Advantage of S-log

[Ron Aerts]

Does S-log 1.5G outputs have any advantage in recording to a Nanoflash recoder? Or does it not work that way

[Chris Medico]

The issue with recording S-log to a nanoflash is that the recording will be 8bit. It can be done and recording onto a nano would be an improvement over recording S-log onto the SxS cards. To get the most benefit and gradability in post you want to record S-log with a 10bit recorder.

In summary, Is it possible - yes, Would I recommended it - no.

[Dan Keaton]

Dear Ron,

Since S-Log is 12 bits, encoded into a log format using 10-Bits, we generally do not recommend recording S-Log with a nanoFlash.

But, for the F3, you can use CineGamma 1 and CineGamma 4, (instead of S-Log) and record it to the nanoFlash and get outstanding results.

And for the ultimate in image quality, we have the Gemini 4:4:4 which fully support Sony S-Log, including applying a S-Log to Rec. 709 LUT for viewing purposes.

[Ron Aerts]

Thanks I'll stick to the normal Cinegamma 1/4 way, which is great for most jobs.
Can the (gemini) S-log data easily be edited in Sony Vegas?

[Dan Keaton]

Dear Ron,

The current version of Sony Vegas does not directly support DPX files.

Since new versions of Sony Vegas sometimes come out at NAB, we can hope for native DPX files in their new version. (I have no actual knowledge of a new version of Sony Vegas, its features or release date.)

For now, one can use Cineform to take our DPX files and then use the Cineform files for editing in the current version of Sony Vegas.

I hope others will join in this conversation.

[Ron Aerts]

With no S-log I use the nanoflash on the Dual link A output which has only a 1.5G 422-video setting. So then I can use the SDI/HDMI/Video out simultaniously for monitoring purposes in any format I want without disturbing the recording to the NF. I assumed that this output is exactly the same as the standard HDSDI output of the camera, with all the picture(profile) settings.
Is that right? I didn't notice any difference between the HDSDI and the DUAL Link A out.
The manual says that on 50i the output wil be 1920/1080 50P. And although I have the camera system settings on 25P it is outputting a signal (25P)

[Chris Medico]

Originally Posted by Ron Aerts

With no S-log I use the nanoflash on the Dual link A output which has only a 1.5G 422-video setting. So then I can use the SDI/HDMI/Video out simultaniously for monitoring purposes in any format I want without disturbing the recording to the NF. I assumed that this output is exactly the same as the standard HDSDI output of the camera, with all the picture(profile) settings.
Is that right? I didn't notice any difference between the HDSDI and the DUAL Link A out.
The manual says that on 50i the output wil be 1920/1080 50P. And although I have the camera system settings on 25P it is outputting a signal (25P)

That is correct, when the F3 is NOT in S-Log mode the SDI A and B ports AND the Monitor ports will follow the same picture profile settings.

[Charles Papert]

I'd like to add my two cents on this discussion. I have mentioned this in other threads but I think it is relevant to this one.

Striving for highest image quality within one's budget is of course the right idea. However I have seen a tendency in recent years as very high quality digital video gear becomes affordable for a certain "nouveaux snobbery" to emerge--users who a few years ago were working in with much more humble gear are now turning down their nose at quite acceptable formats. Largely this is due to the way we receive information, via message boards like this and aggregators and blogs. Sometimes all it takes in one well-read blogger to pass judgement on a piece of gear or a recording format, and within minutes thousands of opinions around the world are solidified. This can be tough on manufacturers, which is why it is imperative in this day and age that they maintain a presence on message boards to answer questions (and perform damage control). Dan is a perfect example of this, he's dialed in and that's great.

On to the question of recording S-log out to the Nanoflash. As Chris noted and Dan echoed, it can be done but it's not ideal. 10 bit is significantly better than 8 bit, I'm sure we'd all agree. But how bad is it? The assumption is that it will "fall apart" in color correction as soon as you start to push and pull the image, right? How can it not--it's 8-bit after all. And of course, you've pushed so much of the information into the middle of the curve, surely it will demonstrate banding and compression artifacts when de-logged...?

From my first-hand experience, this is not the case. When I started shooting my series last year, I had no intention of recording s-log to 8-bit. A traumatic first day with an outboard 10 bit recorder led us to make a quick decision to record to XDCAM 422 disc recorders which were flawless. We color corrected in a high-end Davinci suite, and did plenty of pushing and pulling of the image; power windows, some radically pushed looks etc. We were six episodes in when a casual discussion of the just-released C300 between the colorist and myself led to a question of how much of an issue is it to work with 8 bit vs 10 bit footage. The post supervisor then mentioned that XDCAM 422 (and all flavors of XDCAM) are in fact 8 bit. I refused to believe it until I looked it up myself, and was shocked to learn that I had been recording out to an 8 bit format. I asked the colorist if he had any indication from working with the footage that it was 8 bit, and he said no, hadn't seen any issues when manipulating the image. Suffice to say, the material looked great to all of us, and was broadcast on Comedy Central without any complaints!

One caveat: I had heard that the effects team had had a tough time pulling mattes off the green screen, and this probably explained it. They were successful, it just took more work.

XDCAM422 is the very format that the Nanoflash records, so based on my experience, there's no reason why many levels of work couldn't be recorded with that workflow. By "many levels", I'm referring to the vast majority of users on this board who are seeing their work viewed on the web, and perhaps projected here and there, and maybe even shown in a broadcast environment. It worked for us (and incidentally, the shows "Community, "Happy Endings" and "Up All Night", all of which record out to XDCAM422). Dan, can you chime in and indicate what differences there are, if any, in the way that the Nanoflash records this format vs the disc recorders?

What this experienced proved to me is that one has to really see for oneself what the true differences are in codecs, recording formats, cameras etc. rather than just believe what one is told. We know that 10 bit is better than 8 bit, but 8 bit is actually quite good when coupled with a robust codec (XDCAM422 is quite a bit better than, say, H264, so it's not like working with Canon DSLR footage--but I've been able to push that stuff quite a bit also, much more than many people say "should" be done).

I'd invite you to look through the clips in this channel:
Key & Peele on Vimeo

Of note, due to heavy grading (although they are all pretty damn entertaining and worth watching):
Flicker
What's the Joke
The Bite

Incidentally the first one, "Bitch", was shot on 1DMKIV's so there are some DSLR issues visible in there (moire in the final scene).

Obviously on Vimeo, artifacts are going to present. But having seen all these clips on a 42" reference plasma in the color suite, I can tell you that we never had issues that one would "expect" from 8-bit recording.

[Dan Keaton]

Dear Charles,

I greatly appreciate your thoughtful post.

I agree completely that 8-Bit can be very good.

I try to be completely honest, and never say that 8-Bit is better than 10-Bit, and I never hide the fact that the nanoFlash is an 8-Bit recorder.

But one needs to be very careful. not all 8-Bit codecs are great. The Sony XDCam 422 codec, while 8-Bit is great. It is very efficient and very low noise. While other 8-Bit codecs suffered from banding, this is very rare with the Sony XDCam 422 codec.

To be very candid, this always presents problems for us, as one would assume that a 10-Bit competing recorder is always better since it is 10-bit, but this is not always true.

The Sony XDCam 422 codec is very sophisticated, and very efficient, and 8-Bits. But most importantly, it is a very low noise codec. Some of the other codecs are definitely not as low in noise.

A very technological savy executive at a major camera company told me that there is not much difference between 8-Bit and 10-Bit. I agree. One example is how people use the nanoFlash with the Sony F3 and are very happy with the results.

I agree with you that Sony S-Log can be recorded to the nanoFlash and it can work out.

But, I am being very conservative in stating that this is not the best of all possible options.
I would be very upset with myself if I recommended recording S-Log to the nanoFlash and someone had a problem with their project in post.

I also have experience with other people thinking that they cameras are 10-bit when they are actually 8-Bit. 8-Bit is very common, far more so than most people believe. Panasonic Varicam, HDX900, Sony XDCam, Sony HDCam. The Sony F3 is a great example of a 10-Bit camera that has such low noise that the 10-Bit is actually useful. Some other 10-Bit cameras have way too much noise to make the use of 10-Bit useful.

The image quality of our 50 Mbps 422 files is identical to those produced by the Sony PMW-700 or PMW-F800 cameras. The Sony XDCam Codec Module that we use is the same as the one in these cameras, thus the quality is exactly the same.

So to directly answer you question, there is no difference.

For greenscreen work, we recommend 100 Mbps Long-GOP or 140 Mbps or higher I-Frame Only.
Even with a good 8-bit camera, one can easily pull a key with these bit-rates. I get emails and comments all of the time from people using the nanoFlash stating how easy it was (and how happy they are!).

And we worked with Sony so that our files can be put onto the Sony Professional Disk via a U1 or U2 drive.

With 8-bit recording, I feel that it is always best to capture a good image in the camera so that in post one does not have to manipulate the bits too much to obtain the desired image. Just as with 16-Bit audio, one should set the levels properly, while with 24-Bit audio one can set the levels very low and still get acceptable audio.

Charles, thank you again for your post, and I hope this helps.

[Chris Medico]

I completely agree that 10bit is not always better than 8bit.

One thing that in my opinion 10bit does offer when shooting LOG is more latitude for exposure error. Specifically if you put the luminance of something important in the sharper area of the LOG curve you may see artifacts at 8bit that would not be seen at 10bit.

I demonstrated this to myself when I first got the S-Log key for my camera and was just testing it out at home. I shot some video of a window and a blank wall. I set the exposure to keep a window from blowing out. When I brought the midrange up to normal luminance values and compared the 10bit and 8bit video the wall had visible banding in 8bit that wasn't visible in the 10bit footage.

Note to self - get exposure right when shooting LOG at 8bit.

[Ron Aerts]

Good words Charles. And catching the emotion doesn't come in bits. The NF cuvers easily most of what I need.
Working around the technical sweet spot is what I'm aiming for and it's project depending.

[Charles Papert]

That's a good test Chris and one I would like to see at some point. I guess we never really shot blank walls on our show! Or, underexposed significantly to hold overexposure.

[Alister Chapman]

You also need to consider whether you post workflow is truly 10 bit. 10 bit AVI's on a Mac normally get truncated to 8 bit, media composer truncates many codecs if you import the material (AMA mode is generally OK), not every version of DNxHD is 10 bit (only the ones with a small "x" at the end). A lot of FCP filters and transitions are only processed in 8 bit as there is a YUV -RGB -YUV round trip.

[Ron Aerts]

Good point Allister, so using color correction higher than 8 bits on 8-bit material doesn't make sence?