Gain should go down to 0

Simple as that. At the moment the minimum is somewhere just above 0, like 0.01 or something.

[Peter Chamberlain]

Which gain?

Lift Gamma Gain - Gain

Simple as that. At the moment the minimum is somewhere just above 0, like 0.01 or something.

The numbers displayed are kind of meaningless. It was explained to me years ago by one of the original daVinci Systems engineers that the actual value stored internally is a hexadecimal number, and what's displayed to the user has only a very rough correlation to what's happening to the image: if the number goes up, the level goes up. If the number goes WAY up, then the level changes much more.

Note that there's not a lot of rationality for levels: 0.00 is kind of a "unity gain" (zero change) level for Lift and Gamma, but 1.00 is the unity level for Gain. And 25 is the unity Offset level, which relates to laboratory printer lights used for film printing. Unity for Saturation and Hue is 50 (!), and yet it's 1.0 on the Sat vs. Hue curve. Interestingly, 0.0 is unity on the HDR controls. Unity for Sat on the Color Warper appears to be 1.16 (don't ask), yet for a Power Window, the unity setting for Size/Pan/Aspect/Tilt is 50.00. Frustratingly, with PTZR you can zoom in to "100" (based on a normal of 1), but you can only zoom out to 0. I suspect some of these numbers are legacy stuff carried over from 20 years of daVinci color correctors and 15 years of Resolve. You eventually get used to it, particularly with a control surface.

If it were up to me, I'd say "0 is normal in all modes, then you can go down to -100 and up to +100 for changing them by a hundred percent either way." Maybe go to 500 for position changes.

I generally tell students: "for color, ignore the displayed GUI number and just look at the scopes and a calibrated monitor." Get a feel for how the image looks, and disregard the numbers, assuming the color science and Raw settings are correct. I do pay attention to the numbers if I see a mismatch in consecutive shots, and then I know a knob got inadvertently bumped.

Yes, there's a lot of weird stuff going on in the displayed numbers. for gamma and gain it mathematically would make the most sense if the default was 1. gamma is a power function so x^1 = x. gain is a multiply (x • 1 = x)

the global wheel does a x • 2^(number). so when number is default at 0 it means x • 2^0 = x • 1 = x
having unity at 0 makes total sense, because you're working with stops, actual real life values. like changing the value by 0 stops does nothing. changing to +1 effectively multiplies by 2. +2 --> multiply by 4. thats great, works as expected

actually the problem I see in my original post is not the displayed values, rather the fact that gain cannot reach 0. you cannot multiply by 0 here and therefore make everything a solid black. it get's quite a problem if you wanna work in a linear color space because those small numbers actually make a big difference

actually the problem I see in my original post is not the displayed values, rather the fact that gain cannot reach 0. you cannot multiply by 0 here and therefore make everything a solid black. it get's quite a problem if you wanna work in a linear color space because those small numbers actually make a big difference

I think one of the problems is that sometimes you're dealing with a linear value, sometimes it's a log value, and sometimes it's in a completely different range.

If it's just a case of forcing the image to show a solid black, I rely on scopes and my eyes to show me the adjustment I need. I frequently use a Log Shadows node near the end of my fixed node tree as a final step to fine-tune the black level to make it as deep as I need it to be, though it's rare that I crush the shadows to the extent that it's absolutely black. There are always exceptions. But for me, while the monitor and GUI numbers can sometimes mislead you, the scopes never lie: if it's 0, it's dead black.

If you haven't read it already, check out Steve Hullfish's free Tektronix book "Using Waveform Monitors as Artistic Tools for Color Grading":

https://www.tek.com/en/documents/primer ... resolution

This goes into great deal how to interpret what an image is really doing based on what the waveform monitor, vectorscope, and parade scope tell you.

actually the problem I see in my original post is not the displayed values, rather the fact that gain cannot reach 0. you cannot multiply by 0 here and therefore make everything a solid black. it get's quite a problem if you wanna work in a linear color space because those small numbers actually make a big difference

I think one of the problems is that sometimes you're dealing with a linear value, sometimes it's a log value, and sometimes it's in a completely different range.

If it's just a case of forcing the image to show a solid black, I rely on scopes and my eyes to show me the adjustment I need. I frequently use a Log Shadows node near the end of my fixed node tree as a final step to fine-tune the black level to make it as deep as I need it to be, though it's rare that I crush the shadows to the extent that it's absolutely black. There are always exceptions. But for me, while the monitor and GUI numbers can sometimes mislead you, the scopes never lie: if it's 0, it's dead black.

If you haven't read it already, check out Steve Hullfish's free Tektronix book "Using Waveform Monitors as Artistic Tools for Color Grading":

https://www.tek.com/en/documents/primer ... resolution

This goes into great deal how to interpret what an image is really doing based on what the waveform monitor, vectorscope, and parade scope tell you.

sure there's always a way around it. currtenly I am using the contrast with pivot set at 0 (even with s curve enabled, constrast behaves linearly if smaller than 1).

but like that's a work around. and changing the minimum value of a tool shouldn't be too complicated. I just think, if I where to design a grading tool, where would I set my default and min/max values. minimum at 0 makes sence, you wouldn't want to go negative to avoid artefacts. minimum at 0.01 hmm.. why tho..
like there is no practical benefit of hindering the user to go lower and forcing him to work around it.

imagine opacity would not go lower than 0.1, you'd have to fake it with input opacity or whatever. imagine saturation stops at 5. to achieve greyscale you'd use multiple instances in series and hope for the best

I think one of the problems is that sometimes you're dealing with a linear value, sometimes it's a log value, and sometimes it's in a completely different range.

Multiplication doesn’t care what kind of meta-interpretation user attaches to values. Unless you can tell which of these values is linear and which is log, this is irrelevant: 0.172 and 0.172

Are there any changes in DR18?

[Peter Chamberlain]

No changes planned.

The process, math, and numbers have been this way since, well before we introduced 32bit float processing... which BTW never really gets to 0.

Is there any reason why it doesn't go to 0?

No changes planned.

The process, math, and numbers have been this way since, well before we introduced 32bit float processing... which BTW never really gets to 0.

(It should go to 0)...