Hi Andrea,
I’m a little confused. You say you are doing the gradient in 16 bit RGB but you seem to be giving values in cmyk. At what point are you converting to 4c?
Are you making 4c proofs or so far only referring to the screen? I think ppl. are mentioning the fade so that you don’t get unwanted noise in a channel that should be zero (or 255). Sometimes in drastic cases I apply noise channel by channel.
Also the distance of the gradient is very important. A gradient in the cyan channel from 100c to 50c over 2 inches should be smooth as can be. From 80m to 70m over 10 inches is a percent step every inch. Without a lot of noise that might be noticable. Steps in the yellow channel are far less noticeable.
alan
And don’t forget that 16 bit per channel images don’t display correctly. They will often show banding on screen where none exists in the file. If you’re going to examine an image on screen, look at individual channels in 8 bpc.
16 bit/channel images display correctly.
But they display only 8 bits/channel because that is all your display can show. Without a greater display depth, or dithering for display (which can add more artifacts) you can see banding. That is because the data isn’t dithered for 8 bits, it’s trying to show you as much of the 16 bits as it can.
Since Photoshop can’t display cmyk natively, I guess that it would be showing banding that exists, and I have seen enough examples to know that I can see the exact same bands that we get on press.
Why doesnt photoshop display 10ths of a percent in cmyk files? I have had jobs mismatched by .3% and then fixed them. That would make it easier to deal with these issues.
I use a trick curve to exaggerate very flat images to adjust feathers and blends and I can create a tint of .7% and 1% side by side and they display and print differently but photoshop says they are both 1%.
It must be there because a) I can see it b) 100% divided by 256 steps is less than 1.
Working in cmyk, my experience is that if you see it at 50% view, it will print.
Also, in my experience in cmyk, adding noise cures banding, blur does not, presumably because it uses transparency, is that right? Gaussian blur causes banding.
Chris, you never got back to me about the ‘orphan artifacts’ in masks (emailed you samples). I think it is a serious (and addressable) issue for printers. If there has been any relevant development, please tell me. I am just a humble user, so I might have missed a feature but I am sure these guys in the forum would have picked it up and mentioned it by now, so I guess not.
Phil Taz
Phil – are you talking about the image display, or the info palette? Photoshop shows you the full CMYK image, converted to your display at 8 bits/channel.
Blur exaggerates banding by smoothing out the quantization. Adding noise is similar to dithering, and breaks up the quantization. It has nothing to do with transparency.
I had to hand the files you sent me off to someone else, and I haven’t had time to follow up on it.
I am referring to info palette.
So I am correct in saying that if you can see it – you will print it and that blur exaggerates banding…..everyone reading this should take note because almost without fail, the average (ie. non expert) user turns to blurring firstly to address banding.
Do you mean the quantization of the colour or the opacity? I presume that quantization is a necessary evil required when blurring.
If that is the case, I wish we had more dithering options rather than blurring options, I can’t see much use for it outside of the occasional soft focus effect whereas dithering should have its own palette and presets as a useful tool.
Thanks Chris, I appreciate how busy you are, but I hope you will ask the other person to follow up as it is critically important to everyone who prepares print files…..We have print jobs falling over frequently and being referred to the designers or charged for corrections because of banding issues.
Quantization of the color values – that is what leads to most banding. Quantization is a necessity because of the fixed number of bits per channel (8 bits – you only get 256 values per channel).
You have lots of dithering options. The trick is knowing when to use them and how much.
Thanks, I appreciate that. My point is that it should be an obvious option rather than a ‘trick’. As I mentioned, the logical step for most average users is to use blur to cure banding when that makes it worse….dithering is what works, so I am advocating a dither palette or feature that brings the options together like the blur submenu.
