Luc St-Pierre
Well-known member
Can someone tell me where to find a good technical explanation of typical banding, please no problem solving recipes, just the how and why.
Thank you
Thank you
Banding in vignettes
- Thread starter Luc St-Pierre
- Start date
rich apollo
Well-known member
T
Well-known member
It's the amount of gradient vs. the space it has to fill. I.E. If you have a gradient that is from 0% to 10% spanning across an 11" space, it will likely band in trying to fill the amount of space. If you have that same space filled with a gradient from 0% to 100%, it most likely won't band. It can use more steps to fill the dedicated area. That's how I understand it.
Anybody else??
T
Anybody else??
T
oxburger
Well-known member
think about it this way. say you have a 10" page with a vignette going from 0% color to 10%. If you divide the page width (10") by the ending percentage (10%) you will see that for every 1% increase in darkness of the vignette, you use a whole inch of paper (so basically you have 10 - 1" "strips" each with an increasing percentage of color value. Now, if you have a vignette going from 0% to 100%, the screen will increase in 1% in darkness every 1/10". (so, in this instance, you have 100 - 1/10" strips so the banding is not as noticable to the eye because the change in screen value occurs at smaller intervals: every 1/10" instead of every 1" and the "strips" are narrower).
Luc St-Pierre
Well-known member
Thank you all. I finally found a very usefull document that includes all the maths to predict and/or fix banding issues. Those interested here is the link
Banding Occurs in a Gradient or Feathered Part of a CMYK Photoshop Image
Thanks again.
Banding Occurs in a Gradient or Feathered Part of a CMYK Photoshop Image
Thanks again.
rich apollo
Well-known member
I think I found an error in the article, Luc.
"-- Increase the shades of gray in the gradient. For example, if you create a 0-10% gradient over 7 inches, the 11 shades of gray may be too few to create a smooth transition."
The math is incorrect here. Output is handled on a 0-255 scale, in an 8-bit workflow. 10% would yield 25.6 available levels of gray. If the element is a smooth shade, some workflows can render in 12-bit, or 4096 possible levels of gray. So, 10% would be 409.6 available levels of gray.
Gordo, where are you? Double-check me, please!
"-- Increase the shades of gray in the gradient. For example, if you create a 0-10% gradient over 7 inches, the 11 shades of gray may be too few to create a smooth transition."
The math is incorrect here. Output is handled on a 0-255 scale, in an 8-bit workflow. 10% would yield 25.6 available levels of gray. If the element is a smooth shade, some workflows can render in 12-bit, or 4096 possible levels of gray. So, 10% would be 409.6 available levels of gray.
Gordo, where are you? Double-check me, please!
Last edited:
gordo
Well-known member
You need to separate the shadestepping (banding) in the original PShop image vs shadestepping in the halftone reproduction.
If the original PShop image contains banding then the typical halftone screen (133-200 lpi) will likely render the banding faithfully.
The formula: – [output resolution/screen frequency] (squared) + 1 = The number of possible gray shades – only refers tho the tone values available in a single halftone cell. For example, let's assume a halftone cell that's 2 x 2 pixels. That means that the halftone dot can only have 0, 1, 2, 3, or 4 pixels turned on - thus:
So for each halftone cell, the number of possible gray levels is: 2x2 squared = 4 + 1 = 5 possible levels of gray, in the above case 0%, 25% 50%, 75%, and 100%. Increasing the dpi, i.e. keeping the halftone cell size the same, but dividing it up into smaller pixels will give more gray levels. E.g., if the cell size is 3x3 then there are 10 possible gray levels, a 4x4 cell gives 17 gray levels and so on.
However we do not see single halftone dots but areas of multiple dots instead - and this gives us a way to work around the gray level limitations imposed by the formula.
For example, here is an area made up of (2x2 pixel) halftone cells from the original example.
I can achieve a 50% tone. However, I'm limited to a 75% tone as the next darker one or a 25% as the next lighter one. Big jumps in tone. However, by "dithering" adding a dot or removing a dot from some cells within the tone area I can increase the levels of possible grays:
In the above sample, I can go from just 3 levels: 25%, 50% and 75% to 5 levels: 25%, 40%, 50%, 58%, and 75%.
This dithering strategy to get around gray level limitations has been used by most screening technologies since about 1990. It has also meant that 2400/2540 dpi has become the defacto standard dpi of imaging systems for both film and plate setting even though the dpi/lpi formula suggests that the resolution is not high enough to give you all the gray levels required to avoid shadestepping in gradients.
best, gordo
If the original PShop image contains banding then the typical halftone screen (133-200 lpi) will likely render the banding faithfully.
The formula: – [output resolution/screen frequency] (squared) + 1 = The number of possible gray shades – only refers tho the tone values available in a single halftone cell. For example, let's assume a halftone cell that's 2 x 2 pixels. That means that the halftone dot can only have 0, 1, 2, 3, or 4 pixels turned on - thus:
So for each halftone cell, the number of possible gray levels is: 2x2 squared = 4 + 1 = 5 possible levels of gray, in the above case 0%, 25% 50%, 75%, and 100%. Increasing the dpi, i.e. keeping the halftone cell size the same, but dividing it up into smaller pixels will give more gray levels. E.g., if the cell size is 3x3 then there are 10 possible gray levels, a 4x4 cell gives 17 gray levels and so on.
However we do not see single halftone dots but areas of multiple dots instead - and this gives us a way to work around the gray level limitations imposed by the formula.
For example, here is an area made up of (2x2 pixel) halftone cells from the original example.
I can achieve a 50% tone. However, I'm limited to a 75% tone as the next darker one or a 25% as the next lighter one. Big jumps in tone. However, by "dithering" adding a dot or removing a dot from some cells within the tone area I can increase the levels of possible grays:
In the above sample, I can go from just 3 levels: 25%, 50% and 75% to 5 levels: 25%, 40%, 50%, 58%, and 75%.
This dithering strategy to get around gray level limitations has been used by most screening technologies since about 1990. It has also meant that 2400/2540 dpi has become the defacto standard dpi of imaging systems for both film and plate setting even though the dpi/lpi formula suggests that the resolution is not high enough to give you all the gray levels required to avoid shadestepping in gradients.
best, gordo
Luc St-Pierre
Well-known member
Gordo and Rich
Thank you for the precisions. I have a question. My outputs are always done at 2540 DPI. My workflow and RIP will create a Dithering of 2. If I finf myself in deep trouble with a banding issue, instead of going to 4000 DPI, what would be the issues if I raise the dithering to a higher strength. How much of it could I add to smoothen vignettes without affecting other CT images?
Thank you for the precisions. I have a question. My outputs are always done at 2540 DPI. My workflow and RIP will create a Dithering of 2. If I finf myself in deep trouble with a banding issue, instead of going to 4000 DPI, what would be the issues if I raise the dithering to a higher strength. How much of it could I add to smoothen vignettes without affecting other CT images?
gordo
Well-known member
IMHO, if you've determined that the banding/shadestepping is not in the original art then I would:
a) consult the manual that came with my RIP
b) contact my workflow support group
c) run a press test (single color using vector and PShop gradients of various lengths and tone ranges)
d) consider FM screening as there is normally no shadestepping with FM screens.
best, gordo
a) consult the manual that came with my RIP
b) contact my workflow support group
c) run a press test (single color using vector and PShop gradients of various lengths and tone ranges)
d) consider FM screening as there is normally no shadestepping with FM screens.
best, gordo
juergenroesch
Active member
Luc,
Gordo starts here with a point you might easily miss:
"if you've determined that the banding/shadestepping is not in the original art"
so if you talk vector gradients you should be fine, but if you have background gradients for example in an RGB image file which has then been separated using ICC profiles, you might get some problems there.
GMG has 2 test files called smoothCheck - 1 RGB and 1 CMYK which are nothing than different blends representative of all different areas of these color spaces. They are perfect for testing color space conversions, because if something isn't quite right somewhere will bands, breaks or spots reveal such problems. I think they can be downloaded from their website Home of Color : GMGColor or contact their online support to get these files if you think there might be your problem
Gordo starts here with a point you might easily miss:
"if you've determined that the banding/shadestepping is not in the original art"
so if you talk vector gradients you should be fine, but if you have background gradients for example in an RGB image file which has then been separated using ICC profiles, you might get some problems there.
GMG has 2 test files called smoothCheck - 1 RGB and 1 CMYK which are nothing than different blends representative of all different areas of these color spaces. They are perfect for testing color space conversions, because if something isn't quite right somewhere will bands, breaks or spots reveal such problems. I think they can be downloaded from their website Home of Color : GMGColor or contact their online support to get these files if you think there might be your problem
Similar threads
