 |
Forever on DVD Ultra-high quality film and video transfer
|
| View previous topic :: View next topic |
| Author |
Message |
Shawn Owner

Joined: 29 Apr 2003 Posts: 27 Location: Kitchener, Ontario, Canada
|
Posted: Fri Sep 30, 2005 11:26 am Post subject: Let's get technical |
|
|
Hi everyone:
I thought I might go into a bit of the technical details of what sets Forever on DVD apart from most of the competition. I'd like to discuss how we digitize film, and an honest appraisal of how we compare to the extremely expensive scanners like the Ranks, Spirits and Vialtas, and the low priced simple projector transfers that most of the transfer companies offer.
First of all, let's discuss the equipment and method of transfer here at Forever on DVD:
We use a DV8 Sniper Pro for 8mm gauges and Sniper 16 Pro for 16mm gauges with a 850-line 3CCD broadcast camera. The signal from the camera is output as 10-bit uncompressed 4:2:2 and passed through component to a real time colour corrector, and from there it is either converted to a DV signal and passed through firewire or fed directly through component into a Decklink Extreme SD card. In either case, the footage is digitized directly to a server as either 8-bit 4:1:1 DV files or 8-bit or 10-bit 4:2:2 uncompressed files. The original files are always captured to a server first.
There are 2 main considerations when it comes to digital video - at least in our case. Acquisition format and delivery format. Often, the delivery format determines the acquisition format. For example, if a client wants the highest quality uncompressed footage to edit for a television documentary, it would be a poor idea to digitize to a DV codec. Similarly, if a client wants a simple DVD Video of their family films, digitizing to 10-bit 4:2:2 uncompressed is most likely overkill.
Many people ask if we use an intermediate when digital masters are produced, as many other companies do. We don't of course. The original files are always digitized to a server. Then, based on the delivery format that the client wants, copies are made only from the original files.
This insures that the quality of all video is as high as possible. If it is necessary to transcode video due to a changed or additional delivery format, the original files are always used to minimize any degradation from multiple generations of file transcoding.
So, how do we rate versus the more expensive competition?
The ultimate has always been the million dollar scanners, generally just referred to as Ranks. These machines are made and designed to offer the highest resolution possible when digitizing film. They offer incredible colour correction, scratch and grain reduction, you name it...for a million bucks or so, I would expect nothing but the best. Right?
However, the truth is, they are not always the best alternative. These machines are designed from the ground up for large format film, 16mm and 35mm. Most of the scanners won't even support the smaller gauge films like 8mm and super-8. And, sure they offer super high resolution for the larger gauges of film, but when you consider that a 8mm frame is about 1/16th the size of a 35mm frame, the extra resolution isn't really going to be of benefit to most people.
But, probably the most important factors are how well the Rank is maintained and operator skill. Since the real money for a Rank owner is the larger film formats, many of the Ranks that do smaller gauge film are not necessarily optimized for 8mm transfers. And, the results are not as good as they could/should be. With some extra care and concern, a Rank transfer is hard to beat, but poor optimization and operator indifference can make an expensive Rank transfer a big waste of money.
So, where does Forever on DVD fit in the grand scheme of things?
A good, well maintained Rank transfer by a caring operator will be the best that money can buy - and it will cost a lot of money. Be prepared to spend literally hundreds and hundreds of dollars per hour, not film time, operator time.
Next in line would be Forever on DVD, and companies like us that do frame-by-frame scanning using professional equipment and caring operators. Let's face it, the small gauge formats are our bread-and-butter so we had better specialize in it.
And, next is where I think I would shock many - in my opinion, I would place us and companies like ours above the average, super expensive Ranks. Having all that potential for top notch transfers mean nothing if the machine is not maintained well and the operator doesn't really care.
And, below here is the rest of the industry with the variable speed projectors, and the Elmos and the Gokos.
Feel free to comment or ask questions. I'd like this area to help people decide what is best for them and explain the options and alternatives that exist. Your film is precious, so you might as well get all the facts before you decide to trust a company with your film and hard earned money.
Kindest regards,
Shawn Kimmel
Forever on DVD |
|
| Back to top |
|
 |
cmosher01
Joined: 08 Mar 2006 Posts: 2
|
Posted: Wed Mar 08, 2006 8:38 pm Post subject: |
|
|
| I'd like some further clarification of this process. Starting with the capture: the DV8 Sniper 3CCD has 850 lines, so it will capture and output at this resolution, I assume. After this, when the image gets passed (through either of the two routes you describe) to the server, will the resolution shrink to 480 lines (NTSC standard)? If so, then is there any way to preserve the original 850 lines instead? |
|
| Back to top |
|
 |
Shawn Owner

Joined: 29 Apr 2003 Posts: 27 Location: Kitchener, Ontario, Canada
|
Posted: Wed Mar 08, 2006 9:58 pm Post subject: |
|
|
| cmosher01 wrote: | | I'd like some further clarification of this process. Starting with the capture: the DV8 Sniper 3CCD has 850 lines, so it will capture and output at this resolution, I assume. After this, when the image gets passed (through either of the two routes you describe) to the server, will the resolution shrink to 480 lines (NTSC standard)? If so, then is there any way to preserve the original 850 lines instead? |
Great question. The manufacturers of high-end broadcast cameras give specs for the camera head portions of the cameras. Generally, the better cameras have 800 or higher lines of resolution. However, I'm not really sure if lines of resolution can be directly related to actual NTSC resolution. Now that I think of it, I wonder if they are talking more about resolving power.
Anyway, the short answer is that the camera outputs 720 x 486 for uncompressed transfers and 720 x 480 for DV transfers. The 850 lines that the camera head resolves basically equates to a sharper, higher quality SD image than similar cameras with less resolving power. But, those 850 lines don't directly equate to actual pixels.
I hope that made some sense.
Kindest regards,
Shawn Kimmel
Forever on DVD |
|
| Back to top |
|
 |
cmosher01
Joined: 08 Mar 2006 Posts: 2
|
Posted: Fri Mar 10, 2006 8:10 pm Post subject: |
|
|
Thank you for clarifying that for me. Yes, that does make sense.
I have some more questions regarding the details of the process that I hope you could answer for me.
First, I want to make sure that the each frame is progressive. The web page describing the process states that there is no interlacing, but DV is interlaced, right? Interlaced could be fine, as long as it is done through a simple telecine process, because that can easily be "undone" on my end through software via an inverse-telecine process. This would restore the progressive frames. If so, what is the pulldown applied for an 18 fps source? Normal 3:2 pulldown will convert 24 fps to NTSC's 30 fps, but this doesn't work for 18 fps sources. Any clarification you can provide would be helpful. Of course, my preference would be to receive progressive frames without having to inverse-telecine to get them.
My second question concerns the specific amount of each frame that gets captured. I realize that you use an enlarged gate to get more of the frame than normal, but exactly how much is that? (Specifically, I'm asking about Super 8mm.) As far as I can tell, Super 8mm can image 6.22mm x 4.22 mm, acording to the specifications. I actually measured frames on my films to check, and although I cannot measure that precisely, I could clearly see good image on each frame out to at least 6.1 mm x 4.1 mm. Would all of the frame be imaged? If not, how much would get lost? |
|
| Back to top |
|
 |
Shawn Owner

Joined: 29 Apr 2003 Posts: 27 Location: Kitchener, Ontario, Canada
|
Posted: Sat Mar 11, 2006 10:58 am Post subject: |
|
|
| cmosher01 wrote: | | First, I want to make sure that the each frame is progressive. The web page describing the process states that there is no interlacing, but DV is interlaced, right? |
Yes, you're absolutely right. Since we image each frame individually, both sets of fields are digitized from the "same" frame of film. So, when the uppper and lower fields are combined together, the image formed is from the same frame of film, so there is no aliasing, just a perfect frame. With a projection type of transfer, the fields of a single frame of video are often digitized from two different frames of film, causing aliasing and other visual artifacts.
| cmosher01 wrote: | | Interlaced could be fine, as long as it is done through a simple telecine process, because that can easily be "undone" on my end through software via an inverse-telecine process. This would restore the progressive frames. If so, what is the pulldown applied for an 18 fps source? Normal 3:2 pulldown will convert 24 fps to NTSC's 30 fps, but this doesn't work for 18 fps sources. Any clarification you can provide would be helpful. Of course, my preference would be to receive progressive frames without having to inverse-telecine to get them. |
Since we digitize film at a 1:1 ratio, where each frame of video corresponds to a single frame of film, when the video file is played back it plays at 29.97 fps. Since the film in your example is supposed to play at 18 fps, we have to perform a "pulldown" to adjust the speed to play at the proper speed. You're absolutely right about the 3:2 pulldown for 24 fps film. The pulldown for 18 fps is lesser known, it is 4:3:3. And, I don't know that you can easily perform an inverse telecine to remove the pulldown. However, we can definitely provide the video file without a pulldown applied. We can even perform a non-interlaced pulldown so that the video plays at exactly 18 fps as well.
| cmosher01 wrote: | | My second question concerns the specific amount of each frame that gets captured. I realize that you use an enlarged gate to get more of the frame than normal, but exactly how much is that? (Specifically, I'm asking about Super 8mm.) As far as I can tell, Super 8mm can image 6.22mm x 4.22 mm, acording to the specifications. I actually measured frames on my films to check, and although I cannot measure that precisely, I could clearly see good image on each frame out to at least 6.1 mm x 4.1 mm. Would all of the frame be imaged? If not, how much would get lost? |
Yes, we image the entire super 8 frame. Basically the gate is enlarged to greater than super 8 size and to properly image both regular 8 and super 8 we use 2 different length extension tubes to magnify each size frame. Because the regular 8 frame is smaller, we have to magnify the image a tiny bit more than super 8. One thing to note though is that the film frame is not exactly the same aspect ratio as a video frame, so there is always a tiny bit vertically that is cropped if you don't want to see the sprocket holes, as most don't. Still, the choice is yours.
Kindest regards,
Shawn Kimmel
Forever on DVD |
|
| Back to top |
|
 |
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|