Star Wars Episode III 4:2:2 or 4:4:4?

Mike Brennan · February 27, 2005

What flavour HDCAMSR recording was used on Star Wars Episode III?

4:2:2 @ 2.7:1 compression?

4:4:4 SQ mode @ 4.2:1 compression?

4:4:4 at HQ mode @ 2:1 compression? (880 Mbs mode)

Hype and press releases of using "HDCAMSR 4:4:4 system" without actually quoting what mode the recorder was in seems a common theme with all the movies that have used HDCAMSR, Collateral, Sin City and Star Wars Ep III

(We now know Collateral was in 4:2:2 mode)

Mike Brennan

Landon D. Parks · February 28, 2005

I imagine at was at least recorded @ 4:4:4, or other wise why not just use the HDW-F900? But rather 4:2:1 or 2:1 is a question someone with more experiance will have to answer.

Robert Sanders · March 3, 2005

It's being shot using the Sony F950's using HDCAM-SR (4:4:4) with Fujinon lenses. Rented from Plus8Digital.

Also, don't forget there are other benefits than just 4:4:4 color space. The F950's in SR mode record a full 1920x1080 pixel frame rather than HDCAM's 1440x1080.

Mike Brennan · March 3, 2005

It's being shot using the Sony F950's using HDCAM-SR (4:4:4) with Fujinon lenses. Rented from Plus8Digital.

Also, don't forget there are other benefits than just 4:4:4 color space. The F950's in SR mode record a full 1920x1080 pixel frame rather than HDCAM's 1440x1080.

<{POST_SNAPBACK}>

With greatest respect could you quote the source of your information?

Also is it SQ or HQ mode of 4:4:4

Any word on Sin City? Who provided kit?

I'm getting conflicting reports about Collateral, since they were using prototype 5000. I understand they may have gone 4:2:2 as 444 boards at one point were not ready. There was testing on 444 on a S2 and also HDCAM was used which ended up 4:2:2 in the edit. So plenty of room for confusion and plenty of room to obfuscate.

Many producers will be making decisions about using SR based on what they are hearing about these movies, and other big ones in the wings but since the SRW1 is still not fully configured there are plenty of reasons for Sony to hide or delay info about what flavour of SR format these movies were actually shot on....

No shortage of PR saying X movie shot on "HDCAM SR format"

Seems SR stands for secret recordings.......

Mike Brennan

Edited March 3, 2005 by mike

Mike Brennan · March 3, 2005

Ok the thing that has caused the confusion was that HDSDI output from f900s were used and recorded onto SR.

So initially HDCAM was used then they switched.

So in the edit their were SR tapes with 4:2:2 recordings.

Hence the confusion.

Lucas also went 4:4:4 for latest Star Wars I assume at SQ mode but don't quote me.

Mike Brennan

John Pytlak RIP · March 3, 2005

The F950's in SR mode record a full 1920x1080 pixel frame rather than HDCAM's 1440x1080.

<{POST_SNAPBACK}>

Don't you still lose over half of the resolution with the anti-aliasing filter? Nyquist must be obeyed, unless you are very careful to exclude scene detail that could create aliasing artifacts. Pixel count on the chips is misleading when you talk about real image sharpness.

Mike Brennan · March 5, 2005

Don't you still lose over half of the resolution with the anti-aliasing filter? Nyquist must be obeyed, unless you are very careful to exclude scene detail that could create aliasing artifacts. Pixel count on the chips is misleading when you talk about real image sharpness.

<{POST_SNAPBACK}>

Starway wrote about recording more pixels with an SR than HDCAM.

This thread is drawing attention to the tri mode SR recording format and how it is important for us to know which mode has been used on Sin City, Star Wars III and the new movies shot on Genesis.

I am interested in exploring work flow, reliability and different costs of each of the SR modes of recording and uncompressed recorders as well.

Since Genesis, Arri D20 and 2/3 inch 3 chip cameras all have different camera resolutions (and can all be recorded to 1920x1080) Nyquest would be more relevant in a comparison of cameras, not recording formats?

Mike Brennan

John Pytlak RIP · March 6, 2005

Nyquest would be more relevant in a comparison of cameras, not recording formats?
Mike Brennan

<{POST_SNAPBACK}>

I AM talking about the camera sensors, not the recording formats. Those 1920 horizontal pixels do not produce 1920 lines of horizontal image resolution. Because of the required Nyquist filter (to reduce aliasing), the most you can hope for is 960 lines (480 line pairs or cycles), and probably less in practice.

There are good reasons digital still cameras have quickly moved far beyond a measly 2 megapixels, and getting reasonable sharpness without aliasing is a prime one:

http://www.kodak.com/global/en/professiona...8.22.3.18&lc=en

http://www.kodak.com/global/en/professiona...8.22.3.14&lc=en

http://www.kodak.com/global/plugins/flash/...RnGallery.blind

http://www.kodak.com/global/plugins/flash/...RcGallery.blind

Tenolian Bell · March 6, 2005

Also is it SQ or HQ mode of 4:4:4

<{POST_SNAPBACK}>

What are these modes?

Do they change the sampling rate?

Mike Brennan · March 6, 2005

What are these modes?

Do they change the sampling rate?

<{POST_SNAPBACK}>

4:2:2 2.7:1 440mbs 50 minutes record duration

4:4:4 SQ 4.2:1 440mbs 50 minutes record duration

4:4:4 HQ 2:1 880mbs 25 minutes approx?

It uses intra frame for progressive and intra field for interlace

SRW1 portable has stereo 4:2:2 mode where it records two seperate 4:2:2 HD pictures at 2.7:1 at same time.

Recent upgrade for SRW1 portable will do 720 60p.

Genesis version will do 1920x1080 50p so Sony are promising.

Genesis version will also do time lapse. I think these features will become availabale to all users eventually, but don't lay the foundations for a job until you have seen it work.

Press releases that say that a feature or commercial was made using "SR format 4:4:4 recorder" could mean it was recorded 4:2:2

Mike Brennan

March 7, 2005

I AM talking about the camera sensors, not the recording formats. Those 1920 horizontal pixels do not produce 1920 lines of horizontal image resolution. Because of the required Nyquist filter (to reduce aliasing), the most you can hope for is 960 lines (480 line pairs or cycles), and probably less in practice.

<{POST_SNAPBACK}>

This is one of the least-understood aspects of so-called "oversampling". There are indeed many scenes (or parts of scenes) where the anti-alias filters could actually be removed, producing a noticeably sharper image, but with an actual "physical" glass filter this is simply not practical.

However when you "oversample" (for example, scanning film at 4K for an output at 2K) this filtering can be done in software rather than immutable glass. The software can be "trained" to detect where sampling artifacts will occur with the downconverted image, and apply the filtering to just those parts of the image that need it. The result is a subjectively much sharper-looking image. It's almost as if you could "carve out" a custom anti-alias filter for each scene.

This is one reason downconverted HDTV always looks much better on SDTV screens that video from SD cameras; the apparent resolution can approach twice the maximum resolution that Nyquist's Theorem would suggest is possible.

To the old agage: "Garbage in - Garbage out" we could also perhaps add: "More Detail in - More Detail out". I mean, even plain old VHS recordings look noticeably better if you're recording from a three-chip camera!