Watch the First Sony F3 S-Log Footage!!

[TimurCivan]

http://tinyurl.com/3g8vzlh

My Good friends at Next Level pictures, Khalid Mohtaseb, Jon Bregel and I shot the hell out of some Slog out at NAB!!!

Enjoy!

Visit their site as well! www.nextlevelpictures.com

[plasmasmp]

slog looks fantastic. love how versatile the F3 can be and I'm so jealous you got to use all those toys.

[hunter richards]

OMG- looks like a completely different camera. Color looks strait as an arrow and the dynamic range looks like alexa. What is it now like 13.5+ stops?

[TimurCivan]

My estimate is over 14. its in Alexa Territory.

[hunter richards]

14 stops!? Thats madness!

Timur: Any idea if you can record single link 1.5g to a 10-bit 4:2:2 codec recorder like prores or will you have to do dual link 4:4:4?
I was kinda hoping to be able to just record 10-bit 4:2:2 to prores like a kipromini or something but still use s-log.

[mcgeedigital]

That is ridiculous.

Of course, the one time I DON'T go to NAB.

[Jason Davenport]

Great job! Thanks for making the effort. Great info.

[nyvz]

Thanks for sharing this, very nice to see how much latitude you can really get out of an F3.

It is possible to record in S-Log (which is simply a gamma mode from what I understand) to XDCAMEX/SxS, right?

The F3 upgrade also includes the ability to use custom LUTs in camera right? Did you get to explore that feature at all?

Timur, with your new experience with the Slog option on the F3, do you know why Sony clips 2-3stops of highlights on all of their non-Slog Gamma modes? Juan Martinez from Sony tried to explain the situation to me and confirmed that not using Slog causes the camera to throw away about 3stops of highlight latitude, but while explaining why he got a bit frustrated and I never really understood his explanation. He mentioned something about it being a limitation of gamma-corrected recording in general. He seemed to say that S-log was the only way, but I really do not quite understand why it would not be possible to record close to the full signal range of the sensor with a normal cine-gamma or even standard-gamma curve. I mean if you can arbitrarily gamma-compress 8-11 stops into a 0-255 (8bit per channel) range, why couldn't you just fit 13-14 stops without clipping 2stops of range?

From what I understand, Alexa, which has a LogC gamma mode, is also able to record most of its range (13-15stops of latitude) even in its REC709 gamma mode. And of course for decades, it has been possible to watch movies shot on film with rather wide latitude on 8bit distribution formats and have them look beautiful, so it hardly seems there should be a technical limitation of actually fitting the full sensor latitude into even the smallest container. Does anyone have an F3 and have any idea if there is some way, perhaps through custom picture profiles, to unlock some of that highlight information that Slog unlocks without actually recording in Slog?

I hate to sound like a conspiracy theorist, but it's hard not to think Sony crippled the F3 somewhat by throwing away the brightest 2-3stops of information in all the gamma curves it ships with. Is there any reason this would have been out of their hands, or is it more likely that it is simply a marketing move to increase the sales of the Slog upgrade?

[PDR]

Thanks for posting Timur, looks great !

Can you or Tom make some comments on using S-Log in Davinci ? Any special considerations or "easy-peasy" ?

Any remarks on noise levels when using S-log ?

[Matthew Bennett]

Timur you're living the dream!!! Thanks for showing us this!

[Sage]

At last! This cam is interesting now.

[macgregor]

Very interesting!
1. can the slog be recorded internally?
2. colours look much nicer, but I´m not impressed with DR. Those blown out highlights when the girl is in front of the window look ugly. Was it correclty exposed? Because when switching to XDcam mode it´s clearly overexposed which tells me the aperture might need to be closed 1-2 stops... XD Remmember in Slog the mid gray goes way lower in the waveform!!!

[Noel Evans]

Hot stuff Timur - commented on FB. S-Log is looking like it will deliver what we all thought it COULD be which is great.

[hunter richards]

Mac- I agree all the xdcam stuff looks a bit over exposed- that shot especially looks over- but its probably very good news- none of these guys have ever experienced the f3 with s-log before so im guessing once people figure out how to expose it correctly- its going to be even more amazing looking- there is so much info in the blacks of the f3 its crazy enough already.

[hunter richards]

Maybe using the ire spot meter and setting middle grey at around 37-38 based on s-log white paper would give best results for exposure?

You were not able to see s-log gamma on the on board camera's flip out lcd right Timur?

[tom.wong]

I don't believe a 8 bit 4:2:0 codec can hold enough information to give you any usable information s-log can provide. remember you are cramming 800% worth of luminance information (0-100 in rec709) into a 100% space. 8 bit won't be able to hold it.

yes, some parts were over exposed on the s-log, those guys torture tested the camera. the found the breaking point too, so it's not the say the s-log is impervious to overexposure, but it can def hold a solid stop and half to 2 stops more than the internal recording.

[Jarrett P. Morgan]

slog looks fantastic. love how versatile the F3 can be and I'm so jealous you got to use all those toys.

exactly what I thought, haha

[hunter richards]

If anything seeing the breaking point here on the s-log makes me more hopeful about it than anything, I was very skeptical about a $3k+ upgrade, but this seems like the real deal.

[Fohdeesha]

14 stops!? Thats madness!

Timur: Any idea if you can record single link 1.5g to a 10-bit 4:2:2 codec recorder like prores or will you have to do dual link 4:4:4?
I was kinda hoping to be able to just record 10-bit 4:2:2 to prores like a kipromini or something but still use s-log.


yup, you can, with dual-link output enabled. It's part of the SMPTE 372M spec. Just make sure it's a 10-bit codec or it's going to look like butt. logarithmic curves and 8-bit do NOT mix well. the slight chroma subsampling of 4:2:2 however has minimal, if any impact

Also, Awesome test Timur!

[Nate Weaver]

I don't believe a 8 bit 4:2:0 codec can hold enough information to give you any usable information s-log can provide. remember you are cramming 800% worth of luminance information (0-100 in rec709) into a 100% space. 8 bit won't be able to hold it.

Just my opinion, I think it's a little misleading to say 8 bit "can't hold all that info". You as many stops as you want to 8 bit..it just might look pretty flat. And if you want to get it back to a normal looking contrast, those 8 bits are going to posterize when you "stretch it out".

So technically you're right...I just struggled with that phraseology you used a long time ago because it confused me, and wanted to comment in case other folks are a little murky on the subject as well.

[Nate Weaver]

On another note, that LUT used to record to SxS cards looks like it needs to be modded, because otherwise, if the editor and director are making decisions based on SxS proxies, they might toss some shots because of "overexposure".

I'd say you would just tell them to not do that, but people have been doing silly things with Red footage based on proxies and transcodes for 2 years now.

[nyvz]

I don't believe a 8 bit 4:2:0 codec can hold enough information to give you any usable information s-log can provide. remember you are cramming 800% worth of luminance information (0-100 in rec709) into a 100% space. 8 bit won't be able to hold it.

yes, some parts were over exposed on the s-log, those guys torture tested the camera. the found the breaking point too, so it's not the say the s-log is impervious to overexposure, but it can def hold a solid stop and half to 2 stops more than the internal recording.

Does that mean you cannot record s-log gamma footage on XDCAMEX or just that you don't think you should? So if you had the choice between fitting 14 stops of range into 8bits and clipping out 3 stops to fit 11stops into 8bits, you'd clip off 3stops? Is it that you think 14stops will be too low contrast in most situations and adding contrast might result in greater risk posterization than starting out with higher contrast, and that that less risk of visible posterization is more important than having high latitude to fit high contrast shooting situations?

As for overexposure, it does seem that the fact that S-log unlocks about 3stops of highlight detail may mean that shooting non-Slog XDCAMEX and Slog externally simultaneously would be impractical since S-log does not expand the camera's latitude evenly, only on the top end. If you expose for highlights in Slog, youll always clip your XDCAMEX highlights by nearly 3stops, and if you expose for XDCAMEX, likely underexpose your Slog. Then again shooting for XDCAM and getting 3stops of extra highlight headroom by recording Slog alongside isnt a bad deal, but it may not be that pleasing for some people who are especially sensitive to noise.

Is the XDCAM material shot simultaneously shot with a LUT applied to the Slog, or just with its own separate picture profile (and cinegamma or something).

[Fohdeesha]

You will not be able to record s-log to XDCAM. And yes, even if you could, it would look posterized enough after correction to warrant not doing it in the first place. Just grab a kipro or a PIX 240 or something :)

edit: just remembered...does anyone remember if sony said you would able to load custom LUTS? if you could, you could just load a completely flat LUT dumping the uncorrected s-log to xdcam. But like I and Nate and several others have said, s-log in 8-bit space looks like butt. It's not directly related to the amount of dynamic range you're storing, it's related to how the luma is weighted and stored on the response curve, and the amount of correction it requires to get it back to rec.709. 8-bit, rec.709 formats like xdcam do not take kindly to that amount of correction, and that amount of detail being stored below the midtones

[TimurCivan]

Thanks for sharing this, very nice to see how much latitude you can really get out of an F3.

It is possible to record in S-Log (which is simply a gamma mode from what I understand) to XDCAMEX/SxS, right?

The F3 upgrade also includes the ability to use custom LUTs in camera right? Did you get to explore that feature at all?

Timur, with your new experience with the Slog option on the F3, do you know why Sony clips 2-3stops of highlights on all of their non-Slog Gamma modes? Juan Martinez from Sony tried to explain the situation to me and confirmed that not using Slog causes the camera to throw away about 3stops of highlight latitude, but while explaining why he got a bit frustrated and I never really understood his explanation. He mentioned something about it being a limitation of gamma-corrected recording in general. He seemed to say that S-log was the only way, but I really do not quite understand why it would not be possible to record close to the full signal range of the sensor with a normal cine-gamma or even standard-gamma curve. I mean if you can arbitrarily gamma-compress 8-11 stops into a 0-255 (8bit per channel) range, why couldn't you just fit 13-14 stops without clipping 2stops of range?

From what I understand, Alexa, which has a LogC gamma mode, is also able to record most of its range (13-15stops of latitude) even in its REC709 gamma mode. And of course for decades, it has been possible to watch movies shot on film with rather wide latitude on 8bit distribution formats and have them look beautiful, so it hardly seems there should be a technical limitation of actually fitting the full sensor latitude into even the smallest container. Does anyone have an F3 and have any idea if there is some way, perhaps through custom picture profiles, to unlock some of that highlight information that Slog unlocks without actually recording in Slog?

I hate to sound like a conspiracy theorist, but it's hard not to think Sony crippled the F3 somewhat by throwing away the brightest 2-3stops of information in all the gamma curves it ships with. Is there any reason this would have been out of their hands, or is it more likely that it is simply a marketing move to increase the sales of the Slog upgrade?

Its more to do with the way the camera deals with the masive data rate of the sensor. im not an engineer. Its difficult to explain. But all i know is that it works!


Very interesting!
1. can the slog be recorded internally?
2. colours look much nicer, but I´m not impressed with DR. Those blown out highlights when the girl is in front of the window look ugly. Was it correclty exposed? Because when switching to XDcam mode it´s clearly overexposed which tells me the aperture might need to be closed 1-2 stops... XD Remmember in Slog the mid gray goes way lower in the waveform!!!

Theoretically it can be recorded to SXS, but you wouldnt want to do that. You will be compressing 14 stops into 256 gradations. that leaves like 15 gradations per stop. The sLOG is VERY flat. When you put contrast back into it, you would get severe banding in the mids. While we could have exposed for the highlights, that was overcast sky with the sun directly backlighting it. It was like 10 stops over. Nothing would hold that save the EPIC HDRX and thats a maybe.... I always expose for skintone as a priority. and she was already 2 stops under.


Does that mean you cannot record s-log gamma footage on XDCAMEX or just that you don't think you should? So if you had the choice between fitting 14 stops of range into 8bits and clipping out 3 stops to fit 11stops into 8bits, you'd clip off 3stops? Is it that you think 14stops will be too low contrast in most situations and adding contrast might result in greater risk posterization than starting out with higher contrast, and that that less risk of visible posterization is more important than having high latitude to fit high contrast shooting situations?

As for overexposure, it does seem that the fact that S-log unlocks about 3stops of highlight detail may mean that shooting non-Slog XDCAMEX and Slog externally simultaneously would be impractical since S-log does not expand the camera's latitude evenly, only on the top end. If you expose for highlights in Slog, youll always clip your XDCAMEX highlights by nearly 3stops, and if you expose for XDCAMEX, likely underexpose your Slog. Then again shooting for XDCAM and getting 3stops of extra highlight headroom by recording Slog alongside isnt a bad deal, but it may not be that pleasing for some people who are especially sensitive to noise.

Is the XDCAM material shot simultaneously shot with a LUT applied to the Slog, or just with its own separate picture profile (and cinegamma or something).

All picture profiles are disabled in SLOG. you get rec 709 only.... ( in this pre release version) You can do two things, expose for REC709, and get amazing highlight protection, or expose for Slog, and build a LUT that gives you a rough "grade" that shows some contrast, but still a majority of what SLOG is seeing.

[tom.wong]

Just my opinion, I think it's a little misleading to say 8 bit "can't hold all that info". You as many stops as you want to 8 bit..it just might look pretty flat. And if you want to get it back to a normal looking contrast, those 8 bits are going to posterize when you "stretch it out".

So technically you're right...I just struggled with that phraseology you used a long time ago because it confused me, and wanted to comment in case other folks are a little murky on the subject as well.

Sorry typing on phone on a shaky bus, its been a whirl wind few days getting this thing together. Been speaking caveman all day. So what u said, and fodeesha. Even if sony did let u record log to xdcam, the results wouldnt be good. The posterization alone would make the image unusable... unless thats what u r going for.

Working with the slog was great. Only thing that felt different when compared to c log in pro res was that alexa held a little more in the highlights, and the low con filter in alexa is still nicer on skin. But when u compare the noise floors... u should ask timur what iso they were shooting at the whole time. When the told me my jaw dropped.

[TimurCivan]

:evil:

[Nate Weaver]

The posterization alone would make the image unusable... unless thats what u r going for.


I wouldn't ever be going for "posterization", but what I *might* go for sometimes is a really, really flat looking image with lots of info. Just a low-con look in general. 8 bit would be fine for that. Can get close in the Cinegammas though, so it's kinda academic.

[tom.wong]

I wouldn't ever be going for "posterization", but what I *might* go for sometimes is a really, really flat looking image with lots of info. Just a low-con look in general. 8 bit would be fine for that. Can get close in the Cinegammas though, so it's kinda academic.

of course, and on the reverse side, i'd still prefer a higher quality acquisition so i can really punch in subtleties. super flat can look really awesome depending on the scene. I find it works really great with comedy. it's always about, how dedicated you are to, what you get out of camera is what you stick by. It's a mixed bag, I get some DP's that come into the color grade session, a year after the movie was shot, and they are horrified by what they did a year ago, and wanna go a whole new direction with the film, and others are really hardcore about what they went after in camera, and just need it spit shined.

[dsleep]

yup, you can, with dual-link output enabled. It's part of the SMPTE 372M spec. Just make sure it's a 10-bit codec or it's going to look like butt. logarithmic curves and 8-bit do NOT mix well. the slight chroma subsampling of 4:2:2 however has minimal, if any impact

Also, Awesome test Timur!

I have heard conflicting things about using S-log with a prores 4:2:2 codec. It would be amazing if we can use it with a ki pro mini or one of the new SD recorders.

Timur can you say any more about the upgrade's effect on sensitivity/low light performance? that is very important to me and the s-log upgrade is worth it for that reason alone IF it works with Pro res 4:2:2.

really appreciate you guys doing this!

[tom.wong]

I have heard conflicting things about using S-log with a prores 4:2:2 codec. It would be amazing if we can use it with a ki pro mini or one of the new SD recorders.

Timur can you say any more about the upgrade's effect on sensitivity/low light performance? that is very important to me and the s-log upgrade is worth it for that reason alone IF it works with Pro res 4:2:2.

really appreciate you guys doing this!

log in 10 bit 422 is def possible with no issues other than it's 422. C-Log off Alexa to Pro Res HQ onto cinedeck. worked great.

[Fohdeesha]

Not sure what reports you have heard, but you will be able to capture it to a 10-bit 4:2:2 codec perfectly fine.

[LoganMackay]

Looks insane! Thanks for sharing Timur.

[dsleep]

thanks Tom and Fohdeesha. Now could one of you explain to my wife why I must have this camera?

the F3 is really impressive, especially with the upgrade.

[hunter richards]

I do a redworkflow with log to prores HQ (which is 10-bit 4:2:2) all the time with out issue so I was hoping it would be ok on the f3.

[Fohdeesha]

Yup.

Just watched the whole thing. Man you can see the difference in highlight handling and proper midtone/skintone handling like none other, night and day difference in a 2mbps vimeo stream. Log is so nice.

*puts SEVERAL more dollars into f3 pot*

[HHL]

So much for the "why would anyone buy an F3 when the FS100 looks so good" argument. :evil:

[HHL]

Geez....the footy LITERALLY looks like it came from 2 different cams. WOWSER!!!

[dcloud]

Wow. That looked great. Even at this compression level you ciuld see the big difference.

I wonder if panny will come out with one next year?

[youngindiefilms]

I just wet my pants lol. Me wants shoot on F3 too, had my hands on one at a exhibition recently and now seeing that it's capable of so much more, it's brilliant.

[Ryan Patrick O'Hara]

You had me at:


Arri Master Prime: 25mm T1.3
Arri Ultra Prime: 8R 8mm T2.8
Cooke Panchro 18mm T2.8
Cooke Panchro 100mm T2.8
Angenieux DP Rouge 16-42 T2.8
Angenieux DP Rouge 30-80 T2.8
Angenieux Optimo 45-120 T2.8


That elusive 8R... I've tried to rent it a few times. By gone, so expensive and hard to find. I shall use it someday.

[TimurCivan]

The 8R is magical.

[Lee Saxon]

I'm not buying a $5k Gemini and lugging that around too.

Wake me when this is the onboard/native recording format of the camera.

[Ryan Patrick O'Hara]

I'm not buying a $5k Gemini and lugging that around too.

Wake me when this is the onboard/native recording format of the camera.


“It behooves every man to remember that the work of the critic, is of altogether secondary importance, and that, in the end, progress is accomplished by the man who does things.”

-Theodore Roosevelt

[Nate Weaver]

I'm not buying a $5k Gemini and lugging that around too.

Wake me when this is the onboard/native recording format of the camera.

Ring ring ring. It's the Alexa. You can spend $100 big ones now for that, or lug around a little black box (Hyperdeck, Samurai) for the F3 and get 90% of the same.

I think it will be a LONG time before you can do 220mbit 4:2:2 (internally, SR-R1 doesn't count) on a Sony product for less than 60 grand.

[avro]

i'm not buying a $5k gemini and lugging that around too.

Wake me when this is the onboard/native recording format of the camera.

+1 & evf

[avro]

http://blog.vincentlaforet.com/2011/04/19/nab-redux-pt-1-cameras/?utm_source=feedburner&utm_medium=email&utm_campaign=Feed%3A+wordpress%2FjvaG+%28Vincent+L aforet%27s+Blog%29

[Fohdeesha]

I'm not buying a $5k Gemini and lugging that around too.

Wake me when this is the onboard/native recording format of the camera.

then buy a tiny 1.5k Samurai, a 2k kipro mini, a 2.5k SD PIX 240, etc.......

[macgregor]

Mac- I agree all the xdcam stuff looks a bit over exposed- that shot especially looks over- but its probably very good news- none of these guys have ever experienced the f3 with s-log before so im guessing once people figure out how to expose it correctly- its going to be even more amazing looking- there is so much info in the blacks of the f3 its crazy enough already.


That was exactly my initial reaction. Hey, if the XDcam footage wich is 709 looks clearly overexposed in all or most shots that means they missed the right exposure. The Slog footage looks much nicer when graded basically because it retains more DR so they were able to recover more info in the highlights (Ansel Adams zone system at its maximum!). This tells me that with proper exposure (more balanced --> setting the mid gray lower in the waveform as it should be with Slog) this camera seems very very very attractive.

Last month I shot a commercial where the main camera was a F35 and we got the F3 for B-roll. Nothing we shot with the F3 made it into the final cut and the F35 was clearly better than the F3, but the F3 was in 709 by that time of course. This seems to change things.

[Fohdeesha]

They were exposing correctly for the log format by using a waveform monitor, nothing was "overexposed" unless they wanted it to be. You can't base proper log exposure by looking at a rec.709 proxy of it, the gamma response is nowhere near the same, and due to it's luma weighting it will chop off tons of info at the top. this is why with log you should NEVER just expose by what you see on any monitor.

[macgregor]

Theoretically it can be recorded to SXS, but you wouldnt want to do that. You will be compressing 14 stops into 256 gradations. that leaves like 15 gradations per stop. The sLOG is VERY flat. When you put contrast back into it, you would get severe banding in the mids. While we could have exposed for the highlights, that was overcast sky with the sun directly backlighting it. It was like 10 stops over. Nothing would hold that save the EPIC HDRX and thats a maybe.... I always expose for skintone as a priority. and she was already 2 stops under.




Hi Timur! Were you able to try this? Of course 8bit is 4 times less information than 10 bits but for the situations where you dont want or can´t take an external recorder it might be ok.

Let´s do a quick test to see if we get any banding:

I´ve taken a log arriscanner frame shot on 35mm with a high DR lighting situation. The log files from a film scanner are not mapped in the same exact way as Sony's Slog but the principle is the same: being able to store a high dynamic range image with no banding by the use of a gamma curve that distributes the information of each pixel evenly.

This is a 10bit log DPX file saved as jpg without any correcting LUT:

http://dl.dropbox.com/u/13362047/log.jpg

And this is the same 10bit uncompressed DPX color corrected (just a quick levels adjusment) at 16bit depth. No banding obviously:

http://dl.dropbox.com/u/13362047/10bit%20correction.jpg

Now, let´s save the original log DPX as jpg (JPG is 8bit and introduces compression artifacts as we all know. JPG quality 12/12). Then lets open that 8bit JPG and apply the same grading as the previous DPX. And this is what we get. Hard to distinguish any differences:

http://dl.dropbox.com/u/13362047/jpg8bitcorrection.jpg

Now, you could say, XDCAM compression and final image quality is worse than a 12/12 jpeg image. OK, let´s do the same thing but degrade the JPG quality to 9/12. Here you go:


http://dl.dropbox.com/u/13362047/jpg8bitcorrectionquality9.jpg
I don´t see any banding here. What I do start seeing is compression artifacts in the shadows/lower stops. But that would happen with 709 footage too if it needed some grading..


So, this tells me that Slog straight to XDCAM might be at least worth a try and could make eventually better images than shootting in 709.

[macgregor]

They were exposing correctly for the log format by using a waveform monitor, nothing was "overexposed" unless they wanted it to be. You can't base proper log exposure by looking at a rec.709 proxy of it, the gamma response is nowhere near the same, and due to it's luma weighting it will chop off tons of info at the top. this is why with log you should NEVER just expose by what you see on any monitor.



Let me disagree with the first part of your post. If you were correctly exposing for Slog, the 709 footage would have to look just ok or darker since you are mapping the mid gray into a lower zone (that´s why sony claims an effective increase in sensitivity when using Slog, becuase you are mapping the mid grey into a lower zone and then raising it back with a gamma curve). But never brighter as it is the case in this test.

[tom.wong]

keep in mind guys, the guys behind this only had 24 hours to get everything done. they didn't start shooting until 2 am in the morning, going into the afternoon, and having to return everything asap that they were allowed to borrow. 709 was the only gamma they can view in the camera, and the decision was just to expose for skin tone in 709 and then see what the s-log produced. the goal wasn't to set the exposure perfectly for log at all, they wanted to give xdcam the best fighting chance they kid, and then compare it to the s-log.

that said, i'm STILL surprised that not a single person has even questioned what ISO's they were shooting it. because people are forgetting really fast, and I think neglecting the fact that the F3 is truly the low light king of the industry. Not a single camera can touch it in the high ISO range, save scientific cameras. Even if the Alexa completely destroys the F3, going head to head in log in highlight retention, the noise floor and range in the darks are literally several times better than the Alexa. Alexa's current maximum ISO setting, is the F3's new native rating with log. 1600 iso. you can easily expose for your highlights, and bring up your darks in post. DP's normally do that for HD, but end up putting a lot of fill in to balance out the lows, mids, and highlights. but the girl in front of the window for example in the beginning, had NO fill. not even a bounce card. i brought her up at least a stop and a half with no visible noise without pixel peeping.

that's some food for thought right there.

[eddavid]

all is good in the world. but respect must be given to MacGregor - he is the reason most of us are here right now. What he did with the original DVX100 and lens adapters help change this industry globally.

[Danielvilliers]

The overexposure of the xdcam footage is the only problem as a proper comparison between the two. I really like the s-log footage on it's own because the xdcam footage seems to be so overexposed that it is no use as a comparison tool. Most digital camera tend as the base F3 to have more detail in the shadows than highlight. So need to be underexposed rather than overexpose to protect the highlight and then boost up in post. I am not criticizing here because I know about the time to do the shoot and the test of what is a new camera in the slog F3.

[macgregor]

The overexposure of the xdcam footage is the only problem as a proper comparison between the two. I really like the s-log footage on it's own because the xdcam footage seems to be so overexposed that it is no use as a comparison tool. Most digital camera tend as the base F3 to have more detail in the shadows than highlight. So need to be underexposed rather than overexpose to protect the highlight and then boost up in post. I am not criticizing here because I know about the time to do the shoot and the test of what is a new camera in the slog F3.


Agreed. We must be thankfull to the guys for their test and comparing both recroding formats since it´s very illustrative.

[Shipsides]

This was a great test and really shows what S-Log is capable of in my opinion. I plan to do another Dynamic Range test of the F3 once I have firmware in house.

Great job to all the guys on the team and to Tom Wong for the killer color correction.

[Sam Scoggins]

Very nice test. Well done to everyone concerned for getting it together.

So, I'll be the one to ask. What ISO's were you shooting at?

[Joe Walker]

Hey Timur,

Could you walk me through the post workflow you guys used with the S-Log footage? Like once it was recorded:
1. How was it ingested?
2. What editing platform?
3. Color grading was DaVinci, correct?
4. How was that output, etc.

Just basic stuff, it'll help me out down the road :dankk2:

[hunter richards]

keep in mind guys, the guys behind this only had 24 hours to get everything done. they didn't start shooting until 2 am in the morning, going into the afternoon, and having to return everything asap that they were allowed to borrow. 709 was the only gamma they can view in the camera, and the decision was just to expose for skin tone in 709 and then see what the s-log produced. the goal wasn't to set the exposure perfectly for log at all, they wanted to give xdcam the best fighting chance they kid, and then compare it to the s-log.

that said, i'm STILL surprised that not a single person has even questioned what ISO's they were shooting it. because people are forgetting really fast, and I think neglecting the fact that the F3 is truly the low light king of the industry. Not a single camera can touch it in the high ISO range, save scientific cameras. Even if the Alexa completely destroys the F3, going head to head in log in highlight retention, the noise floor and range in the darks are literally several times better than the Alexa. Alexa's current maximum ISO setting, is the F3's new native rating with log. 1600 iso. you can easily expose for your highlights, and bring up your darks in post. DP's normally do that for HD, but end up putting a lot of fill in to balance out the lows, mids, and highlights. but the girl in front of the window for example in the beginning, had NO fill. not even a bounce card. i brought her up at least a stop and a half with no visible noise without pixel peeping.

that's some food for thought right there.

Not trying to be critical at all- I dont think Macgregor is either. It just seems like for monitoring with s-log on the f3, you would either have to choose to make your xdcam and rec 709 outs look the best it could or make the s-log have the most density (under exposing by 1 stop to match the 800 iso vs 1600 iso). Would that be correct?

So recording s-log to sxs cards, even at 8-bit 4:2:0 would be a good option because the exposures would match for an offline edit.

I already knew it was a high iso master, now that you say that- with uping the max sensitivity to 12,800iso with s-log- its going to be beyond anything else out there. Nice work all around, your color looks great too BTW.

Tom- Would you say that the colors are straiter now with s-log or did you have to shift much around?

[tom.wong]

Not trying to be critical at all- I dont think Macgregor is either. It just seems like for monitoring with s-log on the f3, you would either have to choose to make your xdcam and rec 709 outs look the best it could or make the s-log have the most density (under exposing by 1 stop to match the 800 iso vs 1600 iso). Would that be correct?

So recording s-log to sxs cards, even at 8-bit 4:2:0 would be a good option because the exposures would match for an offline edit.

I already knew it was a high iso master, now that you say that- with uping the max sensitivity to 12,800iso with s-log- its going to be beyond anything else out there. Nice work all around, your color looks great too BTW.

Tom- Would you say that the colors are straiter now with s-log or did you have to shift much around?

not saying anybody is out of line, just trying to explain how they exposed to begin with.


as for the color...

the color from the log files were almost dead on. skin tones were great, highlight rolloff was wonderful, and all I really did was tweak color temp in various ways, and a little touch up with the skin because of a lot of the mixed color temp in a lot of the scenes.

this is the kind of accuracy i'm used to see in Alexa and Phantom by the way.

[tom.wong]

Hey Timur,

Could you walk me through the post workflow you guys used with the S-Log footage? Like once it was recorded:
1. How was it ingested?
2. What editing platform?
3. Color grading was DaVinci, correct?
4. How was that output, etc.

Just basic stuff, it'll help me out down the road :dankk2:

I think I can fill in the blanks here, since I did the Cut nd Color.

1. acquisition was internal xdcam, and cineform to the Cinedeck via dual link.
Wrapped the xdcam in quicktime and cut in Premiere CS5, and I already mapped out in my head how I wanted to do the before and afters. Then I matched the cineform clips to the clips used in xdcam for the edit via timecode, and made a separate timeline for those clips. (cineform kept crashing FCP, so we went with Premiere for the edit)

2. finished the cut, exported EDL of just the timeline with cineform clips, conformed in Da Vinci, and linked to all the files in the cut. (Worked like a charm, not a single hiccup with the timecode and metadata) graded the s log cineform clips how I would typically grade the clips, although I really wanted to do more with it, the team emphasized on just make it look natural, and not to worry about really dialing in a "look"

3. exported the clips out at pro res 444, combined them back in premiere, did the before and afters with my original XDCAM timeline, popped in titles, etc. exported out at pro res 444.



also keep in mind, I had several other workflows in my head while doing all of this that were all possibilities. and there's literally 5 or more ways on set for the acquisition workflow you can use, from job to job.

[Joe Walker]

Cool thanks, that's all i needed to know.

[tom.wong]

This was a great test and really shows what S-Log is capable of in my opinion. I plan to do another Dynamic Range test of the F3 once I have firmware in house.

Great job to all the guys on the team and to Tom Wong for the killer color correction.

thanks andy! without you, and everybody at Abel, this wouldn't even been possible. Thank YOU for the support.

[waketek]

Iso?

[TimurCivan]

The sunlit stuff was 0dB 800 iso, but everything else is +12dB 3200iso....

[Evangelos Achillopoulos]

Since I had lots of experience with log capture, since the time I used my Varicam, then Viper and the last three years with RED always using log LUTs to grade, I can say few thinks...

1. I agree with MacGregor, original footage looked a bit over exposed, the mid gray should be in around 31 IRE.

2. When you grade you need a Log film LUT, only this way you get the maximum.

Next Wednesday in my web under FilmRecording will be a section with Generic LUTs for MAC's and all software platforms for all log targets.

[Sumfun]

Thank you for posting the test. I agree with the others about the Rec709 footage being a little over exposed, but I think that only shows how much highlight details the slog can recover. If you had protected the highlights on the Rec709 footage then brought up the shadows in post, maybe the difference would not be as dramatic, but given the time that you had, I think you did a great job. I wasn't so interested in the F3 before, but now it's definitely on my radar.

[waketek]

Thanks Timur, +12dB at 3200 looked very very good. is it still actually 3200 in S-LOG?

[Lenilenapi]

The overexposure though does give a distorted idea of how bad the 709 is to the SxS card - not nearly as bad as this would indicate - generally pretty damn good in fact. Thus you get a bit of a distorted idea of how great the Slog is in comparison.

[petemack]

That S-Log looks sooooo good! It's playin' tricks on my eyes and that's what it's all about.

[HHL]

The sunlit stuff was 0dB 800 iso, but everything else is +12dB 3200iso....

YOU KNOW it's a new world when cinematographers are CHOOSING to shoot ISO3200 and not even blinking at it. And when it's every single shot (except one), it's time to start saving pennies. :)

[macgregor]

YOU KNOW it's a new world when cinematographers are CHOOSING to shoot ISO3200 and not even blinking at it. And when it's every single shot (except one), it's time to start saving pennies. :)


IN my personal opionion, I don´t like supersensitive cameras. I'm not interested in shooting with street lights or with moonlight. I like setting up the lights myself or just shoot in daylight. With such super sensituve cameras (F3, alexa, new red MX...) it's becoming a real problem to shoot in bright daylight. The ammount of filters you have tu put in front of the lens is immense, and this degrades the image quality and makes the camera heavier. Or I´m forced to shoot at t22... better to shoot with 1/3" then lol!

ND Filters with more than 1.2 density start showing problems because it´s hard to get a pure neutral density with no color cast when making them so dark. We also start seeing far red or IR contamination. If we need to use more than 1 ND filter then we can get reflections, and need a heavier matte box to accomodate more trays.... Imagine a shoot in death valley or the Boneville salt flats at ISO 1600!!! Jeeeesszz I'm going to need a ND 3.3! That´s 11 stops of ND reduction to be able to shoot at something like T4!

[TimurCivan]

IN my personal opionion, I don´t like supersensitive cameras. I'm not interested in shooting with street lights or with moonlight. I like setting up the lights myself or just shoot in daylight. With such super sensituve cameras (F3, alexa, new red MX...) it's becoming a real problem to shoot in bright daylight. The ammount of filters you have tu put in front of the lens is immense, and this degrades the image quality and makes the camera heavier. Or I´m forced to shoot at t22... better to shoot with 1/3" then lol!

ND Filters with more than 1.2 density start showing problems because it´s hard to get a pure neutral density with no color cast when making them so dark. We also start seeing far red or IR contamination. If we need to use more than 1 ND filter then we can get reflections, and need a heavier matte box to accomodate more trays.... Imagine a shoot in death valley or the Boneville salt flats at ISO 1600!!! Jeeeesszz I'm going to need a ND 3.3! That´s 11 stops of ND reduction to be able to shoot at something like T4!

+10

ITs awesome in one way.... but a damn nightmare most of the time.

[tom.wong]

i think the next step for camera companies is gonna be that issue right there. I predict one day you'll be able to change the "mode" of the sensor, and you'll get results similar to what slow and fast film does. the slow sensor mode will let you achieve your max DR at a much lower ISO, throttling the sensor, and the fast mode will give it that crazy sensitivity and low noise floor like we have now.

this is all just conjecture, but I can really see someone figuring that out one day.

[Matthew Bennett]

Maybe a sensor wheel in camera? Have like four sensors in there... and just flip them around... have a nice B/W only sensor too just for fun... ha, this will never happen! But maybe...

[Matthew Bennett]

Oh Tom by the way, lovely grade on the f3 s-log!

I know you've been asked twenty million times but are there any public cineform files that will ever surface for us orphans to play with?

[tom.wong]

Thanks for the kind words Matt, means a lot!

As for the master log footage, keep in mind we didn't get the green light from Sony to do this, it's beta firmware, and the team just kind of snuck out with it and did it (with all good intentions of course) We are wary of flack from Sony if something comes up, and you kno who easily this community can flame things up with something overly critical despite all the factors involved. I've already asked the team about this, and we'll see what we can do. But as of now, sorry no.

[TimurCivan]

I.ll come out and say no. we were given the cam on good faith. We just intended to give a taste of what it can do. Not so much a technical sample. This isnt official and isnt a perfect example of what slog can do. We were trying to figure it out as we went along . We dont want the product mis represented by our user error.

[Harry Lime]

Did you guys turn off the in camera sharpening for the test?

thanks

[TimurCivan]

Slot turns off everyhing

[Harry Lime]

Slot turns off everyhing


thanks

[Ryan Patrick O'Hara]

IN my personal opionion, I don´t like supersensitive cameras. I'm not interested in shooting with street lights or with moonlight. I like setting up the lights myself or just shoot in daylight. With such super sensituve cameras (F3, alexa, new red MX...) it's becoming a real problem to shoot in bright daylight. The ammount of filters you have tu put in front of the lens is immense, and this degrades the image quality and makes the camera heavier. Or I´m forced to shoot at t22... better to shoot with 1/3" then lol!

ND Filters with more than 1.2 density start showing problems because it´s hard to get a pure neutral density with no color cast when making them so dark. We also start seeing far red or IR contamination. If we need to use more than 1 ND filter then we can get reflections, and need a heavier matte box to accomodate more trays.... Imagine a shoot in death valley or the Boneville salt flats at ISO 1600!!! Jeeeesszz I'm going to need a ND 3.3! That´s 11 stops of ND reduction to be able to shoot at something like T4!

yup. These new cams are perfect for me b/c 800iso or 1000 is just about my limit. I don't need or want much more for exactly the reasons mcgregor said. I just shot a commercial last week with the red mx in exterior day sunlight.... And my 1st had forgotten to pack my NDs. Needless to say it was almost game over! My lens was at T16-22 split. Even when the filters were brought to set, we needed so much! My dailies had discernable color shift and we had to color it out. Not fun.

I like the idea of interchangable sensors, like switching film stocks!

[jc5274]

wish S-Log would have came out 3 weeks ago, just finished a 3D feature with 2x F3s and had to use cine gama. Could have really used the extra DR...

[tom.wong]

wish S-Log would have came out 3 weeks ago, just finished a 3D feature with 2x F3s and had to use cine gama. Could have really used the extra DR...

the full 3d linking upgrade would have been handy too!

[Dafilman21]

I finally got around to watching this and I am thoroughly impressed. At first I kind of dismissed this camera, not anymore and luckily I know a DP who just got his in. So now I can get some hands on treatment and really see how I feel about it.

[HHL]

Thanks Guys! I haven't ever shot with anything this light sensitive, so the problems that come with the territory are not apparent to me. Thanks for the perspective. All that said, it looks like my Hacked GH1 will have a use if I get the super light sensitive FS100. Yay GH13!!! (lol) :)

[Sumfun]

yup. These new cams are perfect for me b/c 800iso or 1000 is just about my limit. I don't need or want much more for exactly the reasons mcgregor said. I just shot a commercial last week with the red mx in exterior day sunlight.... And my 1st had forgotten to pack my NDs. Needless to say it was almost game over! My lens was at T16-22 split. Even when the filters were brought to set, we needed so much! My dailies had discernable color shift and we had to color it out. Not fun.


Why not increase the shutter speed a little bit (as well as close down the lens, but not as much) ? Does a few stops increase in shuttle speed look that much different?

[Viddovation]

Why not increase the shutter speed a little bit (as well as close down the lens, but not as much) ? Does a few stops increase in shuttle speed look that much different?

Did NASA give us permission to go ahead and use the "shuttle speed" button? I was concerned we might need a firmware update...

[Nate Weaver]

Why not increase the shutter speed a little bit (as well as close down the lens, but not as much) ? Does a few stops increase in shuttle speed look that much different?

Yes.

[Ryan Patrick O'Hara]

yes, shutter speed changes would be very obvious. To save a simple two stops the action would go from 180* or 1/48 to 45* or 1/192... A big difference in how motion is captured. That was not an option. I added a pola, and shot with the stop I had to. It should be noted that shooting at very small apertures can diminish the resolving power of lenses by scattering light. I almost always stay away from deeeeep stops. I don't know how well cookes perform at full closed aperture, but from past experience with different lenses, i avoid it

[Lenilenapi]

I don't think the sensitivity is such a problem. At -3 as I recall (could be wrong though) its similar in sensitivity to say an EX and so exterior issues don't demand that much filtration. The difference is that you can shoot at +12 and even +18 which you never could before.

[hunter richards]

1/4000th is the new 1/48th

[TimurCivan]

using -3 reduces dynamic range.

[CinemaElectronika]

What about using the built-in ND Filters..?

These can be used along with externals...

#1 Filter is 1/8ND = .9ND = 3 stops less...

#2 Filter is 1/64ND = 1.8ND = reduces approx 6 stops...

Might them help... or am I wrong? Correct me please.

Thank you.

Henry Epstein

[BLUESPIDER]

Good stuff Tim. Things are getting more and more exciting with all these new toys. You should of called me, I was too in the Vegas area. I could of lend you guys my ninja skills.

[Ryan Patrick O'Hara]

1/4000th is the new 1/48th

I'm sure the next avatar film will say the same thing

[David G. Smith]

I'm sure the next avatar film will say the same thing

Yikes, I am getting a headache and eyestrain just thinking about.... "I laughed.... I cried.... I kissed 15 bucks good bye!!".

[Sumfun]

i think the next step for camera companies is gonna be that issue right there. I predict one day you'll be able to change the "mode" of the sensor, and you'll get results similar to what slow and fast film does. the slow sensor mode will let you achieve your max DR at a much lower ISO, throttling the sensor, and the fast mode will give it that crazy sensitivity and low noise floor like we have now.

this is all just conjecture, but I can really see someone figuring that out one day.

Good idea. Maybe Sony can do this with the F65 because it has 8k pixels, but only outputs 4k. So it's conceivable that they can just throw away data from half the pixels and have a sensor that is less sensitive. Just guessing.


Did NASA give us permission to go ahead and use the "shuttle speed" button? I was concerned we might need a firmware update...

Good catch! Since we (US citizens) already pay the bill for NASA, everything they do is public property. So we might need a firmware upgrade, but it will be free.

[Harry Lime]

Apparently the sensor for the digital magazine for the Aaton Penelope will have native dual iso.
From what I remember it's a CCD.

[macgregor]

What about using the built-in ND Filters..?

These can be used along with externals...

#1 Filter is 1/8ND = .9ND = 3 stops less...

#2 Filter is 1/64ND = 1.8ND = reduces approx 6 stops...

Might them help... or am I wrong? Correct me please.

Thank you.

Henry Epstein


Of course we are counting on those! But with a 1600ISO camera 6 stops of ND is not enough.


The only good thing is that I´ll be able to use the Nikon ND400 that I bought and never quite used. XD

[Michael Palmer]

First of all, great job to all those who worked on this test. It is really exciting to know what the capabilities are for this F3.

Question? Does anyone here know why there is an adjustment for ND within the menu? I asked the Sony people at NAB and was finally told it adjusts the iso. My F3 gets in my hands on the 26th and this menu feature is hot on my list to understand.

Thanks
Michael Palmer

[Duke M.]

It seems like 6 stops on the ND filter, a couple stops on a hot mirror and close the aperture to f8 or so ought to work most of the time. (Excluding white sand, snow and water.)

[tom.wong]

Of course we are counting on those! But with a 1600ISO camera 6 stops of ND is not enough.

The only good thing is that I´ll be able to use the Nikon ND400 that I bought and never quite used. XD

the built in ND is very strong, i think they did that on purpose knowing what the camera can do. unfortunately, at the moment there's gonna be no way around the high iso if you want to keep all your DR in there.
but really strong ND's, a really REALLY good hot mirror, and a 3 stage mattebox are gonna be a must have...

[tom.wong]

It seems like 6 stops on the ND filter, a couple stops on a hot mirror and close the aperture to f8 or so ought to work most of the time. (Excluding white sand, snow and water.)

but the major issue I believe is that you're options are limited on a reverse way these days. slow sensors, in tough lighting situations, you had to get faster glass, more lights, etc. and you couldn't stop down as much as you would like to. now on the reverse end, it's so sensitive you're cutting light, cutting light, and it will still become increasingly difficult for you to open up on your lens. what if you WANT that particular shallow depth of field shot, wide open on a t1.3 on a master prime in broad daylight, and still find that all the ND's aren't doing a trick? and if you lower the iso, you might lose your highlights.

[Bertholt4]

What about using something like the Fader ND?
And why do we have to use the high isos to resolve the full dynamic range???

By the way, when will they post more of the footage - I`m so excited to see more? Anybody knows it?

[tom.wong]

J
What about using something like the Fader ND?
And why do we have to use the high isos to resolve the full dynamic range???

By the way, when will they post more of the footage - I`m so excited to see more? Anybody knows it?

All sensors have a iso sweet spot where it maintains equal amount of stops on the top and bottom. Its all based on sensor design and the curve applied to it coming in from non linear to linear sensor data. Thus is why cameras are rated differently, and when you go above or under ur rated iso, ur steps in DR shift up and down within the realm of your DR. So a 12 stop camera arlt 800 iso is its rating, 6 stops on top and bottoms. If you lower your iso, you lose stops on your top end and gain on your bottom. And there are limits to that too where if you lower your iso too much, you crush your bottom to the point where you lose that information too, further decreasing your DR.

[Bertholt4]

Ah! OK, thank you tom wong for the explanation - diddn`t know that the native iso affects the dr.

[TimurCivan]

NO FADER ND's!!!!!!!

[dsleep]

how about Singh Ray ND's, Timur?

[TimurCivan]

No fader nd!!!!!!

ITs a double polarizer.... you are double polarizing your image, and making funny things happen to the light rays. I mean, its good when its not mission critical, but otherwise, i would NEVER use a fader on somethign that was a big paying job.

[EIREHotspur]

J

All sensors have a iso sweet spot where it maintains equal amount of stops on the top and bottom. Its all based on sensor design and the curve applied to it coming in from non linear to linear sensor data. Thus is why cameras are rated differently, and when you go above or under ur rated iso, ur steps in DR shift up and down within the realm of your DR. So a 12 stop camera arlt 800 iso is its rating, 6 stops on top and bottoms. If you lower your iso, you lose stops on your top end and gain on your bottom. And there are limits to that too where if you lower your iso too much, you crush your bottom to the point where you lose that information too, further decreasing your DR.

For me the age old problem (or ART as most would say probably) is light or lack of light.
Shooting Weddings in candlelit churches and dancefloor ballrooms with the main lights switched off are a nightmare.
I don't want to carry around lighting to these kinds of jobs because with too many people moving around it can add up to a Nightmare Boxset.
Also to have the ability to create the same kind of Photography shots with video is fantastic and at great quality.

So how do you solve the high sensitivity of this camera.....because I want one.
If you go the way of adapters to make use of Canon L lens then will the Filters for those work well?

[Lenilenapi]

J

All sensors have a iso sweet spot where it maintains equal amount of stops on the top and bottom. .... So a 12 stop camera if 800 iso is its rating, 6 stops on top and bottoms. If you lower your iso, you lose stops on your top end and gain on your bottom. And there are limits to that too where if you lower your iso too much, you crush your bottom to the point where you lose that information too, further decreasing your DR.

Good explanation Tom,
Has anyone really checked out exactly where the F3 really does begin to lose highlight or shadow latitude? I've been thinking of it as a free ride from +12 to -3 that I can use to pick my DOF - but your post gives me pause.

At this point I'm shooting 709 a to SxS cards alot. In that case we're not using the full DR anyway so might it be more like film negative where you actually have a useable range of a few stops of ISO before there is any loss at either end.

[mrglassworm]

All sensors have a iso sweet spot where it maintains equal amount of stops on the top and bottom. .... sony seo company direct (http://www.facebook.com/pages/Seo-Company-Direct/166223750105572)So a 12 stop camera if 800 iso is its rating, 6 stops on top and bottoms. If you lower your iso, you lose stops on your top end and gain on your bottom. And there are limits to that too where if you lower your iso too much, you crush your bottom to the point where you lose that information too, further decreasing your DR.

Interesting thread, learned a lot from here.