CineSSD strategy

[Paul Nordin]

Right on. Actually I am doing 4444XQ Dlog vs. DNG tests today cuase I have not been seeing epic differences in Dynamic Range.

And you won't. 4444XQ is so close to uncompressed in quality (while not technically being so) that it is visually indistinguishable from uncompressed. I do a lot of color correction of features, and most distributors these days only want ProRes 4444 as their final deliverable. The only time that people really want to deal with the overhead of truly uncompressed footage is for shots that involve a lot of VFX work. Then they typically ask for EXR or DNG image sequences, and for them CinemaDNG can be a good choice when capturing.

All of these high-end 12-bit codecs are designed to preserve high-frequency detail as captured with a true HDR camera (Arri Alexa, or Red). It is very questionable in my mind anyway, that the new X7 captures enough DR to require or benefit from that over ProRes 4444 or XQ. Bear in mind that while the X7 is really cool new tech (which I plan to buy as soon as Camrise gets stock), its priced to match mid-level DSLR cameras...not even close to actual Digital Cinema cameras for which XQ was designed.

 

[rmb]

Just did this down and dirty test. Basically a cloudy day outside shot AL in a window.
Both clips were shot ISO 200 at a 50th f 2.5 Olympus 17mm quick color grade in Resolve
Clip 01 - 4444XQ DLog
Clip-02 - DNG RAW
I must say because you can push the DNG just a bit more, I notice a little less noise in the blacks and grey card. As for me it is barely worth it unless you are in extreme an Dynamic Range environment.
Clip-01


Clip-02

 

[Sarsipius]

Just did this down and dirty test. Basically a cloudy day outside shot AL in a window.
Both clips were shot ISO 200 at a 50th f 2.5 Olympus 17mm quick color grade in Resolve
Clip 01 - 4444XQ DLog
Clip-02 - DNG RAW
I must say because you can push the DNG just a bit more, I notice a little less noise in the blacks and grey card. As for me it is barely worth it unless you are in extreme an Dynamic Range environment.
Clip-01
View attachment 17091

Clip-02
View attachment 17092

Thanks for this, it's very useful.

Just out of interest which did you find was truer to the natural colour?

I noticed the small "yellow" vase on the window sill was noticeably different between the two.

Also, could you comment on the average file sizes per minute?

Thank you.

 

[rmb]

Good eye. That’s hard to say and as a result I think that I will do a similar test again. The reason why it is hard is that its not exactly an apple to apple comparison. The DNG was captured as is (RAW) and the 4444QX was captured in DLOG. The reason that that is significant is that in DaVinci, the DNG was graded freehand and the 4444QX I started with an ARRI LUT which I find adds a bit of red.
I also find that the DNG Raw starts off more dark and has to be treated very differently.
I will also roll 1 minute of each and get back to you. Lastly if you enjoy color grading I can always provide you with a full res tiff frame of the 4444QX and a single DNG frame to play with your self.

Anyway, thanks for reading my stuff.

rb

 

[rmb]

Thanks for this, it's very useful.

Just out of interest which did you find was truer to the natural colour?

I noticed the small "yellow" vase on the window sill was noticeably different between the two.

Also, could you comment on the average file sizes per minute?

Thank you.

I did a new test. Attached are 4 grabs from DaVinci all shot at ISO 200 @ 50th f-4.5 Olympus Zuiko 17mm 3840x2160 with 1 Westcott Daylight Balanced LED Flex light for fill.
I tried to add as many colors and reflective surfaces as I could as well as large diversities in dynamic range and saturation. Also (and this is important) The 30k in cash is Movie Prop Money so please don't come to my house and rob me!
Clip-01 = h.264 (Gross)
Clip-02=422
Clip-03=4444XQ
Clip-04=DNG

Couple of things that I noticed;
H.264 is great until you see a real image!
422 is **** nice
4444XQ is hard to distingush but a tad easier to grade
DNG color space is totally different so I could't match color and dynamics exactly. You don't notice it right off on most of the image, but if you look at the bolder colors like the Magenta monster or especially in the 2 diagonal blues on the color chart, you will really notice it. Also look at the the Blues and notice the massive lack of Noise compared to the 422 and 4444. That is really big for me.

Regarding file size, I found this the most suprising and will make a unique post on this as to get more user input. 4444XQ actually eats 25% MORE space than DNG so I am not sure why I will continue to shoot it.
1 minute clip of the exact static scene as the pictures above
H.264 = 455 MB
422= 5.41 GB
4444QX=12.45 GB
DNG Raw=8.43 GB.

Let me know if any of you find any of this helpful.

Happy Sunday.
rb

 

[Paul Nordin]

Regarding file size, I found this the most suprising and will make a unique post on this as to get more user input. 4444XQ actually eats 25% MORE space than DNG so I am not sure why I will continue to shoot it.

Possible explanation XQ is a 12bit RGB and 16bit A codec.

 

[2nd2non]

I did a new test. Attached are 4 grabs from DaVinci all shot at ISO 200 @ 50th f-4.5 Olympus Zuiko 17mm 3840x2160 with 1 Westcott Daylight Balanced LED Flex light for fill.
I tried to add as many colors and reflective surfaces as I could as well as large diversities in dynamic range and saturation. Also (and this is important) The 30k in cash is Movie Prop Money so please don't come to my house and rob me!
Clip-01 = h.264 (Gross)
Clip-02=422
Clip-03=4444XQ
Clip-04=DNG
View attachment 17271 View attachment 17272 View attachment 17273 View attachment 17274
Happy Sunday.
rb

Thanks for sharing
1) Looks like you C-DNG sequence is missing lens correction. What fw version are you using...wondering if this is another fail by DJI
2) Do you notice color shift using Iso200 vs 100?
3) Did you use AWB or fixed setting for all images?

 

[rmb]

Thanks for sharing
1) Looks like you C-DNG sequence is missing lens correction. What fw version are you using...wondering if this is another fail by DJI
2) Do you notice color shift using Iso200 vs 100?
3) Did you use AWB or fixed setting for all images?

Good questions all.
Lens correction-Not sure how would I confirm that?
Color Shift-Honestly I did not consentrate on that. I will next test. Question, why would you think that would happen BTW 100 and 200ISO? Another firmware thing?
WB. Manual white balance (Actually chose sunny)

 

[2nd2non]

Good questions all.
Lens correction-Not sure how would I confirm that?
Color Shift-Honestly I did not consentrate on that. I will next test. Question, why would you think that would happen BTW 100 and 200ISO? Another firmware thing?
WB. Manual white balance (Actually chose sunny)

I have an older fw version and CDNG has lens correction just like h264. ProRes does not.
I read somewhere that ISO 100 has a bit of megenta shift vs ISO 200. Not sure why this would be the case and haven't necessarily noticed it myself.

 

[Sarsipius]

I did a new test. Attached are 4 grabs from DaVinci all shot at ISO 200 @ 50th f-4.5 Olympus Zuiko 17mm 3840x2160 with 1 Westcott Daylight Balanced LED Flex light for fill.
I tried to add as many colors and reflective surfaces as I could as well as large diversities in dynamic range and saturation. Also (and this is important) The 30k in cash is Movie Prop Money so please don't come to my house and rob me!
Clip-01 = h.264 (Gross)
Clip-02=422
Clip-03=4444XQ
Clip-04=DNG
View attachment 17271 View attachment 17272 View attachment 17273 View attachment 17274
Couple of things that I noticed;
H.264 is great until you see a real image!
422 is **** nice
4444XQ is hard to distingush but a tad easier to grade
DNG color space is totally different so I could't match color and dynamics exactly. You don't notice it right off on most of the image, but if you look at the bolder colors like the Magenta monster or especially in the 2 diagonal blues on the color chart, you will really notice it. Also look at the the Blues and notice the massive lack of Noise compared to the 422 and 4444. That is really big for me.

Regarding file size, I found this the most suprising and will make a unique post on this as to get more user input. 4444XQ actually eats 25% MORE space than DNG so I am not sure why I will continue to shoot it.
1 minute clip of the exact static scene as the pictures above
H.264 = 455 MB
422= 5.41 GB
4444QX=12.45 GB
DNG Raw=8.43 GB.

Let me know if any of you find any of this helpful.

Happy Sunday.
rb

Great posts.

Thank you so much for re-running the test and posting the results. It is hugely helpful to those of us wondering which codec is best for our purposes. I was thinking I'd go 4444QX until you described the file size. Now I'm torn between 422 and DNG and will have to think longer about this.

Cheers though!

 

[rmb]

I think for me the choice is simple. 422 is good enough for most anything and usually all my clients pay for. That said for personal use or when there is a shot that I know that I am going to need to really push in post or do some re-framing I will shoot 5k DNG raw. The 4444XQ seems to be an unnecessary option. That said if you have to do a lot of mixing and matching the 4444XQ and the 422 color grad and match perfectly. That said as I am far from an expert I would like to know from others what 44444XQ provides to them other than la larger file