? Has that been confirmed? Or do you have to drop to 2K for 120fps?

From Mike Curtis, "HD for Indies" on ProVideo Coalition.com: "I asked, and got confirmation, that Scarlet WILL do 120 fps at 3K. Burst mode up to 180 fps." Of course, it always helps to bear the standard Red disclaimer in mind.

 

[Barry Green]

Well, dang!