morgan_moore
Major Contributor
Same. And neither have I. Id go to the shop where it was bought and get a demo of it working.
S
FS7: FS7 and 7Q+ for Raw 4K DCI. Do I have a workflow problem.
- Thread starter Francois Zaidi
- Start date
Same. And neither have I. Id go to the shop where it was bought and get a demo of it working.
S
broughtonfilm
Well-known member
Yes, feels underexposed. Not the exposure I'd expect if I used 800 EI, especially that my understanding is you exposed "normal" for this rating (no underexposure for the scene done in camera), the face should be 1 1/3 stop overexposed (but it is small in the frame, so maybe it is and I just can't see it, plus I can only assume how bright/dark her skin tones are in real life). I was talking about my very first unscientific impression, don't worry about it. Cheers.
Paul schefz
Well-known member
i understand what you are saying and of course the DNG format can handle much more then the camera can throw at it..
but if what you are saying is correct then the histogram is completely useless since it is not a representation of the actual file.....
you can always pull an image (sometimes better, sometimes worse) out of a scrunched histogram but you are not even close to getting the best out of the sensor....
everything about this workflow (if it can even be called one) goes completely against anything i have ever seen or heard about raw workflow....and the fs7 XAVC files are the first non raw files i have ever worked with, all raw before that...
i would shoot the same scene 1, 2, 3 and 4 stops over and look at the file and histogram again....if what you are saying is correct: would the histogram change? if it does, i would like to see the same file probably about 2 stops over which should provide a full histogram and a much better file....
and i just looked at some XAVC files, the one shown looks at least a stop down to me.....
Nick_Shaw
Well-known member
If I remember correctly, with MLUT set to OFF in camera, your XAVC will always be at ISO 2000 (so you need to set 2.0 in Resolve). With MLUT set to S-Log3, it will be whatever you had it set to. But I do not have an FS7, and haven't tested with one since the firmware updates, so can't be certain.
The compensated LUTs are for use in the 7Q when shooting, so you can "overexpose" but see a normal image. Then you "print down" in post to reduce shadow noise.
ChristopherA
Well-known member
Thanks Nick; as always, your contributions are invaluable. I had overlooked your LUTs for RAW, thinking that they were only for adjusting out the WB bug.
Mitch Gross
Veteran
I think Nick has pretty well explained it here, so I'll just give a couple comments.
Given the end result image for which you say you like the look and comparing it to the pre-corrected ones (this is the vacuum cleaner frame), I would say that you are in fact a little underexposed, perhaps by a stop. One cannot rely purely on a lightmeter with any digital sensor until performing some tests to see how they inter-relate. I would say that based on how you seem to expose an image, you may want to shift the rating on the lightmeter to give yourself a "fatter file".
Unlike many other camera systems, the Sony FS7 outputs in a true linear RAW fashion. This will always yield an image that appears very underexposed to the eye as there is no "characteristic curve" applied to the contrast. For those unclear, here's what linear exposure looks like in terms of IRE levels:
100% Top exposure (just before clipping)
50% One stop down
25% Two stops down
12.5% Three stops down
...and so on. You can see that only the very top two stops of exposure are on the upper half of the exposure scale, with everything else piled down into the bottom. That is linear, where double the exposure (one stop increase) is represented by double the information. So the vast amount of information will be shoved down into the lower part of the exposure range. If we say that there are 14 stops of dynamic range on the FS7 (some argue this, but let's just drop that number in for now), that is two stops above the halfway point and 12 stops below it. Now you understand why the histogram appears as it does.
This is also why we do not simply increase exposure in post, but instead adjust the gamma curve. By addressing this (say a gamma of 2.2 instead of linear's 1.0), then the spread across the histogram becomes very different, with the mid range exposures lifted up while the top remains in place.
Not all RAW is the same and certainly not all RAW is linear. Sony is neither right or wrong in using linear. The Phantom high speed cameras use linear as well. From an engineering perspective it is the simplest form, but I will also note that it is as different as can be from what Canon does with the RAW data from the C500 and upcoming C300 mkII. Not better or worse, only different.
Given the end result image for which you say you like the look and comparing it to the pre-corrected ones (this is the vacuum cleaner frame), I would say that you are in fact a little underexposed, perhaps by a stop. One cannot rely purely on a lightmeter with any digital sensor until performing some tests to see how they inter-relate. I would say that based on how you seem to expose an image, you may want to shift the rating on the lightmeter to give yourself a "fatter file".
Unlike many other camera systems, the Sony FS7 outputs in a true linear RAW fashion. This will always yield an image that appears very underexposed to the eye as there is no "characteristic curve" applied to the contrast. For those unclear, here's what linear exposure looks like in terms of IRE levels:
100% Top exposure (just before clipping)
50% One stop down
25% Two stops down
12.5% Three stops down
...and so on. You can see that only the very top two stops of exposure are on the upper half of the exposure scale, with everything else piled down into the bottom. That is linear, where double the exposure (one stop increase) is represented by double the information. So the vast amount of information will be shoved down into the lower part of the exposure range. If we say that there are 14 stops of dynamic range on the FS7 (some argue this, but let's just drop that number in for now), that is two stops above the halfway point and 12 stops below it. Now you understand why the histogram appears as it does.
This is also why we do not simply increase exposure in post, but instead adjust the gamma curve. By addressing this (say a gamma of 2.2 instead of linear's 1.0), then the spread across the histogram becomes very different, with the mid range exposures lifted up while the top remains in place.
Not all RAW is the same and certainly not all RAW is linear. Sony is neither right or wrong in using linear. The Phantom high speed cameras use linear as well. From an engineering perspective it is the simplest form, but I will also note that it is as different as can be from what Canon does with the RAW data from the C500 and upcoming C300 mkII. Not better or worse, only different.
Francois Zaidi
Member
Hello Mitch
We actually did a key light test a week before to rate the lightmeter with the same FS7. Only we did not have the xdca and the 7Q by then (I rented those). Maybe we did it wrong.
So what is new on the july firmware (compensated LUT, what does it mean) ?
I do not currently have the calibrated screens and whatnot to really see that, but is my footage still OK ? I am mainly concerned about the noise in the black. In order to have a stop more, we would have had to push the lights on the set (I did not want to shoot wide open since it tends to be a lot softer at T1.5 or T3.1 depending on the optics), hence maybe have a less "natural" light. I don't know.
I have mixed feelings now. Good machine though (the 7Q).
We actually did a key light test a week before to rate the lightmeter with the same FS7. Only we did not have the xdca and the 7Q by then (I rented those). Maybe we did it wrong.
By that, you mean using Nick's LUT that are on your website, right ?
So what is new on the july firmware (compensated LUT, what does it mean) ?
I do not currently have the calibrated screens and whatnot to really see that, but is my footage still OK ? I am mainly concerned about the noise in the black. In order to have a stop more, we would have had to push the lights on the set (I did not want to shoot wide open since it tends to be a lot softer at T1.5 or T3.1 depending on the optics), hence maybe have a less "natural" light. I don't know.
I have mixed feelings now. Good machine though (the 7Q).
Francois Zaidi
Member
Maybe that is where I made the mistake. My understanding is that in cineEI mode, the camera is 2000ISO on its recording, no matter what. The rerating of 800EI concerns only the viewfinder, the scope and so on, making you in essence push one or two stops without seing it.
I don't remember if the 7Q saw that 800ISO setting or not, and we relied exclusively on the 7Q waveform monitor. So is it probable that instead of having 1,5 stop overexposed (80OISO to 2000ISO), we actually did not ?
And this is another hint of that...
Grug
Veteran
Check your raw settings in Resolve. I find that with Sony footage, it often defaults to seeing the footage at 800 ISO (which is a stop and third under what the camera is supposedly recording - 2000 ISO). If the raw ISO setting is at 800, just bump it up to 2000 ISO and go from there, you'll then be working with the actual exposure you set in camera.
broughtonfilm
Well-known member
First you said you used a light meter, now you say you "relied exclusively on the 7Q waveform monitor". That's a big difference...
Grug
Veteran
Was the 7Q waveform reading the LUTted image? Or the underlying raw footage? That could seriously mess with your exposure.
This is why I always recommend people stick to their meters for measuring exposure, it's fool-proof (a factor which counts for a lot these days when cameras insist on being so complex for no good reason).
This is why I always recommend people stick to their meters for measuring exposure, it's fool-proof (a factor which counts for a lot these days when cameras insist on being so complex for no good reason).
Mitch Gross
Veteran
I don't think your exposures were terribly low, just a little under where I would normally set them for such an image. In general I would rather give extra exposure to an image and then lower it more in post, providing that I knew it would always be going through a color correction anyway. Sony cameras in particular can be sensitive to noise in the lower exposed parts of the image, so overexposing a bit and then darkening in post will push that noise down as well, making it less apparent. I think your images are fine, but I think you might want to test exposing a little brighter in the future to see if you like it better. There's no right or wrong here, but such tests only really have value if you take them all the way through post correction to an end result.
This is effectively what the Exposure Compensation LUTs do. The RAW signal going to the Odyssey cannot be made lower than 2000ISO. That's just how the camera works. So if you overexpose the image by rating the camera at 1000ISO the picture would look rather bright on the Odyssey7Q+ when using the standard LUT. But if you use the one stop compensation LUT it darkens the image by one stop to balance for the overexposure you do by setting your light meter to 1000ISO. We provide two exposure compensation LUTs, one stop (1000ISO) and two stops (500ISO). It's simple enough to create your own if you wish, and Alister Chapman has a variety of such LUTs available for download on his website.
The purpose of the exposure compensation LUT is to allow the user to overexpose and then see what the image will look like once that overexposure is "corrected" in post. The point of overexposure is twofold. By lowering the brightness of all the image, you are also lowering the brightness of any noise in the image, thereby making it less visible. Also by overexposing you are enriching the functional bit-depth of the image. To explain briefly, the brighter the pixel the more bits are used to describe it. So if something in the shadows would normally only have two bits to delineate its color value, that is something very prone to banding. But if you overexpose the entire image then you are taking that same shadowy part of the image and now attributing four bits to it. When you darken it all down later it still has more information in it to begin with, and the part of the image with only two bits has now been pushed down into darkness. This improves color fidelity and lowers the chance of banding.
So overexposure reduces noise and reduces banding. It also reduces dynamic range and can lead to clipping in the highlights, so one must be careful about what is happening in the uppermost part of the exposure. But if a camera has 14 stops of dynamic range then there's a lot of room at the top to play with.
This is effectively what the Exposure Compensation LUTs do. The RAW signal going to the Odyssey cannot be made lower than 2000ISO. That's just how the camera works. So if you overexpose the image by rating the camera at 1000ISO the picture would look rather bright on the Odyssey7Q+ when using the standard LUT. But if you use the one stop compensation LUT it darkens the image by one stop to balance for the overexposure you do by setting your light meter to 1000ISO. We provide two exposure compensation LUTs, one stop (1000ISO) and two stops (500ISO). It's simple enough to create your own if you wish, and Alister Chapman has a variety of such LUTs available for download on his website.
The purpose of the exposure compensation LUT is to allow the user to overexpose and then see what the image will look like once that overexposure is "corrected" in post. The point of overexposure is twofold. By lowering the brightness of all the image, you are also lowering the brightness of any noise in the image, thereby making it less visible. Also by overexposing you are enriching the functional bit-depth of the image. To explain briefly, the brighter the pixel the more bits are used to describe it. So if something in the shadows would normally only have two bits to delineate its color value, that is something very prone to banding. But if you overexpose the entire image then you are taking that same shadowy part of the image and now attributing four bits to it. When you darken it all down later it still has more information in it to begin with, and the part of the image with only two bits has now been pushed down into darkness. This improves color fidelity and lowers the chance of banding.
So overexposure reduces noise and reduces banding. It also reduces dynamic range and can lead to clipping in the highlights, so one must be careful about what is happening in the uppermost part of the exposure. But if a camera has 14 stops of dynamic range then there's a lot of room at the top to play with.
morgan_moore
Major Contributor
Mitch do you have any videos/promo material showing FS7/back/7q raw working in any sensible manner?
Ive only heard of two raw shoots, both have ended on this board with (apparently) super thin files.
S
Ive only heard of two raw shoots, both have ended on this board with (apparently) super thin files.
S
Ben Scott
Veteran
I might shoot a new test this weekend as I was one of the moaners at the start about raw and pro res 7q issues and the latest updates really seemed to have improved everything to a point where it is useable again.
Will do do various resolutions and frame rates.
what I would say is that due to the nature of Sony's raw output Mitch's talk of overexposing or rather, exposing to the right, is of paramount importance. Sony appears to detest dark areas in raw.
Will do do various resolutions and frame rates.
what I would say is that due to the nature of Sony's raw output Mitch's talk of overexposing or rather, exposing to the right, is of paramount importance. Sony appears to detest dark areas in raw.
Francois Zaidi
Member
I mean I "relied exclusively on the 7Q waveform monitor" AS FAR AS WFM goes. I meant I did not use the FS7 WFM (just unusable as is).
So we :
1/ Took several reading with the light meter (that was calibrated with the camera (the speedbooster was a headache in itself))
2/ Set the T stop on the optics accordingly
3/ Checked the 7Q waveform.
I wish I had used 18% gray before each shoot to be on the safe side, but I did not.
We looked at both. You can toggle your WFM with or without LUT.
Well that is it. The camera was set at 800ISO, but I did not notice any overexposition. I assumed the 7Q would do as the sony monitor : adjust. So I was wrong about that. But how come did it not look overexposed ?
I guess my question is : does it make a difference for the 7Q if I rate the camera at 500ISO instead or 1250ISO ? Or the 7Q just "sees" 2000ISO no matter what ?
Yeah, I know ! What I was going for was a 1,5 stop overexposed. I end up with half a stop under
and this is making me nervous for noise and banding, especially with the linear distribution of sony RAW...
Nick_Shaw
Well-known member
Looking at the metadata in your posted DNGs the ISO is 2000. That is consistent with the exposure looking better at 2.0 with my LUT.
You are right that the RAW is the same on the FS7, whatever your ISO is set to. The ISO is just in the Sony metadata, which Resolve does not read. It does affect the viewfinder on the camera if you have an MLUT enabled, but not is you set MLUT OFF. I don't have an FS7 here, so can't confirm whether camera ISO setting affects the 7Q display or not. Mitch?
ChristopherA
Well-known member
Assuming that you are using the RAW output, changes in the Exposure Index (technically you can't change ISO) on the FS7 will not affect the image on the O7Q. If however, you are pulling SDI or HDMI out of the camera body, and you have an MLUT enabled for that output, changes in the ISO will make their way to the O7Q.
were you using a MLUT in the VF.. if you changed your EI to 800.. your VF should have got darker by 1.3 stops.. then you would have opened up the lens iris by the same.. so actually over exposing your recorded footage by 1.3 stops.. as you wanted to do.. or does RAW not work in the same way.. ?
Francois Zaidi
Member
I feel so stupid right now. Probably because I am. I read Alister Chapman's and dennis hindberg's guide many many times prior to this shoot and I still managed to forgut about the LUT requirement. I could kick myself.
So no wonder where my 1,5 stops of overexposure went.
Yep, I see that now.:huh:
So I just should have overexposed, or ETTR the old way. Damn.
I just understood a little more about everything. I hope my footage is gonna be OK though...
Last edited:
Francois Zaidi
Member
No, I did not. Hence the problem for xavc footage. For raw, you need to do it "manually" (ETTR)