Redcode Comparions / Calculations

[Chris Jordan]

So I was laying in bed this morning just feeling super lucky to even own a Dragon, and the fact that I had left over pizza in the fridge just waiting for me to have for breakfast. Somehow that started this thought:

"I'm shooting 6K RC 10:1 then 5K 8:1, would they have the same size file? OR would 6K 10:1 and 5K 10:1 be the same size file....?

Thanks for the input, and if anyone wants some breakfast, i'm serving up in a few.

 

[joshua csehak]

I've actually been meaning to do some redcode comparisons myself. I think probably the best way to stress test it would be to shoot something like a windy wheat field, and a smooth gradient at various ISOs.

Looking forward to seeing yours!

 

[Cosmo D]

I guess you can use that no?
http://www.red.com/tools/recording-time

 

[Chris Jordan]

YES! There it is... you're the man

 

[Gavin Greenwalt]

The math is pretty straight forward.

Width * Height * 1/REDCODE = Datarate. 

So 6k:  6144 (h) x 3160 (v) = 19,415,040 / 10 = 1,941,504

vs 5k: 5120 (h) x 2700 (v) = 13,824,000 / x = 1,941,504

== : 13,824,000(x/x) = 1,941,504x
== : 13,824,000/1,941,504 = 1,941,504x/1,941,504
== : x = 7.12

So Redcode 7:1 5k == REDCode 10:1 6k.

If you want to skip the "showing your work" the equation can be reduced:

(SmallFormat.Height * SmallFormat.Width) / (LargeFormat.Height * LargeFormat.Width / LargeFormat.REDCode) = SmallFormat.REDCode

LargeFormat 6k HD : 5617 (h) x 3160 (v), REDCode 8:1
SmallFormat UHD : 3840 (h) x 2160 (v)

What REDCode would we be able to use to match 6k HD at 8:1 compression?


(3840 * 2160) / (5617 * 3160 / 8) = 3.73 ---> REDCODE 4:1

 

[Chris Jordan]

Thanks so much GAVIN. I honestly don't notice a difference from 9:1 - 11:1.. Does it matter than much if i'm shooting for myself without any need for Motion graphics or special effects?

 

[Gavin Greenwalt]

Here is a spreadsheet version.
http://indiecinerental.com/RedcodeCalc.html

For Comparison of subjective quality:
http://www.reduser.net/forum/showthread.php?124071-Compression-Wedges

^BTW can we get a mod to sticky that. I think it's useful enough and comes up often enough to just sticky at the top.

EDIT: Updated spreadsheet version to be an embedded spreadsheet. Just enter in your two sizes and a redcode value and it'll cross convert for one to the other.

 

[Marcus Friedlander]

+1 on the sticky. Thanks Gavin for the information!

 

[Jeff Kilgroe]

The one thing that should be pointed out here with wavelets is that larger compression ratios hold up better as the resolution increases. We get more efficient compression when there is more data to sample and interpolate into wavelet structures. So 10:1 at 6K is going to be visually comparable to 7:1 or 8:1 at 5K. That's an arbitrary and approximate example, we would have to conduct specific testing or talk to the compression gurus at RED to get a more direct answer. It's just something to consider when shooting Dragon at 6K, knowing that you don't have to stress as much over the higher RedCode numbers we often see.

I also think a sticky is warranted for the wedges.

 

[Gavin Greenwalt]

The one thing that should be pointed out here with wavelets is that larger compression ratios hold up better as the resolution increases. We get more efficient compression when there is more data to sample and interpolate into wavelet structures. So 10:1 at 6K is going to be visually comparable to 7:1 or 8:1 at 5K.

I believed that too until I tested the motion wedges. The problem is that REDCode isn't a straight RGB wavelet encoding it heavily favors the Luma channel and aggressively throws away Hue and Saturation data. The result is that you get really bad 'flicker' at higher encoding ratios. So while you would expect it to generally soften the image from 6k 10:1 down to about the same sharpness of 7:1 5k what really happens is that it's 6k sharpness but like 2k chroma. vs 5k sharpness with 4k chroma.