Scanning 35mm cine, myself

March 21, 2005

I was thinking about actually buying some kind of a film scanner and scanning in 35mm cine myself. Yeh I'm mad aren't I.

I mean, you can buy a pretty descent scanner for a couple hundred quid, I'm just wondering how easy it would be to scan in cine.

My only fear is scanning all the frames perfectly. i.e. it must be hard scanning the image perfectly without chopping bits off e.t.c. I don't want to end up with a load of footage that looks like those old 1900 movies where the film is being thrown about all over the place in the gate.

Would it be possible to buy a flat bed scanner, and scan all the frames in myself? I know it would be extremely time consuming, but I have got a lot more time than money.

Thanks for any advice.

Edited March 21, 2005 by Daniel J. Ashley-Smith

gnasr · March 21, 2005

The main problem is you'd have to chop up your negative.

You usually can insert 6 frame strips in a negative scanner.

So you would have to cut 1 second of film in 4 strips... imagine for an entire 4 minutes reel! Oh man this is a hell of a job.

It should work though, but a think it's better with a negative scanner that has a tray.

Edited March 21, 2005 by gnasr

Dominic Case · March 21, 2005

but I have got a lot more time than money

If you embarked on this venture you'd have a lot less time.

I think positioning each frame would be a bit of an issue. Domestic flatbed scanners don't have 35mm register pins, so I guess you'd have to adjust each frame after scanning so that the perfs lined up. Maybe there's software that can do that . . . but maybe not on your desktop.

Try calculating the time it would take. I think just the scan would take a couple of seconds (probably longer at a decent resolution on an affordable machine): that's after you've positioned the negative as well as you can. Maybe you could find a machine that didn't need cut negative and rig it up on a rewind bench - that way you could maybe get each frame into position inside ten seconds. And the scan time, and we have 5 frames per minute. That's one second of screen time in five minutes. Allowing for sanity breaks you could get a screen minute done in an evening.

If you aren't crazy now you could soon get that way ;)

Rolfe Klement · March 21, 2005

I would question why you want to do this? Depends what you are trying to do. Depends on your final output.

I would agree and I sometimes do the same thing but only for one frame and only for presentation on a web site or print format. BUT only after a neg cut and only one frame - I would take the one frame and get it scanned at 10bit high res @ any high street photo shop. Then you get amazing images - that look like what you see at the cinema.

There are always junior and assistant TK ops who will do anything to get experience on film. The big loss of quailty is not the scan to Digibeta - but rather to conversion from Digibeta to DV or QT then the screen dump. So if you can output a Tiff or Targa file from a digibeta you are half way to a great image.

I am not an expert at the difference between 8 or 10 or floating bit scan color differences of the big TK machines ITK, Spirit etc

IMHO focus more on getting great shots so that TK guys will work with you for free - originating on film is half the job. I am guilty of trying to produce and operate at the same time so I feel your issue

thanks

Rolfe

David Cox · March 21, 2005

Hi,

Yes I think you spotted the most likely issue, and thats frame registration. This is getting the film frame in exactly the same place time after time. If you don't it wobbles like hell! You could try and stabalize it in your post session, but that would lead to positional interpolation (because you have to move the frames at sub-pixel amounts) and this would cause some softness in the image. Also, you might have a similar problem with exposure (or the scanner equivalent of) in that each frame might have slightly different brightness curves.

Having said that, a couple of years ago a major projector manufacturer slapped a camera on the front of one of their units to make a "telecine". It looked like something out of a 60's science fiction film, but I have to say the results were more than passable. So they made a "cheap" telecine at a mere $150,000. So if you can get your scanner to work, you'll be a very rich chap!!

Good Luck

David Cox

Baraka Post Production Ltd

www.baraka.co.uk

Brent J. Craig · March 21, 2005

I don't think it would be very difficult to line up the scanned images because the edges of the film frame are always in the same place. Just scan them at the same size and line up a corner. Not too hard, but very timeconsuming.

David Cox · March 21, 2005

I don't think it would be very difficult to line up the scanned images because the edges of the film frame are always in the same place. Just scan them at the same size and line up a corner. Not too hard, but very timeconsuming.

<{POST_SNAPBACK}>

Hmmm, if you were just 1mm out on your line up, you would be a foot out on a decent size cinema screen every frame. That will make a few people think there's an earthquake!

Stephen Williams · March 21, 2005

Hmmm, if you were just 1mm out on your line up, you would be a foot out on a decent size cinema screen every frame. That will make a few people think there's an earthquake!

<{POST_SNAPBACK}>

Well you could line up the perf and make it sort of Pin registered! I think the perfs are accurate to 1/100 mm. The corners should match much better than 1mm as they are exposed in the camera gate, so for a non pin registered camera they would be very good. Remember TK's are not pin registered, they usually rely on the edge of the film. A very bad idear IMHO.

Stephen Williams DP

Zurich

www.stephenw.com

David Cox · March 21, 2005

Yes, but we are talking about positioning a neg in a flat bed scanner by hand with 1/100mm accuracy? And thats before the neg sticks to your fingers before closing the lid!

Brent J. Craig · March 22, 2005

Yes, but we are talking about positioning a neg in a flat bed scanner by hand with 1/100mm accuracy?

<{POST_SNAPBACK}>

I was thinking they would line up to the frame edges (or perfs - even better) on a computer. Most image editing programs will let you blow up enough that a single perf could fill your screen. Put the new image in a layer at 50% opacity over the old image, jiggle until the perfs line up, repeat.

John Pytlak RIP · March 22, 2005

I was thinking they would line up to the frame edges (or perfs - even better) on a computer. Most image editing programs will let you blow up enough that a single perf could fill your screen. Put the new image in a layer at 50% opacity over the old image, jiggle until the perfs line up, repeat.

<{POST_SNAPBACK}>

Physical registration of the film perforation with a full-fitting registration pin is usually more precise than trying to register using only the optical image of the perforation. Do always use the "reference edge" and "reference perforation" as defined by the appropriate SMPTE standards, so the "principle of cancellation" helps reduce any effect from dimensional variability.

Paul Bruening · March 22, 2005

Hey Buddy,

If you're going to reinvent my thread from yesterday then why not just ask me how it works? The Epson I mentioned spots the frame lines and scans them on that. Adobe After Effects has a stabilization feature that detects and corrects frame jitter. Also, DON"T CHOP YOUR NEGATIVES. Use an intervalometer on a camera to move the film. The rest is a gate (which you will have to hand make or steal out of another camera) to keep the film flat. You will have to get a board capable of SATA drives. Figure on 20 to 30 400GbHDDs to hold all that data. Data is the biggest booga-bear to this approach. While the Epson is not quite as good of a scan device as commercial lab gear, it is cheap, can scan up to nearly 10K, yet, you will have to replace it a few times before you achieve the 133,500 scans needed for a 90 minute run movie. Consider marrying your camera with a show-through pressure plate up to a Canon XL2 w/ EF adapter and macro lens to get 24P telecine (XL2 does the 3-2 pull down for you) for your workprints. The less workprint matched neg you have to scan the faster the 2-4 months of scanning will go. Fortunately, custom written macros on Photoshop can automate the process so you don't have to babysit every scan. How's that for an answer?

Paul Bruening

P.S. John Ptylak doesn't dig me. My ideas cut into their business.:P

Phil Rhodes · March 22, 2005

Hi,

The Spirit scanner does some vertical stabilising based on imaging the perfs, so it's not a new idea.

Phil

David Cox · March 22, 2005

As I mentioned right at the beginning, if you use your software to reposition your frames, you will induce softness. The reason for this is that if you move your frame by anything less than 1 pixel, your software has to interpolate your image between two neighbouring pixels. And yes, you will find that moving your image by one complete pixel either way is not accurate enough to completely remove shake. This is especially true since you will have to twist as well, as your frames will not lie at exactly the same angle each time.

In another thread, it was commented that all DI's make the frames go soft. Well do it this way, and it will!

Stephen Williams · March 22, 2005

Oxberry camera's & optical printers sometimes come up on E-Bay. There is little demand and the price is usually under 1000 USD . IMHO This would be a perfect way to register the film, when building a scanner.

Many years ago I did time lapse with a Nikon and a 250 frame back. As the Nikon was not pin registered I had to line up the images by overlaying the frames with reference to the first frame. Finally I filmed them on an Oxberry. It worked well

Stephen Williams DP

Zurich

www.stephenw.com