Millivolts may describe the size of the container but not the quantity or value of the "information' contained within. It may be black, which is effectively "null" or every pixel could have a different value (maximum data). But even if every pixel has a different value, this may just be noise, also arguably a "null" set.
More importantly, millivolts only quantify the video container after it becomes a video signal. Log operates on the data BEFORE it becomes video. It is a remapping of the raw sensor data into the video signal that maximizes the potential of the optical information represented in the video stream.
Now, using the prior analogy, one can argue that there is less data in a Log signal than a Rec709 signal, at least from the point of view of a human viewer. But from the point of view of software that can manipulate the digital data, there can be information that would not have been captured in a Rec709 signal.
Granted, it may not be a lot more information, and it is not always better. The difference is only meaningful when comparing the two in optimized conditions. Hence, if you don't shoot Log correctly, you just make your life more difficult.
Cheers,
tod
The millivolts are a representation of the data whether it is compressed in a manner like LogC or some other choice or linear. They are a direct representation of the binary numbers that have been stored to represent the image. If there is less range to the millivolts then there is less detail or fewer steps in the actual levels being recorded. It's a Look Up Table so the data values get remapped accordingly. I know nature and humans perceive things in a log manner in both audio and video. The math behind all the transforms is something I have lost the ability to cypher like I could back in my engineering college years and even then I never got to the level of sophistication things like codecs and LogC transforms are doing in this case.
On my most Keep It Simple Stupid mode I would think if the math of the LogC transform used the full 700 millivolt range there would be more inherent detail to be had that could then be interpreted in a manner that is best for the particular system in use. If limited to Rec 709 map accordingly and if going to Resolve or other systems that employ floating point math then remap to a much greater range from the full 700 millivolts of detail/steps of levels.
In my simplistic mind I understand that LogC will lessen the detail in the shadows and bright areas. If I were to shoot in a controlled environment a calibrated Light Box with a perfect black to white ramp that in a pure Rec 709 world as a perfect diagonal line from 0 to 700 millivolts on my scope and then repeated that exact same thing only transforming the signal to LogC and then back to 709 emulating the LogC workflow would I not see more stair stepping in the diagonal line at the lower levels and upper levels as a result of the way the LogC favors detail in the mid tones? I'm not saying that this scientific experiment is a true indication of something the viewer would see as bad when it comes to a real image but this is what the science I'm trying to understand suggests to me.
As you say, "The issue for us is not really whether log contains more information, but rather whether it contains more useful information." This is where it gets into an even more grey area, no pun intended, for me. I know LUTs aren't going to go away so I'm trying to dig as much under the hood so I can better trouble shoot and deal with the various footage I get.
I did stumble on to some actual camera bars from one of our Panasonic Varicam35 VLog shoots and looking at what appeared to be 100% bars I did see full 0 to 700 millivolts on the scope with or without the Vlog to Rec 709 LUT applied. What I did see what a big difference in chroma with the unlutted material low in chroma and displaying a quadrature error. Then with the LUT applied the vector dots were closer to the 100% boxes but there was still a fair amount of quadrature error. I don't know exactly what the Camera Bars levels really are, would there be a different signal for color bars in 20/20 color space as opposed to Rec 709. I would think so based on what I see when I look at my scope and toggle between differing color spaces in Avid projects but I don't know what the Varicam 35 generates as a test signal. It was the first time I got a clip with some bars on it from the Vari 35 so I was excited but still left confused.
---In
avid-l2@yahoogroups.com, <hoplist@...> wrote :
On Aug 19, 2016, at 1:28 AM,
bigfish@... [Avid-L2] <
Avid-L2@yahoogroups.com> wrote:
When I see that I have 700 millivolts of range and it's now condensed into 500 to 550 millivolts of range there is clearly some detail being lost.
You cannot measure data in millivolts. That is a measure of the container, not a measure of the information in the container.
There is a revolution in data storage based on escaping preconceptions of the connection between "information" and the quanta of storage. Consider "a picture contains a thousands words." To fully convey all the information in a photo of a room would require many thousands of words, consuming many, many pages of paper and you could never convey all of the information in the picture. The picture is a more efficient way of conveying a certain type of information. And it is possible to convey that complete picture using a stunningly small amount of energy.
The issue for us is not really whether log contains more information, but rather whether it contains more useful information.
Cheers,
tod
