As gheto said, you are using too little differentiation between the individual levels. The following happens: Your colors are 19:36:44 (bottom right) and 69:82:90 (gradient center). With a little math we arrive at the following: Red: 69-19=50 Green: 82-36=46 Blue: 90-44=46 Therefore the maximum available number of discrete color steps in 8bit is 50x46x46=105800 These colors spread across an area of about 928x555 pixels, ergo 928x555=515040 105800/515040=0.208...- meaning that within that field not each pixel can have a unique color (te result would have to be 1 or higher). The ratio of available colors to required ones is about 1:5. Therefore the only thing the program can do, is to use the same color for multiple pixels, which in your case happens to occur along the radius of the gradient circle. Now, as suggested, adding noise may make it vissually less noticable because it introduces a random element, but even this form of random dithering does not change the fact that you still are running low on colors. The logical conclusions therefore must be: - use the maximum color range you can use, i.e. b/w gradients, wherever possible - color them using gradient tints or by putting layers above them and using blending modes and opacity adjustments - work in 16bit or 32bit to get more color levels - save to 16bit or 32 bit formats - only reduce the gamut/ color range for final output with respect to color profiles and other requirements, e.g. in web graphics - if you work in 8bpc, make sure to not exceeed certain dimensions that cannot cover the limited fidelity Mylenium
... View more