Re: [Digital BW] Tips on Curves [Was Differences in perception of tonality ...]
2001-09-18 by Paul Roark
Howard wrote: >I'm using Paul Roark's curves and no, it's not methodical. >Trial and error. ...I've begun to develop just a little intuition... I'm very pleased that Howard and some others are starting to dive into the curves. Can I retire now? Here are some tips that come to mind on dealing with these curves. Which scale to use -- The Curves dialogue box has two options, the 0% to 100% ink scale and the 0 to 255 light value scale (my terms, I don't know what the proper terminology is). The arrows on the horizontal axis bar switches between them. Use the 0 - 255 scale, which has the black of the horizontal axis bar on the left. The other scale does not allow fine enough adjustments. (Needless-to-say, use the method that allows points to be placed and moved.) To move a point, one must click on it. The problem is that the point tends to move. So, be very careful to hold the mouse still when clicking on the point. I often use both hands on the mouse to hold it. I also have my mouse speed on the second slowest setting. If you think the point has moved, hit Edit and see if the Redo is highlighted. If it is not grayed-out, you've moved the point. Just hit Redo and try again. There is not complete consistency in how far to move a point up or down to achieve a one percent change in the image density. In general, however, the lighter inks can take a 6 unit move and the cyan a 3 unit move to make a 1% change in density. As all the curves approach the black point, the driver starts to feed in black. As such, the amount of movement to affect a 1% change decreases. If all the inks are in the darkest quadrant, a 1 point move will be close to a 1% density change. The toner and lightest gray ink are about the same density in 4-ink printers. As such, one can change tone without much affect on density by moving the two inks in opposite directions. So, if you want less coolness/blue, move the toner up a few points, and move the lightest gray down the same number of points. In 6-ink printers the light toner is about 60% the density of the full strength, so the offsetting moves will vary as you move down the curve, with a 1:1 move working in the bottom/darker half of the graph (good thing, because you don't want the black ink to be affected by offsetting moves of the toner and lightest gray ink). The lightest gray ink has several uses. First, of course, it gives you dotless highlights. Also, however, it can give you control of the black ink indirectly when all the curves are in the bottom left (darkest) quadrant. The lightest gray is so light that it does not have that much effect in the central part of the image, so you can go to a negative slope. You can, in fact, have a very "S" shaped curve without problems. The lightness of the gray also allows fairly radical turns without having visible artifacts, as long as the black ink is not being used yet. Thus, it is good for fine-tuning. With the cyan -- main gray -- ink, keep the curve smooth. It is dark enough that sharp turns can leave artifacts. You'll find that PS only allows so many points on these curves -- fewer than you would like. So, allocate the scarce points were they are needed most. I've left the combined RGB curve black. I think this is the curve that individuals can best use to match the gamma of their monitor to printer. For starters, make a very large 50% spot and 100% spot on a piece of paper in PS, print it, and see how they look compared to your monitor (which I assume has been adjusted with the Photoshop Gamma routine). The combined RGB curve can easily modify the gamma of the print. Put a point in the middle and drag the curve up or down as needed, and try another print. This combined RGB curve may also be a handy place to fine tune curves to individual machines, as it hopefully will not cause the color balance to be thrown off. (I admit I have not done much with this RGB curve yet.) Very steep slopes can leave artifacts -- humps and dots on the 3000. This is especially a problem when trying to get the maximum warmth, because the toner ink must go from totally off to full-on to turn on the black ink -- in a rather short and sensitive (because of the black ink) space. At one point I avoided going for maximum warmth because of this problem. However, most of the warm curves are now max'd out. Use a step-wedge test file to make adjustments. I made one that has both the ink percentages and approximate theoretical (linear) RGB light value units on it. (I'll send a copy to Martin so it can be shared.) Note that the color space affects the actual conversion from the grayscale to RGB values. For example, in sRGB (Photoshop 5 and web graphics defaults) 50% = 128/128/128. However, in Adobe RGB (U.S. Pre-press default) 50% = 147/147/147. At the dark end, both standards "distort" the response. The sRGB space has 95% = 10/10/10. Adobe RGB 95% = 20/20/20. I use a scanner to read the densities and color of the steps of the test file prints. (Be sure the test print is thoroughly dry.) After making an RGB scan, I make a duplicate image and convert it to grayscale. To adjust the grayscale image I use first Auto levels, and then I carefully set the 100% and 0% points manually if Auto levels didn't hit it right. The g/s image is used to read the density values. From the RGB image, I am interested only in the differences between the red and blue values for each step -- that is, red minus blue (not percentages). So, the "nc" curve is about 4 units cool (blue - red = 4). Cold is about 9 units cool. Medium warm is about 6 units warm. (Since the warm curve is usually as warm as I can go, I don't record the values, but it is about 13 units warm at 50%.) These values should be very close through the midtones, and taper off at the ends. I have a form that has the same numbers as are on the test file across the top, with horizontal and vertical lines so that I can write down the numbers I read with my eyedropper. It's handy to keep these so you can see what you did the last time around. I only record the patches that are off, and then check off the changes I made -- sometimes noting what they were. This helps you see how much affect a movement has for the next time around. I'm sure I have a lot more to learn about making these curves efficiently. It would be great if some of the list's brilliant software gurus could figure out a way to automate this. However, if we all share our work, it will sure help make a cheap and easy system to use. Good luck. Paul http://www.PaulRoark.com