Nikon Z6 does pixel skipping in FX ProRes RAW

[falconeyes]

Hi,

I eventually concluded my analysis about how Nikon actually produces its 4k raw feed for external ProRes RAW video (FX mode).

-> http://blog.falklumo.com/2020/01/the-conundrum-of-nikon-z6-prores-raw.html

To summarize:

The Z6 does pixel skipping, unfortunately.

If I must speculate, I'd say it drops every third block of two lines and two rows, leaving 44% of pixels.

Maybe, it is irrelevant in the field, let me know.

 

[Horshack]

Nice work Falk. Just read the article and your conclusions seem reasonable. Will read it again more thoroughly when I have time later today.

 

[falconeyes]

Nice work Falk. Just read the article and your conclusions seem reasonable. Will read it again more thoroughly when I have time later today.

Thank You so much for cross-reading.

Something is going on in ProRes RAW mode for sure but I may still be off with my conclusion. Another pair of eyeballs is highly appreciated

 

[Gollan]

I appreciate your analysis and insights. My perception is that much of the hype over Z6 ProRes RAW is being generated by Nikon, so it would not make sense for them to deliberately nerf it to protect the Z7. I hope that Nikon and Atomos engineers have simply concluded that pixel skipping to reduce bandwidth doesn't have any practical impact on video. However, if there are image quality impacts, do you have any thoughts about mitigating them in post processing?

 

[falconeyes]

My perception is that much of the hype over Z6 ProRes RAW is being generated by Nikon, so it would not make sense for them to deliberately nerf it to protect the Z7.

Yes, right. But it wasn't so at CES 2019 when the announcement was made and 4k was already being decided.

However, if there are image quality impacts, do you have any thoughts about mitigating them in post processing?

I don't currently own a Ninja V (part of the reason I did the analysis was the decision making about whether or not I want to own one ). Therefore, I'd like to leave comments about possible implications in the field, and possible workarounds, to people using ProRes RAW in actual projects. I'd really like to hear from them and hope, practical implications are minor.

As with any source of aliasing artifacts, mild blur (motion and/or focus) should help a lot. I guess that handheld filming with a 180° shutter (1/50s) suppresses most of the aliasing issues already.

 

[robertfel]

Did you test the DX crop?

 

[falconeyes]

No.

But the DX crop uses a 4k readout mode to produce 4k raw. Nothing to test for ...

(I.e., there is no pixel skipping in 4k DX raw.)

--
Falk Lumo

 

[beatboxa]

Hi,

I eventually concluded my analysis about how Nikon actually produces its 4k raw feed for external ProRes RAW video (FX mode).

-> http://blog.falklumo.com/2020/01/the-conundrum-of-nikon-z6-prores-raw.html

To summarize:

The Z6 does pixel skipping, unfortunately.

If I must speculate, I'd say it drops every third block of two lines and two rows, leaving 44% of pixels.

Maybe, it is irrelevant in the field, let me know.

Looks valid to me. Thanks for posting!

Of important note (maybe an addition to the conclusion) would be related to dynamic range or really tonality: dropping ~1 stop worth of pixels will also drop the latitude / tonality by about 1 stop.

So I can't imagine much practical improvement of this (12-bit then reduced a stop) linear raw output over the 10-bit logarithmic Nlog, which as I've previously demonstrated will concentrate space where tonality is required. I actually cannot think off the top of my head any scenario where I'd prefer this raw implementation to nlog--or even to an 8-bit log picture profile (though the internal would obviously have significantly more compression). This appears to be a marketing gimmick.

So 6K raw, and it's something worth doing. Reading only some of the pixels as raw and entirely skipping others doesn't quite do most people would expect raw to do.

 

[Tavarino]

Really interesting coverage of this topic - much over my head, but I get the gist. Thank you! As you say:

"I haven't yet have the chance to evaluate all this in practice. It may turn out to be totally irrelevant in practice and ProRes RAW is the way to go."

In practice, my interest is in image detail data for compositing - using plug-ins like Mocha Pro to track objects or cameras. How many keyframes do I need to track a dynamic object on a dynamic background? I'm already seeing major improvements in my workflow with 10-bit recording to the Ninja V vs. internal 8-bit. I'm assuming that 12-bit RAW will be even better - significantly better? Marginally better? A simple metric for me would to record the same action with the three formats and track the same object in each one in Mocha Pro. Once Adobe supports ProRes RAW, I can share my results, if anyone is interested.

 

[beatboxa]

Really interesting coverage of this topic - much over my head, but I get the gist. Thank you! As you say:

"I haven't yet have the chance to evaluate all this in practice. It may turn out to be totally irrelevant in practice and ProRes RAW is the way to go."

In practice, my interest is in image detail data for compositing - using plug-ins like Mocha Pro to track objects or cameras. How many keyframes do I need to track a dynamic object on a dynamic background? I'm already seeing major improvements in my workflow with 10-bit recording to the Ninja V vs. internal 8-bit. I'm assuming that 12-bit RAW will be even better - significantly better? Marginally better? A simple metric for me would to record the same action with the three formats and track the same object in each one in Mocha Pro. Once Adobe supports ProRes RAW, I can share my results, if anyone is interested.

Please do.

Based on this post (which seems valid to me), my guess is the 10-bit log would do the best, then raw, then internal.

 

[Stephen Brenner]

Based on this post (which seems valid to me), my guess is the 10-bit log would do the best, then raw, then internal.

Other than losing a stop of light, would the full pixel readout of the DX 4k output provide more detail?

Excuse my ignorance, but if they are line skipping, is it still raw data? Why couldn’t they do down sampling 6k to 4k instead? Not enough processing power?

 

[falconeyes]

I actually cannot think off the top of my head any scenario where I'd prefer this raw implementation to nlog

Well, one such scenario may be producing HDR (UHD TV) output (Rec2020 rather than Rec709).

Also, you‘re right to loose dynamic range when skipping pixels. But the Z6 has around 14 stops to start with, so there should be dynamic range advantages left (the 12 Bit readout is NOT limiting dynamic range to 12 stops). Marc Jungbluth did some dynamic range tests in his Youtube channel and got some impressive results for highlight recovery (which of course is only half the story).

I‘d say, non practioneers of Z6 prores raw, please don‘t jump to conclusions just yet

 

[falconeyes]

Because of pixel skipping, in 4k prores raw, FX and DX modes would use the same amount of light despite the crop.

With a good lens, DX and FX modes would have about the same amount of detail, with DX possibly being a bit cleaner, i.e. with less aliasing artifacts (of course, only when using equivalent focal lengths).

4k prores raw can‘t be downsampled from 6k because it is still bayered (unmosaiced) data. To downsample, you would have to first demosaice and you can‘t mosaice back. Like you can‘t create a NEF from a 16 Bit Tiff.

The way to go is to output 6k raw and let Apple (the codec) downsample to, e.g., 4k. Which is the last paragraph of my article.

--
Falk Lumo

 

[Discombobulated]

is it as bad as Z7 video?

 

[Tavarino]

Really interesting coverage of this topic - much over my head, but I get the gist. Thank you! As you say:

"I haven't yet have the chance to evaluate all this in practice. It may turn out to be totally irrelevant in practice and ProRes RAW is the way to go."

In practice, my interest is in image detail data for compositing - using plug-ins like Mocha Pro to track objects or cameras. How many keyframes do I need to track a dynamic object on a dynamic background? I'm already seeing major improvements in my workflow with 10-bit recording to the Ninja V vs. internal 8-bit. I'm assuming that 12-bit RAW will be even better - significantly better? Marginally better? A simple metric for me would to record the same action with the three formats and track the same object in each one in Mocha Pro. Once Adobe supports ProRes RAW, I can share my results, if anyone is interested.

Please do.

Based on this post (which seems valid to me), my guess is the 10-bit log would do the best, then raw, then internal.

I have no idea what the raw footage is going to look like. There are all kinds of luma and chroma adjustments I can make in post to help the tracker. Perhaps optimizing those adjustments for different situations will be the key for me to get the most out of Z6 raw recording. Lots to think about.

 

[beatboxa]

I actually cannot think off the top of my head any scenario where I'd prefer this raw implementation to nlog

Well, one such scenario may be producing HDR (UHD TV) output (Rec2020 rather than Rec709).

Also, you‘re right to loose dynamic range when skipping pixels. But the Z6 has around 14 stops to start with, so there should be dynamic range advantages left (the 12 Bit readout is NOT limiting dynamic range to 12 stops). Marc Jungbluth did some dynamic range tests in his Youtube channel and got some impressive results for highlight recovery (which of course is only half the story).

I‘d say, non practioneers of Z6 prores raw, please don‘t jump to conclusions just yet

Sure--and I didn't want to imply that 12-bit encoding limits dynamic range, particularly when I did a post on this very subject myself .

But remember that 12-bit raw should be linear or close to linear (since it's raw)--and being linear, it is more directly related to dynamic range than a logarithmic profile is, where the dynamic range can vary drastically from the bit depth. I haven't found Marc Jungbluth's video dynamic range test but I assume it was for a log profile?

In other words, is a 12-bit linear raw better or worse than a 10-bit log encoded video? Log itself being a form of compressing a larger dynamic range into a lower bit depth (that's its entire purpose). And in response to this question, I would say that dynamic range shouldn't be a significant difference between the two in practice, particularly in the way you've posited (I think correctly) that Nikon has implemented its raw here.

What is different in raw, however, is (or should be) access to the pre-demosaiced information. So if there's something you need this for, then I could see this method of raw being differentiated from NLog. But not for dynamic range.

 

[falconeyes]

But remember that 12-bit raw should be linear or close to linear (since it's raw)--and being linear, it is more directly related to dynamic range than a logarithmic profile is, where the dynamic range can vary drastically from the bit depth. I haven't found Marc Jungbluth's video dynamic range test but I assume it was for a log profile?

Hi beatboxa

Here are N-Logs encoded 12-Bit values (from a quick calculation while the eggs cooked in the oven ):

0,0 0,1 0,1 0,2 0,2 0,3 0,3 0,4 0,5 0,5 0,6 0,6 0,7 0,8 0,8 0,9 0,9 1,0 1,1 1,1 1,2 1,3 1,3 1,4 1,5 1,5 1,6 1,7 1,8 1,8 1,9 2,0 2,1 2,1 2,2 2,3 2,4 2,5 2,5 2,6 2,7 2,8 2,9 3,0 3,1 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 4,0 4,1 4,2 4,3 4,4 4,5 4,6 4,7 4,8 5,0 5,1 5,2 5,3 5,4 5,5 5,6 5,8 5,9 6,0 6,1 6,2 6,4 6,5 6,6 6,7 6,9 7,0 7,1 7,3 7,4 7,5 7,7 7,8 7,9 8,1 8,2 8,4 8,5 8,7 8,8 9,0 9,1 9,3 9,4 9,6 9,7 9,9 10,0 10,2 10,4 10,5 10,7 10,8 11,0 11,2 11,4 11,5 11,7 11,9 12,0 12,2 12,4 12,6 12,8 12,9 13,1 13,3 13,5 13,7 13,9 14,1 14,3 14,5 14,6 14,8 15,0 15,2 15,4 15,7 15,9 16,1 16,3 16,5 16,7 16,9 17,1 17,3 17,6 17,8 18,0 18,2 18,4 18,7 18,9 19,1 19,4 19,6 19,8 20,1 20,3 20,5 20,8 21,0 21,3 21,5 21,8 22,0 22,3 22,5 22,8 23,0 23,3 23,6 23,8 24,1 24,4 24,6 24,9 25,2 25,4 25,7 26,0 26,3 26,5 26,8 27,1 27,4 27,7 28,0 28,3 28,6 28,9 29,2 29,5 29,8 30,1 30,4 30,7 31,0 31,3 31,6 31,9 32,2 32,6 32,9 33,2 33,5 33,8 34,2 34,5 34,8 35,2 35,5 35,8 36,2 36,5 36,9 37,2 37,6 37,9 38,3 38,6 39,0 39,3 39,7 40,1 40,4 40,8 41,2 41,5 41,9 42,3 42,7 43,1 43,4 43,8 44,2 44,6 45,0 45,4 45,8 46,2 46,6 47,0 47,4 47,8 48,2 48,6 49,0 49,4 49,9 50,3 50,7 51,1 51,5 52,0 52,4 52,8 53,3 53,7 54,2 54,6 55,0 55,5 55,9 56,4 56,8 57,3 57,8 58,2 58,7 59,2 59,6 60,1 60,6 61,0 61,5 62,0 62,5 63,0 63,5 64,0 64,4 64,9 65,4 65,9 66,4 66,9 67,5 68,0 68,5 69,0 69,5 70,0 70,5 71,1 71,6 72,1 72,7 73,2 73,7 74,3 74,8 75,4 75,9 76,5 77,0 77,6 78,1 78,7 79,3 79,8 80,4 81,0 81,5 82,1 82,7 83,3 83,9 84,4 85,0 85,6 86,2 86,8 87,4 88,0 88,6 89,2 89,9 90,5 91,0 91,6 92,2 92,8 93,5 94,1 94,7 95,3 96,0 96,6 97,3 97,9 98,6 99,2 100 101 101 102 103 103 104 105 105 106 107 108 108 109 110 110 111 112 113 113 114 115 116 116 117 118 119 120 120 121 122 123 124 124 125 126 127 128 129 130 130 131 132 133 134 135 136 137 138 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 163 164 165 166 167 168 169 170 171 173 174 175 176 177 178 180 181 182 183 184 186 187 188 189 191 192 193 195 196 197 199 200 201 203 204 205 207 208 209 211 212 214 215 216 218 219 221 222 224 225 227 228 230 231 233 235 236 238 239 241 242 244 246 247 249 251 252 254 256 257 259 261 263 264 266 268 270 272 273 275 277 279 281 283 285 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 317 319 321 323 325 327 329 332 334 336 338 341 343 345 348 350 352 355 357 359 362 364 367 369 371 374 376 379 382 384 387 389 392 394 397 400 402 405 408 411 413 416 419 422 424 427 430 433 436 439 442 445 448 451 454 457 460 463 466 469 472 475 479 482 485 488 492 495 498 501 505 508 512 515 518 522 525 529 532 536 540 543 547 551 554 558 562 565 569 573 577 581 585 588 592 596 600 604 608 612 617 621 625 629 633 637 642 646 650 655 659 663 668 672 677 681 686 691 695 700 705 709 714 719 724 728 733 738 743 748 753 758 763 768 773 779 784 789 794 800 805 810 816 821 827 832 838 843 849 855 861 866 872 878 884 890 896 902 908 914 920 926 932 938 945 951 957 964 970 977 983 990 996 1003 1010 1017 1023 1030 1037 1044 1051 1058 1065 1072 1079 1087 1094 1101 1109 1116 1123 1131 1139 1146 1154 1162 1169 1177 1185 1193 1201 1209 1217 1225 1233 1242 1250 1258 1267 1275 1284 1292 1301 1310 1318 1327 1336 1345 1354 1363 1372 1381 1391 1400 1409 1419 1428 1438 1447 1457 1467 1477 1487 1496 1506 1517 1527 1537 1547 1558 1568 1578 1589 1600 1610 1621 1632 1643 1654 1665 1676 1687 1699 1710 1721 1733 1744 1756 1768 1780 1792 1804 1816 1828 1840 1852 1865 1877 1890 1902 1915 1928 1941 1954 1967 1980 1993 2007 2020 2034 2047 2061 2075 2088 2102 2117 2131 2145 2159 2174 2188 2203 2218 2232 2247 2262 2278 2293 2308 2324 2339 2355 2370 2386 2402 2418 2435 2451 2467 2484 2500 2517 2534 2551 2568 2585 2602 2620 2637 2655 2673 2691 2709 2727 2745 2763 2782 2800 2819 2838 2857 2876 2895 2915 2934 2954 2974 2993 3013 3034 3054 3074 3095 3116 3136 3157 3179 3200 3221 3243 3264 3286 3308 3330 3353 3375 3398 3420 3443 3466 3490 3513 3536 3560 3584 3608 3632 3656 3681 3705 3730 3755 3780 3805 3831 3857 3882 3908 3934 3961 3987 4014 4041 4068 4095

As you can see and as you say indeed, there is plenty of information to encode shadow, and probably enough in the highlights, considering shot noise. Using multiple encoding values at the low end makes sense, because there are rounding errors due to color space conversion.

So, possible benefits of raw may be from implementation details, like bitrate, the change of field of view in N-Log, etc. Or there may be none.

Marc's video is here:

he seems to have more than one channel ...

 

[beatboxa]

But remember that 12-bit raw should be linear or close to linear (since it's raw)--and being linear, it is more directly related to dynamic range than a logarithmic profile is, where the dynamic range can vary drastically from the bit depth. I haven't found Marc Jungbluth's video dynamic range test but I assume it was for a log profile?

Hi beatboxa

Here are N-Logs encoded 12-Bit values (from a quick calculation while the eggs cooked in the oven ):

0,0 0,1 0,1 0,2 0,2 0,3 0,3 0,4 0,5 0,5 0,6 0,6 0,7 0,8 0,8 0,9 0,9 1,0 1,1 1,1 1,2 1,3 1,3 1,4 1,5 1,5 1,6 1,7 1,8 1,8 1,9 2,0 2,1 2,1 2,2 2,3 2,4 2,5 2,5 2,6 2,7 2,8 2,9 3,0 3,1 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 4,0 4,1 4,2 4,3 4,4 4,5 4,6 4,7 4,8 5,0 5,1 5,2 5,3 5,4 5,5 5,6 5,8 5,9 6,0 6,1 6,2 6,4 6,5 6,6 6,7 6,9 7,0 7,1 7,3 7,4 7,5 7,7 7,8 7,9 8,1 8,2 8,4 8,5 8,7 8,8 9,0 9,1 9,3 9,4 9,6 9,7 9,9 10,0 10,2 10,4 10,5 10,7 10,8 11,0 11,2 11,4 11,5 11,7 11,9 12,0 12,2 12,4 12,6 12,8 12,9 13,1 13,3 13,5 13,7 13,9 14,1 14,3 14,5 14,6 14,8 15,0 15,2 15,4 15,7 15,9 16,1 16,3 16,5 16,7 16,9 17,1 17,3 17,6 17,8 18,0 18,2 18,4 18,7 18,9 19,1 19,4 19,6 19,8 20,1 20,3 20,5 20,8 21,0 21,3 21,5 21,8 22,0 22,3 22,5 22,8 23,0 23,3 23,6 23,8 24,1 24,4 24,6 24,9 25,2 25,4 25,7 26,0 26,3 26,5 26,8 27,1 27,4 27,7 28,0 28,3 28,6 28,9 29,2 29,5 29,8 30,1 30,4 30,7 31,0 31,3 31,6 31,9 32,2 32,6 32,9 33,2 33,5 33,8 34,2 34,5 34,8 35,2 35,5 35,8 36,2 36,5 36,9 37,2 37,6 37,9 38,3 38,6 39,0 39,3 39,7 40,1 40,4 40,8 41,2 41,5 41,9 42,3 42,7 43,1 43,4 43,8 44,2 44,6 45,0 45,4 45,8 46,2 46,6 47,0 47,4 47,8 48,2 48,6 49,0 49,4 49,9 50,3 50,7 51,1 51,5 52,0 52,4 52,8 53,3 53,7 54,2 54,6 55,0 55,5 55,9 56,4 56,8 57,3 57,8 58,2 58,7 59,2 59,6 60,1 60,6 61,0 61,5 62,0 62,5 63,0 63,5 64,0 64,4 64,9 65,4 65,9 66,4 66,9 67,5 68,0 68,5 69,0 69,5 70,0 70,5 71,1 71,6 72,1 72,7 73,2 73,7 74,3 74,8 75,4 75,9 76,5 77,0 77,6 78,1 78,7 79,3 79,8 80,4 81,0 81,5 82,1 82,7 83,3 83,9 84,4 85,0 85,6 86,2 86,8 87,4 88,0 88,6 89,2 89,9 90,5 91,0 91,6 92,2 92,8 93,5 94,1 94,7 95,3 96,0 96,6 97,3 97,9 98,6 99,2 100 101 101 102 103 103 104 105 105 106 107 108 108 109 110 110 111 112 113 113 114 115 116 116 117 118 119 120 120 121 122 123 124 124 125 126 127 128 129 130 130 131 132 133 134 135 136 137 138 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 163 164 165 166 167 168 169 170 171 173 174 175 176 177 178 180 181 182 183 184 186 187 188 189 191 192 193 195 196 197 199 200 201 203 204 205 207 208 209 211 212 214 215 216 218 219 221 222 224 225 227 228 230 231 233 235 236 238 239 241 242 244 246 247 249 251 252 254 256 257 259 261 263 264 266 268 270 272 273 275 277 279 281 283 285 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 317 319 321 323 325 327 329 332 334 336 338 341 343 345 348 350 352 355 357 359 362 364 367 369 371 374 376 379 382 384 387 389 392 394 397 400 402 405 408 411 413 416 419 422 424 427 430 433 436 439 442 445 448 451 454 457 460 463 466 469 472 475 479 482 485 488 492 495 498 501 505 508 512 515 518 522 525 529 532 536 540 543 547 551 554 558 562 565 569 573 577 581 585 588 592 596 600 604 608 612 617 621 625 629 633 637 642 646 650 655 659 663 668 672 677 681 686 691 695 700 705 709 714 719 724 728 733 738 743 748 753 758 763 768 773 779 784 789 794 800 805 810 816 821 827 832 838 843 849 855 861 866 872 878 884 890 896 902 908 914 920 926 932 938 945 951 957 964 970 977 983 990 996 1003 1010 1017 1023 1030 1037 1044 1051 1058 1065 1072 1079 1087 1094 1101 1109 1116 1123 1131 1139 1146 1154 1162 1169 1177 1185 1193 1201 1209 1217 1225 1233 1242 1250 1258 1267 1275 1284 1292 1301 1310 1318 1327 1336 1345 1354 1363 1372 1381 1391 1400 1409 1419 1428 1438 1447 1457 1467 1477 1487 1496 1506 1517 1527 1537 1547 1558 1568 1578 1589 1600 1610 1621 1632 1643 1654 1665 1676 1687 1699 1710 1721 1733 1744 1756 1768 1780 1792 1804 1816 1828 1840 1852 1865 1877 1890 1902 1915 1928 1941 1954 1967 1980 1993 2007 2020 2034 2047 2061 2075 2088 2102 2117 2131 2145 2159 2174 2188 2203 2218 2232 2247 2262 2278 2293 2308 2324 2339 2355 2370 2386 2402 2418 2435 2451 2467 2484 2500 2517 2534 2551 2568 2585 2602 2620 2637 2655 2673 2691 2709 2727 2745 2763 2782 2800 2819 2838 2857 2876 2895 2915 2934 2954 2974 2993 3013 3034 3054 3074 3095 3116 3136 3157 3179 3200 3221 3243 3264 3286 3308 3330 3353 3375 3398 3420 3443 3466 3490 3513 3536 3560 3584 3608 3632 3656 3681 3705 3730 3755 3780 3805 3831 3857 3882 3908 3934 3961 3987 4014 4041 4068 4095

As you can see and as you say indeed, there is plenty of information to encode shadow, and probably enough in the highlights, considering shot noise. Using multiple encoding values at the low end makes sense, because there are rounding errors due to color space conversion.

So, possible benefits of raw may be from implementation details, like bitrate, the change of field of view in N-Log, etc. Or there may be none.

Marc's video is here:

he seems to have more than one channel ...

Thanks for including this. This looks right. It certainly shows the increased tonal precision in the shadows, where it looks like there will be almost no difference between the raw & N-log in the deep shadows. (And yes, considering shot noise, you probably don't want to push these anyway).

Somewhere around 150 is the point where the N-Log curve is closest to linear (where this would essentially have the precision of a linear 10-bit. Below this, it has somewhere between what a 10-bit & 12-bit linear would have).

And then above this 150 value, it slowly starts losing some precision relative to what a linear raw would capture. For the highlights, it's probably enough to avoid most posterization in almost all cases.

In the brightest highlights--somewhere around 2402--it equates to the precision that a linear 8-bit would have. This should be less than the last stop of highlights, which you'd probably want to avoid anyway. There is rarely much useful information here. In practice, this is often the exact point of trying not to blow the brightest parts of the sky. Your main concern is exposing everything else properly, and you don't really care about the exact tonal precision of the sky. Only that the brightest bits of the sky are not blown or posterized--and a linear 8-bit (or almost 7-bit) is plenty.

BTW, this got me to finally look into Nikon's N-Log specification document--and it notes that:

• The gamut for N-Log is same as the wide color gamut known as “ITU-R BT.2020”.

So don't get me wrong: dynamic range or tonality might be technically slightly better in raw, but I think it would be so slight that there are usually no real differences in practice.

I think we agree that the main possible benefits of raw would be the implementation details like bitrate, field of view, perhaps lens corrections (I would think these are in N-log but not in raw), etc. which will have more of a difference in practice than tonality.

 

[Horshack]

But remember that 12-bit raw should be linear or close to linear (since it's raw)--and being linear, it is more directly related to dynamic range than a logarithmic profile is, where the dynamic range can vary drastically from the bit depth. I haven't found Marc Jungbluth's video dynamic range test but I assume it was for a log profile?

Hi beatboxa

Here are N-Logs encoded 12-Bit values (from a quick calculation while the eggs cooked in the oven ):

0,0 0,1 0,1 0,2 0,2 0,3 0,3 0,4 0,5 0,5 0,6 0,6 0,7 0,8 0,8 0,9 0,9 1,0 1,1 1,1 1,2 1,3 1,3 1,4 1,5 1,5 1,6 1,7 1,8 1,8 1,9 2,0 2,1 2,1 2,2 2,3 2,4 2,5 2,5 2,6 2,7 2,8 2,9 3,0 3,1 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 4,0 4,1 4,2 4,3 4,4 4,5 4,6 4,7 4,8 5,0 5,1 5,2 5,3 5,4 5,5 5,6 5,8 5,9 6,0 6,1 6,2 6,4 6,5 6,6 6,7 6,9 7,0 7,1 7,3 7,4 7,5 7,7 7,8 7,9 8,1 8,2 8,4 8,5 8,7 8,8 9,0 9,1 9,3 9,4 9,6 9,7 9,9 10,0 10,2 10,4 10,5 10,7 10,8 11,0 11,2 11,4 11,5 11,7 11,9 12,0 12,2 12,4 12,6 12,8 12,9 13,1 13,3 13,5 13,7 13,9 14,1 14,3 14,5 14,6 14,8 15,0 15,2 15,4 15,7 15,9 16,1 16,3 16,5 16,7 16,9 17,1 17,3 17,6 17,8 18,0 18,2 18,4 18,7 18,9 19,1 19,4 19,6 19,8 20,1 20,3 20,5 20,8 21,0 21,3 21,5 21,8 22,0 22,3 22,5 22,8 23,0 23,3 23,6 23,8 24,1 24,4 24,6 24,9 25,2 25,4 25,7 26,0 26,3 26,5 26,8 27,1 27,4 27,7 28,0 28,3 28,6 28,9 29,2 29,5 29,8 30,1 30,4 30,7 31,0 31,3 31,6 31,9 32,2 32,6 32,9 33,2 33,5 33,8 34,2 34,5 34,8 35,2 35,5 35,8 36,2 36,5 36,9 37,2 37,6 37,9 38,3 38,6 39,0 39,3 39,7 40,1 40,4 40,8 41,2 41,5 41,9 42,3 42,7 43,1 43,4 43,8 44,2 44,6 45,0 45,4 45,8 46,2 46,6 47,0 47,4 47,8 48,2 48,6 49,0 49,4 49,9 50,3 50,7 51,1 51,5 52,0 52,4 52,8 53,3 53,7 54,2 54,6 55,0 55,5 55,9 56,4 56,8 57,3 57,8 58,2 58,7 59,2 59,6 60,1 60,6 61,0 61,5 62,0 62,5 63,0 63,5 64,0 64,4 64,9 65,4 65,9 66,4 66,9 67,5 68,0 68,5 69,0 69,5 70,0 70,5 71,1 71,6 72,1 72,7 73,2 73,7 74,3 74,8 75,4 75,9 76,5 77,0 77,6 78,1 78,7 79,3 79,8 80,4 81,0 81,5 82,1 82,7 83,3 83,9 84,4 85,0 85,6 86,2 86,8 87,4 88,0 88,6 89,2 89,9 90,5 91,0 91,6 92,2 92,8 93,5 94,1 94,7 95,3 96,0 96,6 97,3 97,9 98,6 99,2 100 101 101 102 103 103 104 105 105 106 107 108 108 109 110 110 111 112 113 113 114 115 116 116 117 118 119 120 120 121 122 123 124 124 125 126 127 128 129 130 130 131 132 133 134 135 136 137 138 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 163 164 165 166 167 168 169 170 171 173 174 175 176 177 178 180 181 182 183 184 186 187 188 189 191 192 193 195 196 197 199 200 201 203 204 205 207 208 209 211 212 214 215 216 218 219 221 222 224 225 227 228 230 231 233 235 236 238 239 241 242 244 246 247 249 251 252 254 256 257 259 261 263 264 266 268 270 272 273 275 277 279 281 283 285 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 317 319 321 323 325 327 329 332 334 336 338 341 343 345 348 350 352 355 357 359 362 364 367 369 371 374 376 379 382 384 387 389 392 394 397 400 402 405 408 411 413 416 419 422 424 427 430 433 436 439 442 445 448 451 454 457 460 463 466 469 472 475 479 482 485 488 492 495 498 501 505 508 512 515 518 522 525 529 532 536 540 543 547 551 554 558 562 565 569 573 577 581 585 588 592 596 600 604 608 612 617 621 625 629 633 637 642 646 650 655 659 663 668 672 677 681 686 691 695 700 705 709 714 719 724 728 733 738 743 748 753 758 763 768 773 779 784 789 794 800 805 810 816 821 827 832 838 843 849 855 861 866 872 878 884 890 896 902 908 914 920 926 932 938 945 951 957 964 970 977 983 990 996 1003 1010 1017 1023 1030 1037 1044 1051 1058 1065 1072 1079 1087 1094 1101 1109 1116 1123 1131 1139 1146 1154 1162 1169 1177 1185 1193 1201 1209 1217 1225 1233 1242 1250 1258 1267 1275 1284 1292 1301 1310 1318 1327 1336 1345 1354 1363 1372 1381 1391 1400 1409 1419 1428 1438 1447 1457 1467 1477 1487 1496 1506 1517 1527 1537 1547 1558 1568 1578 1589 1600 1610 1621 1632 1643 1654 1665 1676 1687 1699 1710 1721 1733 1744 1756 1768 1780 1792 1804 1816 1828 1840 1852 1865 1877 1890 1902 1915 1928 1941 1954 1967 1980 1993 2007 2020 2034 2047 2061 2075 2088 2102 2117 2131 2145 2159 2174 2188 2203 2218 2232 2247 2262 2278 2293 2308 2324 2339 2355 2370 2386 2402 2418 2435 2451 2467 2484 2500 2517 2534 2551 2568 2585 2602 2620 2637 2655 2673 2691 2709 2727 2745 2763 2782 2800 2819 2838 2857 2876 2895 2915 2934 2954 2974 2993 3013 3034 3054 3074 3095 3116 3136 3157 3179 3200 3221 3243 3264 3286 3308 3330 3353 3375 3398 3420 3443 3466 3490 3513 3536 3560 3584 3608 3632 3656 3681 3705 3730 3755 3780 3805 3831 3857 3882 3908 3934 3961 3987 4014 4041 4068 4095

As you can see and as you say indeed, there is plenty of information to encode shadow, and probably enough in the highlights, considering shot noise. Using multiple encoding values at the low end makes sense, because there are rounding errors due to color space conversion.

So, possible benefits of raw may be from implementation details, like bitrate, the change of field of view in N-Log, etc. Or there may be none.

Marc's video is here:

he seems to have more than one channel ...

There are 896 values in that table (2^9.81 values), encoding to a range of 2^12 values.. What is it encoding from?

 

[falconeyes]

There are 896 values in that table (2^9.81 values), encoding to a range of 2^12 values.. What is it encoding from?

What eagle-eyed readers we have

Nikon specifies an N-Log function which decodes to negative "reflectance" values for the first ~100 parameter values between 0 and 1023. I dropped them as not meaningful. Ask Nikon about what they were thinking ...

--
Falk Lumo