FS7: New firmware for the FS7 will allow you to bake in gain/offset in Cine EI
- Thread starter starcentral
- Start date
morgan_moore
Major Contributor
Its just obvious - that we cannot dontrol scene DR and illumination level- we could be photographing the moon or a city scape or a theatre lit by someone else or.. anything!
Of course it is important not to be lazy and to be properly equipped and also manage client expectations.
(Yes corporal your film 'night raid exercise' - I might need to use/rent some specialst kit)
As it happens the Fs7 and a booster 400 2.8 (yes I own both) or suchlike is pretty up there as one of the best low light tool for theatres etc.
Of course it is important not to be lazy and to be properly equipped and also manage client expectations.
(Yes corporal your film 'night raid exercise' - I might need to use/rent some specialst kit)
As it happens the Fs7 and a booster 400 2.8 (yes I own both) or suchlike is pretty up there as one of the best low light tool for theatres etc.
Liam Hall
Veteran
I think you've misunderstood Sam's method here and how this new version of CineEI will work.
morgan_moore
Major Contributor
Not really a method - just conjecture about why this might be 'a good thing' in certain conditions (dark+low dr)
gkantziper
Esteemed Member
If the "bake in gain/offset in Cine EI" is the only addition in this update, than I'll pass for many, many months so everyone else can "beta" test, ha!
I'm a little confused by this. Will the native ISO still be locked at 2,000?
When you over expose S-Log by a stop, you are increasing the light falling on the sensor, and thereby improving the signal strength in relation to the noise. You bring it down in post and have a "cleaner" image.
But here, it seems as if they are actually allowing you to decrease the ISO by a stop, and reducing the noise but not increasing the signal. But dropping the ISO in half, by a full stop, also pulls down the DR. And of course, if you lower the ISO, it is baked in. Unfortunately, in my mind, I associate baked in with an MLUT, but is that what this is? A glorified LUT?
But that's not true if the ISO is still locked at 2,000, which, if that's the case, then I'm confused how there is less noise achieved. Decreasing ISO would lower noise, but if the ISO is locked, how does it achieve lower noise?
P.S. Does anyone have the link to AC's post? I can't seem to find.
When you over expose S-Log by a stop, you are increasing the light falling on the sensor, and thereby improving the signal strength in relation to the noise. You bring it down in post and have a "cleaner" image.
But here, it seems as if they are actually allowing you to decrease the ISO by a stop, and reducing the noise but not increasing the signal. But dropping the ISO in half, by a full stop, also pulls down the DR. And of course, if you lower the ISO, it is baked in. Unfortunately, in my mind, I associate baked in with an MLUT, but is that what this is? A glorified LUT?
But that's not true if the ISO is still locked at 2,000, which, if that's the case, then I'm confused how there is less noise achieved. Decreasing ISO would lower noise, but if the ISO is locked, how does it achieve lower noise?
P.S. Does anyone have the link to AC's post? I can't seem to find.
Last edited:
morgan_moore
Major Contributor
Follow this. It might make sense.
To get a less noisy in normal Slog you expose right and pull down in post (doing this does risk clipping the highlights)
Consider a dark scene where you have opened the iris as wide as you dare/can.
With NormalSlog Your brightest point is nowhere near clip. Your mids are maybe sitting, non ideally at 20. Your full image range from 10-50 (a thin exposure!)
You take this into post and boost the exposure to where you want, your mids gain noise as you lift them.
Now the quality of the file is not just Signal to Noiise, but also compression artifacts.
In XAVC compression artifacts are harsher in the low zones, say 0-30 (because artifacts in shadow areas are not considered bad in baked codecs and compression is all about throwing away 'unimportant' data)
Your NormalSlog exposure was not only noisy but full of compression artifacts.
Now Engage ISOSlog. You make the same exposure but the 'ISO boost' writes the data not at 10-50 but at 30-80
Take that file to post and you still have noise but have lost some compression artifacts - which is good ?
--
Any wide DR scene you will be chasing your tail I think as the ISO boost will lift the highs forcing you to stop down, to avoid clip and pushing your shadows back into the highly compressed are of the file. Gain in image quality? None.
To get a less noisy in normal Slog you expose right and pull down in post (doing this does risk clipping the highlights)
Consider a dark scene where you have opened the iris as wide as you dare/can.
With NormalSlog Your brightest point is nowhere near clip. Your mids are maybe sitting, non ideally at 20. Your full image range from 10-50 (a thin exposure!)
You take this into post and boost the exposure to where you want, your mids gain noise as you lift them.
Now the quality of the file is not just Signal to Noiise, but also compression artifacts.
In XAVC compression artifacts are harsher in the low zones, say 0-30 (because artifacts in shadow areas are not considered bad in baked codecs and compression is all about throwing away 'unimportant' data)
Your NormalSlog exposure was not only noisy but full of compression artifacts.
Now Engage ISOSlog. You make the same exposure but the 'ISO boost' writes the data not at 10-50 but at 30-80
Take that file to post and you still have noise but have lost some compression artifacts - which is good ?
--
Any wide DR scene you will be chasing your tail I think as the ISO boost will lift the highs forcing you to stop down, to avoid clip and pushing your shadows back into the highly compressed are of the file. Gain in image quality? None.
I am following what you are saying until the "ISO boost" part.
I am confused because this doesn't seem to be an ISO boost, but an ISO decrease. I've always understood that decreasing ISO decreases noise, but with CINE-EI, the ISO was locked, so decreasing was never an option. I can understand an ISO boost that is rewriting values to a better range and decreasing compression artifacts.
But the whole EI Applied thing seems to be an ISO decrease. Are you saying that the ISO decrease is rewriting values somehow and not actually decreasing amplification of the sensor's signal?
Just trying to understand.
Thanks!!
I am confused because this doesn't seem to be an ISO boost, but an ISO decrease. I've always understood that decreasing ISO decreases noise, but with CINE-EI, the ISO was locked, so decreasing was never an option. I can understand an ISO boost that is rewriting values to a better range and decreasing compression artifacts.
But the whole EI Applied thing seems to be an ISO decrease. Are you saying that the ISO decrease is rewriting values somehow and not actually decreasing amplification of the sensor's signal?
Just trying to understand.
Thanks!!
morgan_moore
Major Contributor
in 'normal slog' one could 'change the ISO' by 'over exposing' and pulling back in post leading to a less noisy file with less highlight headroom than the same scene exposed as per sony grey card guidance ('correctly exposed/correct ISO')
cyvideo
Veteran
Is it maybe easier to put it this way? My explanation may be as clear as mud but here goes. I think I'm right.
Rate the camera in Cine EI at let's say 1000 ISO. Remember in EI the camera is always operating at 2000 ISO regardless of what lower ISO you rate the sensor at. Currently at 1000 when you look at at the image it looks dark. It is dark because the image is under exposed by one stop. So what do you do? You adjust exposure until it looks correct on your Waveform, VF, monitor or whatever. But by doing this you have now over exposed the sensor by one stop above the correct exposure for 2000 ISO. How do you correct this? In post your reduce the over exposure by one stop to bring your levels back to the correct exposure.
From what I am understanding, hopefully correctly, from the info shown in the first post is that you can rate the camera at let's say 1000 but the image won't look dark in the viewfinder, it will look like a correct exposure for native 2000 ISO. In fact the camera will over expose by one stop but as the file is recorded in camera this one stop over exposure will be adjusted back down as the file is saved. At the same time reducing noise levels by one stop. For editors not used to handling over exposed LOG I think this will be less confusing for them. The files will look to have been exposed correctly.
Am I on the right track?
Chris Young
Rate the camera in Cine EI at let's say 1000 ISO. Remember in EI the camera is always operating at 2000 ISO regardless of what lower ISO you rate the sensor at. Currently at 1000 when you look at at the image it looks dark. It is dark because the image is under exposed by one stop. So what do you do? You adjust exposure until it looks correct on your Waveform, VF, monitor or whatever. But by doing this you have now over exposed the sensor by one stop above the correct exposure for 2000 ISO. How do you correct this? In post your reduce the over exposure by one stop to bring your levels back to the correct exposure.
From what I am understanding, hopefully correctly, from the info shown in the first post is that you can rate the camera at let's say 1000 but the image won't look dark in the viewfinder, it will look like a correct exposure for native 2000 ISO. In fact the camera will over expose by one stop but as the file is recorded in camera this one stop over exposure will be adjusted back down as the file is saved. At the same time reducing noise levels by one stop. For editors not used to handling over exposed LOG I think this will be less confusing for them. The files will look to have been exposed correctly.
Am I on the right track?
Chris Young
^ Chris - that's a good explanation.... but until we hear more from Sony - or someone in the know - your guess is as good as mine. I assumed, perhaps wrongly, because of the term "baked" - that this would involve burning in an offset type lut to internal recordings. But I hope you are right - that this is still shooting Slog - just passing along the "intended metadata" (if that makes sense).
As an often FCPX user - or if handing off footage to editor who works the system that reads the metadata, if this metadata is passed along (as it is now) - then it would make very quick rough cuts very very fast - since FPCX applies Slog to Rec709 Lut automatically.
As an often FCPX user - or if handing off footage to editor who works the system that reads the metadata, if this metadata is passed along (as it is now) - then it would make very quick rough cuts very very fast - since FPCX applies Slog to Rec709 Lut automatically.
cyvideo
Veteran
I hope I'm right as that would work nicely for me. As you say though we are all 'awaiting the news from the big four letter word company.
Chris Young
If this functionality ends up just baking in an exposure offset lut – can't you do the same by baking in the P5:S-Log3 LUT in the current firmware, and then setting to an EI other than 2000?
I've considered doing this to rate the camera below 2000, but skip the exposure offset step in post. I monitor off of a 502/702, so I can drop a 709 LUT in there when necessary and free up the camera's MLUT feature for this. I'm sure I'm missing something?
I've attempted to search for anyone else talking about this but "slog3 to slog3 lut" is not very googleable.
I've considered doing this to rate the camera below 2000, but skip the exposure offset step in post. I monitor off of a 502/702, so I can drop a 709 LUT in there when necessary and free up the camera's MLUT feature for this. I'm sure I'm missing something?
I've attempted to search for anyone else talking about this but "slog3 to slog3 lut" is not very googleable.
Last edited:
chappelldigital
Well-known member
Had thought through that before myself as well. Didn't know one was already available (P5:S-Log3 LUT ?) - but simply generating a LUT with no transformation to anything and baking it in (just straight Slog3) would cause the EI changes to be recorded, would it not?
starcentral
Veteran
Remember the LUT you bake in is only 33 matrice, and not going to be the same quality as what the camera does to offset gain internally. You can bake in a REC709 LUT and it's one thing, but to bake in a slog curve and then still having to grade it later can lead to quality problems
Cheers
cameramanben
Well-known member
A 33 point LUT would be plenty for an exposure adjustment of log material. The differences between before and after are smaller and more consistent than with the big colour and tone curve changes usually applied in LUTs, so rounding errors from linear or tetrahedral interpolation (the kinds likely done in camera with a 3D LUT) should be insignificant to the final result. It certainly can't hurt to try it out. What issues have you come across Dennis?
This is not to say I'm advocating baking-in exposure adjusted log LUTs - I don't really have a particularly strong opinion myself (other than not wanting to view log in the viewfinder!),
Ben
This is not to say I'm advocating baking-in exposure adjusted log LUTs - I don't really have a particularly strong opinion myself (other than not wanting to view log in the viewfinder!),
Ben
Last edited:
It does sound to me like this is like baking in an exposure adjusted SLog3 LUT. However the difference would be with the new CINE EI you still can use a 709A or similar LUT for monitoring or even for baking in if you like, but if you used an adjusted SLOG-3 LUT you've taken that option away.