Don't turn around... I am just one step behind you...

Dietmar

It's difficult to tell what you're talking about.
If you're talking about the delineation between the colors within a gradient, which sometimes cause lines in the gradient, then adaptive supersamping should work to somewhat correct this.
Or maybe we/I need a better example of what you're describing to us.

No, that's exactly what I mean.
With adaptive supersamping, the lines disappeared a little bit, but not complete - you can still see them

Will have a look at the videos...

Dietmar

If you mean the 2 videos I made for Molly, then you're going to be at a loss. There is nothing about this issue in either of the videos I made.
Or maybe I misunderstood what you're asking?

Hm, then I think Molly misunderstood me...

Dietmar

What!?
Now I really don't understand...

@Dietmar
Try merging your two gradient layers (radial and bilinear) and then applying a large-radius blur eg. Filters>Blur>Gaussian Blur with radius 50px or even 100px.

I think the problem is that you are hitting the limits of what is possible with 8-bits-per-channel graphics.
Your bilinear gradient goes from white to black in about 320 pixels vertically; since 320 is not a multiple of 256, some values are present more than others, eg. there may be two rows with color #808080 but only one row with color #7f7f7f.
Meanwhile your radial gradient goes from white to black in about 1000 pixels, so there may be four rows or columns with color #808080 but only three with #7f7f7f.
Individually these gradients appear perfectly smooth, but when overlaid, where the gradients meet at an angle the slight irregularities in the gradients combine to produce a noticeable stepping.
Supersampling will reduce the undesirable effect but cannot eliminate it altogether.

You can see the problem and the difference that blurring makes in the histogram in Colors>Curves:
..histogram of the radial gradient (nice and smooth)


..histogram of the bilinear gradient (small peaks - some values present more than others)


..histogram of the two gradients merged (some nasty peaks)


..after applying 50px Gaussian Blur (much better)


..and after applying 100px Gaussian Blur (also much better)

It's difficult for me to explain. Sorry for my bad english.

I think jontait got it... will see.

Dietmar

Hi jontait2,

thanks for explaining this. I was quite near, but not near enough...

After playing with colors, gradients, blurring etc., I finally decided to "not use" the bi-linear gradient, because this is the main problem.

I finally did the radial gradient with adaptive supersampling and blurring.
Now it's ok for me

Dietmar