F5: F5 successor revealing (really a F65 successor)

[combatentropy]

a stop is a doubleing of light, 1,2,4,8,16,32,64,128,256,512,1024,2048,4096 (13 stops) 8192, 16384 (stops 14 and 15)

very quickly representing one stop (with technical perfection) can eat half of the 444 space.

True, up to a point. 10-bit log, meager as it sounds, is enough for any scene.

- 10 bits is 2 to the 10th power: 1,024.
- Log means logarithmic, which means spreading the bits evenly per f-stop
- The human eye can distinguish, at best, 70 steps per f-stop, and only in the brightest regions (in the darker zones, it's less than half that)
- This applies to the number of f-stops in the final viewing screen, not the original scene. TVs, I would say, are usually 7 stops, especially when you consider ambient room light. Even a movie screen in a theater --- isn't that like 10 stops? A 15-stop viewing surface would hurt your eyes. You want it reduced. That's the raison d'etre for the S-curve. It compresses the brightest stops, gradually from, say, 5 stops to 2. Likewise the shadows, maybe less severely. Meanwhile the mids have realistic contrast (unless you're talking about most of the digital mis-grading these days. "Plastic," "video," "digital," are all synonyms for low-contrast mids. Try it. Does it look to video-ey? Apply a stronger S-curve. It will turn into film.

1,024 / 70 = 14.6 (worst-case scenario; it's more like 1,024 / 30 = 34 max stops). Still plenty, even for HDR TV, even with pushing in pulling in post. So, record in 10-bit log (which will look washed out), and grade into a nice S-curve.

 

[starcentral]

Mark is the DP Sony should be collecting feedback from. :beer:

 

[morgan_moore]

True, up to a point. 10-bit log, meager as it sounds, is enough for any scene.

- 10 bits is 2 to the 10th power: 1,024.
- Log means logarithmic, which means spreading the bits evenly per f-stop
- The human eye can distinguish, at best, 70 steps per f-stop, and only in the brightest regions (in the darker zones, it's less than half that)
- This applies to the number of f-stops in the final viewing screen, not the original scene. TVs, I would say, are usually 7 stops, especially when you consider ambient room light. Even a movie screen in a theater --- isn't that like 10 stops? A 15-stop viewing surface would hurt your eyes. You want it reduced. That's the raison d'etre for the S-curve. It compresses the brightest stops, gradually from, say, 5 stops to 2. Likewise the shadows, maybe less severely. Meanwhile the mids have realistic contrast (unless you're talking about most of the digital mis-grading these days. "Plastic," "video," "digital," are all synonyms for low-contrast mids. Try it. Does it look to video-ey? Apply a stronger S-curve. It will turn into film.

1,024 / 70 = 14.6 (worst-case scenario; it's more like 1,024 / 30 = 34 max stops). Still plenty, even for HDR TV, even with pushing in pulling in post. So, record in 10-bit log (which will look washed out), and grade into a nice S-curve.

Interesting info on the "spec" of the eyes.

Clearly a 12bit file has a lot of fat and true uncompressed 16bit raw is not a requirement for pretty pictures. I do however think tha for example a 12bit Alexa file has more chunk than a 10bit slog file and it shows.

Also as we examine a a scene our eyes ramp iris - so thier spec may be a little better than the eye spec you gave when they are "locked off"

Now our delivery media might only do 7-10 stops but that is where lighting (lifting shadows) or Nd grads come into play to compress a wide dr scene into a smal delivery space.

Well lighting and grads used to be used - but now we expect the camera to hold the detail and pull it back using secondaries in post - and that is where a 15 stop camera becomes useful..

 

[morgan_moore]

I would admit that my suspicion of 10bit might be somewhat clouded by compression - I've never used and uncompressed 10bit camera
We rarely see the combination of low bit depth and low compression ratio - so in the "real world" aka my experience of red, some Alexa, lots of fs7, some sr444 amd a million raw stills the 10bit does not shine

 

[starcentral]

Sam,

Here's my take regarding 10-bit;

If the image does not undergo a lot of preprocessing IN-CAMERA for highlight handling and decent color science, then yeah you're going to need a much bigger bucket and container for it all to manipulate the crap out of it later on. RAW prevails here because the extra data in the highlights will let you add some roll off in post and color science is also applied later in post which has a chance to continually be improved. ie. We saw RED do this (how many combinations of redlogfilm and redgamma are there now?) and Sony has also done it by releasing 3 log gamma curves and 3 color sciences (s-gamuts) to date and maybe one more coming soon.

Now take a look at the cameras which actually do a lot of IN-CAMERA preprocessing before the image is recorded, mainly that would be any ARRI camera and the F35. Specifically highlight wrangling and decent color science, so much so that you could honestly record images in 8-bit and with the little wiggle room you had left to grade still produce absolutely beautiful and stunning images. The image is 95% there, not much left to do.

About codec since you mentioned uncompressed shooting in 10-bit; I tested 10-bit uncompressed before with the hyperdeck shuttle and while you get no dancing jaggies or macro blocking artifacts, the image just looks a bit cleaner and that's it. Color doesn't suddenly get transferred to ARRI color and make it a night and day difference. That's been my experience with it anyway.

 

[DLD]

Mark is the DP Sony should be collecting feedback from. :beer:

I was jowkeeng ... abetted by my visiting from Melbourne relative.

... This applies to the number of f-stops in the final viewing screen, not the original scene. TVs, I would say, are usually 7 stops, especially when you consider ambient room light. Even a movie screen in a theater --- isn't that like 10 stops? A 15-stop viewing surface would hurt your eyes...

Rec. 709 is 6 stops (or so I've read in a lot of places). Current HDR sets might get to 1,000 nits and 10 stops. 1,000 nits is far too bright for most folks in a room with controlled ambient light. The current aim in displays is 13 stops, which on a scale from 0 to 100 will probably have their brightness at around 20.

 

[Sttratos]

It's just different style of shutter, there's no inherent light loss to it. If you've got the mirror spinning at 180 degrees, you'll have the exact same exposure as turning it off and having the electronic shutter on the sensor expose at 180 degrees.

Using particularly funky shutter angles is another time when swithing to an EVF can help. But it's very rare for me to go to longer shutter angles than 180 degrees (things start to look 'digital' when you do that), I use short angles somewhat frequently, but rarely longer ones. Plus the camera's great up to 1600 ISO and very usable up to about 2000 ISO (at which point you can light things with the screen from your iPhone) so I don't really need more sensitivity.

Price-wise you pay a premium for the camera (I certainly did), but I think the feature set makes it worthwhile. It's also the only big Alexa with internal ND, which is super handy.

OK. Thanks for clarifying. I thought it used the same shutter style as in film cameras. Maybe I was thinking of the D21 or mistaking all together.

 

[Sttratos]

The 65 didn't catch on because of the convoluted post workflow and if I remember correctly the recoding system wasn't fully available and very expensive. Mitch Gross posted something a while back detailing why things went off the rails. If may be in the F35/F65 section... or on CML.

By comparison the Alexa offered an idiot proof direct-to-edit Prores workflow for commercials / TV and a pretty straightforward ArriRAW pipeline. Also the data footprint is a lot more manageable from the Alexa when shooting RAW. Remember that it's been about 6-7 years and hard drive space etc was lot more expensive than today. Also the Gemini444 that allowed you to record ArriRAW was tiny in comparison to the Sony rig and cheaper.

I really haven't heard a lot of complaints about F65 IQ. If anything I've heard people say it's too sharp and perfect, but I was blown away by the RAW files I got to handle a few years back.

All this talk of the F65 IQ. I don't think anybody here is complaining or even talking about the IQ. Alexa IQ is not better. The image is just more pleasing. Nothing to do with IQ.

I know the workflow helped to sink the F65 ship. But I don't buy that price and size did. They were comparable to the Alexa.

 

[morgan_moore]

Sam,

mentioned uncompressed shooting in 10-bit

Yep I only mentioned that as I try to keep things 'scientific' and if you have two elements degrading the image, tonal compression and spacial compression then it sound to mention that the other one exists.

Working from a self gernerated PSD file representing an 'ungraded sky' I was able to emulate various compressions and how they reacted to a grade, and certainly was surprised that compression ratio rather than bit depth was 'the killer' - contrary to what I was trying to show!! https://vimeo.com/163942387

My overall point was that I think comparing the 2.5k big pixel, 444 recording Alexa to any 10bit small pixel 4k camera of equal sensor dimensions is not a comparison that (for the consideration of DR and tone - not spacial resolution) one would reasonably expect to be a flat playing field. Clearly the Arri holds all the cards to win.

The only way to beat Arri for look and also have 4k resoluton is to move in the dirction of Vista. Something Arri know, and Sony now acknowledge

S

 

[DLD]

Well, it's Sony who makes both the full frame and the medium format sensors (ARRI "makes" its own Alef or 3xAlef for Alexa65).

 

[Bern Caughey]

Mark is the DP Sony should be collecting feedback from. :beer:

Couldn't agree more.

 

[Sttratos]

from a personal experience: my wife took me to the cinema for Luc Besson's Lucy. I was so impressed about the colors and crisp look that I googled what camera it was shot on. Car chase was on Red Epics and other shots Arri Alexa but most of all on a F65. If it's what you call Sony look: I like it!

(I am just watching, far from an assignment on Alexa, F65 or Weapon)

Talking about Lucy again, interestingly, Luc Besson and his Lucy DP Thierry Arbogast chose to go with the Alexa for their next project together, Valerian and the City of a Thousand Planets.

Thierry Arbogast's prior few projects to Lucy had been Alexa and he went back to the Alexa for all his post Lucy projects.