The Sony F65, thoughts and feelings of an F35 Addict

Thank you to those who do not capitalize raw. For some reason it bugs the sh** out of me. It's r a w and not the pretentious "all-caps" method that every Reduser and amateur camera guy loves to yell at those lesser people. (no offense star!)
 
Im not as technical as some of you guys here, and I agree that there shouldn't be more latitude or DR with RAW compared to log footage, but wouldn't it be true that grading RAW gives you more flexibility to grade, as apposed to grading a compressed/codec version of the same footage?

To expose correctly is the real point here tho :)


Also has anyone ever denoised RAW? as apposed to denoising a compressed/codec version of the same footage? I have, and the difference is night and day.
 
Viddovation said:
And again, if everyone would just expose properly in the first place...
Click to expand...

Shame this log vs raw chat is muddy-ing the info on the 65!

Video Vovation - should I call you Mr Video or Mr Vovation? - Ill go with Mr Video :)

Now if you always expose properly, well, why not save some DIT work and use an H264 8bit 420 camera? Simply most of them have enough colour and little enough compression to out do most delivery media - web and broadcast (which I think rarely get above about 12mbs)

OK all we have to out do is our final delivery media. BUT we probably are not going to 'expose properly' or at least we are going to get our lighting ratios, or colour off a little - we will want to grade our footage, if we are delivering to two o/p (maybe 16-235 and 0-255) we will want to grade our footage. Ok we may not want to 'grade it' but we will want to perform a mathematical transform of some nature on it.

So with a camera that collects more data than the o/p (output) requires our job as DP/Lights/CamOp is to collect that data in a manner that offers the post chain most choice, or to fit choices that have been pre-agreed, it may also be our job to do it efficiently in terms of data load transcode times et al.

All post is mathematical transforms

And when performing mathematical transforms 'oversampling' is the key to avoiding degradation of the image

For example if my first transform is to take all pixels from 245 to 250 and put them in a box labeled 'nearly white' and I then want to changhe them again, well Ive lost which ones were 246, 247 etc. Ive just got a lump of pixels labelled 'nearly white'

And that is what the flattening of the log curve does - it closes your options at higher values

Now that is not a bad thing - but it does close options - not only at the top end but it feeds through to the bottom end too, for example if you had a 10stop scene and a 12stop camera, shooting raw you could choose to shoot with 1stop head and foot room, 2 stops of head room (dumb) or with 2 stops of foot room and no stops of head room (clever) - these choices would give you different flex in post on different sections of the image.

Anyway - this is a kind of impossible argument for me to follow through as it has many aspects but Im convinced that..

-Bit depth is king. (tonal sampling)
-Raw gives more options and flex than log
-log is a good way to monitor exposure
-log often presents enough options in a fairly efficient workflow manner..
 
morgan_moore said:
-Bit depth is king. (tonal sampling)
-Raw gives more options and flex than log
-log is a good way to monitor exposure
-log often presents enough options in a fairly efficient workflow manner..
Click to expand...

I can't agree with you, at least not with the way you've structured your argument. It certainly would be true to say that *some* raw cameras have more flexibility than *some* log cameras, and certainly more so than video cameras recording directly to 8bit 420 and such.

But a statement like "Raw gives more options and flex than log" is misleading. For example, consider the Sony F35 and the Red One MX. The Red One MX A/D converter is 12 bit. That's 4096 luma values per photosite. The Sony F35 has a 14bit A/D converter - that's 16,384 luma values. When the Sony F35 processes the user selected gamma curve and other image settings, it's got 16,384 luma values to assign from 0% to 110% in a video signal that will ultimately be a maximum of 12bit (4096 luma values). While the log curve does sacrifice highlights to give greater emphasis to the darks and mids, if you put 12 bit F35 log footage against 12 bit Red One MX the F35 will have far more nuanced values in the mids and darks to work with than the Red. If you exposed correctly, this will be valuable, as you won't be looking to do highlight recovery in post, so much as make tonal shifts within the given exposure.

Yes - the newer Red Cameras have a 16 bit A/D, as do the newer Sony cameras. But my point is that raw itself isn't giving more options and flex than log. Or really, it gives you some options (highlight recovery) at the expense of others (more nuanced mids and darks).

Also, I think the math above makes a pretty clear case for part of why the F35 looks far more natural on skin tones and such than the Red One MX ever did.
Meanwhile the Red One MX A/D only ever collected 4096 level
 
Your 'argument' assumes that the scene employs the full DR of the sensor and/or a 'need' to retain the shadows/highlights.

On that basis (for a given bit depth) you are correct that log is equal or in terms of flex of the lows of the image potentially better.

Now if your scene is narrower than the DR or the sensor and/or you are prepared to blow highlights (or lose shadows) then raw gives you 'choice' aka 'flexibility' as the info in recorded in a linear manner and where you expose becomes irrelvant to where you can go in post.

Consider an unlit dark room with a window - if you decide to keep the outside window detail then you are probably going to stretch the sensor to the limit. In post you will probably need to lift the room, inducing noise.

Now if you decide the window can blow, you can lift the raw exposure so the room is exposed maybe at 60 instead of maybe 20, and then you get a clean image

Now if you are in log and you lift the exposure you push 'valid' detail zone into the highly compressed log section of the file and lose information pushing the file towards banding.

Simply ETTR in raw works and creates a clean file, but ETTR doesnt work in log as you push stuff into the 'compressed zone' at the top and it all goes funky (!)

When the scene occupies the full DR of the sensor there is no space to ETTR (as you will clip) but with a smaller DR the scene can be ETTR'd

ergo - "Raw gives more options and flex than log"

Now those options may or may not be useful, to your scene and they may also be bad options (noise induction vs log) - so IMO the statement is true, but you have read that I meant that the statement also meant 'better' - I did not.

==
Meanwhile in the real world Ive just been testing the BMC4k in log and raw, log is super simple to expose and Davinci has the perfect LUT, raw is very wonky and poorly handled by Davinci. My next project will be shot in Log - while raw may offer some advantages they are not advantages that I feel the post chain will have the skill/budget to fully utilise, ergo Im gonna get it right 'in camera' and save some HD space. Probably.

S
 
Last edited:
Guys

I already started a thread some time ago about exposing for S-Log, can I suggest the conversation goes there?

I am happy for F65-specific questions to be in this thread, but I don't want it veering wildly off-topic.

For the record, I have posted two DPX files already, I will see about posting an MXF, but it will have to be very short as the file sizes are frightening!

Anyone that wants to post me a hard drive I will put footage on for them.

James
 
Hi James, sorry your thread got all muddied up and even more sorry if it was not clear enough in post #12 that I was specifically talking about the F65.

From post #12
starcentral said:
Not to state the obvious but regarding exposure RAW will give you much more exposure latitude as opposed to S-Log on these cameras
Click to expand...

Anyway my only suggestion or request would be you conduct some bracketed exposures in both RAW and S-Log (10-bit and 12-bit) and after having the images restored to more normal levels observe any differences in image quality and judge for yourself.

If anyone wants to continue the banter on a new or existing thread it would be great.

Cheers,
 
Starcentral - in the very narrow element of your argument that for a given camera using the 16bit raw will give more exposure lattitude than the 10 or 12 bit Log recordings will - yes, I agree with you. You made other claims and assertions that were broader and that I cannot concur with, and those I was addressing.
 
I feel these discusions are proper for all of us here. If it wasnt for all of you GREAT PEOPLE adding your knowledge to this amazing camera experience knowledge pool, then we wouldnt be here NOW.

James, continue doing your 65 evaluation on here. we are all listening
 
What I would actually like to know more on - and I'll read the doco on it - is decent exposure tools for the F65.

With the 65, there are so many options for viewing LUTs, working out what works is going to be tricky. Too many of the LUTs add too much contrast, where S-Log always looks dull (and it was never supposed to be a viewing solution anyway)
 
A7s to the left of me, F65 to my right... Here I am: stuck in the middle with you, F35.

Thanks for for the incredible depth so far, James, on a camera that has not gotten much of this kind of detail from an actual user.

As as for the other silliness here: I prefer to make decisions on set and compose in camera, so s-log gives all the possible option I could need. I'm not sure my workflow would benefit from all the extra data of raw.
 
Last edited:
RAW will never compensate for improper exposure, really. I am out and about today, but i'll be working with the camera in the week and I'll try and get some more pics for you all
 
JMtheDP said:
What I would actually like to know more on - and I'll read the doco on it - is decent exposure tools for the F65.

With the 65, there are so many options for viewing LUTs, working out what works is going to be tricky. Too many of the LUTs add too much contrast, where S-Log always looks dull (and it was never supposed to be a viewing solution anyway)
Click to expand...

I think monitoring, luts, even exposure and codec selection ( :) ) have to very much revolve around clear intention from yourself.

For example I would want to be running 3 Luts, simply a punchy saturated 709 style lut and also 'highlights I can get back' and 'shadows I can get back' (because that punchy lut will clip stuff)

But that is me - I would expect to shoot for secondaries because that is how I think nice images are made - I guess you need to build a system that works for you and you can clearly communicate to director, crew and post.

S
 
morgan_moore said:
For example I would want to be running 3 Luts, simply a punchy saturated 709 style lut and also 'highlights I can get back' and 'shadows I can get back' (because that punchy lut will clip stuff)
Click to expand...

That's exactly where I want to be. Fortunately, you will LOVE the F65's "hi-key/lo-key" button - I am just trying to find that middle-ground LUT.
 
A shot of the camera in the wild...

2rneyag.jpg
 
It's the sony EVF-DL100 or whatever they call it - the OLED number from the F5/55.

It is a big beast but fairly easy to move about and deal with - nice not being tied down with batteries etc...

At Sony we worked out that an F5/55 with all the modules added on was actually longer than an F65!
 
macgregor said:
Do you need some sort of adapter to attach it to the F65?
Click to expand...

From memory the camera needs to be sent in and modified to have the new EVF port added to it. It'd be worth it for the OLED though, I have one on my F5 and it's the first time I've ever felt truly comfortable judging a digital image I'm recording (particularly focus) by-eye.
 
You must log in or register to reply here.
Back
Top