Autor: Uli Mors

Working with the S-Log color space (2 / 4)

The basics of log curves and gamma.

All good things

Many years ago S-Log was implemented by Sony in some cameras with 12 f-stops dynamic range. With the development of sensors offering 14 f-stops dynamic range, the "old" S-Log became obsolete and was replaced by a new one: S-Log2. Some time later Sony modernized the cameras with a firmware update and another curve: S-Log3, which is also designed for the 14 f-stops of the F5/F55/FS7/FS5. Sony wanted to take S-Log2 off the market, but then decided to leave the user the choice. So why two S-Logs?

Pay attention: BLUE is the "old" S-Log2. What exactly happens here? Coming out of the highlights, S-Log2 has an excellent start. With over 110 values for the first f-stop, we can call it  "luxury greyscales". This level is maintained over about four more f-stops and then begins to "crash" slowly but increasingly.

At the 10th f-stop below maximum, only 35 grey values are available.
We can see from this curve that it is primarily concerned with mid-range and highlights, less with the details in the shadows. Sony places it very low in this gamma curve, which on a REC709 reproduction looks more contrasty than S-Log3 (here in red).

From the highlights, S-Log3 starts "moderately strong" - almost 80 values in the first f-stop below maximum is still a very good start, but S-Log2 offers about 50 percent more. S-Log3 keeps this level over about ten f-stops , without any significant restrictions. Ten f-stops.
Below the top ten f-stops the value ranges of S-Log3 slowly but surely decrease: At -12th f-stop still with about 30 greyscales, at the sensor minimum still with about 10 values, which then almost only have to handle the noise of the sensor.


An experiment. The light meter manufacturer Sekonic sells a measuring and calibration software for the large light meters L758 and L748. For the measuring procedure
a measuring panel with steps of 1/6 f-stops is filmed in a series of exposures, from which TIFFs are exported and then evaluated. This was the result of S-Log3 and the Sony PXW-FS7:

Well, I'm not the ultimate measurement professional, but what the software has presented to us actually corresponds to the S-Log3 curve published by Sony. You can take advantage of the different characteristics of S-Log:

* S-Log2: When it comes to mid-range and highlights

* S-Log3: When it comes to the contrast in the lower mid-range and shadow

S-Log3 is based on the Cineon (Kodak) film scanning gamma curve and has some similarities to ARRI's Log-C. Bottom line: S-Log3 should be more oriented towards common workflows from common film productions.

Now, Sony informs us in various documents that both S-Log2 and S-Log3 can handle 14 f-stops dynamic range and that a grey card is correctly exposed if it is applied six f-stops UNDER maximum white (sensor burns away).This point can be found (let the sensor nearly burn away, then stop down six f-stops). Because of the different value distributions Sony defines this point for ISO - 2000 as follows:

* S-Log2: Grey card on a waveform monitor at 32% IRE.

* S-Log3: Grey chart on a waveform monitor at 41% IRE.

Here's the bad news: You buy an 18% grey card. It reflects 18 percent of the incident light, which the human eye interprets as the mean value between black and white ("medium grey", neutral grey). These values can hardly be seen on Sony's own viewfinder WFM, but since V3.0 the zebra pattern can be set to such low values. This would be an exposure assistance for S-Log2 and S-Log3.


Since the introduction of the 14-f-stops sensors there has been controversy about it. My opinion: S-Log is a gamma curve, and gamma curves don't produce noise. If it happens, it is the sensor, and the gamma curve (here: S-Log) makes the noise visible.

However, a compromise has proven itself in practice. The point is, aren't +5 f-stops sufficient for overexposure and maximum highlights in practical use?

Or perhaps 4 f-stops? If we increase the light level on the sensor by one or two f-stops, the noise of the recording will be reduced by the same amount.

Let's stay with S-Log2 at first:

You can already do a lot with these zebra values in practice: If you deliberately expose two f-stops higher than originally intended by Sony, you can "target" a grey card with the camera and display with Zebra 55 percent. Also, bright image parts (white shirt, very bright elements in the image) could be displayed with Zebra 85 percent. In the linear part of S-Log2 about 11 percent correspond to one f-stop increase. For S-Log3 somewhat different values apply, since the characteristic is different: In the linear part of S-Log3 about 8 percent correspond to one f-stop increase.

These values also explain why a waveform monitor makes sense when shooting. A waveform monitor provides information at any given time about the exposure level of my subject and where in the picture there are still reserves upwards or downwards.

1. S-Log (especially S-Log3) is a technical means of covering the entire sensor scope to record in 10 bit. It is not intended to be presented without colour correction.

2. S-Log must be graded. This also means that image portions that the sensor has "seen" at -10 f-stops, accordingly need to be set low again in grading.

3. For this grading, log processing tools or targeted LUTs are recommended, a simple contrast enhancement may not be sufficient.


There is one thing to keep in mind for both S-Log2 and S-Log3: S-Log does not offer any smoothly burning highlights.

At +6 stops it's over, definitely. "Save the highlights" is the order of the day - the reserves for "nice" highlights or a nice roll-off have to be created by yourself or you use LUTs (Look up Tables) to expose. The F5/F55 and the FS7 have different LUTs in the Cine-EI mode, which make working with S-Log much easier.

The most "brutal" LUT in these cameras is called the "REC709 (800)" and grades the S-Log image down to about seven f-stops for the viewfinder or external monitors. Similarly, the Alpha 7R/7S II and the FS5 also have a contrast correction for the display, but the processing here is much different. LUTs and "Display Gamma Assist" are dealt with later in Part 3 in the next issue.

Related Articles: