1 person found this helpful
Photoshop has dither, that's a big advantage. And the option of adding noise.
And square pixels by nature help to break transitions between stripes, if the stripes are angled or curved.
In Photoshop you also have the option of working in 16 bit if you need to avoid banding on an extremely critical job. Even though the output is 8 bit CMYK, there is much less banding when you actually create the gradients in 16 bit, then convert to 8 bit as opposed to just making them in 8 bit.
If it's not really critical you might try making the gradients in Illustrator, then just rasterizing them in Illustrator. This can soften banding quite a lot sometimes.
Thanks, I understand the principle and I'll keep the bit info on-hand. In the meantime, I'll rasterize in AI before going into PS, unless the printer gives me grief.
I'm tired of being told to avoid banding I have to do my gradients in PS.
Banding is just as possible in a raster-based grad as in a vector-based grad. Far more pertinent are the parameters of the grad: What are the colors involved, and how great a distance must they span; what are the printing parameters (levels of gray, halftone ruling).
If you think it necessary to resort to rasterization and adding noise to disguise the banding, that can be applied as a live raster effect in Illustrator.
It's different for each piece of artwork. If someone is telling you:
- Photoshop grads ensure against banding.
- Always use raster grads.
They're suffering under misconception and are parroting gross over-simplifications and over-generalizations they've heard.
Another factor to consider is the number of inks in the gradation and how much each of those changes. If Cyan changes by 15% and Magenta changes by 25% then the bands, if any, don’t coincide and are less noticeable. But if bot Cyan and Magenta change by the same amount, the bands will line up and be more visible. Just changing one or or both colours by 1% (one higher, one lower) can help.
I'm printing your reply and gluing it to my office wall.