5D Mark III vs D800 Dynamic Range? Is the D800 Superior in this area?

definitely not - I'm not seeing all that noise, except with Flaat_11 and Flaat_12, which have much more DR than CineStyle (hence the noise in the heavily pushed-up shadows)
from my tests, Flaat_10 (the evolution of Flaat_2) leads to less noise in the final images than CineStyle (and both have the same DR: 10 stops)

no idea why it wouldn't work for skital; maybe it was the settings used: some users have reported that after installing Flaat in the camera, weird things happened to the picture style settings (e.g. sharpness went to 7); it has never hapened to me, so I don't know the cause
contrast has to be =-4; for the others, it's only a recommendation: Saturation=-2, Tone=1
 
it looks like Nikon is going to win the DR war in the DSLR arena...

Canon DSLRs have a hard limit in video mode, so they won't deliver more than 11.5 stops of DR unless ML can move that hard limit (the other alternative wold be Canon giving Technicolor the inside info to move that hard limit; what's clear is that, with the PSE and without ML, I can't get beyond 11.5 stops)

on the other hand... I'm looking for D800 and D7000 testers for a beta version of a picture control that should get slightly over 13 stops of DR on the D800 (no idea yet on whether those stops are useable, or just full of noise): http://www.dvxuser.com/V6/showthrea...te-of-picture-controls-for-Nikon-DSLRs-beta-1
 
Samuel H said:
definitely not - I'm not seeing all that noise, except with Flaat_11 and Flaat_12, which have much more DR than CineStyle (hence the noise in the heavily pushed-up shadows)
from my tests, Flaat_10 (the evolution of Flaat_2) leads to less noise in the final images than CineStyle (and both have the same DR: 10 stops)

no idea why it wouldn't work for skital; maybe it was the settings used: some users have reported that after installing Flaat in the camera, weird things happened to the picture style settings (e.g. sharpness went to 7); it has never hapened to me, so I don't know the cause
contrast has to be =-4; for the others, it's only a recommendation: Saturation=-2, Tone=1
Click to expand...

I was using Flaat_12. Using zebras and histogram the Flaat_12 clipped the highlights faster and when avoiding the clipped highlights seemed to clip the blacks before Cinestyle as well (and I calibrated Zebras for Cinestyle when in use....put black clipping at 9%/23. It starts off at 5%, but since Cinestyle blacks start at 16 I put it higher as ML stated to do).

I will try to do some tests. I just looked at the LCD. I'll try to upload some footage and see how they look. It just looked like whatever dynamic range improvement the Flaat_12 had was not useable dynamic range, but I'll test it out more.
 
not surprising at all, but it's just an effect of your zebra settings

this is CineStyle:
technicolor_4.jpg
and this is Flaat_4:
flaat_4.jpg

the white clipping point is exactly the same, but zebras won't tell you that unless you set them at 99%
your result appears because CineStyle reserves 22% of its color space for the brightest stop of light (those wedges are a third of a stop apart)
I consider that a bad thing

and if you look at the black clipping point, CineStyle has the 4 bars on the right clipped at its minimum brightness level, and another 3 pretty much indistinguishable from that, whereas Flaat_12 always shows some slope; yes, it will display lots of noise on those heavily pushed-up shadows (just look at how thick the horizontal lines get), but at least they're not clipped; whether that's useable DR or not, is subjective; I totally understand that some people are not interested in working with that noisy data at all


edit:
in any case, notice that the price of having this extra headroom in the shadows is basically zero: the slope of both light-response curves is, for the most part, exactly the same; I don't like CineStyle because it delivers somewhat noisy final images; this is because, for the most part, it uses only 28 brightness values per stop of light; that's exactly the same with Flaat_12, but at least it gives you something: a clipping point in the shadows that's over a stop further down
if you stick to the Flaat picture style that has the same DR as CineStyle (that would be Flaat_10p), you get, for the most part, 37 brightness values per stop of light (i.e., the slope is much higher)
flaat_2.jpg
 
Last edited:
Samuel H said:
not surprising at all, but it's just an effect of your zebra settings

the white clipping point is exactly the same, but zebras won't tell you that unless you set them at 99%
your result appears because CineStyle reserves 22% of its color space for the brightest stop of light (those wedges are a third of a stop apart)
I consider that a bad thing
Click to expand...
Ah. I had them set at 95%.

and if you look at the black clipping point, CineStyle has the 4 bars on the right clipped at its minimum brightness level, and another 3 pretty much indistinguishable from that, whereas Flaat_12 always shows some slope; yes, it will display lots of noise on those heavily pushed-up shadows (just look at how thick the horizontal lines get), but at least they're not clipped; whether that's useable DR or not, is subjective; I totally understand that some people are not interested in working with that noisy data at all
Click to expand...
Yeah, a lot of noise for sure. I'd have to run it through Neat's noise reducer. Maybe then it may work.
 
A1ex convinced me to run some tests on a ML feature that may (or may not) lead to cleaner images with something like Flaat_12
that's what I'm working on today; let's see how this goes...
if it works, it could make wonders on the video DR of an eventual 5D3+ML
 
Samuel H said:
A1ex convinced me to run some tests on a ML feature that may (or may not) lead to cleaner images with something like Flaat_12
that's what I'm working on today; let's see how this goes...
if it works, it could make wonders on the video DR of an eventual 5D3+ML
Click to expand...

Interesting. One very noticeable problem I have discovered with Flaat_12 vs Cinestyle or Marvel's Cinema, etc is extreme highlight banding. I just shot some footage in my bathroom of the light bulb above the mirror and the Flaat_12 utterly fell apart. The banding on the slightly off white wall just behind the light bulb had virtually no highlight roll off until the exposure was high enough that the light bulb was completely blooming past the glass fixture. It was just a large splotch and as I lowered exposure for the bulb to become less bloomy the splotchy banding around the light on the wall area got worse and worse. I'll have to post some pictures of what I was seeing. Cinestyle had a nice roll off with none of that nasty banding. I could expose the bulb all the way to the point where I could see its letters while on and the reflected light just behind the bulb on the wall was nice and smooth. I could adjust exposure just the the point of blooming or below and the wall highlight still looked nice while the bulb looked nice as well. Got the complete opposite with the Flaat_12....it was unusable until the bulb's light was exploding past the glass fixture. The highlight roll off needs fixing majorly.
 
well, Flaat_12 is always going to have banding everywhere, it's storing 11.5 stops of DR on only 256 values for brightness...

the proper comparison would have been with Flaat_10, which is the one that has the same DR as CineStyle

but I'll take it that even Flaat_10 will have more issues with this than, say, CineStyle, because Technicolor decided to reserve 22% of the codec's color space just for the highlights, and Flaat spreads it out evenly - the downside for CineStyle is that there's a much smaller color space left for the midtones, which makes them noisy as soon as you get your contrast reshuffled in post; I'd rather have clean skin tones in my final images than absolutely perfect highlights

this short film was shot with Flaat_10 (actually, its older brother Flaat_2), and it's pushing the highlights pretty hard without showing many issues: http://www.similaar.com/foto/flaat-picture-styles/real-world-tests-3.html
CineStyle might have done even better in the highlights, but would have suffered a lot everywhere else

Flaat, specially the Portrait-based "p" option, is designed for skin tones first, controlled DR second
 
Samuel H said:
well, Flaat_12 is always going to have banding everywhere, it's storing 11.5 stops of DR on only 256 values for brightness...

the proper comparison would have been with Flaat_10, which is the one that has the same DR as CineStyle

but I'll take it that even Flaat_10 will have more issues with this than, say, CineStyle, because Technicolor decided to reserve 22% of the codec's color space just for the highlights, and Flaat spreads it out evenly - the downside for CineStyle is that there's a much smaller color space left for the midtones, which makes them noisy as soon as you get your contrast reshuffled in post; I'd rather have clean skin tones in my final images than absolutely perfect highlights

this short film was shot with Flaat_10 (actually, its older brother Flaat_2), and it's pushing the highlights pretty hard without showing many issues: http://www.similaar.com/foto/flaat-picture-styles/real-world-tests-3.html
CineStyle might have done even better in the highlights, but would have suffered a lot everywhere else

Flaat, specially the Portrait-based "p" option, is designed for skin tones first, controlled DR second
Click to expand...

I will give Flaat_2/10 a try. Which version is best? Based on Neutral or Portrait? BTW, another thing I noticed about 12 was the changing of the color red. I wore a Bull's Jersey which is very blood red and it turns to very orange with flaat_12 and even have adding the post S gamma curve the red is just destroyed compared to pre or post grading with Cinestyle or Marvel's Cine. Does Flaat_10 mess up the reds? It seems really difficult to grade and this is with photos btw.
 
* based on portrait: much better skin tones, but some color shifts (e.g. blue goes visibly towards cyan, and also the red one you noticed)
* based on neutral: similar color to what you're used to (Neutral, CineStyle, Marvels Cine), which means no color shifts, but also that you'll need good color grading skills to get nice skin tones in your final images (not any more difficult than with the others; but a lot more difficult than with Flaat_10p)

I ALWAYS use the ones based on portrait: to me, skin goes first
if these color shifts are going to bug you, go for the ones based on neutral; but don't do so without running a test of Flaat_10p vs Flaat_10n with fome flesh in it! I know the color shifts are annoying, but I'm happy to live with them, skin in Flaat_10p is so much nicer
 
I have picked my winner: in video mode, the Nikon D800 has more DR than the 5D3

at similar noise-in-the-highly-pushed-up-shadows issues, the Nikon seems to have about a stop higher DR than any of the Canons, including the 5D3

* if you're not afraid of running heavy noise reduction, the 5D3 can give you 11.5 stops of DR (with Flaat_12p), the D800 will give you 12.7 (with Flaat_13p)
* if you want extended DR but don't want so much noise, the 5D3 can give you 11 stops of DR (with Flaat_11p), the D800 will give you 12 (with Flaat_12p)
* if you're just going to use the workhorse option, with little noise at low ISO values, the 5D3 will give you 10 stops of DR (Flaat_10p, CineStyle, Marvels Cine), the D800 will give you 11 (Flaat_11p, TassinFlat, Milk)
* if you're only willing to play with the in-camera picture styles (e.g. Neutral), the 5D3 will give you 9 stops of DR, the Nikon will give you a bit more than 10

just compare these two sets of tests:
http://www.similaar.com/foto/flaat-picture-controls/technical-tests-1.html
http://www.similaar.com/foto/flaat-picture-styles/technical-tests-1.html


also, new thread for everything picture-control related, and first full version of my Flaat suite of picture controls for Nikon DSLRs (no longer looking for crazy DR, just nice, easy to grade pictures): http://www.dvxuser.com/V6/showthread.php?279408-Adventures-in-the-PCU
 
Samuel H said:
well, Flaat_12 is always going to have banding everywhere, it's storing 11.5 stops of DR on only 256 values for brightness...

the proper comparison would have been with Flaat_10, which is the one that has the same DR as CineStyle

but I'll take it that even Flaat_10 will have more issues with this than, say, CineStyle, because Technicolor decided to reserve 22% of the codec's color space just for the highlights, and Flaat spreads it out evenly - the downside for CineStyle is that there's a much smaller color space left for the midtones, which makes them noisy as soon as you get your contrast reshuffled in post; I'd rather have clean skin tones in my final images than absolutely perfect highlights

this short film was shot with Flaat_10 (actually, its older brother Flaat_2), and it's pushing the highlights pretty hard without showing many issues: http://www.similaar.com/foto/flaat-picture-styles/real-world-tests-3.html
CineStyle might have done even better in the highlights, but would have suffered a lot everywhere else

Flaat, specially the Portrait-based "p" option, is designed for skin tones first, controlled DR second
Click to expand...

This is f'ing beautiful, dude. Reminds me of Cronenweth's work on the Red.
Was each and every interior, exterior, scene done in Flaat_10, or did you switch it up according to the light's intensity as the shoot progressed? Very very impressive.
 
Last edited:
I didn't shoot it - I don't even know the guys who did - but my understanding is that everything is Flaat_10p (actually, its older brother, Flaat_2, but they're nearly the same) on a couple of 60D bodies, and L zooms; they'll love to hear your comments and queries in the vimeo page: http://vimeo.com/37993764



edit:
new test footage, with Flaat_10, Flaat_11, Flaat_12: http://vimeo.com/39937102
it's still mostly a DR test, but I found it very useful: I'm very glad to see Flaat_12 seems definitely useable, I wouldn't say it's any more noisy than Flaat_11 on the Canons; the D800 seems like the DR king in the DSLR world right now
 
Last edited:
Nice going with the Vimeo clip! Thanks a lot for that! Do you happen to know how dynamic range on the D800 compares to a Sony EX1? Supposedly the EX1 also has excellent dynamic range when the cinestyle gamma curve is used.


Samuel H said:
I didn't shoot it - I don't even know the guys who did - but my understanding is that everything is Flaat_10p (actually, its older brother, Flaat_2, but they're nearly the same) on a couple of 60D bodies, and L zooms; they'll love to hear your comments and queries in the vimeo page: http://vimeo.com/37993764



edit:
new test footage, with Flaat_10, Flaat_11, Flaat_12: http://vimeo.com/39937102
it's still mostly a DR test, but I found it very useful: I'm very glad to see Flaat_12 seems definitely useable, I wouldn't say it's any more noisy than Flaat_11 on the Canons; the D800 seems like the DR king in the DSLR world right now
Click to expand...
 
You must log in or register to reply here.
Back
Top