Jump to content

sunyata

Members
  • Posts

    391
  • Joined

  • Last visited

About sunyata

Profile Information

  • Gender
    Male
  • Location
    Los Angeles

Recent Profile Visitors

4,197 profile views

sunyata's Achievements

Frequent member

Frequent member (4/5)

120

Reputation

  1. sunyata

    Great article

    You're right, I want the bicycle. I'm sorry.
  2. sunyata

    Great article

    Ah, bang goes the dynamite. I was actually just laughing at how cantankerous these forums are, not you specifically. But if you want to talk about screw ups, these are your words: This is not correct but I understand how one could think that from that Shogun test, to be fair. The radial I used was not initially monochrome btw, it was de-saturated to try to closely match the source test you referenced. "nice try to decompress with that Marcie one haha".. I'm not even sure what that means. I said "maybe" 3 times in a sentence to underline how many possible variables could be introduced in the Shogun test, i.e. "Maybe this, maybe that, who knows." "Or have hard chance to publicly accept they are as much ordinary as the others." " We all suffer of such need of love. Everyone in the same boat." Okay, now I'm in a 12 step program?
  3. sunyata

    Great article

    I'm am genuinely laughing out loud, not fake LOL.
  4. Yea, it's a 4k 10bit cin file (although I think it was originally scanned at 2k), maybe the last of the film neg leader ladies? As time goes on, she starts to look lower and lower quality You can still push the hell out of her hair. http://www.northwestchicagofilmsociety.org/wp-content/uploads/2011/09/DSCN2295-e1316147819183.jpg
  5. lol, the marcie is just FPO. This is a gamma test of S-Log2 vs S-Log3 with an 8bit intermediate pipeline. I think the banding in the toe for the S-Log3 looks better than S-Log2 (top right ramps) and clearly more contrast below like 6% is retained (bottom chart). Highlights aren't really as different as I was expecting. S-Log3 should be a little brighter but I can barely notice it.
  6. Emanuel, it's a really simple test to replicate, please feel free to try. I was working on a different test more related to this topic of the A7sII and S-Log3 when I got distracted by that Vimeo link. Here's the S-Log-3 test: I was rendering a 3D project, I swear I don't just do tests all day.
  7. Check out the results of the updated image if you want to see a different radial used to start with... both radials have more banding as you would expect, but the 4:2:2 is marginally better; I think quite a few people wouldn't know the difference.
  8. It's true that some colors degrade more than others, but because the source radial was generated in RGB space and converted to Y'CbCr, you will still see chroma subsampling artifacts (it still get's subsampled in Cb and Cr to compose the final image). I was comparing with the other test, which looks grayscale, but in the past I've preferred to test this with random animated colors just to see where most breakup happens. I've updated the image to use a lot of red, typically the worst color.
  9. Oh my, I never saw that one. There is a little flaw in his test that probably explains the huge difference between the demonstrated quality of 8bit 4:2:0 vs 8bit 4:2:2: he's recording externally to the Shogun with the 8bit 4:2:2 and internally to the GH4 for the 4:2:0. I'm gonna bet a dollar that's why they look so different. Maybe it's a data rate thing, maybe its smoothing happening with the Shogun, Maybe it's the GH4's implementation of 8bit 4:2:2, I don't know. Below is just a radial saved as 8bit dpx, then two conversions to respective chroma subsampling schemes. One is yuv420p 8bit and the other is yuv422p 8bit, Same very high h264 constant bitrate settings for both. Re-imported, gamma curve applied and saved out as a sRGB png.
  10. Just to address both of these comments, even though they are disagreeing, the total number of colors per pixel is kind of a useless metric. Nobody will ever ask you "Hey, can you save that file out 1,073,741,824 colors please?", they'll just ask for 10bit, 12bit etc. This is because the total number of possible colors per pixel is not only impossible to remember, is not really what determines noticeable quality issues; that's contrast. You get 256 or less for 8bit, 1024 for 10bit etc.
  11. If you're mainly interested in color that you can actually correct for sRGB display space, then the inclusion of S-Gamut3.Cine is likely the big news. "S-Gamut3.Cine is similar to negative film scan which was used for TV production, film out digital cinema. Color reproduction is designed slightly wider than DCI-P3 to provide ample room for grading. Thus manual grading for P3 becomes easier. S-Gamut3.Cine is more convenient to grade than S-Gamut3 and S-Gamut as camera digital negative. If you would like to apply print film emulation, please increase saturation to about 1.4." Can never get enough plots.
  12. Slog3 might put a little more strain on the highlight noise level in 8bit 4:2:2, but I think that will largely be forgotten in exchange for the color benefits of encoding in S-Gamut3.Cine, which people should find much easier to color correct; It's closer to a negative film scan.
  13. Under the "preview" button http://cameramanben.github.io/LUTCalc/LUTCalc/index.html there's also a waveform, vectorscope, and rgb parade option. Cursor over the test image to get 10bit color values.
  14. Assuming you haven't seen this.. it looks like cameramanben has done something really useful here: he's compiled a ton of research and written a JavaScript LUT calculator tool released under a GNU freesoftware license (I think this is about as free as it gets). This could be used for 1D shaper LUT's, or 3D LUT's with gamma/gamut baked together - if that's more your thing - and even custom code value ranges. I think you can upload a sample image too. I see an A7s profile as well as Rec2020 gamut, DCI-P3 gamut... it's on github. Online tool: http://cameramanben.github.io/LUTCalc/LUTCalc/index.html Link to browser based apps: https://cameramanben.github.io/LUTCalc/ Objective-C wrapper for MacOS: https://github.com/cameramanben/LUTCalc-For-Mac He needs a donation button somewhere.
  15. Strange, they had 37.6mm diagonal for the D800, but even if you compensate for a vertical 16:9 crop when shooting 1.78, you get 41.1. from the width of 35.9, If you just calculate the full sensor you get 43.2mm diagonal. I'm going to put it in at 41.1, not sure how they got the 37.6mm diagonal. Do you get the full width of the sensor when shooting video?
×
×
  • Create New...