FS7: RAW SDI output recorded in 4KDCI@50p on Shogun Inferno: good news!

Ben is from the UK.. and therefore ratty and wicked.. its a cultural thing..

Hahaha. It's not acting superior if you're ACTUALLY better ;)

Piotr, don't worry yourself I shan't intrude in your threads any longer. And a quick note for the future; don't bother to apologise directly after a few sentences continuing to have a dig at me and complain about me and how I'm not helpful enough for you. It sort of rings a bit hollow if you know what I mean.
 
You really think I should have done it in opposite sequence?

And BTW - here you go again: you did have the XDCA unit and problems with "linear RAW", why didn't you just answer my question in the post directly before the last one? Here it goes again for your convenience:

"So please guys - don't play offended, but tell me: has this changed compared to the early version that all this rave about RAW being linear, and as such holding too little information in the shadows? Or am I still missing the point...|

Ben - you're making my point for me...

Nevertheless, my apologies were sincere as I'm not a kind of person who would be crossed with anyone for longer than a couple of minutes; you didn't accept them. Pity.

Piotr
 
No, Atobit - I thought exactly along this line at first, until I found it on Page 52 of the *newest* User Manual for FS7: in order to activate RAW output the camera must be in CineEI mode, and the only options for Gamma are S-log2 and 3. You cannot turn Gamma OFF, either (this option is grayed out). So it's not the Shogun that bakes in the log look; it's FS7 RAW which is either S-Log2 or S-log3 (but interestingly enough, you *can* change color space between S-Gamut for S-log2, and S-Gamut3 or S-Gamut3.Cine for S-log2).

So please guys - don't play offended, but tell me: has this changed compared to the early version that all this rave about RAW being linear, and as such holding too little information in the shadows? Or am I still missing the point... Thanks,

Piotr

Again, you are drawing conclusions left and right. What is said in the manual about having to select one of the log gammas does not mean that the RAW signal has a baked in gamma. It could be that in order to also have internal recording in addition to RAW output, one of these gammas need to be selected (and it may be a requirement even when internal recording is not selected). I'm surprised that you insist on that even though you actually have not recorded RAW (as RAW ) to neither the Inferno nor Odyssey, and ones who have are telling you that the signal is linear, without a baked in gamma.
 
Perhaps I'm drawing conclusions - and your theory of S-log curve having nothing to do with RAW, but required by internal recording when e.g. "RAW+XAVC-I" format is selected, is true. But then, why would it be limited to -Slog only? OK - the answer could be "because setting RAW external requires putting the FS7 in the Cine EI mode". Fine, by why is it a prerequisite? If RAW isn't somehow "biased" towards Log, but the most neutral linear picture - it should work with the camera in Custom mode as well, with HGs for internal recording available!

So please don't make "drawing conclusion left and right" an accusation; I openly admitted "I simply don't know" and politely asked those who do to share their knowledge, so that everyone (me included) learns something and doesn't need to "draw conclusions". If I can admit my lack of knowledge, why is everoby else jumping on me instead of saying "I don't know, either" or - if he/she does know - explaining things to me in a professional and convincing way?

Last night I tried to remove Shogun encoding from the equation, and connected XDCA RAW SDI output straight to my Decklink 12G SDI input; the picture looks exacly like the S-log picture from HDMI (as opposed to RAW SDI). But the washed-out picture is how RAW should look - so it'snot been a conclusive experiment, either...

All in all: I have difficulties assuming it's the Shogun that applies S-Log curve before encoding RAW to DNxHR. The RAW signal from FS7 hasn't got the same abundance of metadata that HDMI-out does; enough to say that when fed from HDMI, the shogun recognizes the source as Sony FS7; when getting the SDI signal it says "source unknown". So not even knowing the camera it's getting the RAW signal from, is it supposed to know: "aha - the operator has chosen S-log3 on the camera; let's apply the S-log3 curve before encoding his RAW signal into DNxHR"...

I guess it'd be an interesting exercise to compare what the current FS7 manual says on Page 52 with what the original version said about prerequisites for external RAW recording; does anyone have an old manual available? Thanks,

Piotr
 
The old manual from December of 2014 does not say anything about Slog2 and Slog3 being the only gamma curves available but it does say that RAW output is available only in Cine EI mode, so in a way it does say that only Slog curves are available. The manual is available as a sticky on Sony's FS7 forum main page.
 
BTW, Piotr, what is the date of the instructions manual that has that info and where did you find it? I can't seem to find anything newer than version from June 5th 2015.
 
OK - so now that we know RAW is still indeed linear, and there is no gamma baked in (either internally, or by the Shogun - the latter I assume, but why would it do it?) - has anyone taken a look at the .drx screen grab I posted? Any comments, please?

Piotr
 
There has to be some gamma applied to even view RAW no..? its not a video signal by itself I believe..

PS I think maybe you take, (particularly British) humor to be an insult.. its not meant to be..
 
Wow. This thread exploded. In the hopes that Sony might have fixed the RAW, I upgraded all of my gear to the latest firmware and re-ran a basic shadow noise test. The results are the same as the last time I did this. The noise in the RAW output (whether recorded as RAW or as ProRes) is dramatically higher and is extremely ugly. I'm attaching LUT'd jpegs of a side-by-side. If you're interested, I'm also putting single frame uncorrected DPX files up for anyone to download.
First the XAVC-I Image:
XAVC-I LUT_1.2.2.jpg

Now the RAW Image:
RAW-LUT_1.1.2.jpg

Editing because the built in compression on this forum destroyed any chance you might have of seeing the noise. Here is an imgur album that is a little better, but if you are really interested, download the DPX files.
 
Wow. This thread exploded. In the hopes that Sony might have fixed the RAW, I upgraded all of my gear to the latest firmware and re-ran a basic shadow noise test. The results are the same as the last time I did this. The noise in the RAW output (whether recorded as RAW or as ProRes) is dramatically higher and is extremely ugly. I'm attaching LUT'd jpegs of a side-by-side.

Wow - indeed not nice. But since my Shogun Inferno's DHxHR HQX footage is indeed very clean (much cleaner than XAVC-I) - I'd venture to say that perhaps DNxHR codec (and its Atomos implementation) is the key to this puzzle?

I haven't got the CDNG functionality yet, But I will record Prores instead of DNxHR and compare the two; will report back... Cheers,

Piotr
 
broughtonfilm—I'm shooting with an FS7 (obviously), recording to an Odyssey 7Q+, and pulling the footage up in Davinci Resolve.

dreamXpress— Here is a download link. I spent a good bit of time setting up this test frame, and since yesterday's results were basically the same as the ones I did a year ago, I'm just posting the RAW frame from that. In this shot, 90% white is exposed at 61% as is recommended for SLOG3. I'm also going to link to the same shot exposed up by 2 stops. You can see that while this reduces the noise threshold, it doesn't solve it, and it introduces other problems.

moldcad— while I'm interested in the results of your ProRes test, I'm not expecting anything different than what you're seeing in DHxHR. On my system (and the systems of everyone I've spoken with that owns FS7's) the failings in the RAW show up in the actual signal. It wouldn't matter what format you transcode a bad signal into—it will always look bad.

I looked at the DPX clip you uploaded. It's hard to tell, but I did see some noise in the dark areas. Setup a test scene that includes at least one 100+ IRE bright spot, a field of dark (that should resolve at or near 10% IRE), and a set of colorful things (a color chart, flowers, children's toys). Then expose so that white is at or near 61%, and take a look at the noise profile in the darks. I'd really appreciate it if you posted that DPX. If it's significantly better than mine, I'll be renting an Atomos to determine if the Odyssey 7Q+ is the problem.
 
Christopher, your results mirror my own from a while back but since the FS7 came out a lot of us have found Sony's target of 61% to be wildly optimistic on their part and myself and others have had exponentially better images by targeting 75% for 90% reflectance instead.
 
Last edited:
I agree. That's why I included the +2 Stops shot. However, in my testing, when I go +2 stops, I really miss the extra headroom (especially if I'm shooting outside).
 
Ah cool.

Yep, this is another point where the claimed stops vs 'useable stops' is such a hotbed of contention.
 
Ive also come to expose less hot lately as I seem to like the images with more headroom and a bit more noise.
 
Last edited:
Chiming in here as well - mostly to thank both Ben & Christopher for their posts about this early on in this forum. It saved me from going down the raw path.
Around the time of FW 3.0 (wasn't that when Sony "improved" the raw signal ?) - I rented the ext unit along with the 7Q+ for the first time to test it out for myself. And my conclusions were exactly the same as those posted by Christopher.
I'm sure there are many of us, Piotr, that would be interested to see your different results.
 
Last edited:
Ive also come to expose less hot lately as I seem to like the images with more headroom and a bit more noise.

It's definitely a balancing act.

If I am completely honest I've somewhat fallen out of love with the image recently, although it may just be the nature of the recent work I've been doing...
 
Back
Top