In our shop we frequently encounter banding issues and wonder if there are other solutions we can use to eliminate them besides adding noise. Suggestions?
Adding noise to the final image will fool the eye into not seeing the banding.
The issue can stem from a number of places:
1) it might not actually be in the file- with 8 bit monitors there are only so many versions of a color to go round. So if you have a blue gradient that crosses the entire screen and you are using the sRGB or adobeRGB gamut for your screen set up, there just aren't that many versions of blue in that gamut to allow for each pixel across 1920 of them to get its own unique color. So you end up with, say, 10 pixels in a row getting the same exact tone and hue before the next hue gets put in on the next 10 pixels - voila! stripes!
So you can fiddle with your screen/gfx card colorspace or get a higher bit screen. But thankfully, when printed this may not show up if this is the cause.
2) same issue as above but the color space of the file is insufficient to allow enough colors. This becomes doubly (or trebley) true when you start gradating from one color to another. There are only so many cyans in a specific color space and only so many yellows etc so you can start getting what amounts to moires building up as these two stepping patterns become interference.
3) some effects (like glows) can compress the rate at which a grad spreads out and interacts with a grad below (when using a non "normal" transfer function)
4) The mask or clippingmask of a layer may not have enough bit depth to prevent that layer interacting, causing banding with the layers below. This goes double for images brought in that are used as masks or content - check what bit depth they are.
5) some layering transfer functions can cause the colors to roll around the "end" of the color spectrum/gamut and depending on the function can begin to clip. Culprits include the grouping that starts with Softlight and ends with Pinlight. Darken and lighten also have funny math that causes clipping or "compression" depending on whats below the layer - in these cases try screen, add or multiply - the worst these do is add up to white or black (the end of all clipping)
6) any compression put on the image at the end (jpg for instance) immediately reduces the color available to the file - see above for number of colors available problem.
7) make sure you aren't looking at the preview proxy - its always horrible
We realize it could be the monitors and working to get calibrators to compensate. The solution to that is to zoom in quite a bit and if the banding is still showing then its in the file. We do not have any kind of proofing system and printing will assume the file is correct and run almost the entire job before noticing there is a problem. It still is like throwing darts in the dark.
The other thing that likely was causing it was that the RIP color profile was different from the software profile. We have now standardized both to Adobe U.S. Web Coated SWOP v2. This has helped, but there still seem to be situations where despite out best efforts the banding still shows.
As far as gradient color we generally do not mix colors. We will create a gradient with one side 100% fill color and the other the same color set to 0. If another color is needed a second background object is created with 100% of the second color. This has worked decently, but not always.
We generally use the default preview, but I have been working in Overprint Preview because we will get customer files with transparencies that are not supposed to be there, but without OP view turned on you can't see it and then it prints incorrectly. This has been real problem lately.
As for placed images they generally are right in Photoshop and then embedded so there usually isn't a problem. Although if making gradients in Photoshop similar banding issues occur and then you have to sort that out as well.
And we ALWAYS work in CMYK, and work in the highest bit depth our system will allow.
What I don't understand is Adobe used to have a method to assign how many steps you want in a gradient fill which eliminated banding altogether. This is no longer available so you have to stop and figure out which needle in a haystack is causing the problem. It should NOT be this difficult.
On the (halftone) printing side of things there are some additional considerations:
1) The three places where banding most likely occurs in gradients are (in halftone printing):
i) where a given plate goes to pure white (from small dots to no dots at all)
ii) where a given plate goes over the 50% line (from black dots to reversed out white dots)
iii) where a given plate goes to 100% (small reversed out dots to flat color)
i and iii above can be avoided by not ever going to full 0% or100% on any plate. It is a bit funky in that you don't get a pure white or a pure black, nevertheless this approach can be used.
Anyways, knowing that, you can try and avoid situations where a number of colors go over those thresholds at the same point in the gradient.
2) The smoothness is proportional to halftone dot resolution AND actual printer resolution (those are two different things):
- Generating halftone dots at PRINTER resolution of 300, 600, 1200 and 2400dpi native printer resolution give you different results. Higher dpi is better.
- Larger halftone dot sizes compensate for lower printer output resolution, so 72dpi is smoother than 150 dpi. Of course, you're more likely to see the dots at 72 dpi, so there's an optimum value between the printer resolution and the halftone dot size. Up to a point, larger is better for smoother gradients. If the printer native resolution is very high 2400dpi or more, this hardly matters.
- Not all digital printers use halftone dots, so this may or may not directly apply to your printing scenario.
3) As a result of the combination of the pattern generated by the printer, whether it is halftone or some other repeating pattern and the printer resolution you get banding at specific points along the gradient. Print out a white to black gradient and you will see where your printer has the biggest issues. The issues will be in the same locations on all the colors. The only one where it doesn't really matter is yellow.
Similar to point (1), avoiding getting those visible breaks stacking up in the same place on multiple plates can help this. (Unfortunately, the handling you're trying to avoid, adding some noise to the image, is the easier way to handle this.)
4) Output profiles can interfere with this as well, making things more complex. Putting a smooth 0 to 100% gradient through an output profile will shift the balance of the output, as 50% will not be output at 50%, etc. If you put an 8-bits-per-channel image through any profiles or color conversions, you're already starting to introduce imperfections to the originally (mathematically) perfect gradation. Again, the higher the native resolution of the output device, the smaller the issue.
- Converting from RGB to CMYK at output stage rather than working in CMYK to begin with causes similar issues. Solution here is simple enough: the artwork should be created in CMYK.
That's what comes to mind.
Senior Graphic Designer