dreamXpress
Member
there is a bit of noise in the blacks.. yes. but i dont think it is too bad..
Last edited:
FS7: RAW SDI output recorded in 4KDCI@50p on Shogun Inferno: good news!
- Thread starter moldcad
- Start date
moldcad
Veteran
@dreamXpress - display Christopher's picture on a large display, and pay close attention to the darkish background around and behind the 2 bulbs; in my opinion this is the most offending area with what looks like chroma noise, but what actually is banding / color "peeling" due to not enough info bits in shadows.
Speaking of which I can say this: Christopher - no, I'm not going to post my screen grabs and pixel-peep like this. I'm perfectly able to create a similar nasty effect with my XAVC-I pictures is not exposed correctly, or - after opening in Resolve - I use grading tools in the wrong way. In fact, I can also manipulate my super-clean DNxHR Inferno footage in Resolve in such a away that areas like this start to appear. But what's the point of this? Why would one compose his scene content, lighting and camera exposure in purpose to produce this sort of artifacts (and this is - more or less - what you would like me to do)?
For me what counts is that in real-life scenes, I'm not seeing this and the picture is so much cleaner with DNxHR from RAW SDI output than similar scenes can ever be when recorded internally to XAVC-I. This makes my investment in the Shogun and the XDCA extension absolutely worth-while, and this is all that counts for me
BTW, I am able to manipulate my own, clean picture in such a away that local areas of exactly the same "peeling"/banding colors appear in shadows. So what? With proper grading, I can't see them - plus there is no noise whatsoever! While no grading could repair the compression artifacts in blue sky I came across so often in my XAVC-I footage...
Thanks for your cooperation,
Piotr
moldcad
Veteran
Yes - I get the idea, but cannot do any tests with scenes arranged as per Christopher's suggestions because the Shogun dock station is away for repair. I'll do it when it comes back; also it would be worth more if I could record RAW as RAW which will only be possible once Atomos releases a new firmware.
Piotr
Piotr
ChristopherA
Well-known member
there is a bit of noise in the blacks.. yes. but i dont think it is too bad..
dreamXpress—that image is way too dark. You can't make out where the hair ends and the background starts. You need to bring up the mids, and when you do, you will start to see noise.
Speaking of which I can say this: Christopher - no, I'm not going to post my screen grabs and pixel-peep like this. I'm perfectly able to create a similar nasty effect with my XAVC-I pictures is not exposed correctly, or - after opening in Resolve - I use grading tools in the wrong way. In fact, I can also manipulate my super-clean DNxHR Inferno footage in Resolve in such a away that areas like this start to appear. But what's the point of this? Why would one compose his scene content, lighting and camera exposure in purpose to produce this sort of artifacts (and this is - more or less - what you would like me to do)?
For me what counts is that in real-life scenes, I'm not seeing this and the picture is so much cleaner with DNxHR from RAW SDI output than similar scenes can ever be when recorded internally to XAVC-I. This makes my investment in the Shogun and the XDCA extension absolutely worth-while, and this is all that counts for me
BTW, I am able to manipulate my own, clean picture in such a away that local areas of exactly the same "peeling"/banding colors appear in shadows. So what? With proper grading, I can't see them - plus there is no noise whatsoever! While no grading could repair the compression artifacts in blue sky I came across so often in my XAVC-I footage...
Thanks for your cooperation,
Piotr
I guess I'm kind of confused. There was nothing special that was "done" to this image to create that noise. I literally setup a scene that had both very bright areas and very dark areas, lit a human and a color chart to the Sony recommended level, and captured to RAW. I then brought it into resolve and converted from linear to Rec.709. That is about as "real world" as I can possibly make a test. To control, I took the same image, with the same exposure using XAVC-I (and a bunch of other formats), and ran them through the EXACT SAME WORKFLOW. The XAVC-I produced a substantially cleaner image.
My guess is that either 1) you are exposing to the right with your RAW (the Shogun could actually be doing this internally without you knowing), I have found that 1.5-2 stops is enough to reduce the noise level to an acceptable point, but you then lose that on the top, or 2) you are incorrectly identifying low noise because you haven't put your camera in a position to really tell, or 3) the Shogun is actually doing a better interpretation of the FS7's RAW.
If it's the first two, then we haven't learned anything new as a community. The RAW implementation is, as we have known for about a year and a half, critically flawed. If on the other hand, it's the third, I'll be selling my Odyssey in the next couple of months and buying a Shogun. This is why we design and build meaningful tests—to learn meaningful answers. If the purpose of your original post wasn't to help the community then why did you post here? Did you only wish to brag about a new purchase?
If however, you wish to believe that your camera is magic, that testing to find out it's true capabilities is blasphemy, and that real-life will only contain shades of gray that will never challenge the limits of your equipment, I wish you nothing but luck.
broughtonfilm
Well-known member
The comment below is in regards to the statement in parentheses.
How is that possible, or rather how relevant? Can you elaborate on that? Are you saying that it is shifting levels of the signal before recording? How valid would that be in this evaluation since by doing so it can't change the actual exposure (how much light hits the sensor)? So even if it does shift signal levels it has to base it on the same exact RAW signal - the very thing that is being evaluated here. Or are you saying that Shogun is sort of applying a compensating LUT for viewing only which forces user to overexpose ? Sort of like shooting with Exposure Index of 500 ISO on FS7. I'm vert confused by your statement that I quoted.
Last edited:
ChristopherA
Well-known member
It would work like a transformation LUT that re-rates the camera's base ISO and requires the operator to overexpose the image by 1-2 stops. This would push the noise level down at the expense of headroom in the highlights. Basically, instead of saying that you have 14 stops of usable dynamic range, you rate the camera at 12 stops and throw away the bottom 2 (because they are noisy garbage).
The downside (and why I wouldn't expect Atomos to do this) is that your in-camera exposure wouldn't match the exposure on your external recorder. That being said, if you're taking your exposure measurements using the recorder, it would work just fine.
This exactly. It's an easy thing to test, if you have access to an FS7, Extension back, and Shogun. It would account for everything in this thread, as well as the fact that the folks at convergent have been unable to parse the FS7's RAW signal into a usable image. As I mention above, it's not something that I would expect given the mismatch in exposure, but it is a possibility.
broughtonfilm
Well-known member
Well, I wish I had Inferno handy so I could test it. Out of all the people who visit this forum, is Piotr really the only person who owns FS7, XDCA and Inferno?
moldcad
Veteran
If I'm indeed the only one here who owns all the "ingredients", than I may consider myself a lucky one (and Christopher's comment that I perhaps started this thread with the sole intention of bragging might be justified, in a way ). But none of these is true, and we all know it (I hope).
Christopher - I really am willing to run and post tests that would be oranges to oranges with yours, but currently:
- as I said earlier, I have no means of transferring footage from Shogun to my edit system (the Shogun dock station should return to me soon, replaced or not)
- if you read my last exchange of posts here, you will notice that I'm still unsure which S-log curve Shogun is applying to my FS7 RAW before encoding it to DNxHR. So I'm not sure how excatly I should transform those DHxHR clips so that what I'm seeing is directly compared to your picture, which resulted from transforming linear (as you operated on linear RAW directly, which I'm unable to do as the CDNG option is not yet active on te Shogun) to Rec.709.
So please tell me: should I transform my footage that I already managed to offload into my PC (before I had to send my dock for servicing) from Sony's S-Log3/S-Gamut3.cine (as this is how my FS7 was set up when shooting, and in the discussion I linked you to Alister says that S-log3 is applied by default before encoding DNxHR) to Rec.709? I guess this is as close to your workflow as possible, but I'm not 100% positive... The only screen grab I posted so far in this thread was clearly apples to oranges, as - using my usual HDR worflow - I transformed from S-Log3/S-Gamut3.cine to Rec.2020/BT.2084, and only then grabbed the still I later posted here...
Piotr
). But none of these is true, and we all know it (I hope).Christopher - I really am willing to run and post tests that would be oranges to oranges with yours, but currently:
- as I said earlier, I have no means of transferring footage from Shogun to my edit system (the Shogun dock station should return to me soon, replaced or not)
- if you read my last exchange of posts here, you will notice that I'm still unsure which S-log curve Shogun is applying to my FS7 RAW before encoding it to DNxHR. So I'm not sure how excatly I should transform those DHxHR clips so that what I'm seeing is directly compared to your picture, which resulted from transforming linear (as you operated on linear RAW directly, which I'm unable to do as the CDNG option is not yet active on te Shogun) to Rec.709.
So please tell me: should I transform my footage that I already managed to offload into my PC (before I had to send my dock for servicing) from Sony's S-Log3/S-Gamut3.cine (as this is how my FS7 was set up when shooting, and in the discussion I linked you to Alister says that S-log3 is applied by default before encoding DNxHR) to Rec.709? I guess this is as close to your workflow as possible, but I'm not 100% positive... The only screen grab I posted so far in this thread was clearly apples to oranges, as - using my usual HDR worflow - I transformed from S-Log3/S-Gamut3.cine to Rec.2020/BT.2084, and only then grabbed the still I later posted here...
Piotr
broughtonfilm
Well-known member
Just post some footage, both straight from Inferno and internal XAVC-I (I assume the internal recording is either Slog-2 or Slog-3). This way it is easy to cross reference and compare, XAVC-I recording will also eliminate speculation about what Inferno (and if) does to influence exposure.
EDIT: just realized that if exposure was evaluated based on Inferno and it indeed somehow was forcing overexposure (as per Christopher's speculation) this would affect internal recording too.
ChristopherA
Well-known member
Your camera should have the ability to record XAVC-I as well as output RAW at the same time. The quickest way to tell would be to put the camera into SLOG3 and record, then compare the clips. If the contrast appears to be the same, then it's probably using SLOG3, if not, you can repeat the test with the FS7 in SLOG2. This test would also allow us to determine if the Shogun is forcing you to overexpose the image: if the image in Camera appears to be brighter than the image you record on the Shogun, that's how they're handling the noise issue.
Without the ability to shoot new footage, testing will be difficult. In the material you've already shot and transferred, do you have any shots that have areas below ~15IRE (prior to color correction/LUT)? Theoretically, that should show the noise that I'm seeing on my end. Don't think that you need a particularly large dark area (although that make this much easier); a simple shot that has a dark shadow will work just fine.
I'm not a fan of Sony's LC709A LUT, but let's use it for the conversion here. It will allow us to have a standard conversion that is easy to replicate. It also, by default, has an elevated black point which should make searching for shadow noise easier.
broughtonfilm
Well-known member
This is not what the theory was, this would obviously not create a cleaner image as it would require applying post gain to the raw signal in Inferno. What Christopher suggested was that perhaps the way Inferno displays the image for monitoring prompts user to overexposing (by displaying the image darker, only for monitoring, sort of like a compensating LUT). If one used it as a guide though, it would equally affect both external and internal recordings in terms of exposure.
remsyatassi
Well-known member
The reality is that these differences are so extremely negligible, it's not worth eight pages of discussion or any more consideration!! It is a waste of time.
By the time you add in the Odyssey/Inferno, XDCA, and batteries you are easily in the range of a used Red or Red Raven. If Raw is important to you then the FS7 is simply not the camera to buy.
Here's a simple question I can for you. If you improve image quality by 1.5%, does it actually matter? NOPE!
By the time you add in the Odyssey/Inferno, XDCA, and batteries you are easily in the range of a used Red or Red Raven. If Raw is important to you then the FS7 is simply not the camera to buy.
Here's a simple question I can for you. If you improve image quality by 1.5%, does it actually matter? NOPE!
broughtonfilm
Well-known member
The XDCA is needed for any multi cam shot where timecode syncing is necessary. Even in single camera shoots ability to have TC in / out option is a plus and sometimes a deal breaker. Then Prores recording which many productions favor. This means you have to get batteries anyway. Then it comes down to the difference in price between a monitor and a monitor/recorder which in most cases is less than $1000, sometimes (overpriced SmallHD) monitors are actually just as expensive, plus the RAW license and recorder drives (storage is another thing but I consider it a post cost).
While I don't see myself using RAW on FS7 in a near future because I agree that the difference in quality is likely not worth the hassle, I can see how that can fit someone's workflow better and actually matter. I'm curious to find out if Inferno offers any improvement over Odyssey.
moldcad
Veteran
Yes I understand - that's why I mentioned "disregarding Inferno's exposure "suggestions""...
Piotr
moldcad
Veteran
Well - I felt obliged to comment on the above post. Mainly because I'm the one who started this thread, and also because of suppositions by some that I did so for bragging around how good equipment I have. Well - being an engineer by trade, not a videographer - I'm indeed a technology geek
But I also am extremely sensitive to everything fine art - from music through poetry through (motion) picture, so "making movies" - as the common ground of those areas - has always been an exciting world to me, and I've always been ready to spend my last penny on it...However - and I want to stress it - bragging was not the reason for setting up this thread: I just wanted to share with you guys the great news that the picture I'm getting from Inferno (in DNxHR HQX codec so far) is indeed of much higher quality than the internal XAVC-I codec based footage. I was struck by the absence of noise and any compression artifacts (remember my thread on "the thin line across the blue sky"?) I got used to with XAVC...
So to summarize, I'll repeat: perhaps FS7 RAW does indeed leave much to be desired, and nothing has been improved in this area (contrary to what Ben suggested in his answer to my specific question about it, asked just before deciding whether to buy or not to buy the Inferno) - but when encoded to DNxHR, the PQ increase in video I'm getting from RAW SDI output of my XDCA has indeed been worth all this expenditures to me. And this PQ increase has most certainly been higher than 1.5% mentioned in the quote above!
Piotr
I would shoot a clip in internal XAVC-I both in slog2 and slog3 togheter with the recorders footage. My guess is that the recorder use slog2 with less raised blacks for a cleaner look. If you can post the clips untouched I would be happy. Then I could apply some of the conversion luts I usually work with and see what it looks like
Erik Wittbusch
Well-known member
As long as there's no side by side of the Shogun vs. internal SLog it's a shot in the dark for all of us except Piotr.
I'm happy to believe that there's some magic sauce involved in the Shogund recording that other manufacturers have missed before.
But why not let us see the actual difference?
I'm happy to believe that there's some magic sauce involved in the Shogund recording that other manufacturers have missed before.
But why not let us see the actual difference?