F3: Noise in Slog?

So Ive just run a test as good as I can do not in a lab.

Ill post the full images at a later point.

Slog versus Cine1, It seems clear to me that Slog retains the highlights better at a cost of shadow noise (and less hunger for light) .. This is IMO entirely to be expected, but to shoot Slog in all situations would IMO not be the best use of the camera..

noise1.jpg
 
basically if I was shooting where there might be a lot of contrast or lots of highlights I would choose slog. For flat type lighting or very controlled lighting where your scene contrast might fit into 5-7 stops would opt for cine 1, 3, or 4.

But that's just my opinion and preference, obviously others are free to do as they wish.
 
Indeed, and that is what I am saying too which should keep the noise out of the shadows; relevant to "F3: Noise in Slog?"

:)

S
 
In your black curtain test what was your curtain exposed to in IRE, did you have a reference grey card, and what codec were you using?

Something about the F3 that's fantastic and quite defying is the fact it has a -64db signal to noise ratio, compared to the EX1 which is -54db, and the F5/F55 which is -57db so in reality it should be the "cleanest" of all cams. But as we all know bench specs don't always translate to "real world".
 
Very interesting thread. As someone who has used an F3 hundreds of times over several years, one of the benefits discussed for S-Log is that it is free of the "Sony look" present on the cinnegammas. I like the paid upgrade version because then you can get a LUT to view filming, record that version on the SxS cards, and have a pure S-Log version for grading to an external recorder. The backup version has saved me a few times both on audio and video lost from my external recorder. Everyone talks about underexposing S-Log, but I find that as long as you keep the exposure range between 25 (dark) to about 75 (white or highlight on faces) [i really like false color to help with this] then the file grades well without plastic faces or noise in the dark areas. I use a light meter often (Sekonic) -- much better than a card for exposure. The histogram is also a useful "gestalt" that most of the image is where I want it. Sometimes it is nice to use the non-paid S-Log, when there is light far from 5600 or 3200 and white balance (such as sunsets) but you only get the same image everywhere (internal and external and EVF).

No one talks much about this, but I suspect the larger/fatter sensors is why the -64db S/N is actually better than the F5. I am not aware of a head to head comparison of F3 to F5 and I know that the F5 has a better internal codec, but it might actually not produce a nicer image. I have a Nikon D3s which is 12 megapixels and a D4 which is 16 megapixels and both have the same specs in general, but I can't help but sometimes thinking that the images are actually nicer out of the D3s, especially in low light or high contrast situations. the Alexa is much cleverer in how they handle highlights in my opinion.

In any event, I am sticking with Morgan and going to stick with the F3 for at least another year.

I am definitely not a pro, so might not be applicable to the rest of you.

Jon
 
Jon thanks for the comments, an interesting read.

Stills wise I do use and D3 and do many of my chums, even ones who own D4 cameras.

I tested the F5 against the D3 as my 'reference imager' - dont ask about that one

S
 
BTW I think the sony look is simply defined.

Assuming 10bit recording one has a ratio of 1:4 data to 709,

Adding contrast in post one may want high contrast (steeper curve than 709) let us now say we need a ratio of 1:3

Any point on a gamma curve flatter than a 1:3 ratio versus the 709 curve looks 'Sony'. Simple.

S
 
As to the name on this thread - noise. The F3 has many shortcomings, but this isn't one of them. If you can find a quieter camera (S/N 68) let me know what it is - the F55 is 55, I believe, not as clean as the F3. Hard to find the Alexa numbers. As one poster commented: "We all know that few good examples of F3 Slog footage is posted online but these tests prove to me that many users who have bothered to do so simply don't understand how this camera works."

Slog results are wonderful if you expose properly - most overexpose. And agree that IR polltiuon not a problem, even with moderate external NDs.

A good solution for both exposure and general monitoring is the Gemini. It has a simple and quite accurate button to allow a Sony LUT to be inserted in the monitoring only. The Blade has yet to have this feature, but I'd expect it soon.

LUTs can speed up grading, but high contrast scenes need to be graded without a LUT or you'll lose your highlights.

I'm on Baselight and find grading F3 scenes easy.

Last, the way to get rid of the "Sony" look is to get low contrast high res lenses, like Cookes. the new Schneiders, or Leicas. While the Rokinon/Samyangs are very sharp - and cheap, they make everything people hate about Sony';s look even worse. Good glass can make your F3 a very different and lovely camera.

Good luck with the camera - it requires an investment of time and learning;it's not an EX1. Remember the wise words posted earlier: "We all know that few good examples of F3 Slog footage is posted online but these tests prove to me that many users who have bothered to do so simply don't understand how this camera works."
 
Barons said:
LUTs can speed up grading, but high contrast scenes need to be graded without a LUT or you'll lose your highlights.
Click to expand...

Or you scale the video before it hits the LUT. Or use ACES. I feel there's a lot of things you can do to avoid that predicament.

Some LUTs are nice enough in the highlight rolloff, you can let the highlights go, and it will be very nice looking.
 
Here is a frame from a shoot today that I think demonstrates no noise in the dark parts of the image. This was F3 with internal ND (hi) using Sony kit 85 PL lens, exposure f/8 natural light S-Log at sunset. In post I used magic bullet looks to add an S-curve to delog, increased saturation slightly, and increased exposure 0.3 stops. No color correction. I also used the new beta version of Portraiture for video, at a very light setting which matches skin tone color. There was no evidence of overexposure of the skin and the dark shades held up well. Obviously people could do much better with Resolve, but I am in a hurry. It actually matches quite well with photos I shot.Suede Dress.jpg
 
Thanks for all the posts. Here's a question concerning noise again. I pointed the F3 at the western horizon tonight just as the light was reaching its last 10 minutes. A dark hillside took up the bottom 40% of the frame and the sky above was at about 50 in the brightest spot. There were a few houses that were light enough in color to be seen in the monitor on the hillside but were well under 20. I had noise in those spots and in shadows in general. What else can you do here to get rid of it? Crush the blacks in post? I suppose going the cine gamma path would apply here.
 
Shoots56 said:
Thanks for all the posts. Here's a question concerning noise again. I pointed the F3 at the western horizon tonight just as the light was reaching its last 10 minutes. A dark hillside took up the bottom 40% of the frame and the sky above was at about 50 in the brightest spot. There were a few houses that were light enough in color to be seen in the monitor on the hillside but were well under 20. I had noise in those spots and in shadows in general. What else can you do here to get rid of it? Crush the blacks in post? I suppose going the cine gamma path would apply here.
Click to expand...

A grad-ND is the best thing you can do for shots like that. It really is. Getting detail into a heavily silhouetted image is a really hard ask for any camera - still or motion. A grad-ND could allow you to brighten up the foreground and minimise the level of noise in those shadows (by having them brighter).

If you want to get rid of noise though - crush those blacks. It's pretty much a necessary requirement with any log footage I find.
 
A grad would not do anything for this shot. The yards around the barely visible houses on the hillside were black and this shot was wide open at a 1.4. If I run into shooting this kind of thing it seems like I will have to either crush, or switch off slog.
 
Crush??? Why? Just use a de-noise plugin like neat video or Magic bullet denoiser and limit it only to shadows.
 
Shoots56 said:
A grad would not do anything for this shot. The yards around the barely visible houses on the hillside were black and this shot was wide open at a 1.4. If I run into shooting this kind of thing it seems like I will have to either crush, or switch off slog.
Click to expand...

Ah, well when you run out of exposure you run out of exposure. That said, I've never seen an image from a cinegamma on the F3 that I've thought wouldn't be better were it shot in SLOG. Noise be damned, the smoother tonality that SLOG gives you makes all the difference IMO.
 
Shoot56.

I have been trying to make a 'delog' curve using my nikon D3 as a 'reference image' - why? because the nikon D3 takes pretty pictures, which is the end goal - I hate tech.

See the images and scopes at http://www.dvxuser.com/V6/showthrea...-an-F3&p=1986377114&viewfull=1#post1986377114

To expose the F3 I put a very light grey at about 80.

When this curve is applied to a 'real world' image it heavily crushes the blacks, and is of course very noise free.

I am happy with this, finding that Slog made a much closer scope reading to my nikon D3 than CineGammas

The upshot of this is that if you are monitoring log you are probably going to lose the bottom 20% of the image with the application of a suitably hard gamma curve

(of course you can dig that dark out again, with a cost of noise, just like digging into a stills camera file)

I would suggest that the monitor look you are using is over building your expectations.

Either you will have to make a mental adjustment for that or get some LUT monitor action going, be that an onboard LUT (do you have the paid upgrade?) or a LUT generated by your monitor or a LUT box.

Seeing the gamma match my nikon D3 was a happy moment.

A side note - I tried recording uncompressed 10bit (10gb per min) and could clearly see that the image was cleaner in the shadows, an attractive option if you have the data capability..

S
 
morgan_moore said:
Either you will have to make a mental adjustment for that or get some LUT monitor action going, be that an onboard LUT (do you have the paid upgrade?) or a LUT generated by your monitor or a LUT box.
Click to expand...

If you don't have onboard 1D LUT capability you can just toggle between REC709 and slog PP but don't forget to switch it before recording :) . As mentioned by a few others just use the tools already on the F3 or get a good monitor that has the tools to make life easy exposing.


By the way Sam;

Your findings regarding cleaner shadows using uncompressed on the hyperdeck is very consistent with what I observed testing my F3 and hyperdeck earlier this year. When I compared 8-bit 420 and 10-bit 422 ProRes the noise in the shadows were practically the same. It wasn't until I switched the deck to uncompressed that all the tiny noise jumpies completely went away. Back then I didn't really make anything out of it but it did leave me even more convinced that there really was not much image quality difference between the XDCAM 8-bit footage versus ProRes 10-bit 422 with the stock F3 before any grading of course. Over a year ago I was quite the 8-bit advocate who swore that good exposure and light control were far more important than bit-depth if you read my older posts. :happy:
 
You must log in or register to reply here.
Back
Top