Cinematography Mailing List - CML
    advanced

2K Transfer on a Spirit

 

If one, let's say, shot 4-perf Super-35 but only wanted to transfer a 2.35 : 1 area of the frame -- to be converted to an anamorphic image for output to 35mm scope -- can one allot the full 2048 x 1556 pixels (?) to just the 2.35 picture information, as opposed to have to scan the whole Super-35 frame at 2048 x 1556 but only end up using approx. 2048 x 870 pixels to achieve 2.35, or only scanning 2048 x 870?

I'm really asking about films like "O Brother Where Art Thou" and "Lord of the Rings" that were shot in Super-35 but output to 2.35 anamorphic. It seems to achieve maximum resolution allowable in a Spirit datacine transfer, you would only transfer the 2.35 area of the Super-35 frame and fill the full 2048 x 1556 pixel array with the 2.35 image (I'm really sorry if these numbers are wrong -- I'm just sort of guessing here.)

But someone just told me that a Spirit can't resize a frame in order to use all of its pixel array, so even though a 2.35 area on a Super-35 frame would receive the full horizontal resolution, 2048 pixels, it would only use up about 870 pixels vertically as opposed to squeeze 2.35 across 2048 x 1556 or whatever.

If that's true, then the resolution difference between a Spirit 2K transfer of a 2.35 area of the Super-35 frame at 2048 x 870, as opposed to a transfer to 24P HD-D5 (1920 x 1080) and then cropping down to 2.35 (approx. 1920 x 800) wouldn't be that much.

So exactly how was the 2.35 area of the Super-35 frame handled in the case of the digital intermediate process used for "O Brother Where Art Thou", "Amelie", "Lord of the Rings", and "The One" -- all recorded out to 2.35 anamorphic?

It seems to me that if the Spirit can't use all of its potential pixels for the 2.35 area of Super-35, one would still be better off shooting with anamorphic lenses and thus being able to scan more film negative vertically in the Spirit.

David Mullen


David Mullen wrote:

It seems to me that if the Spirit can't use all of its potential pixels for the 2.35 area of >Super-35, one would still be better off shooting with anamorphic lenses and thus >being able to scan more film negative vertically in the Spirit.

Yep, I think that makes sense unless you have good reason to shoot S35. And there are good reasons -- ease of a common-topline TV version, smaller and faster and shorter lenses, etc.

I did like the film Go, seen only on video, which was shot S35 (I assume) with an Aaton 35. No camera noise problems that I noticed . Pretty funny film, and I liked the way it was made.

Jeff "933 channels and still nothing on" Kreines


David Mullen wrote :

But someone just told me that a Spirit can't resize a frame in order to use all of its >pixel array, so even though a 2.35 area on a Super-35 frame would receive the full >horizontal resolution, 2048 pixels, it would only use up about 870 pixels vertically as >opposed to squeeze 2.35 across 2048 x 1556 or whatever.

First, the Spirit doesn't have a "pixel array" -- it's a line array telecine -- meaning that it scans the film line by line as opposed to frame by frame (which of course you know).

Since you are maintaining the aspect ratio of the image as photographed (i.e. it's not anamorphic) you will of course want to use the full width of the CCDs -- meaning, as you suspect, that the vertical information in extracting a 2:35 image from S35 will of course be less than that of a 1.85 or 1.33 capture, due to frame geometry, pure and simple.

I can imagine that there are ways to modify the Spirit (factory job) so that the "gearing" (metaphorically) between the sprocket and scanning is changed, and vertical res is increased.

A different option would be to do 4K scans, which I believe a Millenium can do.

I did see some occasional aliasing in "O Brother" as well as some limitations in terms of color (obviously, it was heavily manipulated in timing, but some of the color looked -- not unpleasantly -- a bit like a hand-painted postcard). But these were pretty subtle things.

Jeff "wants to do 2K timing at home in his spare time" Kreines


someone just told me that a Spirit can't resize a frame in order to use all of its pixel >array

I was in a telecine suite today and asked this question and was told that the very first machines had a limited ability to resize but most models out there have no problems with any format.

Tom Gleeson D.O.P.

Sydney Australia

http://users.bigpond.net.au-gleeson


someone just told me that a Spirit can't resize a frame in order to use all of its pixel >array

CRT based telecines such as C-Reality are very flexible for various aspect ratios filling the data files. That was the design philosophy, to do the ratio/size change optically before sampling, just like in camera. Also the front end is full RGB bandwidth and I believe 4K DATA output is now available on the street.

When choosing a telecine for a project there are many factors to consider, this is just one. All the telecines on the market have their strengths and weaknesses.


Sorry to keep pestering people on how the Spirit works...

But if the film is in a fixed position to the line array scanning the frame, so that 2048 pixels is the maximum horizontal resolution for Super-35 / Full Aperture and 1828 pixels for Academy -- does that mean that the 2.35 : 1 picture information off of a Super-35 negative will ultimately be limited to 2048 x 870 (approx.) resolution?

And how does the Spirit scan a Super-16 frame then -- does it only end up being scanned at 1024 x 620 pixels?

David Mullen

Cinematographer / L.A.


The CCD array in the Spirit is fixed at 1920x1080 (not 2048x1556 which is true 2k). So if your going to go 2:35:1 your numbers of 2048x870 sound about right. The line array CCD can't change it's shape to fit the aspect ratio. Facilities that need 2048x1556 from the Spirit do a up rez with interpolation software. All blow ups by the way are done in the

frame store. This sound a bit undesirable but in reality the Spirit does a excellent blow up.

An advantage of the CRT telecine is that it can change it's sample size at will.

Jim Mann

Freelance Colorist


David Mullen wrote:

And how does the Spirit scan a Super-16 frame then -- does it only end up being >scanned at 1024 x 620 pixels?

The Spirit has different gates for S16 and S35, with different lenses -- so S16 also uses the full width of the CCDs.

does that mean that the 2.35 : 1 picture information off of a Super-35 negative will >ultimately be limited to 2048 x 870 (approx.) resolution?

Unless there's some clever oversampling tricks we don't know about, I'd guess that's the case. The question is, can a Millenium or C-Reality, using tube tricks of some sort, get a higher-res scan of this material, since they're not limited to fixed CCD pixels.

Jeff "better than broken pixels" Kreines


The CCD array in the Spirit is fixed at 1920x1080 (not 2048x1556 which is true 2k).

So other than color information and compression issues, is there any real resolution difference between a 24P HD transfer on a Spirit to HD-D5 Ð and a 1.78 : 1 "2K" transfer on a Spirit? I'd always assumed that a 2K transfer on a Spirit had more resolution than an HDTV transfer.

So how exactly does it handle Super-16 in terms of not being limited to a small area of the line array? An optical system?

David Mullen

Cinematographer / L.A.


David Mullen wrote :

So how exactly does it handle Super-16 in terms of not being limited to a small area >of the line array? An optical system?

A different lens/gate combo, which spreads the S16 image over the entire width of the CCD.


Yea, roughly the same CCD area is used for both 35mm and 16mm by optically filling the line array with a lens change, much like an optical printer would. You need to look out for field sharpness on that s16 gate however.

In terms of hardware up-res from 1920 to 2048, even and "excellent" blow-up is often undesirable for film work. You gotta think that to achieve even a near-real-time uprez via hardware, there must be some cheating going on by using a lower quality image filter (sinc, anyone know?). Did some tests on a high-profile feature looking at laser output from a Spirit and saw significant aliasing in near-horizontal lines in the frame - stairs, wall dressings, etc. Now not all Spirits are tuned the same, so not all necessarily have this "problem". And perhaps it's not a Spirit issue but rather a 2k issue .

>The question is, can a Millenium or C-Reality ... get a higher-res scan of this material

In theory, but resolution is not just a number. Spot size comes into play with CRTs. You can produce a 6k CRT, but if the spot is blooming and yielding a 3.5k equivalent (for example), then...?

Doug Delaney

Imaging

Pacific Title


Yea, roughly the same CCD area is used for both 35mm and 16mm by optically >filling the line array with a lens change.......

How does say 2 perf or even 3 perf fit into this ? For a 1.85 extraction to 16x9 would you lose anything ? I am imagining the loss if any would be marginal ? Also would this issue make any difference if your output is in 2k data or the HD format ?

Tom Gleeson D.O.P.


I would think the "problem" would be more of a 2k issue than a Spirit issue. But there's no question, if I had to blowup a scene that had potential aliasing problems, and I needed to play with it in the Inferno, I would scan on a C-Reality or ITK at 4K , output to inferno and Arri laser it back to film.

Jim Mann

Freelance Colorist


>>other than color information and compression issues<<<

Oh thanks David

I really enjoy a good laugh in the morning!

It's a What's the difference between Rembrandt and a house painter, other than talent, type question :-)

I've felt for a long time that we're too tied up in resolution, the main concerns should be, apart from bandwidth recordable on the original material, exactly colour depth and compression. Oh and you could add motion artefacts to that as well.

Geoff Boyle FBKS

Director of Photography

www.cinematography.net


I've felt for a long time that we're too tied up in resolution, the main concerns should >be, apart from bandwidth recordable on the original material, exactly colour depth >and compression.

I sort of agree in principle, but as someone who generally shoots for print projection, not telecine transfer, resolution can't be ignored either. In some shots in a movie, resolution or color space or compression might vie with each other as the most prominent feature. In some shots, for example, compression is less obvious than lower resolution.

I like the idea of the digital intermediate system -- it's just ending up with an 2.35 scope image from a Super-35 negative that ends up being only 2048 x 870 pixels... Since I shot "Jackpot" in 24P HD and cropped it to scope -- basically 1920 x 800 -- I was hoping to follow that up with something more significantly better in resolution, not just slightly better.

Anyway, eventually we'll have those 4K C-Realities, Millenium-TK's, and new 4K Spirits up and running, right? I'd hate for the digital intermediate technology to become stuck at 2K.

The other advantage I see to the Spirit is that the method of scanning, the light source, tends to minimize grain and dirt & dust problems, which seems why these digital intermediate Super-35 films out in theaters this Christmas ("Amelie", "Lord of the Rings") seem less grainy than the other Super-35 blow-ups made using the same Kodak stocks. Some people have been complaining about the softness of "Lord of the Rings" but it didn't seem too bad to me.

Just saw "Black Hawk Down", which looked like a conventional optical printer blow-up from Super-35 to anamorphic - really crisp but gritty, which probably suited that film better than something softer but less grainy. Great photography & directing, I thought.

Actually, the main reason why I've been questioning the pixel resolution of the Spirit transfer isn't so much because I'm weighing D.I. versus Optical Printer -- but doing a D.I. with a Super-35 negative versus an anamorphic negative. I can't get the look I want without the D.I. anyway, so what I'm debating in my mind is the camera format, not the post approach. At first I thought that the extra size of the anamorphic negative would sort of be offset by the "equalizing" factor of a 2K limit to the resolution, but now it seems that even with a Spirit transfer, there is more pixel information devoted to an anamorphic frame versus the 2.35 area of a Super-35 frame.

David Mullen

Cinematographer / L.A.


David Mullen wrote:

but now it seems that even with a Spirit transfer, there is more pixel information >devoted to an anamorphic frame versus the 2.35 area of a Super-35 frame.

Lots more.

Since I shot "Jackpot" in 24P HD and cropped it to scope -- basically 1920 x 800 -- I >was hoping to follow that up with something more significantly better in resolution, >not just slightly better.

I'm not sure it's fair to look at this as a resolution-only thing. The Spirit transfer from film is going to have better color depth and a little nice random grain that HDCam lacks -- but, hmm... I wonder if the Anamorphic S16 that JP (the lurker) mentioned wouldn't have a similar number of pixels devoted to a frame as S35 2.35 extraction?

Jeff "pixels, schmixels" Kreines


Oh no question David, you'll still get much more data if you go anamorphic as an origination format.

I need help here but I believe that at the distances people normally sit away from a screen that the differences between a 2K and optical finish, in 1:1.85, are undetectable.

At a recent conference on of the panel said that there were only 6 people in the auditorium that would be able to tell the difference.

Most of the audience thought he was referring to eyesight, in fact he was referring to the fact that the panel, 6 people, were the only ones close enough to tell!

Geoff Boyle FBKS

>Director of Photography

>EU based


It always amazes me that even at this late date, how a Spirit class scanner gets info off of film seems to be so poorly understood. The machine has one "detail" channel with 1920 pixels, and three color sensors for RGB at about 1000 pixels, and makes SD, HD or "data" out of this. Any repositions, zooms, ect, are done by math. You can imagine for yourselves the limitations for doing film work this implies, and where the impetus for spirit II comes from

I have been involved in a number of evals of 2k vs all film, and there are a number of factors that seem to interplay as to what the outcome seems. One is if it really is 2k. One is the bit depth. One is the nature of the material and how well it's photographed (note:this holds for 24p capture as well) . One is the sophistication of the audience doing the judging. As always, the quality of the display systems is a factor.

If we were to take the consumer sensibilities as our guide, TI projectors at 1280x and an HD production pipeline would be fine.

I have yet to see a critical audience see images that have gone through a digital process and back out as completely "transparent", although most of us would agree that the process is very good when done well.

Lou


Jeff wrote:

A different lens/gate combo, which spreads the S16 image over the entire width of >the CCD.

If I may piggyback my own question onto this thread.

It would seem the Spirit is doing an optical "blowup" here, so to speak. So it would be scanning at the same res as in 35mm but scanning an optically magnified image.

Does a C-Reality or a Millennium then, in 2K or 4K mode scan with the same resolution i.e. in "scans/mm" (what is the term I should be using here ?) with 16 and S16 as it (C-Reality or Millennium) does with 35mm ?

Is choosing 4K instead of 2K perhaps even more important if the camera negative is a 16mm format ?

P.S. Maybe it's just me, but I think I'd chose information (sharpness, grey scale, colorspace) over "hiding grain" in a 16/S16 DI any day: grain I can deal with on the front end (choice of stock, light & frame a certain way etc) or even live with it - (what it is) but a plastic looking image: I don't think so...

-Sam Wells

film/nj/usa


How does say 2 perf or even 3 perf fit into this ?

Roughly the same as it would for a s35 finish. The unused image area north and south of image is not printed when duping s35 for anamorphic release. So, in a way, you're only really getting, say, 2048x872 (2.35AR) of resolution when dealing with this digitally. I think there may be a way of slowing the film thru a Spirit gate to get a non-square pixel and thus increasing vertical resolution, ie. 2048x1556 from a 2.35 neg, or a 3-perf neg, or whatever. Anyone confirm?

>I would scan on a C-Reality or ITK at 4K , output to inferno and Arri laser

>it back to film.

Funny how the term "scan" is more and more loosely used. I guess I think of a "scan" as an uncompressed, full-bandwidth digital capture of an image at 12-bit minimum as opposed to a sub sampled (with respect to chroma as Lou pointed out) capture. I must admit I'm not as familiar with the ITK or C-Reality as I should be, so please correct me if I'm wrong in suggesting that these machines are not capable of this yet.

The first thing I look for when screening a film gone thru DI is color quality rather than resolution or apparent sharpness. For the average scene, 2k is sufficient as has been the case in the VFX world for a number of years. Even if top-on-the-line imaging tools are used, 4k film scan, 16-bit color-correction (or better), laser output via ARRI or Kodak, there's plenty of room for slop. Hand a young DP an Arricam and a set of S4 primes and see what you get compared to, say, Mr. Storaro. It ain't the tools at a certain point in the curve.

Doug "not sure if Storaro likes Cooke" Delaney

Motion Picture Imaging

Pacific Title


>P.S. Maybe it's just me, but I think I'd chose information (sharpness, grey scale, >colorspace) over "hiding grain" in a 16/S16 DI any day: grain I can deal with on the >front end

Mainly by shooting on slower-speed film stocks, which may not always be practical. I think it comes down (as always) to WHAT DO YOU WANT IT TO LOOK LIKE?

Optical printing Super-16 to 35mm through dupes has a look; converting Super-16 to 35mm using a digital intermediate has a look -- which approach you choose should depend on which gives your project the look it needs. You might have to do color-correction tricks only possible digitally. You might want as little grain as possible.

In the case of Super-35 and digital intermediates with an output to 2.35 scope, my concern is the resolution loss from cropping that scan to 2.35. But in the case of using a Spirit transfer to "blow-up" Super-16 to 35mm, I see more pluses than minuses compared to optical printer blow-ups, which have all sorts of problems connected with them.

I did one Super-16 feature, for example, and we had to do all of our optical work and titles in 35mm and intercut them with our 35mm blow-up, because doing dupes in Super-16 and THEN blowing up the image through duping was just too big of a quality loss. And while we blew-up directly from Super-16 neg to a 35mm I.P. to get a little extra quality, we now had dirt & dust blown-up and permanently photographed into our 35mm I.P., making our final transfer to home video rather tricky to keep acceptably clean (a lot of dirt fixes later). So handling all of that post work digitally makes more sense to me.

And it seems to me that the 2K resolution limit of the Spirit is more than enough for a Super-16 negative; so compared to an optical printer blow-up through dupes, I actually expect to see better sharpness, not less, PLUS less graininess. And all the control over the image that digital color-correction allows. I actually think that in the case of Super-16, a D.I. finally allows you to hold onto some of the quality of the original negative rather than suffer all that loss from the duping & optical blow-up.

And dupes and optical printing ALSO have an effect on color, so it's not like there's a clear case for superior color rendition by sticking to film throughout, as opposed to a digital intermediate.

The key seems to be to weigh all the strengths and weaknesses for each particular project between format, film stock, post approaches, etc., and chose the one that gets you the look you need (assuming you can afford it). Like I said, there are situationswhere digital color-correction is the only way to get the effect you want.

David Mullen

Cinematographer / L.A.


David wrote:

>Anyway, eventually we'll have those 4K C-Realities, Millenium-TK's, and new 4K >Spirits up and running, right? I'd hate for the digital intermediate technology to >become stuck at 2K.

Hi David,

I don't think we're in any danger of being stuck. CRT telecines can scan 4k and are wondering (dreaming) about 6k. The Spirit 2 will soon be rolling. Of course the bottle neck is moving all these zeros and ones around. HSDL has us moving 2K data off the ITK at 16 fps.(the current speed champ.) The telecines will do real time, we just need bigger pipes, and in time they will come. If we want to do entire features in 2k or 4k we're talking terabytes for storage and those prices are getting better.

I don't thing film is stuck either, as Fuji is about to unwrap 4 dye layer stock. Have faith things are changing and will continue to do so.

Jim Mann

Freelance Colorist


>It seems to me that if the Spirit can't use all of its potential pixels for the 2.35 area of >Super-35, one would still be better off shooting with anamorphic lenses and thus >being able to scan more film negative vertically in the Spirit.

Yes, that is the better choice.

The Spirit grabs the entire full aperture 35mm image in a internal frame store. This is 25mm x 18.8 mm and is stored in 1920 x 1600 pixel (internal format, non square pixel). This image then can be resized in the internal spatial converter to whatever output format. If you only want to use the 2.35 area your output image is derived from 1920 x 900 scanned pixel.

If that's true, then the resolution difference between a Spirit 2K transfer of a 2.35 area of the Super-35 frame at 2048 x 870, as opposed to a transfer to 24P HD-D5 (1920 x 1080) and then cropping down to 2.35 (approx. 1920 x 800) wouldn't be that much.

Pixelwise you are right, but HD-D5 is 4:2:2 only and has a 1:5 compression. Data is 10 bit RGB with no compression. Thats the better choice for a digital intermediate.

Regards

Uwe Braehler

Thomson Multimedia


>And it seems to me that the 2K resolution limit of the Spirit is more than enough for >a Super-16 negative;

OK here is where I might have a problem with the conventional wisdom of "2K, it'sgood enough for Super-16" Is it "more than enough" for a 35mm negative ?, If not, how could it be more than enough for a Super-16 negative ? I don't get it...

Here is my point: Does a given 2K scan capture ALL the useable information in (for the sake of argument) a 35mm Academy negative (let's say for comparison 1.85:1 area of it)? If you were to say it does not, then might I say that using the same scanning resolution on an S16 negative does not capture all the information either? - i.e. there is information that is missed.

Now, which will suffer more from missed information, the 35mm negative which isusing about three times as many grains to represent the same image (and you could say has ~ 2.5 X the 'total' information) OR, the S16 negative which will need all the help it can get if it is to be ultimately magnified 2.5 X it's size? (or in the case of the Spirit DataCine, has been magnified that much before scanning?)

Look at the 35mm negative, and mentally draw a rectangle 12.35 x 7.5mmm within it. (This is the area of a S16 negative). Now, within that 12.35 x 7.5mm rectangle of the 35mm negative, can a given scan capture all the useful information in that rectangle? If not, then a scan of a 35mm negative at the same scanning resolution is not capturing all the useful information in an equivalent S16 neg either.

Or so it would seem to me. Comments?

-Sam Wells

film/nj/usa


Well, all I can say is that now that I've seen the digital blow-up of "Conspiracy" projected onto a large screen, the 2K resolution of the digitial intermediate produced a sharper, finer-grained blow-up from S-16 to 35mm than any conventional optical printer blow-up using dupes that I've seen from something shot on 7274 before.

So whether or not that digital intermediate captured all the information on the Super-16 neg is sort of moot since conventional method of dupes and optical printing do an even poorer job of capturing that info!

The quality was similar to a DIRECT blow-up from Super-16 neg to 35mm print -- if you can guarantee that all your 35mm prints will be made off of the S-16 neg, then I suppose there's little reason for the expense of a D.I. unless you need to use digital color-correction tools, or have to integrate digital effects & electronic titles into the image.

David Mullen

Cinematographer / L.A.



Copyright © CML. All rights reserved.


 

 

 

 

 

 

 

t>