F3 Dynamic range tests?

[Harry Lime]

Has anyone conducted a dynamic range test of the F3, yet?

I am very curious to see if the F3 really does deliver the 10-12 useable stops of range that has been mentioned in various articles.


But I suppose I may be jumping the gun, because the 4:4:4 / log firmware upgrade has not yet been released.


thanks

[Duke M.]

Also, people are going to need to learn the best profile settings to get the best out of a new camera. Hyper-gammas for instance.

[rbilsbor]

In REC709 it's probably not going to be impressive; 8 stops or thereabouts. Of course, Sony is claiming that S-LOG boosts dynamic range by 800%, so....

[matthew77]

In REC709 it's probably not going to be impressive; 8 stops or thereabouts. Of course, Sony is claiming that S-LOG boosts dynamic range by 800%, so....

What a weird way for Sony to spec dynamic range differences! Sounds like they were trying to use the largest number they could find. 800% = 8x more range (presumably measured linearly) = 3 stops.

[Steve Castle]

What a weird way for Sony to spec dynamic range differences! Sounds like they were trying to use the largest number they could find. 800% = 8x more range (presumably measured linearly) = 3 stops.

?

Dynamic range is limited by its gamma profile, REC709's DR is limited due to the characteristics of that profile. You make it sound like Sony is trying to artificially inflate their DR numbers. To that regard, I would say that there is a more colorful company in this industry that is more guilty of that than any other. Quite simply, DR claims are generally useless. Non-standardized and easily manipulated.

In the case of the F3, based on how the profile is setup, you'll get different DR characteristics. When the sLog firmware is available you'll get 12 stops based on the profile. There is nothing to suggest that a 12 micron photosite couldn't achieve that.

[TimurCivan]

800% is Four stops no? 100% 200% 400% 800%

That would coincide well with 12 stops of DR since 709 gives around 8 stops.

[Harry Lime]

I guess we'll have to wait a little longer.

Keeping my fingers crossed.


thanks

[nyvz]

800% is Four stops no? 100% 200% 400% 800%

That would coincide well with 12 stops of DR since 709 gives around 8 stops.

100%=+0stops 200%=+1stop 400%=+2stops 800%=+3stops

[TimurCivan]

Ahh, thank you.

The old school video terms are a mystery to some.

[G.P.]

Doing a practical dynamic range test tomorrow with my gaffer. Practical meaning not with a chart but with gradual lighting differences on a subject and foreground/background. I'm not one for charts, I don't shoot images of charts, I shoot images of people. If people look good on the screen I'm happy :)

[Pietro Impagliazzo]

Doing a practical dynamic range test tomorrow with my gaffer. Practical meaning not with a chart but with gradual lighting differences on a subject and foreground/background. I'm not one for charts, I don't shoot images of charts, I shoot images of people. If people look good on the screen I'm happy :)

Cool, numbers really don't mean jack if it looks plain ugly.

[Fohdeesha]

Doing a practical dynamic range test tomorrow with my gaffer. Practical meaning not with a chart but with gradual lighting differences on a subject and foreground/background. I'm not one for charts, I don't shoot images of charts, I shoot images of people. If people look good on the screen I'm happy :)


use cine/hypergamma 1! black balance it as well

[TimurCivan]

Fohdeesha; You know the EX series well. which of the cine gammas allows the signal to go up to 109%? is there one? i know there are 4 and one of them is "braodcast" safe stopping at 99%.

[David Williams]

Enable the spot meter, point at a bright source and it will peak at 100% or 109%. Cine 2 is 100% on the EXs. All the rest hit 109%.

These are the notes I lifted from Adam Wilt's EX1 test years ago.

CINE1 – “deep cine”. Compression starts around 80%; midgray at 37%. Highlights stretched, blacks crushed.
CINE2 – “broadcast safe cine”. CINE1 rescaled with whites limited to 100%.
CINE3 – “brighter cine”. Compression starts around 65%; midgray at 45%. Blacks somewhat stretched, highlights slightly crushed.
CINE4 – “video-bright cine”. Compression starts around 65%; midgray at 50%. Blacks stretched, highlights crushed.

[Fohdeesha]

Well on the ex series, all cine curves 1-4 go up to 109%, except for #2, which limits highlights to about 100%. #1 has the most even distribution, the best for grading and capturing the most information in my book, especially when you're trying to cram everything in 8-bit space.

taking the gain down to -3 limits all the cine curves highlights to 100%. bringing the gamma level up about +4 will bring the highlights back up to 109%. However whether this is bringing back the clipped information, or simply raising the level of the already limited information, I'm not sure, I haven't figured out a test to find out. Negative gain is not to be trusted in my book. with the std curves, the auto knee will automatically bring the clipped highlights back up into superwhite space when dealing with negative gain. But again, I don't trust it.

Keep in mind this is all on the ex1/ex1r, I'd imagine most would be the same on the f3. Before doing any sort of conclusive testing with an f3 I highly recommend black balancing it, all the EX series Ive worked with have been pretty off. My ex1r came with absolute black hovering around 7-10IRE. that's almost a 10% reduction in bit depth! To get black back where it should be, hook it up to a waveform monitor you trust, cap off the lens and close your aperture all the way down, and lower the black setting in the picture profile until it's right at 0 (or 7.5 setup IRE, you'll be able to tell pretty easily). On the ex1r it comes out usually with black at -3 or -4. The on camera histogram is useless for this test. If you don't have access to a WF monitor, a last resort is taking it down to about -3, hitting record for a couple second with everything capped off, and bringing that footage into your editor, and having a looksey there.

And while were on the topic of random picture profile information - Believe it or not, just about nothing in the PP menus effects noise. The noise that you see is there because it's being picked up on the way to the sensors ADC's, and there's not much to do about it. I've done some pretty extensive testing, noise wise there is no discernible difference between detail on at 0, off, matrix on, hisat, and matrix off. The Curves do however have a slight impact on visible noise, with std being the nosiest, and of course cine being the quietest.

Also, this should be common knowledge, so excuse me if I'm explaining something you're well aware of - 5600k is your friend. That is the native WB of the sensor, where it isn't having to add or subtract gain for the R G or B channels. when you shoot at 3200k, there is almost no blue light for the camera to pick up, so it compensates by adding a substantial amount of gain to the blue channel. The blue channel is already the noisiest one to begin with, so it does a number or visible noise. If you just shoot at 5600k with 3200k light sources, besides looking completely wrong, you still aren't getting any advantage. It isn't having to add gain to the blue channel, which is nice, but there is no blue light coming in to stomp out the noise hanging out in the blue channels circuitry. Pump some 5600k light down there, and suddenly you have voltage to overcome that noise. All this is probably not nearly as much of a concern with the f3 however, with those mac truck sized photosites. I'd really love to get my hands on one and run it through the usual tests.


edit: per adam wilts findings posted above - crushing is a poor word choice. There is no "crushing" going on, the different curves simply record the same range of information, but assign each brightness range more or less bit depth. The dynamic range of the ex1, when tested, was the exact same across all cine curves. I believe this was discovered in the same test by adam wilt quoted above. Cine 1 assigns shadows, midtones, and highlights pretty uniformly across the 0-110IRE range.

[David Williams]

That's what I've always understood the term crushing to mean it this context. Allocating fewer bits in the range. The reverse of stretch, allocating more bits. These are old terms from analogue video days.

[Fohdeesha]

Ah yes. To me, when I hear crushing, it implies clipping information. Good ole analog voltages!

[TimurCivan]

Thank you Gentlemen.

So it seems CINE1, 0dB gain, and Detail off, is where to live till we get sLOG working.

[G.P.]

Thank you Gentlemen.

So it seems CINE1, 0dB gain, and Detail off, is where to live till we get sLOG working.

I have been tweaking and working on a cine wide gamma curve I'm calling "G-Log"... yes pun intended... It will look washed out, but retain as much information as possible... Still in the works, but testing has been done today and will continue for the next 3 days. Still working on getting the noise in the blacks down more some how.

[Fohdeesha]

If you're using black gamma in large quantities...I'd recommend against it. It doesn't recover any shadow information, it just stretches out what's already there, which is exactly what cine 3 does, and cine 4 to an even larger extent. However they do it without raising noise so much.

[G.P.]

I wasn't really touching the black gamma so much as finding the right combination of settings that gave the brightest yet cleanest image. I'm going through all the different color spaces, and all the different levels and just trying each one against each other until I find something that pleases my eye. My re-shoot of the test on my site looks much better than the first one, and I'm tweaking that one so it can be a really great low light setting.

[Shipsides]

I posted a short test here - http://blog.abelcine.com/2011/03/31/abelcine-tests-the-sony-f3-dynamic-range/

[G.P.]

Hey Andy, thanks for the test, I'm not much of a chart guy, I go for the look, but its nice to prove how many like in your test. I'm shooting in extreme light differences here and getting GREAT results, the dynamic range is great with the F3 even without S-Log, I can only wonder what it will be like with it.

[henryolonga]

Great one Andy.

[Shipsides]

Giuseppe, there is a lot to be said for a practical test outside or with lighting. Charts are nice but I'd love to see some practical examples as well. I'm glad if the video can be of some help though.

Andy

[Harry Lime]

12 useable stops in rec709. That's pretty amazing. I assume that with s-log / 4:4:4 there may be a little more to squeeze out of that sensor.

It's a like a poor man's Alexa...

[yoclay]

How much more DR might we expect typically with the S-log?

[BobbyMurcerFan]

I posted a short test here - http://blog.abelcine.com/2011/03/31/abelcine-tests-the-sony-f3-dynamic-range/

Andy,

Is Rec 709 out of the camera extended range, e.g. 4 to 1019 (assuming a ten bit scale)? I ask b/c looking at the waveform monitor, the Rec 709 range is wider than 7.5 to 100 IRE.

Also, while Cine 2 highlights end at 100 IRE, it looks like its shadows are below the lower limit of what's been generally accepted as legal broadcast, 7.5 IRE. So wouldn't the black level have to be brought up to make Cine 2 truly straight to broadcast compliant? Perhaps this is b/c Japan uses NTSC-J, which I believe does not have 7.5 IRE setup?

Thanks much.