FS700 raw output and SLOG-3

[Cliff Totten]

I have a question about the Odyssey with the raw option and the Sony FS700.

As you know, the FS700 has a gamma option for SLOG-2 in it's picture profile but has no SLOG-3 option. For raw to ProRes recordings, it has been said in your instructions to set the FS700 for SLOG-2 gamma. This brings up some questions about the Sony raw protocol and how the Odyssey is receiving it.

1.) True raw sensor data is captured at the sensor level and should have no gamma applied by the camera. So why does the FS700's picture profile have to be set for SLOG-2?

2.) If SLOG-2 must be chosen, does this affect the Odyssey's ability to take that raw data deBayer it, convert it to video and apply an SLOG-3 gamma curve for ProRes recording?

3.) Given that we are talking about raw sensor data with no gamma or color processing yet, what is the role of the FS700's picture profile (or gamma settings) settings at all?

My concern is the Odyssey's ability to record proper SLOG-3 from the FS700's raw output. And it would seem to me that the FS700's in-camera gamma setting/processing shouldn't matter to it's raw sensor data output which is way upstream from internal camera gamma/color processing???

I must be wrong,....but how?

CT

[Dan Keaton]

Dear Cliff,

I will attempt to answer your questions, please feel free to ask follow up questions.

Originally Posted by Cliff Totten

I have a question about the Odyssey with the raw option and the Sony FS700.

As you know, the FS700 has a gamma option for SLOG-2 in it's picture profile but has no SLOG-3 option. For raw to ProRes recordings, it has been said in your instructions to set the FS700 for SLOG-2 gamma. This brings up some questions about the Sony raw protocol and how the Odyssey is receiving it.

1.) True raw sensor data is captured at the sensor level and should have no gamma applied by the camera. So why does the FS700's picture profile have to be set for SLOG-2?

First, as you noted, the FS700 has S-Log 2 but no S-Log 3.

For us to properly work with the Raw data from the FS700, we have to have a known starting point, so we can process the Raw Data appropriately. Since a Rec. 709 output would limit the dynamic range of the data, we require that the Gamma = S-Log 2. In addition, we recommend (require for the best results) that you use Picture Profile 7 set to Sony's Factory Defaults.

Background, While it is possible to design a camera with zero image processing of the true raw data from the sensor, most cameras (as I understand it) perform some processing of the raw sensor data.

Applying Log encoding to the Raw data is a way to effectively record higher bit-depth into fewer bits.

For example using Log, one could record more than 12 Bits of data into 12-Bits. Using Linear, one can only record 12-Bits of data into 12-Bits. The Log encoding is a way to pack the most important part of the "more than 12-Bits" into 12-Bits.

Our eyes are more sensitive to shadow details and mid-range details, and far less sensitive to highlights. Thus, the Log curve is adjusted to provide the most data for the shadow details and mid-range details..

As you know, for Sony, they originally came out with S-Log, then S-Log 2, then S-Log 3. S-Log 3 is a tweaking of S-Log 2, but the FS700 camera does not support S-Log 3, at least not at this time.

2.) If SLOG-2 must be chosen, does this affect the Odyssey's ability to take that raw data deBayer it, convert it to video and apply an SLOG-3 gamma curve for ProRes recording?

Yes, S-Log 2 must be chosen in the FS700 to properly record a high dynamic range signal in the Odyssey. But this does not limit your ability to color correct your footage in post, in fact it enhances your ability to do so, since if we recorded in Rec. 709, you would only have a little more than 6 stops of Dynamic Range (which is inherent in Rec. 709).

Stating this differently, since we are using S-Log 2, we preserve the higher dynamic range of the FS700, and then you can use Post tools to use that higher dynamic range, as appropriate, for your project.

3.) Given that we are talking about raw sensor data with no gamma or color processing yet, what is the role of the FS700's picture profile (or gamma settings) settings at all?

As with most Raw camera, the actual, true (unadjusted) Raw data is not output from the camera. It is adjusted for various reasons, and Gamma=S-Log 2 is applied to record more than 12-Bits of data into 12-Bits.

ARRIRAW in the ALEXA, and Canon Raw, from the C300 Mark II and C500 are also processed, thus they are not output with "no gamma and no color processing".

My concern is the Odyssey's ability to record proper SLOG-3 from the FS700's raw output. And it would seem to me that the FS700's in-camera gamma setting/processing shouldn't matter to it's raw sensor data output which is way upstream from internal camera gamma/color processing???

For the Odyssey to properly record S-Log 3 data, the best way to do so, would be for the FS700 to output S-Log 3.

Technically, the Odyssey could be programmed to accept S-Log 2 data from the FS700, then convert it to Linear (in more than 12-Bits) and then apply a S-Log 3 gamma curve to the Linear data, and then record it.

But, since S-Log 2 and S-Log 3 are similar (or very similar) this would be a lot of processing for very little gain, unless I am missing something.

I have attempted to answer all of your questions to the best of my technical knowledge at this moment.
I can review my answers with our Chief Engineer and see if he has better or different answers for you.

Others may critique my answers, I welcome their comments and suggestions.

I must be wrong,....but how?

CT

I hope this helps.

Respectfully,

[Cliff Totten]

Thank you Dan for your reply!

This Sony raw protocol is somewhat of a mystery as they don't really have any white papers published as to what is actually in the protocol. They say it is "raw" but just how "raw" is it? I always thought that pixel intensity or brightness would be mapped from 0% to 100% (pure black to pure white) in a linear way.

So it seems that Sony does map it's raw pixel (brightness) values in a "curved" gamma? It's not linear or proportional? (Because a linear read-out at 12 bits couldn't hold the 14stops that the FS700 has)

This would mean that Cinema DNG files would also be affected in the same way? Seems so.

In it's most fundamental form, Sony raw protocol contains this data:

1.) Individual pixel address. (It's physical location on the sensor grid)
2.) Individual pixel voltage reading. (or, it's intensity amount for that read cycle)
3.) Color filter flag. (all pixels are black and white, grayscale. This marker identifies what R,G or B optical color filter it sits under)
4.) Lot's of meta-data about the condition the camera is in. (ISO/Gain and so forth)

I realize that this is some serious "under the hood" talk but I'm sure that allot people would benefit from a deeper understanding of what "raw" data actually is. I think it is an industry term that we all throw around allot but it is one that is very commonly misunderstood, or not fully understood by a large number of us.

Since raw recording is such a big feature on the Odyssey, I would love to see Mitch do a CD video with a title of "Mysterys of raw data revealed"...lol...or something like that. Something where you folks can really describe the protocol itself, what it includes and what it doesn't have in it. Bayer pattern processing is a wonderful topic in the Odyssey itself. Mitch was telling me at NAB that there are different ways to do it and that CD believes they do it best.

The way I describe "raw" to people is this: Raw is not "video" as you know it. Raw is not a baked chocolate cake. Instead, "raw" data is all the ingredients in a grocery cart that "can" become a chocolate cake someday. It's the Odyssey that takes the flour, sugar, butter, salt and chocolate and bakes it carefully into an actual cake. (For ProRes)

It's a great educational video idea for CD! Hell, I think the entire industry can benefit from it. "Raw" is so misunderstood, industry wide at all levels from engineers up to executives.

So my take away is that the FS700 does not have a linear brightness intensity and will map an SLOG2 gamma curve into it's raw data. (this should also affect Cinema DNG too)

Hmmmm....very very interesting!

[Dan Keaton]

Dear Cliff,

We created a video that explains S-Log. We created this for the Sony F3 camera and our Gemini 4:4:4, but the principles still apply.

S-Log 2 is a refinement of S-Log.

Better Living Through S-Log - YouTube

You wrote:

In it's most fundamental form, Sony raw protocol contains this data:

1.) Individual pixel address. (It's physical location on the sensor grid)
2.) Individual pixel voltage reading. (or, it's intensity amount for that read cycle)
3.) Color filter flag. (all pixels are black and white, grayscale. This marker identifies what R,G or B optical color filter it sits under)
4.) Lot's of meta-data about the condition the camera is in. (ISO/Gain and so forth)

To save space, (1) is not included in the Raw data, the relative position in the file determines this information. The same applies for (3).

For (4), there is a metadata passed from the camera that provides us with certain data, and this includes information on how the Raw data values are to be interpreted. But, it takes much more than using this metadata. You used "Lots of Metadata", but the reality is there is very little Metadata compared to the amount of Raw data.

================

DeBayering a Raw data stream is very much of an Art.

For competitive reasons, I will not go into how we do this, but the results speak for themselves. (Or to be more bold, we do a very good job of DeBayering the data. I agree with Mitch's statement.)

And we use very special techniques to take the 4K Raw data, deBayer it, and convert it to HD, when you use our 4K to HD ProRes or 4K to 2K ProRes modes. These modes provide you with outstanding HD or 2K.

When one uses our Raw Bundle, or just the Sony FS part of our Raw Bundle, you are getting the benefit of many years of work on our part. And in general, the processing is different for most cameras, even for cameras from the same manufacturer (typically).

I am sorry, but I can't go into more details, but I recommend you watch our S-Log video (above).

Respectfully,

[Cliff Totten]

Wow!

Excellent post Dan, and excellent video!

Looking at SLOG-2 and SLOG-3. I color grade them both and I use very different LUTS for each. Also, on waveform scopes, they take up very different areas. SLOG-2 can go right to the top at 109 IRE but SLOG-3 will clip at around 94 or so on an FS5 or A7s.

I have never tested this but from your experience do you find that SLOG-3 holds a bit more dynamic range than SLOG-2? Obviously, the shadows, midtones and highlights are allocated or mapped differently between the two. I have read other say they still hold the same actual DR "information" from top to bottom.

Anyway, thank you so much for your writings on this topic. SLOG is extremely well documented for years on the web but so little has ever been discussed about the Sony raw protocol. I wrongly thought that it was a linear read out and that CD built the gamma curve in the Odyssey.

Excellent stuff! Thanks again for turning some light bulbs on in my head! I'm going to forward your video to friends of mine too.

CT

[Dan Keaton]

Dear Cliff,

Yes, S-Log 3 in an FS7 can clip before 100% IRE, but I do not know all of the details. Some say at 94%, some say 96%.

S-Log 2 can go higher, up to 109% IRE in the FS700.

The ARRI ALEXA has special techniques to prevent highlight clipping. When watching the ARRIRAW output on a calibrated Waveform Monitor (including the Odyssey), as you increase the exposure, the values will never go above a certain level. It could be that S-Log 3 is using similar techniques.

On S-Log 2, normal, general highlights (when looking at the data in a calibrated Waveform Monitor, the target is 59% (for S-Log 2). This allows for a great deal of headroom, from 60% IRE to 109% IRE.

This allows one to overexpose by two stops, when using S-Log 2 in the FS700, and still have some headroom.

I have never tested this but from your experience do you find that SLOG-3 holds a bit more dynamic range than SLOG-2? Obviously, the shadows, midtones and highlights are allocated or mapped differently between the two. I have read other say they still hold the same actual DR "information" from top to bottom.

I have not personally tested this.

================

As promised, I asked our Chief Engineer to review all of my previous posts in this thread for accuracy.

He agrees with all of my comments.

================

When using the Sony FS700 with the camera in 2K Raw or 4K Raw, there are techniques you can use to greatly reduce the noise in your images.

We make this very easy, and extremely effective if you have an Odyssey7Q or Odyssey7Q+, with our Raw Bundle and using our latest firmware version.

Many cameras are noisy in the shadow details, and for the FS700, in S-Log 2, you can overexpose by two stops.

You may select our Sony_EE_SL2_L709A-2 LUT in the Odyssey, which will allow you to see a normal image (not overexposed by two stops), while recording the image overexposed by two stops.

Then you can use our Sony_EE_SL2_L709A-2 LUT in Post to obtain great images without the noise.

This LUT will reduce the exposure in your images, in Post, by two stops, restoring the image to normal for Post, and the shadow detail noise will be eliminated.

If you and others send me a Private Message, I will send you a document on how to obtain low noise images from the FS700.

To do so, just click on my name, then click on Send Mail to Dan Keaton.

Respectfully,

[David Peterson]

Originally Posted by Dan Keaton

First, as you noted, the FS700 has S-Log 2 but no S-Log 3.

For us to properly work with the Raw data from the FS700, we have to have a known starting point, so we can process the Raw Data appropriately. Since a Rec. 709 output would limit the dynamic range of the data, we require that the Gamma = S-Log 2. In addition, we recommend (require for the best results) that you use Picture Profile 7 set to Sony's Factory Defaults.

Background, While it is possible to design a camera with zero image processing of the true raw data from the sensor, most cameras (as I understand it) perform some processing of the raw sensor data.

Applying Log encoding to the Raw data is a way to effectively record higher bit-depth into fewer bits.

For example using Log, one could record more than 12 Bits of data into 12-Bits. Using Linear, one can only record 12-Bits of data into 12-Bits. The Log encoding is a way to pack the most important part of the "more than 12-Bits" into 12-Bits.

Our eyes are more sensitive to shadow details and mid-range details, and far less sensitive to highlights. Thus, the Log curve is adjusted to provide the most data for the shadow details and mid-range details..

As you know, for Sony, they originally came out with S-Log, then S-Log 2, then S-Log 3. S-Log 3 is a tweaking of S-Log 2, but the FS700 camera does not support S-Log 3, at least not at this time.

This is good to hear that a gamma curve is applied by the FS700 before outputting its "raw" signal, which is how it squeezes more than 12 stops into 12 bits.

However I wonder what is happening in the details of when a 10bit ProRes HQ file is being recorded in the Odyssey from a FS raw signal from the FS700?

Is a further extra curve being applied to squeeze those 12bits of into down into 10bits?

[Gary Huff]

Originally Posted by David Peterson

This is good to hear that a gamma curve is applied by the FS700 before outputting its "raw" signal, which is how it squeezes more than 12 stops into 12 bits.

The color bit depth has almost nothing to do with the amount of dynamic range, outside of being better for retaining color fidelity if you're stretching the image severely, which 12 stops isn't much at all. Vlog-L does 12 stops.

[David Peterson]

It is an issue if you're trying to cram in more stops of dynamic range than the bits of space there are available to fit it.