FS7: New Sony F7 XDCAM 4K Super 35mm Camcorder

This Log vs linear raw conversation is one that I've seen butchered numerous times before. As starcentral has partially alluded to, it's important to understand that the image pathway is composed of multiple pieces. When photons (light) hits the sensor photosites it's converted into an electrical charge. More photons (light), a greater electrical charge. But the photosite is like a bucket - too much light and it overflows. These photosite voltages are analog. The next step is a A/D converter that converts the electrical charge into a number for each photosite. Typically these numbers are generated in a linear fashion, meaning that a doubling in voltage produces a number twice as large. But the A/D converter is limited by two factors: how precisely it can read the voltage, and how many bits it has to store the results.

Practically speaking, the A/D converter is always going to use a higher bit depth than the final signal path or recorder outputs. In some cases for linear raw the bit depth of the A/D converter and the record path can be equal, but for video or log output this simply doesn't make sense. The reason being is that our eyes don't perceive light in a linear fashion. In fact, the downside of linear raw is that that it allocates a tremendous portion of it's storage bits to highlights and very little to the shadows. While this can be useful in situations with extreme highlights, it's not terribly useful for recovering from the shadows. Log by comparison makes an effort to use a DSP (digital signal processor) to reallocate the linear values from A/D converter into log encoded values.

Absolutely - 16 bit raw is incredible. But when you're looking at cameras storing log and raw values at 10bits and 12bits, it really deserves a more careful look at the math. For instance, the Sony F35 has a max output of 12bit RGB 444, but it uses a 14bit A/D converter, where the RED One MX used a 12bit A/D and 12bit linear raw encoding. When you look at the math that way, and remember that the linear raw file is over half highlight information, it's a bit easier to understand why people though the F35 had more pleasing skin tones - it was using more pixels to store where the skin tones were exposed.

Of course the F35 is dated now, but I suspect it's 12bit RGB 444 will be a very competitive image against the FS7 12bit RAW.

So, in my opinion, if it's coming from a good enough A/D converter, I do think that Log gives you the majority of the raw benefits without the hassle.
 
Maybe a separate sub-forum as this forum already has 3 prefixes - F3, F5, F55.

If the FS7 does have to share a forum, maybe it's better with the FS700? Less cameras, they are closer in price range and many of the people who own FS700 are moving to the FS7.

Thanks!



^ Agreed with this.
 
But as explained in my post, if you are still getting your 2.5 stops above 90% white and the only difference are data levels which are mostly important for grading - is it really going to make a difference? Does anyone here really grade highlights that they think the need all that data? You still have 14.5 levels of white above 90% white with slog3 or slog2 - you just have less data per stop but not much less.

I personally have not done any extensive analysis or testing to conclude one way or another but if anyone has anything similar in terms of tests I'd love to see it. I have a hard time believing there would be a big difference. Most tests people have done didn't account for exposing the slogs differently so it caused slog3 to clip earlier than it would with slog2 at the same f-stop. If you clip slog3 its going to look clippy, but that becomes an exposure issue not an slog3 issue.

By the way check out this comparison video on slog2 versus slog3. See if you can find faults with any highlights. Note however that the shooter did not actually expose for the two slog gammas differently as he might have wanted to. ie. He shot both gammas at the same f-stop but in fact by doing that you are actually overexposing slog3. you need to stop down or put on 1 stop ND filter when shooting with slog3 to really compare it to slog2 when using the same stop. So ignore the "side by side" summary at the end. It's wrong.

Go to 0:37 which is slog3 f5.6, then jump to 1:37 which is slog2 f4. This is the same middle grey properly exposed for both. Look at the glass in the background and light. Really? Any mega differences here?

http://wolfcrow.com/blog/sony-f55-s-log-3-vs-s-log-2-comparison/

Thanks for all the useful info as usual, Dennis.

So what you are saying is if you expose SLOG 3, 1 stop less than the SLOG 2 equivalent, you get about the same highlights?

One thing I noticed from the Wolfcrow.com test you posted is that SLOG 3 was noticeably less contrasty, which I like. The shadow detail in SLOG 3 was noticeably more perceivable as well. At least in that indoor test, I preferred the SLOG3.

Is there any outdoor tests between the SLOG 2 and 3? I'd love to see that.
 
Do you folks want this as a separate sub-forum, or as a prefix in this existing sub-forum?

I know it would cause some short term confusion, but perhaps if there were just "Sony HD 35mm Cameras" and "Sony 4k 35mm Cameras" forums, with prefixes available for the most popular cameras in each... that might be the easiest in the long run.
 
This Log vs linear raw conversation is one that I've seen butchered numerous times before. As starcentral has partially alluded to, it's important to understand that the image pathway is composed of multiple pieces. When photons (light) hits the sensor photosites it's converted into an electrical charge. More photons (light), a greater electrical charge. But the photosite is like a bucket - too much light and it overflows. These photosite voltages are analog. The next step is a A/D converter that converts the electrical charge into a number for each photosite. Typically these numbers are generated in a linear fashion, meaning that a doubling in voltage produces a number twice as large. But the A/D converter is limited by two factors: how precisely it can read the voltage, and how many bits it has to store the results.

Practically speaking, the A/D converter is always going to use a higher bit depth than the final signal path or recorder outputs. In some cases for linear raw the bit depth of the A/D converter and the record path can be equal, but for video or log output this simply doesn't make sense. The reason being is that our eyes don't perceive light in a linear fashion. In fact, the downside of linear raw is that that it allocates a tremendous portion of it's storage bits to highlights and very little to the shadows. While this can be useful in situations with extreme highlights, it's not terribly useful for recovering from the shadows. Log by comparison makes an effort to use a DSP (digital signal processor) to reallocate the linear values from A/D converter into log encoded values.

Absolutely - 16 bit raw is incredible. But when you're looking at cameras storing log and raw values at 10bits and 12bits, it really deserves a more careful look at the math. For instance, the Sony F35 has a max output of 12bit RGB 444, but it uses a 14bit A/D converter, where the RED One MX used a 12bit A/D and 12bit linear raw encoding. When you look at the math that way, and remember that the linear raw file is over half highlight information, it's a bit easier to understand why people though the F35 had more pleasing skin tones - it was using more pixels to store where the skin tones were exposed.

Of course the F35 is dated now, but I suspect it's 12bit RGB 444 will be a very competitive image against the FS7 12bit RAW.

So, in my opinion, if it's coming from a good enough A/D converter, I do think that Log gives you the majority of the raw benefits without the hassle.

Crap, as I was driving to the office I thought about this exact thing, but you beat me to it. Yes, absolutely.

The entire chain needs to be considered, and I was thinking mostly about cameras that are more exclusive to RAW versus thinking about Alexa and F35 as well.

I had this same discussion on why I didn't think the GH4 was any better than the A7S, 8-bit or 10-bit, recently. I believe the GH4's only putting 10 > 10, anyway, so it's really not all that much better than having a really good 8-bit representation of the same content.

THere's no real argument to be made when comparing say REDCODE to a simple 10 422 HQ from Alexa... the latter's industry leading for many reasons and that's one of them.
 
Dear Gene,

If you want to expose for slog properly always go by the white paper value for middle grey.

ScreenShot2014-02-15at65331PM_zps438a4526.jpg


Yes contrast is a whole other thing but that's because of the curve differences between slog2 and slog3,, its not a good or bad thing, it just is what it is.

Here's an outdoor video of slog2 vs slog3 but again, there is no reason highlights should be clipped other than the gammas were exposed the same and shouldn't have been.

http://vimeo.com/102398147
 
Last edited:
Regarding slog2 vs slog3

But as explained in my post, if you are still getting your 2.5 stops above 90% white and the only difference are data levels which are mostly important for grading - is it really going to make a difference? Does anyone here really grade highlights that they think the need all that data? You still have 14.5 levels of white above 90% white with slog3 or slog2 - you just have less data per stop but not much less.

I personally have not done any extensive analysis or testing to conclude one way or another but if anyone has anything similar in terms of tests I'd love to see it. I have a hard time believing there would be a big difference. Most tests people have done didn't account for exposing the slogs differently so it caused slog3 to clip earlier than it would with slog2 at the same f-stop. If you clip slog3 its going to look clippy, but that becomes an exposure issue not an slog3 issue.

By the way check out this comparison video on slog2 versus slog3. See if you can find faults with any highlights. Note however that the shooter did not actually expose for the two slog gammas differently as he might have wanted to. ie. He shot both gammas at the same f-stop but in fact by doing that you are actually overexposing slog3. you need to stop down or put on 1 stop ND filter when shooting with slog3 to really compare it to slog2 when using the same stop. So ignore the "side by side" summary at the end. It's wrong.

Go to 0:37 which is slog3 f5.6, then jump to 1:37 which is slog2 f4. This is the same middle grey properly exposed for both. Look at the glass in the background and light. Really? Any mega differences here?

http://wolfcrow.com/blog/sony-f55-s-log-3-vs-s-log-2-comparison/

You won't see major differences until you've gotten into extremes, either extreme contrast or extreme lowlight.. The value of SLog2's greater highlight range would probably be color latitude in upper mids and highs. That's purely a color thing, where you could easily fake a clipped channel, losing two means a really ugly blow out. BUT....

BUT... if Slog3's giving you pretty dense mids and upper mids, with more latitude in shadows, it's basically telling you to protect highlights and create your own rolloff... I don't see how that's a bad way to use the DR.

If I'm reading it correctly (I swear Im gonna get some waffles soon) then Slog3 just wants to be under for best results.
 
Here's an outdoor video of slog2 vs slog3 but again, there is no reason highlights should be clipped other than the gammas were exposed the same and shouldn't have been.

http://vimeo.com/102398147

I get the feeling that a lot of the issues Slog2 users are having with Slog3 is they are treating them the same not exposing to protect highlights in Slog3.

Being new to Slog I can't wait to get my hands dirty with the FS7 and find out what works for me and my clients.
 
I suggest that one of you guys summarize the important points in this recent discussion and just put it in a new thread called Slog 2 & Slog3 . It is relevant to all the cameras now using those gammas.

Without going into all the technical stuff here all I can say is that the best techs I know were very excited by Slog 3 and felt it came close to making Sony look like an Alexa and that it was easier to expose for. Loved what they could do with flesh exposure. Haven't tried it myself yet.
 
This is LC709-A applied to slog3 on the F55 that I baked-in for a fast turn around shoot where I was not allowed to grade it for "free" even if I wanted to.

This image is not graded and straight out of camera, maybe black pedestal pulled down a bit, but straight out of the file. It seems to have some kind of grey'ish wash over it which can be graded out pretty easy.

Does it look like Alexa? I dunno.

 
This is LC709-A applied to slog3 on the F55 that I baked-in for a fast turn around shoot where I was not allowed to grade it for "free" even if I wanted to.

This image is not graded and straight out of camera, maybe black pedestal pulled down a bit, but straight out of the file. It seems to have some kind of grey'ish wash over it which can be graded out pretty easy.

Does it look like Alexa? I dunno.

F55-LC709A-June2014.jpg

Looks like something you could hand off.

Everyone wants to look like Alexa, though. Nothing actually LOOKS like ALexa, however. xD I dropped that a long time ago, if you want an Alexa look then shoot Alexa. Film, Film.

SLog3 looks really promising to me.
 
I have a few questions about S-Log 2 vs S-Log 3

1) Doug Jensen mentioned that Slog 3 clips at 95 IRE while Slog 2 clips at 110 IRE. On paper that seems huge, but is that a huge difference in highlights in a REAL WORLD sense?

I never said any such thing, so please don't attribute incorrect information to me. (I would appreciate it if you would edit your post because it is a couple of pages back)
S-LOG3 clips at 95 IRE and S-LOG2 clips at 106. And yes, it does make a difference in the real world.
For more information, see my F55 book. http://www.vortexmedia.com/Guide-to-Sony-F55.html
 
I never said any such thing, so please don't attribute incorrect information to me. (I would appreciate it if you would edit your post because it is a couple of pages back)
S-LOG3 clips at 95 IRE and S-LOG2 clips at 106. And yes, it does make a difference in the real world.
For more information, see my F55 book. http://www.vortexmedia.com/Guide-to-Sony-F55.html

Sorry Doug, I meant the 95 IRE part that you mentioned in another thread. The 110 IRE on S-Log was my own miss-information. I should not have put them in the same sentence.
 
1) On the F5 (and I believe the F55 as well) SLOG3 clips at 100 IRE (not 95 IRE), SLOG2 clips at 106 IRE.

That is incorrect. S-LOG3 clips at 95 IRE. It is easy to see if you have a real waveform monitor and aim the camera at a naked light bulb. There is nothing you can do to get an highlight brighter than 95 IRE with S-LOG3.
 
Sorry Doug, I meant the 95 IRE part that you mentioned in another thread. The 110 IRE on S-Log was my own miss-information. I should not have put them in the same sentence.

No problem. I just don't want to get blamed for being wrong when I don't deserve it. There are plenty of times when I do deserve it!
 
Everybody is talking about gammas these days and tend to forget about colour space.

SLog3 also introduces a "new" colour space which is called S-Gamut3.cine. This is exactly why I like
the ability of SLog3 more than SLog2. The SLog3 SGamut3.cine colour space is tighter or smaller than S-Gamut3,
but it's also much more "natural".
What does it mean? It means that the colour space is more "usable" as it is closer to what is used in the output
format which will be rec709 for broadcast or P3 for cinema anyways.

I'm definitely not an experienced expert in grading as I come from stills photography and know linear RAW the best.
It's just easy to grade as it respondes linear to most grades.

Log is different in terms of a very complicated science how the different values are stored in the codec.
Why on earth should anyone want that complicated science?

Have you ever shot a feature length project with a tight budget in RAW? It needs a lot of time for backup
and lots of space, too. Big budget - no problem.
Small budget - RAW does make a difference because it changes your whole post workflow as you need huge
and fast storage at the same time.

Log tries to get the most information out of the sensor data squeezed into a relatively small bandwith/codec.
That means more science in post for delogging.

My experience with SLog3/SGamut3.cine is that it is much easier to delog and grade than SLog2/SGamut.
At least for me. I just need the corresponding LUT, that Sony is delivering and can do my grades easily after that.

Here's some input from Art Adams (even if it's SLog3/SGamut3 vs. SLog3/SGamut3.cine):
http://www.dvinfo.net/article/acquisition/sonyxdcam/sony_sgamut_vs_sgamut3.html

I was testing how to get the look I want out of the FS100 and achieved this with highlight desaturation in post.
This led me to the fact that I don't need much colour information in the highlights but a nice rolloff without
clipping abruptly. We don't wanna see fine details in the highlights but want to avoid clipping!

I did not send the FS7 through the chambers of shaolin...
But the SLog3/SGamut3.cine is very nice even with the LC709 Type A LUT baked in. At least for me this is
something that works even without grading for me.
 
Last edited:
Back
Top