Spirit 4K DataCine Scan Question

[Frank Chang]

Thanks, David. I agree with you and I believe that is why Photoshop says 4096 for the DPX.

Do you think the scaling engine will make the image a little less clearer since it is being upscaled?

I know upscale is never a good thing to do, but I am also thinking that since the image was cropped

during scanning with both left and right at the start of the curve (area that we don't need, but still part

of the Academy area), perhaps this cropping (i.e. lessor area to scan and with 3656) should help with

the upscale quality?

 

Just curious. Is 3656 the max or best pixels one would get from the Academy area and anything higher won't

enhance the image quality any further? (i.e. If one would use a different method or scanner that

actually scans at 4096 with the Academy area)

[Frank Chang]

Just got off the phone with the tech support line with 3 different scanning manufactures.

All 3 of them said Academy area is smaller than the full 1.33, and therefore will have less

pixels to scan. So it is limited to 3656 at the max, although the auto scale will save DPX

at 4096.

 

So the main question is if the scaling engine making the image a little less clearer since

it is being upscaled from 3656 to 4096?

PS: I think I should mention that 2/3 of the reel are black and white scenes and only

1/3 are color scenes.

Edited September 19, 2016 by Frank Chang

[David Mullen ASC]

It would be easy for them to send you a frame either way, scaled or unscaled, to check... but the truth is that your old 35mm Academy footage, especially duped to an IP and then IN, isn't resolving 4K worth of detail anyway so I doubt it matters much one way or the other. Even the original negative wouldn't resolve 4K in terms of image detail and sharpness. So I'd be surprised if you saw a difference unless the scaling was badly done, which I doubt otherwise the Spirit wouldn't offer that function.

[Frank Chang]

I agree. Thanks, David.

 

And thanks for all others for the info posted earlier.

[Craig Nichols]

The Spirit 4k does have a fixed lens. The maximum Academy width without scaler will be either 3606 for Cineon, or 3616 for Smpte/Din. For Full Aperture Super 35, the normal size without scaler is 4096 wide.

[Frank Chang]

Thanks, Craig. In your experience, do you think/or seem if using auto scaler in Spirit 4K to 4096 DPX 16 bits,

will it show any loss in noticeable image quality*?

 

*2/3 of the reel are black and white scenes and only 1/3 are color scenes.

[Perry Paolantonio]

I would point out once again that there is a difference between digital resolution and optical resolution. The question should not be "is a higher resolution scan getting me more information from the film frame?" Instead the question of digital resolution (the size of your output file) should be "what do I need?"

 

Upscaling is never good and should be avoided. There's nothing inherently wrong with 3656 as a scan size. But if someone needs it to be 4096, they will have to scale it up to that size, and that's going to mean the image will be slightly softer. It's never a bad thing to scan larger than you need, because you don't have this same effect when you downsample (unless the downsample factor is very small, then you might see some artifacting). This is why modern film scanners allow you to scan the film directly to the size you want (up to the limits of the sensor) without having to scale the image up.

 

Treating digital resolution (The number of pixels in the image) with optical resolution (the ability of an optical system - lens, film, camera, operator) as if they're the same doesn't make sense because they're two different things. Yes, there are limits to what a given film stock, lens and camera combination can capture. But if you scan a film at a lower res than you need, and then scale it up, you're making it worse by softening it. Why do that?

[Frank Chang]

Thanks Perry.

[David Cunningham]

Exactly what Perry said.

 

I always overscan (both frame and resolution wise) so that I can frame and down-sample as I please. For example, there is not 2K (2048x1536) of image on a Super 8 frame. However, I always scan at that resolution for my HD Super 8 work so that I can stabilize/zoom in and put my 1920x1080 frame wherever I want. It makes a huge difference in the final result.

 

The second you start scaling up a digital image it looks terrible quickly, especially grainy film.

[Frank Chang]

Thanks David Cunningham.

[Ted Keaton III]

Exactly what Perry said.

 

I always overscan (both frame and resolution wise) so that I can frame and down-sample as I please. For example, there is not 2K (2048x1536) of image on a Super 8 frame. However, I always scan at that resolution for my HD Super 8 work so that I can stabilize/zoom in and put my 1920x1080 frame wherever I want. It makes a huge difference in the final result.

 

The second you start scaling up a digital image it looks terrible quickly, especially grainy film.

 

Wonderful topic!

 

I really enjoy your Super 8 work and input Cunningham!

Watched your recent postings on Vimeo and I am curious why you choose 500T on one of your latest videos... were you pleased with results and would you recommend using again? Noticed 50D does seem to be a "gold standard" or "go-to" for you :)

I understand the benefits of scanning 2k. How often do you scan at 2k compared to 5k? In other words, is your decision based on costs or does 2k seem to "get the job done" (IYO)?

I do feel that I am staying on topic by also asking a workflow question... do you color grade and/or frame before you proxy (the larger resolution scan) or is the transfer immediate.

Lastly, what are you rendering out to currently... you always seem to have a "trick up your sleeve" when it comes to best quality efforts with Youtube and I wonder if you have any magic with Vimeo settings.

Thanks for your time and thanks for posting your work especially.

[David Cunningham]

Thanks Ted!

 

So, assuming you mean this one:

 

 

With Kristen and Drew, I did use quite a bit of 500T because of the heavy overcast to start the day and the extremely yellow light from the tent. When using light meters I was getting color in the 3000 to 4000 range pretty consistently.

 

That said, I do try to avoid 500T as much as possible. Obviously, the bright sun scenes are all 50D and the color, saturation and grain are excellent! The 500T is a necessary evil sometimes, but the result is just SOOO much better with 50D. The 500T is just too grainy in this small format. In fact, moving forward, I am probably going to do exclusively outdoor weddings because the on-camera LED light is just too intrusive and the 500T just never turns out as good as I'd hope.

 

I'm actually in the process of completely re-coloring, stabilizing and framing this footage as I was just not happy with the way it came out the first time. I hope to post that soon.

 

I have shot indoors with a bright light using 50D. There are two major problems with that.

 

1.) Your subject is the only thing that shows up in the footage. All the background is completely black even if the room isn't.

 

2.) You need to open up your f-stop to full or nearly fully open. These Super 8 cameras, even the good ones, have very soft lenses at wide open. I try to keep my lens stopped down to at least f4 and if possible 5.6. The ideal image comes in the f5.6-8 range on most Super 8 cameras.

 

So, in super 8 I aways get a 2K output because that yield a 2048x1536 image in 4x3 Super 8. This is a MUCH higher resolution than HD in this case, 1920x1080. If it were "max 8" or super 16 (widescreen) then you would only get a 2048x1080, so no real advantage over HD.

 

For all my wedding shoots I sample at 5K, however. This is not for resolution gains but to properly render the grain which is critical in these small and grainy formats. The result is a perceivably "sharper" image. The resolution doesn't end up higher, but the grain is better resolved and results in a perceived increased sharpness. It's very subtle.

 

In Super16 I also always sample at 5K and, depending on whether I intend to stabilize or reframe, I will either take this raw 5K overscanned output or down-sample to 4K with a locked-in 4K cinema crop.

 

 

My workflow for Super 8 is to take the 2K (2048x1536) Prores 4444 overscanned image and first apply a color grade. Then, in the same pass, I stabilize and crop the image on a scene-by-scene basis. At this point I render out to a Prores 422 2K file.

 

Then I use that graded and stabilized 2K file to edit my films, then add a window mask around the outside and export to my final format, be it HD or 2K.

 

With both youtube and vimeo one of the best tricks to get Super 8 to look as good as possible is to upload the highest resolution possible. In fact, I sometimes up-res my HD and 2K files to 4K before upload. This forces vimeo and youtube to increase your bit-rate if you select the 4K stream. This helps to keep the film grain from getting blocked up too much due to compression.

 

Thanks for all the compliments! I'm blushing. :)

 

Dave

[Frank Chang]

Just an update. Good news. Our lab tech confirmed that the reel was scanned as FA-1.33:1 with scanning width 1.32:1

with no spatial processed. I suppose, this also explained why the left and right were cut off at the frameline/curves

(scanning width 1.32:1) Thus, it was not scanned as the Academy format of 1.37 at all. Reason is actually because

(and I was not aware of this until an hour ago) that the internegative have scenes done in 1.33 and some done in 1.37

and there are some frames of certain scene have mixed 1.33 and 1.37. Very weird. But because of this, the lab tech

decided to scan it as FA-1.33:1 with scanning width 1.32:1.

 

PS: The lab tech also said the reel is an optical duplicate negative. I didn't get into too much details with the tech, but I assumed

it is optically printed with all the needed optical effects, etc. Not sure if this means the same as duplicate negative/internegative.

Edited September 21, 2016 by Frank Chang

[David Cunningham]

Very interesting...

[Tom Tomlinson]

The Spirit 4k does have a fixed lens. The maximum Academy width without scaler will be either 3606 for Cineon, or 3616 for Smpte/Din. For Full Aperture Super 35, the normal size without scaler is 4096 wide.

 

Hi Craig,

 

just wondering...Didn't DFT make academy centered film gates for the Spirit2k/4k. I recall considering that when we got our first Spirits.

 

TomT

NiceShoes

[Robert Houllahan]

I can post some pics of our Spirit-2K (all later HD, 2K and 4K Spirits are the same with the same optics, sensors etc. a 2K can be turned into a 4K with a license code) tomorrow.

 

The film path of the Spirit (rollers, follower sprocket etc.) are fixed and the gate is a curved surface with a precision line aperture at the curves apex. The only difference between a Acad gate and a Full frame would be the cutout of the slit aperture.

 

There is no way to move the Tri-Linear 4096 line CCD array relative to the film to use the full 4096 pixel line to see academy, thus the reduced number of pixels scanned for Acad.

 

I will post pics of the 2008 Spirit-2K gate, lens and CCD array tomorrow ;-)

[Tom Tomlinson]

Thanks for all the info.

I guess they stopped that when they went from original 1920 spirit to spirit2k.

 

TomT

[Robert Houllahan]

Spirit/2K/4K Gate and CCD's

[Chris Burke]

rob that's fantastic

[Craig Nichols]

 

Hi Craig,

 

just wondering...Didn't DFT make academy centered film gates for the Spirit2k/4k. I recall considering that when we got our first Spirits.

 

TomT

NiceShoes

Hi Tom,

 

Yes, DFT made an academy centered film gate for the Spirit2k/4k. Universal Studios was one of the few facilities who purchased this option. We experimented with metallic air conditioning tape on the underside of S35mm gate block, and this approach was used successfully a few times at Warner Brothers.

 

Craig

PHI