GH5 Waveform Display Calibration in VLog

[Roger Keay]

The waveform monitor in the GH5 seems to be incorrectly calibrated in Vlog mode. The waveform display never exceeds the second dashed line below the solid top line. Increasing exposure from a normal level shows compression and hard clipping at the second dashed line with no transients above. This observation suggests that the second dashed line corresponds to 1023 bits.

Has anyone compared the internal GH5 WFM display with an external monitor?

[Christopher Young]

This is sounding very much like the same queries relating to the GH4 a while back. Check out Adam Wilt's V-Log article surrounding these levels issues. I think it will give you some pointers.

Not dissimilar to working with S-Log on the Sony cameras with their white levels of 59 and 61 depending on whether it's S-LOG 2 or 3. Which in turn means your skin tone levels also come down.

On the GH5 you can display a 709 LUT to the screen:

"Select “V-LogL View Assist”
Select “LUT Select”
Then choose “Vlog_709”

I haven't checked but does this LUT not re-scale the WFM levels?

For example on the some of the Sony's when you monitor via their Rec 709-800 LUT it re-scales the WFM to your more standardized 70 IRE skin and 90 IRE peak levels.

https://www.provideocoalition.com/v-...4-don-t-panic/

Chris Young
CYV Productions
Sydney

[Cary Knoop]

Originally Posted by Roger Keay

The waveform monitor in the GH5 seems to be incorrectly calibrated in Vlog mode. The waveform display never exceeds the second dashed line below the solid top line. Increasing exposure from a normal level shows compression and hard clipping at the second dashed line with no transients above. This observation suggests that the second dashed line corresponds to 1023 bits.

Has anyone compared the internal GH5 WFM display with an external monitor?

The waveform is correct, V-LOG-L is ranged between 7.3 and 80 IRE which translates in to 10 bit code values of 128 and 765

[Christopher Young]

Right.

So it is just a matter of understanding this and exposing accordingly. Which is exactly the same as on the Sony's if you are running your own 709 monitoring LUTs the WFM levels you are seeing are reflecting the true LOG exposure levels.

Thanks good to know.

Chris Young
CYV Productions
Sydney

[Roger Keay]

Thank you for your explanations of V-Log L levels.

It seems to me that the WFM display as implemented wastes space on the EVF in V-Log L mode. The waveform will never enter the top third (roughly) of the WFM display but the virtual graticule remains overlaid on part of the image. Perhaps a 0 to 100% scale (765 code value) would be more useful with lines to represent 90% diffuse white and 18% grey. A relational display of this type could also be used for HLG when it becomes available on the camera. The 100%, diffuse white and grey lines would have to be matched to the HLG curve.

Fiddling with the display as suggested may complicate use of the GH5 in multi-camera settings but the relational display should help the large number of users who shoot single camera.

[Tom Roper]

You're not looking at the specular transients.

In Resolve:

Color Science to ACES.
No input transform.
No output transform.
Luts disabled.
Bypass all grades.
Waveform monitor brightness full. (So you can see the transients)

V-Log-L transients are written to code value 965 and may even go higher. You have to consider the signal to noise ratio for specular highlights is not identical to the saturation level.

The waveform monitor in the GH5 LCD/Viewfinder is coarse and not showing you the specular transients.

My Sony F55 in S-Log2/3 behaves similarly.

[Cary Knoop]

Originally Posted by Tom Roper

You're not looking at the specular transients.

In Resolve:

Color Science to ACES.
No input transform.
No output transform.
Luts disabled.
Bypass all grades.
Waveform monitor brightness full. (So you can see the transients)

V-Log-L transients are written to code value 965 and may even go higher. You have to consider the signal to noise ratio for specular highlights is not identical to the saturation level.

The waveform monitor in the GH5 LCD/Viewfinder is coarse and not showing you the specular transients.

My Sony F55 in S-Log2/3 behaves similarly.

Sorry Tom but you are wrong.
V-Log does extend beyond this but V-Log-L does not.

Make sure you explicitly set the clip attribute to data levels otherwise Resolve interprets the clip as video levels.

[Tom Roper]

Hi Cary,
Levels were set to data explicitly. If you are looking at the code values anywhere else but inside the written mp4 files it will clipped. For example, if you monitor Log levels over hdmi, you are only seeing log inside broadcast 2.2 - 2.4 gamma. Levels will be clipped. If you are recording log externally that way, recorded levels are also clipped.

[Cary Knoop]

Originally Posted by Tom Roper

Hi Cary,
Levels were set to data explicitly. If you are looking at the code values anywhere else but inside the written mp4 files it will clipped. For example, if you monitor Log levels over hdmi, you are only seeing log inside broadcast 2.2 - 2.4 gamma. Levels will be clipped. If you are recording log externally that way, recorded levels are also clipped.

Sorry but this is just nonsense, you are simply wrong and apparently have a hard time admitting it!

Did you even try your own suggestion using ACES?

Here is a screen shot using the GH5 in V-Log-L fully blown out, using ACES with all transforms at bypass.

[Cary Knoop]

And here is one with minimum exposure.

As you can see code values range between 128 and 768, as designed!

[Tom Roper]

Cary, the sensor is in saturation when the whole image or a large part are blown out. The transient peaks of a properly exposed image are not limited to 765 by VLogL. When the sensor is in saturation, it will only write up to the code value of the sensor saturation point.

[Cary Knoop]

Originally Posted by Tom Roper

Cary, the sensor is in saturation when the whole image or a large part are blown out. The transient peaks of a properly exposed image are not limited to 765 by VLogL. When the sensor is in saturation, it will only write up to the code value of the sensor saturation point.

More nonsense!
I give up, there seems to be no point arguing with you!

[Tom Roper]

Cary,

The GH5 is recording specular highlights above 80 IRE, the waveform monitor is clearly showing this. If you clip the entire sensor then yes 80 IRE.

The attached V-Log-L 10 bit 422 image has bypassed grade, no lut, no idt or odt.

https://drive.google.com/file/d/0B7w...ew?usp=sharing

[Cary Knoop]

If you upload the out of camera clip I will take a look at it.

You may have inadvertently setup all kind of things in the Color Management page (transforms, LUTs) that misleads you in thinking V-Log L is not limited at 80IRE.

[Cary Knoop]

Here is a DCTL to see what, if anything, is coded at or above 80IRE (full range):

__DEVICE__ float3 transform(int p_Width, int p_Height, int p_X, int p_Y, float p_R, float p_G, float p_B)
{
const float min = 0.8f;

const float r = (p_R < min ? 0.0f : p_R);
const float g = (p_G < min ? 0.0f : p_G);
const float b = (p_B < min ? 0.0f : p_B);

return make_float3(r, g, b);
}

Replace min with .75f for 80IRE full range equivalent.

What noise level do you select for V-Log L in camera?