Questions about Spirit telecine systems

[Jake Emmerson]

(I make a lot of assumptions on the way to my questions... Feel free to correct me on any of them.)

 

We just completed shooting a feature on Super16mm. Our plan is to do a poor-man?s DI by telecine-ing to HDCAM-SR 4:4:4 RGB 10-bit log, 1920 x1080, 24p fps.

 

It sounds like Spirit is the best Telecine DI option out there (the next step up would be a frame-by-frame scanner...too expensive). So I had a few questions about the Spirit systems...

• I hear a lot about new vs. old Spirit scanners and how the new ones are so much better. What do I go by? What question can I ask to know if the Spirit is ?new? or ?old??
  
• When people say a Spirit DataCine. Is that a system older than 2k and 4k? Or is that the full name that?s rarely used and they just haven?t specified what "k"? Ie Spirit DataCine 4k.
  
• I?ve read Marc Wielage say in these forums that Spirit 2k is 12bit linear, Spirit 4k is 16bit linear. Is there a mathematical difference (as opposed to debatable visible difference) between the two once they?re collapsed into 10-bit log?
  
• Since we?re working with 16mm, grain reduction is going to be important. The Spirit 2k has a built in grain reduction system. (correct me if i?m wrong) There?s also ?software? grain reduction in the color correction systems. What is the best way to deal with grain reduction? Is it good to use the Spirit grain reduction or wait for the end to deal with that. Or just a combination of Spirit and grading software?
  
• My post supervisor says we?re going to telecine/scan for ?full information?. It is my understanding that film negative has 15 stops of latitude from whitest white to blackest black. A film print is something like 9 stops. An Arriscan (thanks to its 2-light system) supposedly captures ALL the information (ie. All 15 stops). What range of that does a Spirit 2k capture if it's set up to scan for ?full information?.
  
• If I do final color correction tape-tape in a daVinci...I?ve heard that?s not as good as a ?real? DI box. In what way?
  

Thanks for your help.

 

Cycad

[Stephen Williams]

(I make a lot of assumptions on the way to my questions... Feel free to correct me on any of them.)

 

We just completed shooting a feature on Super16mm. Our plan is to do a poor-man?s DI by telecine-ing to HDCAM-SR 4:4:4 RGB 10-bit log, 1920 x1080, 24p fps.

 

It sounds like Spirit is the best Telecine DI option out there (the next step up would be a frame-by-frame scanner...too expensive). So I had a few questions about the Spirit systems...

• I hear a lot about new vs. old Spirit scanners and how the new ones are so much better. What do I go by? What question can I ask to know if the Spirit is ?new? or ?old??
  
• When people say a Spirit DataCine. Is that a system older than 2k and 4k? Or is that the full name that?s rarely used and they just haven?t specified what "k"? Ie Spirit DataCine 4k.
  
• I?ve read Marc Wielage say in these forums that Spirit 2k is 12bit linear, Spirit 4k is 16bit linear. Is there a mathematical difference (as opposed to debatable visible difference) between the two once they?re collapsed into 10-bit log?
  
• Since we?re working with 16mm, grain reduction is going to be important. The Spirit 2k has a built in grain reduction system. (correct me if i?m wrong) There?s also ?software? grain reduction in the color correction systems. What is the best way to deal with grain reduction? Is it good to use the Spirit grain reduction or wait for the end to deal with that. Or just a combination of Spirit and grading software?
  
• My post supervisor says we?re going to telecine/scan for ?full information?. It is my understanding that film negative has 15 stops of latitude from whitest white to blackest black. A film print is something like 9 stops. An Arriscan (thanks to its 2-light system) supposedly captures ALL the information (ie. All 15 stops). What range of that does a Spirit 2k capture if it's set up to scan for ?full information?.
  
• If I do final color correction tape-tape in a daVinci...I?ve heard that?s not as good as a ?real? DI box. In what way?
  

Thanks for your help.

 

Cycad

 

Hi,

 

The 4K Spirit is better but I think it may be overkill for S16. There are many more 2K Spirits out there.

 

Datacine is an option as is picture rotate!. Data output is not real time.

 

I would try to grade direct from the telecine, more information will be available from the original neg leaving minor tweaks to later.

 

A high end DI will cost many times the amount of a Spirit to HDCAM SR DI. Its down to budget do you want to spend, less than $80,000 or more than $200,000? Budget will dictate your workflow!

 

Stephen

[Michael Most]

We just completed shooting a feature on Super16mm. Our plan is to do a poor-man?s DI by telecine-ing to HDCAM-SR 4:4:4 RGB 10-bit log, 1920 x1080, 24p fps.

 

You cannot do a dual link, log format transfer on any Spirit machine other than the 2K and 4K. At least not natively.

 

[*]I hear a lot about new vs. old Spirit scanners and how the new ones are so much better. What do I go by? What question can I ask to know if the Spirit is ?new? or ?old??

 

The original Spirit, commonly referred to as the Spirit 1, was limited in both resolution and color depth. The second generation, commonly referred to as the Spirit 2, was less limited. The current generation, known as the 2K and 4K, is not resolution limited and has far less color depth limitations. It is by far a superior machine to either of the first two models.

 

[*]When people say a Spirit DataCine. Is that a system older than 2k and 4k? Or is that the full name that?s rarely used and they just haven?t specified what "k"? Ie Spirit DataCine 4k.

 

Spirit DataCine has always been the name by which the Spirit line has been called. It was a term coined by Philips to differentiate the machine when it was first released from telecines that could not and would not offer data output, such as the Rank Cintel Ursa line.

 

[*]I?ve read Marc Wielage say in these forums that Spirit 2k is 12bit linear, Spirit 4k is 16bit linear. Is there a mathematical difference (as opposed to debatable visible difference) between the two once they?re collapsed into 10-bit log?

 

No. A 10 bit log image is a 10 bit log image. However, the results in terms of the detail and color depth information retained from the original linear image are different. "How" different is debatable.

 

[*]Since we?re working with 16mm, grain reduction is going to be important. The Spirit 2k has a built in grain reduction system. (correct me if i?m wrong) There?s also ?software? grain reduction in the color correction systems. What is the best way to deal with grain reduction? Is it good to use the Spirit grain reduction or wait for the end to deal with that. Or just a combination of Spirit and grading software?

 

You don't want to use any grain reduction during a scanning process. You also don't want to do it prior to color correction, because color correction itself is inevitably going to add noise. Any noise reduction processing should be done at the end of the processing path, although I would not recommend using it unless the grain is really objectionable. And even then, only very lightly.

 

[*]My post supervisor says we?re going to telecine/scan for ?full information?. It is my understanding that film negative has 15 stops of latitude from whitest white to blackest black. A film print is something like 9 stops. An Arriscan (thanks to its 2-light system) supposedly captures ALL the information (ie. All 15 stops). What range of that does a Spirit 2k capture if it's set up to scan for ?full information?.

 

You "compress" the amount of information you're capturing by setting the telecine to hold everything between Dmin and Dmax. This means you're not throwing away any information, but it also means you're going to have a "flatter" image as a result. How much of that "flat" image you keep is determined in final color correction.

 

[*]If I do final color correction tape-tape in a daVinci...I?ve heard that?s not as good as a ?real? DI box. In what way?

 

Some of the top DI facilities in the world - including Technicolor, Company 3, Modern Videofilm, and a number of others - use DaVinci 2K Plus units to color correct digital intermediates. So I'm not exactly sure what you're talking about. If you're referring to "tape to tape" correction vs. a "DI" type correction, you're probably referring more to the source material than the color corrector, in other words, a telecine transfer vs. a true film scan. In that case, the scan is going to have a lot - and I mean a lot - more information in it than a telecine transfer. Going to a dual link, RGB transfer mitigates this somewhat, making it somewhat equivalent to a 2K scan, but today many true 2K scans are done by actually scanning 3K or 4K and downsampling to 2K - providing an element with more detail and better color fidelity due to numerous factors. If you're referring to color correction flexibility, or the quality of the result, that's often determined far more by the human being operating the color correction device than it is by the device itself.

[Jake Emmerson]

Thank you very much for your response mmost. Very informative. Of course, every answer leads to more questions...

 

You "compress" the amount of information you're capturing by setting the telecine to hold everything between Dmin and Dmax. This means you're not throwing away any information, but it also means you're going to have a "flatter" image as a result. How much of that "flat" image you keep is determined in final color correction.

I guess my question then is does Dmin to Dmax on a Spirit 2/4k go as high into the brights and and as low into the darks as an Arriscan (with it's 2-light system)? Or am I drinking Arri's PR?

 

Also:

 

Obviously I wouldn't want to use that flat image for final, I'd color-grade it to be more contrasty. Now, if I take an image into photoshop and look at the histogram I'll get a nice full curve of information. (follow me here...)

 

I then lower the contrast in the image using level. Then I open the levels adjustment and increase the contrast again to make it look the same. When I look at the histogram, the curve is the same, but now it's full of gaps.

 

Now, I can't tell the difference with my eye between the image I started out with and the 'processed' image, but it looks like I'm losing shades of 'color'? 'luminance'?.

 

Isn't the same thing happening if I scan Dmin-Dmax and then boost the contrast in the colorgrade? Or is this information that doesn't really affect the image quality? (like I said, I can't see the difference).

[Phil Rhodes]

Hi,

 

> I guess my question then is does Dmin to Dmax on a Spirit 2/4k go as high into the brights and and as low

> into the darks as an Arriscan (with it's 2-light system)? Or am I drinking Arri's PR?

 

I would suspect not, but probably not by an enormously objectionable amount. Film doesn't have anything like 15 stops of range anyway - it's more like 10 if you don't want to pump up the grain in the shadows too much.

 

> Now, I can't tell the difference with my eye between the image I started out with and the 'processed' image,

> but it looks like I'm losing shades of 'color'? 'luminance'?.

 

Yes. Pushed too far, it make steps visible in graduated colour, a phenomenon called banding or contouring.

 

> Isn't the same thing happening if I scan Dmin-Dmax and then boost the contrast in the colorgrade?

 

Yes. This is an unavoidable, intrinsic factor in doing any sort of manipulation on a digital image. It can be disguised by error-diffusion dithering (randomly modifying pixel values within the range of uncertainty), but that eventually makes things look noisy.

 

> Or is this information that doesn't really affect the image quality? (like I said, I can't see the difference).

 

It's a matter of degree. Any competent colour correction device works at a higher bit depth than the Photoshop image you were working with (unless you were working with a 16-bit/channel file such as a camera raw image, in which case there's a very real possibility that you were working at a similar or higher bit depth...) which mitigates the problem to a degree. Also it is usually very hard to push a film-originated image into banding because the very high noise level caused by film grains has the same effect of diffusing errors that dithering can have. We certainly accept grades applied to film originated images that we would never accept on cleaner source data because it would contour horribly.

 

As far as I know no current colour corrector implements diffusion dithering. I know one company did consider using it but dismissed the idea when people became nervous about creating software that wouldn't necessarily produce the same bit-identical result twice from the same data. You can simulate that sort of thing in anything that offers add-noise or film-grain effects filters, such as Baselight or Lustre.

 

Phil

[Tenolian Bell]

I guess my question then is does Dmin to Dmax on a Spirit 2/4k go as high into the brights and and as low into the darks as an Arriscan (with it's 2-light system)? Or am I drinking Arri's PR?

 

The Arriscanner is a scanner, while the Spirit is a telecine. Telecine is a realtime process while scanning is not. Scanners use higher resolution sensors than telecines and yes they also record more luminance information.

 

The Arriscanner uses separate red, green, and blue LED's to flash the sensor. It takes one RGB picture for highlights, another RGB picture for shadows, and combines both pictures together. The good part is you have a high resolution image with most of the detail captured on the negative. The bad part is its a slow process.

[Phil Rhodes]

Hi,

 

Is it using red, green and blue LEDs, or white LEDs with red, green and blue filters over them?

 

The latter technique would provide better spectral characteristics at the expense of some switchoff lag, but there's probably a way to correct for it.

 

Phil

[Tenolian Bell]

From what I've been told they are actual red, green, and blue LED's.

[Guest ecce_rodrigo]

Hi,

 

I thought there were no blue leds!!!.

I read something about a scientist trying to make one just the other day?

And I think Spirit is only telecine, (in this case the most correct term would be real time film scanner, for it has the quality of a data-cine) in the 2k mode. 4K makes it a bit slower.

 

Rodrigo

[John Pytlak RIP]

Hi,

 

I thought there were no blue leds!!!.

I read something about a scientist trying to make one just the other day?

 

Rodrigo

 

LEDs and OLEDs are available in a wide variety of colors:

 

 

http://www.ledtronics.com/pages/news_articles/07-26-02lr.htm

 

http://www.kodak.com/eknec/PageQuerier.jht...requestid=16111

 

http://www.kodak.com/eknec/PageQuerier.jht...pq-locale=en_US

 

Organic light emitting diode (OLED) technology uses substances that emit red, green, blue or white light. Without any other source of illumination, OLED materials present bright, clear video and images that are easy to see at almost any angle.