Jump to content

Questions re scanning 8mm film


Guy Burns

Recommended Posts

Thanks to this forum, I think I now have three superb options for scanning my Standard 8mm films, three companies in the USA. Regarding scanning, I have some general questions. But first, some background.

 

Background

I'm digitising my multi-screen slide shows from the 1980s, and turning them into a Blu-ray. I wasn't expecting to include any movies, but recently I came across some amateur 8mm footage of bushwalking in Tasmania in the 1930s, and it just has to be included.

 

 

My search for a scanner

I thought my Epson V700 would be up to the job, but after comparing scans from the Epson and scans from an ageing Nikon movie scanner done by a company in Sydney (80 minutes to scan a 50 foot reel), I moved on from the Epson – and moved on from Sydney.

 

Next I approached a dedicated enthusiast at Nano (http://nanolab.com.au/). A top bloke. Built his own scanner. But he has so much work he won't scan unless he develops your film as well. He suggested Movie Stuff.

 

So I checked out the Retro Scan Universal (http://www.moviestuff.tv/). To this newbie, it looked pretty good. And the owner is certainly keen to talk on the phone. Gave me his cell phone number. But wouldn't answer most of questions by email. Why not? And I need answers.

 

Several hours on the net gradually led me to this forum, and names such as Lasergraphics, Flashscan, Kinetta, Spirit, Oxberry, and Xena. And three companies which I just know will scan at world's best. I must say I was surprised at the scanning cost. In my mind I had never considered top of the range stuff, thinking it would cost thousands of dollars to scan a ten-minute reel. But times are changing.

 

Ques 1

Does anyone know of an Australian company that can scan Standard 8 on one of the machines mentioned?

 

Ques 2

Is it correct that exposure should be constant when scanning?

If I was setting up a machine, I would expect to set exposure off an 18% grey-scale card, or similar technique, to get maximum lattitude to black and white. What put me off the Retro Scan, for example, was minimal exposure control. It's either automatic – changing frame by frame – or a manual setting by changing numbers. My Epson calibrates itself between every scan.

 

Ques 3

How accurate is the frame-to-frame registration on the top-notch scanners I mentioned?

When I did my test scans on the Epson, I took the scans into After Effects, and tracked the frame movement via the sprocket holes. Worked quite well. But will I need to track, and correct, sprocket movement from the top-notch scanners, or has that already been done by the scanner?

 

Ques 4

Is Digital ICE – the infra-red dust-removal technique used on my Nikon Coolscan – possible on the top-notch scanners? It gives superb results when scanning slides.

 

Ques 5

Is it normal, without extra cost, to receive frame images in tiff, cropped to just outside the actual image? Preferably LZW compressed.

The reason for this question is that a friend received a 32 GB avi file for a 3 minute silent film. It was 1920 x 1080, had black bars each side, with beautifully digitised (but silent) PCM audio at 1.5 Mbps. Nothing was compressed. If they had been LZW Tiff, without the black bars, the file would have been about 5-6 GB

 

Ques 6

Is is sensible to scan 8mm film above 1080P?

My experience after scanning about 7000 slides on my Coolscan, and some on the Epson, backed up by numbers from other sources, is that slide film has an resolution of about 4000 dpi. For Standard 8 frame (4.5mm x 3.3 mm) that would mean about 700 x 500 pixels.

 

 

Thanks in advance for any comments.

Link to comment
Share on other sites

Ques 2

Is it correct that exposure should be constant when scanning?

If I was setting up a machine, I would expect to set exposure off an 18% grey-scale card, or similar technique, to get maximum lattitude to black and white. What put me off the Retro Scan, for example, was minimal exposure control. It's either automatic – changing frame by frame – or a manual setting by changing numbers. My Epson calibrates itself between every scan.

 

You would not want to vary the exposure frame by frame. Our ScanStation can do this as an option, but we never use it. You have to remember that you're not talking about a single image, but thousands of images shown one after another. In that context, variations in exposure from one frame to the next, to compensate for variations in exposure on the film, can result in odd flickering and other artifacts.
The correct way to do this is to set an exposure for the film based on the film's base. A Flat scan will get you a scan where black is elevated and white is reduced, and your midtones are low contrast. From this, you can pretty easily color correct it back to the way it should be. It's done this way for flexibility - if you color correct while scanning, you bake that correction in, and in some cases that's irreversible. So the process is two-steps: Scan then Grade.

Ques 3

How accurate is the frame-to-frame registration on the top-notch scanners I mentioned?

When I did my test scans on the Epson, I took the scans into After Effects, and tracked the frame movement via the sprocket holes. Worked quite well. But will I need to track, and correct, sprocket movement from the top-notch scanners, or has that already been done by the scanner?

 

Depends on the scanner. I can only speak for our ScanStation, which uses the perforations as registration points, aligning the frames to a fixed position in the final scan. It is arguably more accurate than a mechanical pin, since it can handle variable levels of shrinkage.

Ques 4

Is Digital ICE – the infra-red dust-removal technique used on my Nikon Coolscan – possible on the top-notch scanners? It gives superb results when scanning slides.

 

Variations of Digital ICE exist on some scanners. None of them can do 8mm, as far as I know. The Lasergraphics Director, for example, has this as an option, but that's for 16mm and 35mm only. We do our restoration work in an additional pass after the scan. It can be done manually or automatically, but the best results are generally obtained with a combination of the two. Bear in mind that motion picture restoration is expensive. The software to do this kind of work starts at about $6000, on the low end.

Ques 5

Is it normal, without extra cost, to receive frame images in tiff, cropped to just outside the actual image? Preferably LZW compressed.

The reason for this question is that a friend received a 32 GB avi file for a 3 minute silent film. It was 1920 x 1080, had black bars each side, with beautifully digitised (but silent) PCM audio at 1.5 Mbps. Nothing was compressed. If they had been LZW Tiff, without the black bars, the file would have been about 5-6 GB

 

Cropping depends on the scanner, but most can overscan the frame. Ours can overscan almost to the edges of the film.
LZW compressed TIFFs are unusual in film scanning. Frankly, TIFFs are unusual in film scanning. We can scan to them, but they're either 8-bit (don't scan to 8 bit) or 16bit (massive and clunky files). A better format is DPX, which was designed for motion picture scanning, is uncompressed, and can be either 10 or 16bit. Or ProRes, which can be 10 or 12 bits, and is much more convenient to work with in editing software.

Ques 6

Is is sensible to scan 8mm film above 1080P?

My experience after scanning about 7000 slides on my Coolscan, and some on the Epson, backed up by numbers from other sources, is that slide film has an resolution of about 4000 dpi. For Standard 8 frame (4.5mm x 3.3 mm) that would mean about 700 x 500 pixels.

Very much so. It depends somewhat on your ultimate goals, but a modern film scanner that adjusts its optics so that the entire frame fills the sensor (rather than just using a small area of the sensor for smaller gauges) will get you outstanding results. We regularly scan 8mm home movies at 4k.
Bear in mind that 4k television is here, and 8k isn't far off. The more you have to scale an image up to fit the larger screen, the worse it will look. So if you scan to HD, you're getting an image area of about 1440x1080. To make that 4k, you have to scale it up 8x, which will dramatically soften the image. If you scan at 4k, you don't do any scaling at all. For lower resolutions, you lose nothing by scaling down. But you *do* want to avoid scaling up because that forces you to make up image data that wasn't there before.
Also, DPI has nothing to do with pixels. That's a number that's applicable to the print world, so it shouldn't be confused. The notion that there's a limited amount of data on the film is based on studies done at a time when the available scanners were limited to 4k. It was assumed that 4k worked for 35mm, 2k for 16mm and HDish resolutions for 8mm. That was because the scanners worked that way at the time, and because the thinking was that you would want the grain to remain the same size on the display medium for 8mm as for 35mm.
Think of it this way: If you're a scientist looking at something tiny, do you want a magnifier or a microscope? You're going to resolve more detail with a microscope, and that's the equivalent of a higher resolution scan. The more detail on the film you can resolve, the more faithful a representation of the image you'll get.

 

Link to comment
Share on other sites

  • Site Sponsor

IMO a 2K scan (from a 4K scan head oversampled) at 2048x1556 12bit gets just about all of the detail and grain structure from a 8mm film, 4K is also nice.

 

A 10min 8mm film is about 200ft. of film and, for example, we scan that to 2K at $0.50/ft so around $100.00 to scan on the Xena Dynamic Perf scanner.

 

The Xena and machines like the Scan Station are basically machine vision locked to the perf while scanning so the frame is as steady as the camera that shot the film will allow.

 

You can use a flatbed scanner but that is a tedious and extremely slow process and the optics on those tend to be cheap unless you are using a very high end one.

Link to comment
Share on other sites

Thanks for the responses.

 

Rob said: "You can use a flat bed scanner". But the results will be quite poor for 8mm. The Epson V700 is a well-regarded scanner. With a small amount of post-scan processing, it can give results for slides almost as sharp as the Nikon Coolscans. See pp75-138 of my PDF…

 

http://www.mediafire.com/download/qc67n2gkdz3viyp/Scanning_Kodachrome_%28Incomplete%29.pdf.zip

 

Colour is often better, and it gives demonstrably superior scans of colour negatives. But it is quite blurry for 8mm. The frames are too small, 1/50 of the area of a slide.

 

Tiffs

I'm coming from slide scanning on my Coolscan and movie-scanning on my Epson, thus my emphasis on Tiff. I am obviously wrong about tiff for movie scans. However, there are two reasons I like Tiff.

 

Reason 1

The scan people can't foul up the fps

 

What I didn't say about my friend's scan, was that the frames inside the AVI were in a 3:4 cadence. Every third frame was repeated. Exactly. No change at all. And from a company that boasts in bold letters on it's home page:

 

"NOBODY IN THE WORLD DOES MEDIA DIGITISING LIKE ***"

 

That's quite an accurate statement when you read what they did to a movie that was shot at 16 fps. I don't know how they did it, but this is the effect (I'll use Premiere terminology):

 

1. Thinking the film was Super 8, they interpreted the frames at 18 fps.

 

2. The movie then went into a 24 fps timeline, introducing the 3:4 cadence.

 

3. Somehow that 24 fps was reinterpreted at 25 fps, and that's how my friend got his AVI file.

 

 

Reason 2

Extracting and keeping only the desirable frames – without re-encoding

 

If I choose DPX or ProRes for my scanned movies, instead of Tiff, can I crop those files without re-encoding? With Tiff, its easy. I just bin them. It's also easy with AVCHD from my GH3. The utility MP4Tools allows me to crop the start and finish without re-encoding. It's fairly important to me to crop – to get rid of frames I don't need to save space.

 

 

Two more questions.

 

QUES 2

Typically, how does a scanning company deal with various frame rates? Say they receive a batch of reels of different frame rates from the same customer. Is it a non-issue because the scanner itself will assume the correct frame rate from the size of the sprocket holes?

 

I ask because I don't want a repeat of my friend's experience with frame rates.

 

QUES 3

How does a scanning company deal with different-size holes in film reels? I'm looking at two now, and they're different. Do the scanners have spindles of different sizes?

 

 

Retro Scan

I must put in a good word for Roger Evans, maker of the Retro Scan Universal. He reminds me of Steve Jobs, originally making stuff in his garage. He did relent and get back to me by email, and even offered to ring me regarding my thorny questions about the Retro Scan. I'm sure it gives adequate results. I just can't see how it could give quality results without the expertise in colour, software, mechanics, and general R&D that the big boys have.

 

I will be running all the films through one of his machines in Australia, before I send the films to the USA. I have to do that in case they're lost in transit. And when I have received all the scans, I'll be posting a detailed comparison.

 

Watch out you big fellas with your fancy machines!

Link to comment
Share on other sites

Reason 1

The scan people can't foul up the fps

 

Whoever is doing the scan should ask what you want. If they're using old equipment (limited to broadcast standard frame rates such as 25 and 29.97) you would have no choice but to get the repeat frames you're seeing, unless you wanted the film run much faster than it should be (that is, running your 16fps film through the machine at 25 or 29.97fps).
Our scanner, like the Xena I believe, can scan the film at its native frame rate to a container format like Quicktime or AVI. With those formats, as long as you're not interpolating new frames or doubling frames to pull it up to a higher frame rate, changing the frame rate of the film is a simple matter of bringing into software that can do that (such as Adobe After Effects, among many others), and telling it what frame rate to treat the file as. If, however, the pullup was done in the scan, it's harder to back out of that. It can be done with AVISynth scripts, but it's kind of a pain.
What film scanner did they use? If it's designed to go to a video format, you should move along to a service that can scan it frame by frame.

Reason 2

Extracting and keeping only the desirable frames – without re-encoding

If I choose DPX or ProRes for my scanned movies, instead of Tiff, can I crop those files without re-encoding? With Tiff, its easy. I just bin them. It's also easy with AVCHD from my GH3. The utility MP4Tools allows me to crop the start and finish without re-encoding. It's fairly important to me to crop – to get rid of frames I don't need to save space.

 

With any container format, you would read from one file and write to another. With an image sequence, you could theoretically read from one file and then write back to that file, but why would you want to do that since what you're doing is destructive. Better to keep the full version of the film around, just in case, and do your cropping upon export to another copy. That way you have both. Hard drives are cheap!

QUES 2

Typically, how does a scanning company deal with various frame rates? Say they receive a batch of reels of different frame rates from the same customer. Is it a non-issue because the scanner itself will assume the correct frame rate from the size of the sprocket holes?

I ask because I don't want a repeat of my friend's experience with frame rates.

 

There's no way to know the exact frame rate of the film by looking at it. There's no metadata in the film to tell you, and you can only go by the way it looks, or what the client says. As a general rule though, it's usually a safe assumption that footage shot by the same person would have been shot at the same frame rate. Otherwise they'd have to constantly fiddle with playback speed on their projector for each reel. We typically assume 16fps for old 16mm, 18fps for 8mm and Super 8. If it was shot by a professional, 24 is usually a good guess (or 25 outside North America, Japan and Brazil). Regardless, it can be changed, whether it's an image sequence or a container format, as long as each frame of film is in its own frame in the scan.
The size of the sprocket holes has no relationship to the frame rate. With Super8 for example, you could shoot 18 or 24 as standard speeds. most people did 18 for home movies because you got more running time. But it's the same film either way.

QUES 3

How does a scanning company deal with different-size holes in film reels? I'm looking at two now, and they're different. Do the scanners have spindles of different sizes?

 

The smallest diameter is typically for Regular 8 film. Super 8 uses a larger diameter. Even though the film is identical in width, the sprocket size and placement are different. On the ScanStation the hubs are removable and we swap them out depending on the reel type. On other scanners it may use an adapter to fit various reels.

I just can't see how it could give quality results without the expertise in colour, software, mechanics, and general R&D that the big boys have.

The quality of the camera, the quality of the sensor, the quality of the optics and the precision of the machining all come into play here. I don't have any experience with this scanner so I can't say for sure but it's a tiny little camera in there, which means it's probably a tiny and inexpensive little sensor. The camera that's in our ScanStation, and most of the cameras available for the Xena, cost much more than the entire Retro8 scanner. And that's because they're much higher quality cameras.
Link to comment
Share on other sites

  • Site Sponsor

Frame rates:

 

There are only a few video frame rates, i.e. 24fps 25fps 30fos and 60fps. There is no 16fps or 18fps video and the cameras which shot the film used spring motors or wild electric motors so you really have a very wide range of camera-film fps and if you want to play those as video you will have to some way fit the 10fps or 13.5fps or 16fps or 19fps into the standard video frame rates.

 

The lens on the 4K Dynamic Perf Xena is a $5500.00 Schnieder reproducing lens, it's better than that flatbed.

 

Scanners can digitize to DPX frame sequences or Tiff frame sequences, then you can sort it out.

 

Pro video formats like ProRes444 have much of the performance, or indistinguishable performance compared to image sequences.

 

Retro Scan is a ok machine but it's not a data scanner really and does not have a high end imager, a high performance lens and high CRi and RGB+IR controllable Illumination.

 

Your Mileage May Vary.

Link to comment
Share on other sites

Thanks for the responses.

What film scanner did they use? If it's designed to go to a video format, you should move along to a service that can scan it frame by frame.

I've think I've found out the machines that occupy the two top positions in Australian commercial movie scanning.

One is a Rank Cintel Ursa Diamond tele-cine that the woman on the switchboard told me has been in the company for years, purchased secondhand from overseas. Takes up a whole room she said, "and looks rock solid".

The other, the one that scanned my friend's movie, was a "Nikon film scanner" according to the rep, that had been modified in-house, software and hardware, dating back more than 10 years. The rep wouldn't tell me the model. I asked whether it was a modified Coolscan, but he was quick to say it wasn't. The machine takes 80 minutes to scan a 50' reel, and 5 hours overnight to scan a 200' reel. He told me the reason for my friend's duplicate frames: that the customer must have requested a 25 fps transfer.

Looks like the whole of Australia can't support a top-end scanner.

With an image sequence, you could theoretically read from one file and then write back to that file, but why would you want to do that since what you're doing is destructive. Better to keep the full version of the film around, just in case, and do your cropping upon export to another copy. That way you have both. Hard drives are cheap!

I thought someone would say hard drives are cheap. Here's what was originally in my post, under Reason 2, but I edited it out. Should have left it in.

Before anyone says "Storage is cheap. Get a bigger drive." I run three firewire backups on my desk, one in another room, and two in the shed. If I exceed 2TB and need new 3TB drives, it won't be cheap. About $1200 (A$1600). So, I keep file size down.

Using MP4Tools is not destructive. I can remove unwanted video from a 30-second AVCHD clip (about 100 MB) and end up with a 6-second clip, say, in a couple of seconds. Quick and lossless. It just cuts into the data stream. Do it all the time.

 

 

Two out of three ain't bad

Only two of the three companies I chose responded by email. And both of them in this thread. I wasn't expecting that. The one that didn't respond is the company that – I'm guessing here – is referenced by "Hollywood quality, without the attitude".

 

Anyway, they miss out on my small amount of business. Rob and Perry – you've both the job. If you've looked at my Kodachrome document, I'll be doing a similar thing for movie scans, but on a much smaller scale. I want the movies scanned, but I am very interested in scanning itself, so I want comparisons.

 

Once I round up the films (I have to convince the owner to part with them first), they'll go to Melbourne for scanning on a Retro Scan Universal, then back to me and on to Rob. Then I'm hoping Rob is agreeable to forward them to Perry, instead of coming back to Tasmania first.

 

I'll let you know by email when the films are ready. And I'll have a few more questions in those emails, if that's okay.

 

You've both been very helpful. Thanks a lot. And so was Roger Evans.

 

Edited by Guy Burns
Link to comment
Share on other sites

One is a Rank Cintel Ursa Diamond tele-cine that the woman on the switchboard told me has been in the company for years, purchased secondhand from overseas. Takes up a whole room she said, "and looks rock solid".

 

That's a late-1990s era telecine. Pretty good for its day, but way outdated by todays standards. Any machine that's designed to output to video, that is, a telecine, is going to have to either pull up or speed up your 16fps film. There's no way around that because it's inherently tied to the output side, which requires that the signal conform to broadcast standards. Of course, it could be rebuilt - the people who make the Xena scanner sell kits for this, but it's not very common.

The other scanner, the Nikon, must be custom built. Nikon never made a motion picture scanner as far as I'm aware. It's probably an Oxberry or similar optical printer, with a scanner on it in place of a film camera. Those are fairly common, very slow, but can be high quality. It all depends on the specs of the sensor or scanner.

There's nothing inherently wrong with that type of scanner - I'm rebuilding an ancient Imagica scanner in my spare time, using the transport and mechanicals but replacing the entire lighting and imaging system with much more modern hardware that's a hell of a lot faster (but still slow) and much more capable. But a lot depends on how it was done and the quality of the optics and sensor. The lens in my Imagica alone sells for like $3500 on the used market (a 95mm Nikkor printing lens), and is worth more than I paid for the whole scanner. Optics and sensors matter, so on any custom scanner, it'd be important to know that it's all high quality.

 

 

 

Using MP4Tools is not destructive. I can remove unwanted video from a 30-second AVCHD clip (about 100 MB) and end up with a 6-second clip, say, in a couple of seconds. Quick and lossless. It just cuts into the data stream. Do it all the time.

 

 

 

MP4Tools is just cutting an MPEG file (AVC is MPEG4), which you can do with command line file manipulation tools if you know what you're doing. But it has to be done at the boundary between GOPs, which is less than ideal. If it's able to cut at any arbitrary frame, then it's re-encoding that GOP where the cut is. That's just the nature of that type of file, and it's one of the reasons those types of files shouldn't be edited directly, but should be converted to something like ProRes or DNxHD or some other intermediate frame-based file format for editing.

 

If you cut a ProRes file in a tool such as Final Cut Pro, and then export out to exactly the same file format, it's doing the edit non-destructively, just copying the bits from one file into another with no recompression. You're generating another file, but it's not particularly big. Other tools, even on Windows, can do this as well.

 

I thought you were talking about cropping the frame - there is no format where you could do a crop "in-place" without making a new file or recompressing, other than with image sequences, and even then your software has to support it. I doubt most tools that give you much of a GUI will do this, but there are definitely command line tools such as GraphicsMagick or ImageMagick that can do it.

 

 

Link to comment
Share on other sites

But it has to be done at the boundary between GOPs, which is less than ideal. If it's able to cut at any arbitrary frame, then it's re-encoding that GOP where the cut is.

 

The GOP thing is a bit beyond me really. However… I've been using MP4Tools for several years, and early on I commented to the developer that if I selected a trim from the 6-sec point to the 12-sec point, for example, I always lost the start. He was right on the ball, and in a few days came back with a new version that started one second early. So asking for 6-12 gives me 5-12.

 

He also mentioned something about GOP and that he could only cut at approximately one-second intervals, so maybe he was talking about the same thing as you.

Link to comment
Share on other sites

MPEG files don't store each frame as a complete frame. This is a big part of how they're able to be so much smaller than a frame-based file format like ProRes or Uncompressed. Basically, the footage is broken up into Groups Of Pictures (GOPs), which consist of a complete frame, then intermediate frames that tell the decoder what has changed since the last complete frame. Sometimes the structure uses fixed GOP sizes, sometimes it's variable. There are also intermediate frames that are pretty much complete frames, but don't sit on the boundary between GOPs (but they're there to deal with big changes in the image, such as scene changes.

 

In any case, you can only make a non-destructive cut at the boundary between GOPs. If you need to cut at a different frame, then the GOP containing that frame has to be split into two GOPs, with a new boundary placed at that frame. That requires re-encoding the picture data for up to a couple seconds worth of footage.

 

This is a file format that was never designed to be edited. It was meant for final presentation.

 

What the guy from MP4Tools was telling you is basically the same thing I'm saying.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...