I have been trying the achieve a smooth colour gradient in one of my PS drawings using the gradient tool. No matter what I do I alway get banding. The worst scenario is when I try to use radial gradient.
I’m using large () 8bit psd files on a colornavigator calibrated Eizo CG301w. (Vistax64)
Greyscale gradients are really smooth but the colour gradients are not….arrgh.
Any tips or advice as it’s making progress a bit difficult.
Tried the noise thing….doesn’t really work in this case.
What I have found is that colour gradients seem perfect at lowish file sizes up to 50cm, anything above that and gradient breaks down…..in most colours but is more predominant in the blue / dark blue(almost black) zone.
I should also add that I am using multiple gradients to create ‘realistic’, interesting skys and what I get is a colour distortion similiar to the ‘web only’ colour palette ( quite oblivious seperations in tone and colour ).
I’ve made much larger images than this before with gradients and don’t seem to remember any banding…….but, maybe that’s a colour range thing!!!
……I had a thought, I am going to try it in a different way with multiple layers (one colour on each) and have them blur into one another and then do a flatten visible layer and see if that makes for a better, smoother, end result….so much work for something really simple. 🙁
If you have a high exposure adjustment layer or other adjustment layer? If the image have high exposure, the gradient will have band. Try to reset the preference if still reproduce.
There is a reduction when viewing at 100%. To fit the whole image onto the screen I’m working on it at 12.5 – 33.33% range so I guess the image is compressed in some way.
Upon closer inspection it looks like some of the banding is ‘locked ‘ into the image… perhaps during a flattening process while some layers were blurred ( Gaussian blur can cause some nasty banding in some colour / layer blend mode combination ).
I have since smudged out the worst of it as I need to crack on for a deadline.
I do wonder if it’s a profile issue – I say that as the banding shift, but not eliminated, when I adjust the RGB/CMY colours in the Eizo profile software.
Next step is to print out a section at 100% to see if it’s there in a hard copy….important as these are for print.
I’m using a CRT monitor and I only see the bands when I look at the individual channels (G and B). Even then, the differences in the RGB values are increments of 1 RGB value (ex 39, 39, 39 to 40,40,40) So the problem does not have to do with too large a difference between grad steps. The actual width of those steps, however, could be what is causing your problem, and the steps do look very patchy, jpeg artifact-y. Combined with the fact that you’re looking at it on an LCD monitor could be making it seem like more of a problem than it really is. I played with it by adding some noise and it at least helped break up the patchy look.
The banding in the sky looks a bit like compression artifacts to me too. And if it was a monitor profile problem, would we see the banding?
I would like a crop of the psd, to see if I get the banding. I can check it in a couple of hours from now at home, on an iMac and a laptop, and tomorrow on my big Eizo in the studio.
The Eizo should be able to handle colour gradients as it has a 12bit hardware LUT….But that’s a good point about the profile there Rob.
The file has changed a fair bit but I can dig out a crop on an earlier version that will show you what is happening. I appreciate you checking it for me.
The file is quite big so I’ll be careful to crop a ‘good’ part of the banding while keeping the overall file size down and I’ll it have as a download from a link.
I’ve been playing with different profiles and I noticed with my Eizo set to D65 & 6500, 2.2, 80cd, during validation, I have one of my blue colours (0,0,255) at DeltaE 4.7….
With the monitor calibrated to D50,2.2, 80cd the same colour is DeltaE 0.8.
Could the colour temperature have an effect on the way gradient are perceived?
Do you think it’s safe to use the monitor at D50? Colours and gradients are better but I’m concerned about colour matching as all my artwork is for print output.
I wouldn’t do that. It’s contrary to every standard I know. (for monitors, not evaluation light)
I opened your psd file in my PS and I have banding. I try to fool around with it (extreme Gaussian Blur) and I just seem to reshuffle the banding.
However, if I make a new file with the same colour gradient etc. I get a smooth gradient.
I’ve tried this with PS CS3 and CS4, both at 8 and 16 bitch (sorry, bit/channel)
I’m too old for this. I should be sipping Pinacolada and spend big revenue money on some warm island, but the studio heating collapsed, and it’s minus 20 Celsius outside and soon inside.
-20c Holy [insert appropriate religious symbol]….that’s cold – London is coollldddd at 2c day and -8c at night…so I would not like to have no heating at your temps – I hope you get it fixed soon.
Thanks for checking the file…I guess it’s just one of those things.
I have sort of worked around the issue by working over it. Gaussian blurs never seem to resolve banding for me either…..
It’s good to know that D50 is not a good idea……just a little concerned about the high 0,0,255 deltaE at D65, that’s the only one, other readings are good…..I shall revert to D65 despite this.
You can’t blur banding away, it usually makes it worse. I hate making a call on banding on anything but a CRT because all the LCDs I have access to are prone to show banding when it isn’t there.
Here’s a note from my tips folder on banding in Photoshop. I have no idea where it came from so I can’t credit the source.
Regarding banding.
You should not be seeing the banding on screen.
If your output is to a postscript 3 imagesetter/CTP device, then you will have nothing to worry about. PS3 can achieve 4096 gray levels.
If printing to a PS2 device, then you need to incorporate adding noise to break up the banding visually, reduce the line screen or use formulas to adjust the gradient.
Reduce line screen or print out on a higher resolution device.
Basically the formula for ps2 output device to retain 256 gray levels is the Printer’s resolution divided by 16= maximum line screen.
So, if you are outputting to a 2400 dpi imagesetter, your line sceen can be no higher than 150 or you start to lose gray levels upon output.
Another formula that needs to be considered is to determine maximum gradient length based upon percentage change.
Formula: percentage change (256) = number of steps 256 is number of gray levels percentage change is difference from the greatest to least percentage in the gradient. If your gradient goes from 30% to 80%, the percentage change is 50%. If you are using cmyk colors, the you use the greatest channel difference. If c10,m20,y30,k40 to c70, m60,y50,k60, the the greatest difference channel is the cyan with a 60% difference.
After figuring out the number of steps, multiply the step # by desired measurement unit.
inches: .03 cms: .0762 points: 2.16
so for the sample cmyk color gradient with the greatest channel difference of 60%
60% (256) or .6(256) =153.6 steps
153.6(.03) = 4.608 inches.
So upon output, if we print to a 2400 dpi imagesetter, the linescreen can be no greater than 150 lpi and the size of the gradient cannot exceed 4.608 inches otherwise reduced gray levels will occur which will increase banding.
I have an older CRT knocking around I might hook it up and proof images via that screen for a while to see what’s going on……Eizo’s are supposed to be the one for digital colour repro work though.
If it’s the monitor, I wouldn’t see the Info Panel’s RGB numbers jump when I mouse/pen over the banding transitions: I’d see a gradient function.
And, when I start a new file, I don’t get this. I often dramatise skies using gradient layers/masks etc. but I only get banding when I get too far, and when not still in 16 bit. But then usually only the shadows show banding and noise blobs.
Got an Eizo too here. BTW he doesn’t shift, or slows down at cold temps like my bicycle CatEye.
Heating’s repaired. frozen pipe. Thank goodness. Hate editing in ski clothes. so:
I’m too old for this. I should be sipping Pinacolada and spend big revenue money on some warm island, but the studio heating collapsed
I’m a bit younger again, got my senses back and f..k Pinacolada.
There are a lot of broken up patch areas. It looks like there were gradients made with different colours from different directions and they added up to a bunch of stepped shapes. I disagree that blurring cannot solve banding issues. Generally, adding noise helps break up banding, BUT if the banding is due to RGB steps that are greater than 1 value, blurring can help introduce the in-between values that are missing. I found one way to fix the banding in the sky. Switch to 16-bit mode. To get rid of the stepped patchy areas and give them a little more uniform gradient look, you could do a large 40 pixel (or more, or less, to taste) gaussian blur or median (I think blur works better though). This will fix the patchy shapes and give you more uniformly-shaped bands in your sky. In your colour settings, make sure that "use dither(8-bit/channel images)" is checked on, then convert back to 8-bit. The dithering that’s introduced will further break up the edges and give you a nice smooth-looking sky. (history-fill back in the mask on your layer to it’s original state)
Thanks for the tips and input…..’use dither(8-bit/channel images’ is alway enabled.
I think it’s an issue with the image size as they are big(ish) at 80x50ins. On smaller images (10x15ins) the gradients are good, not perfect, but you have to look close to find stepping in some parts.
I tried the 16bit path and they are better but I still see banding. I have just resigned myself to the fact I need draw the skies and not rely on the gradient tool too much. The end result is more interesting that way anyway..
If you have a spare moment, in the psd, how does the lower gradient look? I produced that with just the gradient tool – on my screen it did have minor stepping….If I remember correctly the layer above was the result of two layers merged which could account for the directional stepping.
Chris Sorry for late response I can get the same bands as yours on my imac on both Photoshop CS3 and CS4 It seems CS4 is better than CS3 and drawing on RGB mode is better than drawing on CMYK mode Ping
That’s a very big file. Understand that you only have a certain number of steps that can define a gradient. If you are pulling a grad across a very large distance, bands will be wider and more noticeable. If you are starting with a colour that is very light in any of the channels, you have fewer steps than if you started with a darker colour. (example, take a solid 100% cyan and pull a grad to 0%, you will have smaller and smoother-looking grad steps than if you pulled a grad the same distance going 10% cyan to 0%) Each channel in 8bit will actually give you up to 256 grad steps. Creating a grad in 16 bit mode will help smooth things out. If you have to convert back to 8 bit, make sure dither is turned on in your colour settings as it will help put a little noise into the bands’ transitions and give a smoother appearance.
Learn how to optimize Photoshop for maximum speed, troubleshoot common issues, and keep your projects organized so that you can work faster than ever before!
Related Discussion Topics
Nice and short text about related topics in discussion sections