Single Chip 3 Scan Or 3 Chip Single Scan Capture?

[S8 Booster]

Since quite some guys here are capable of programming/setting up transfer scanning system,
can a single HQ chip triple scan do a better job than the 3 chipper like no use of a beam splitter.

Just some thoughts...

Shoot.....

[carllooper]

There is some marginal signal improvement you'll manage through not using a beam splitter.

But if you scan a frame of film 3 times, through the same single chip colour sensor, you'll just get three scans that are the same. No extra information. You would need to scan the frame at different angles/positions to improve a 3X scan. But any signal improvement you might have obtained by not using a beam splitter would be completely lost by the physical complications of angling, re-registering, and the otherwise lower definition of the colour sensors.

Note that if you decide to include a different filter on each scan it won't help, because the single chip sensor has it's own filters (on each sensor pixel) that will cancel out the filtration you are attempting to achieve.

To get any improvement you need to use a monochromatic sensor, ie. a sensor without any filters on the sensor pixels. All sensors are monochromatic. Colour sensors are just monochromatic sensors with filters on each sensor pixel. Using a monochromatic sensor just means putting a filter on each scan yourself.

This is effectively what a three chip sensor system does. The only difference is that the three chip sensor system uses a beam splitter to do the scan at the same time. And uses custom circuitry to combine the data from each sensor. Some systems have the sensors at different relative angles to each other to avoid moire effects. An idea borrowed from the days of offset printing.

If you otherwise do it three times (through 3 filters) one after the other, you'll also need to combine the images in a way that registers each scan correctly with respect to the others. Typically you'll just hold the sensor/film stationary with respect to each other during each scan - altering only the filter (or filtered light) on each scan.

This is how VideoFred and friends do it. And it's one of the the best systems on the planet. Completely DIY.

[Andreas Wideroe]

Why not use a linescan camera and a framegrabber? Just keep your film in constant motion and software will take care of the rest.

[carllooper]

Why not use a linescan camera and a framegrabber? Just keep your film in constant motion and software will take care of the rest.

A linescan camera has the same potential limitations as an area scanner. If there is no beamsplitter it means each sensor pixel of the linescan sensor requires it's own filter (a Bayer filter). This would limit the definition of such linescan cameras.

For this reason the Spirit uses three (mono) linescan sensors and an RGB beamsplitter.

See page 3: http://www.dft-film.com/downloads/datas ... -11-09.pdf

You can avoid using a beamsplitter by multiplexing the filters in time, but the position of the film will be different from one scan to the next. This can be alleviated by having the film move more slowly and increasing the frequency of the filter multiplexing. The definition of the result (perpendicular to the linescanner) will be proportional to how slowly the film moves past the scanner, the scanning frequency and filter mutiplexing frequency. Linescanners have a much higher scanning frequency than area scanners. Coupled with control over how fast the film moves, etc. you therefore have some additional control over the definition of the results - at least in one dimension.


The highest definition scan possible is achieved by replacing the optics and sensors in a system with your head. Telecine operators can sometimes be found doing this during downtime. This might also explain why many cameras have camera operators attached to the eyepiece.

[PyrodsTechnology]

Awand went very close to goal: the best scanning possible is via monochromatic line scan sensor AND 3 RGB filters. One scan with Red filter, one scan with Green filter and one scan with Blue filter, of the same frame. Then you have to mix these 3 scans to get colour image You can do this in photoshop for example. This system is superior to 3 Chip prism system, because the prism, despite its perfection made , introduces always a bit of distortion.
Regards.
Rob

[carllooper]

Then you have to mix these 3 scans to get colour image You can do this in photoshop for example

Reassembling an image line by line in Photoshop sounds like fun, but I know what you mean.

If you could find somewhere that can do a three pass scan (with a different filter for each scan) using a single mono linescan, the rest can be done in software. For sure.

You could even build your own. Linescan sensors are cheaper than area scan sensors. You may need to put some more resources into the film transport system. A critical factor will be ensuring each pass goes across the linescan in a consistent manner, although software is a great help in this area. The better the software the more relaxed you can be with the consistency of the mechanics. The correct software can accurately compensate for otherwise inadequate mechanics.

One of the greatest things to come out of the digital world in recent years is cheap affordable 3D printers. I mentioned this elsewhere but you can design and build your own mechanics in plastic using a 3D printer. Couple what you can build with those, with clever software, you could very well put together your own scanner that does something equal to or more likely better than a commercial service.

Commercial services require more robust systems that work fast because they do a lot of transfers (or need to do a lot if they are to remain commercial). For the amateur working on their own projects it's a different story. You can have your own scanner taking it's time, on your own project, and producing exceptional outcomes.

Carl

[PyrodsTechnology]

carlooper wrote
Reassembling an image line by line in Photoshop sounds like fun, but I know what you mean

wait, there is a misunderstanding, perhaps i am wrong due to my bad english, but a linescan sensor scanning will end up with a xy image area . So you have not to reassemble line per line but only 3 already complete b/w images, and photoshop do this in a fraction.

[carllooper]

carlooper wrote
Reassembling an image line by line in Photoshop sounds like fun, but I know what you mean

wait, there is a misunderstanding, perhaps i am wrong due to my bad english, but a linescan sensor scanning will end up with a xy image area . So you have not to reassemble line per line but only 3 already complete b/w images, and photoshop do this in a fraction.

I was only joking