An Alternate Method of Producing the Dunthorn Grayscale
Why Printer Grayscales Sometimes Fail
The principal problem reported is that the grayscales are not gray. There is a color shift along the grayscale and often not a consistent shift. There are reasons for this (although not what we consider to be good reasons). In printing, "white" is (usually but not always) the color of the paper. A glance across your desk will show there can be a considerable variation in paper white. Dye-based inks act as filters, so the color you see is actually the white of the paper filtered through the dye dots, mixed with the unfiltered white of the paper where it not covered by dye. Thus the colors are automatically adjusted to the paper white, whatever it happens to be. Pigment-based ink is less accommodating, since the pigments obscure the white of the paper, substituting their own color rather than filtering the paper white. Both systems also use black ink. Ideally the black ink does not reflect any light, but actually it reflects a small amount which can become important in very dark grays. Finally, some printers use pastel inks, light cyan, light gray, etc. Here an entirely new "white" may come in when these colors effectively add a white pigment which may not be the same as paper white. Since the dye-based inks are naturally and automatically referenced to paper white, they present less a problem in balance, but with pigment-based inks there is a problem being sure the printed grays are reductions of the paper white that maintain color integrity. The colorimetric quantities of pigments must be adjusted to match paper color. The white pigment added to make light cyan, light gray, etc., should be selected to match paper white or the papers should be manufactured to match the pigment white. Otherwise colored pigments must be added to the gray scale to make it agree with paper white. This is, of course, what printer manufacturers should be doing when, having complete control over each, they match their own ink, paper, and profile. If you have tried printing a grayscale using the printer manufacturer's recommended paper, ink, and color management profiles and you are not getting a gray grayscale, it is quite clear who is to blame.
The Strategy
Here we will try to make an end run around this situation, combining strengths where we see them. In this we will necessarily assume that the printer is capable of producing a good black, by which we mean that the black is of very low reflectance. If we can persuade the printer to use just black ink, then the color of any area consisting of white paper spattered by black dots will be simply the reflection of the paper times the fraction of the paper that is not covered by dots. The color intensities reflecting from the resulting gray mixture will thus directly be a percentage of color intensities reflected from the pure white paper as we require for color integrity; in this case the color integrity of the grays. This will be true until so little of the white paper remains showing that the color of the black – which is supposed to be non-reflective – will start to become significant. We can hope that will show up only in the darkest steps, if at all. The trick here is in forcing the printer to print in all black ink. Typically this cannot be done under color management. On many printers even "black only" mode will not work – light gray or even colors will be used as well as black.
Making Gray
We will turn to the halftone capability of Photoshop. Photoshop can do an excellent job of halftoning or dithering with a result that is correct for a computer display. (Unfortunately, this is not the same as being correct for printing, but we will deal with that.) To see this, load DunthornGrayScale.tif into Photoshop. First change from RGB to grayscale: Image→Mode→Grayscale. Then halftone the result: Image→Mode→Bitmap. For now enter 300 for pixels/inch. Traditional halftone uses a pattern for the dots, but that is a prescription for producing a moiré effect here. Select Diffusion Dither from the pulldown, and OK. The result will be a very good reduction of the gray scale to random black dots on a white field.
Now it is necessary to print this image so that it remains all black dots. Make a print on high-quality paper suitable for the printer, preferably glossy so that the black will resolve well and dark. Unfortunately, the method of getting the image to print in only black will vary considerably with the specific printer and printer driver, so we can only give general guidelines here. We are printing a bitmap, so we have lost all color management – or black and white management, for that matter. In Photoshop, the "profile" should be "Same As Source." Use Page Setup to burrow into the printer options. Select the paper type and the dpi. We found that with the Epson 2200 it was necessary to use 1440 rather than 2880 to get reasonable results. This may require experimentation for your printer. Select Black Ink Only or Black (versus color) anywhere you find it in the setup tree. Deselect anything that suggests it is smoothing or altering the image – that may mean that gray or other colors will be added. Print the image.
Examine the resulting print with a magnifier. We use an illuminated 30X magnifier. These are available inexpensively at places like Radio Shack and Brookstone and should be in every photographer's toolkit. The dots should be easily seen as all black dots. Look at all the grayscale steps. If your printer has gray ink and there are dots which appear less black than others you are likely seeing gray ink. For example, with the Epson 2200 we found that selecting too fine a pixels/inch setting for the Diffusion Dither caused gray ink to be used. At 300 dpi diffusion dither we appeared to get pure black and the scanner results confirmed this.
Now you have a true gray grayscale, or at worst one that can be made true gray using a single color correction filter, which meets our requirements. Unfortunately, it is not the right grayscale. While the Photoshop halftone does a quite credible job of halftoning for the computer display, the image is not halftoned correctly for printing and the print you have made will likely be too light or too dark – normally the latter. Details like dot gain have to be taken into account for proper printing. Photoshop does not appear to be of much direct help in preparing halftones for printing, but with a few experiments we should be able to produce a suitable approximation to the Dunthorn Grayscale.
First, it is important to be as close as the printer will take us. Look at the grayscale you have produced. If the two blackest steps are close to the same tone but are distinguishable from one another you are on the right track. However, if there is a marked difference between the two darkest steps or if the darkest three or four steps are indistinguishable from one another or very hard to distinguish, experiment with other printer settings. We have found the dpi setting to be very important in our limited tests, and there is interaction between the dpi setting and the pixels/inch setting in Photoshop Bitmap. Try to find printer settings where the two blackest steps are close to the same tone but are distinguishable from one another, or as close to that as you can get. Stick to those printer settings in the following (but be sure that the dots are all black in the setting that you finally choose).
Getting a Dunthorn Grayscale
To get the grayscale to match the tonal gradations of the Dunthorn Grayscale, we will need a reference. There are two commonly available possibilities and, frankly, it is a good idea to make use of both if possible, as a cross-check. First, even if the printer was unable to make a truly gray grayscale there is reasonable hope that the steps on the tonal scale are approximately correct. Therefore, make a print as we suggested previously, a direct color print of DunthornGrayScale.tif with the manufacturer's preferred paper, ink, and color management profile. This grayscale can be cut straight down the middle, with half of it placed over a trial grayscale for direct visual comparison. Second, if a flatbed scanner is available, it also may sufficiently accurate for a check, although this is less straightforward than one might first think. Trial grayscales can be scanned and checked numerically and/or compared with the "color" grayscale.
To obtain a trial grayscale, start with DunthornGrayScale.tif in Photoshop and as before change from RGB to grayscale: Image→Mode→Grayscale. Now adjust the gamma using Image→Adjust→Levels. (This is the only control in Photoshop that readily approximates the proper printer correction.) Adjust the "middle gray" slider or enter a number in the center box for middle gray. If the uncorrected grayscale print is too dark, you will need a "middle gray" setting greater than 1; a value of 1.4 or 1.5 is a good starting point. If the uncorrected grayscale is too light, you will need a middle gray setting less than 1 (we believe this will be unusual) and a value of 0.7 is a good starting point. Then halftone the result, as before: Image→Mode→Bitmap. Enter the same pixels/inch used previously, select Diffusion Dither from the pulldown, and OK. Print exactly as you printed the "straight" version. Compare this with the color grayscale, looking for a match in tonal values, and/or scan and compare.
If the result of the comparison shows the trial to be too dark or too light, make another trial by repeating the procedure of the above paragraph. Be sure to revert at least to where DunthornGrayScale.tif has just been converted to grayscale – do not make gamma adjustments to a grayscale that has already been gamma adjusted (it is a poor procedure for numerical reasons). If the trial was too light, decrease the middle gray setting used the previous time; if too dark, increase the middle gray setting used the previous time. Repeat this until you are satisfied with the resulting grayscale. We do not mean to downplay the work involved here; it may take a number of trials to get a satisfactory grayscale.
Testing the Grayscale
The simplest and most direct test is to cover the trial grayscale halfway with the color print grayscale. Standing back enough so that any texture in the trial grayscale disappears, compare the two grayscales. Do the gray levels match? Does the trial grayscale fade into black more slowly or more quickly across the scale? This works best if you have made several trial grayscales using different "middle gray" settings and compare the group together.
With caution, the comparison can be assisted using a flatbed scanner. The problem here is that the "grayscale" we have produced is composed of black dots on white paper. To get gray, these tiny black and white areas have to be averaged correctly. Briefly, the averaging of black and white reflective areas needs to be applied to light intensities – it will not be correct if the average is applied after the image is loaded into Photoshop as system-gamma-encoded light intensities. When using the grayscale in practice, you will insure the averages are correct by taking an image that is slightly defocused so that the fine black-white detail does not show, thus averaging the intensities. The eye at a distance uses this same defocusing to average the intensities. Unlike cameras, however, scanner systems do not defocus gracefully. Yet it is important that by the time the image is scanned into Photoshop, the fine black-white detail should be mostly gone.
Fortunately, scanners typically do part of this by accident. Flatbed scanners are generally quite poor at edge resolution – enough so that they typically default to some form of edge enhancement, sharpening, or whatever. Be sure the sharpening is turned off and scan in a trial grayscale at full native scanner resolution (you know your scanner's native resolution, right?). Examine the lightest step in Photoshop so that you can easily see the pixels and measure a few of them. Very likely you will see the "black" dots on a "white" background, but your measurements will show that while the "white" areas are reasonably close to white, the "black" areas are nowhere near black. In fact, they will typically measure as a fairly light gray. Examine the darkest step and you will find nearly the opposite to be true – the black will be reasonably close to black, but the white dots will measure as fairly dark gray.
This means that a lot of the averaging has already been done by a defocusing of sorts. Help this along by performing a scan at a low resolution; that is, setting the scanner for 50 dpi, 75 dpi, whatever low resolution it will allow so that the scanner is averaging together several readings for each pixel, and read that image into Photoshop. There will still be some fine detail in the low-resolution scanned image, but the gray values within each step should be close enough together that a further blending within Photoshop will produce valid averages. Check to be sure the values in each step really are fairly close together.
Using the Scanner to Check the Grayscale
With the grayscale image scanned into Photoshop as described above, select the grayscale area so that only the grayscale will be affected and use Filter→Blur→Gaussian blur. Select a radius that is sufficient to make the grayscale steps fairly uniform, typically a radius of 4 or 5. This blurring will sharpen the peaks in the histogram. Set the eyedropper by clicking the eyedropper tool button in the main tools palette and setting its dropper size to 5x5. Now use the Photoshop Levels tool to adjust the grayscale so it can be compared. Click the right, highlight eyedropper in the levels tool:
Then double-click the same dropper button and the Color Picker dialog box will appear:
R, G, and B, show the value of 255 above. Enter a value of 242 for each of these and click OK. The mouse cursor appears as a dropper. Use it to select areas in the lightest step of the grayscale image. Pick several points and stop when the Levels histogram seems to bounce about the least.
Now repeat the above, clicking and then double-clicking the left, shadow eyedropper. This time when the color picker appears, R, G, and B will show values of 0. Enter 38 for each of R, G, and B and click OK. The mouse cursor appears as a dropper. Use it to select areas in the second darkest step of the grayscale image. Pick several points and stop when the Levels histogram seems to bounce about the least.
Now OK the Levels dialog. Probably a message box will inquire whether you wish to save the new target colors as defaults. Answer No.
Start the Levels dialog again. The steps should show up like a picket fence, although it may deteriorate somewhat toward the dark end. In particular, the pickets should stand together for the three colors R, G, and B. If the pickets fail to coincide in any but very minor ways as you flip through the colors, it is very likely that gray or colored ink has been used in the grayscale along with the black. Alternatively, the scan may be poor – try another scanner if possible. If two scanners agree, the problem is that other than black ink is being used.
If the pickets match, it becomes a matter of getting the tones of the steps right. Compare the scanned and adjusted image to the set pixel values, 13, 38, 64, 89, 115, 140, 166, 191, 217, 242. If the tones are too dark, make another trial grayscale using a Levels middle gray setting a little larger in value than used to make the current trial; if the tones are too light, use a Levels middle gray setting a little smaller in value than used to make the current trial. We found that with an Epson 2200 and good glossy paper, printing a 300 dpi diffusion dither at 1440 dpi on the printer we got black ink and a good Dunthorn Grayscale using a middle gray adjustment of 1.2. We offer this as a ballpark example of what might be expected, but we do not expect this exact setup will work for you even if you have an Epson 2200.
We used the second grayscale step, 38, instead of the first step, 13, to set up the grayscale above because it is very difficult to get the darkest step in the grayscale to read 13. If the other grayscale values are close to correct and show no trends, the calibration is probably good even when the darkest grayscale step is somewhat off.
It is wise to cross-check the flatbed scanner against the color printed grayscale. Use the same means described above to check the color grayscale, setting the 242 and the 38 points. We would not be doing all this if the color grayscale had not been found wanting, so we may expect that the picket fence will not coincide well for the three colors. To eliminate the confusing color variation, convert the color grayscale to "grayscale" and then back to RGB and use the single numbers for comparison. These should agree reasonably well with the numerical Dunthorn Grayscale if the scanner is working properly. Significant variation warrants a check against another scanner of different type.
The Dunthorn Grayscales resulting from this method will be quite good, but it is necessary to remember to defocus test images of them enough to hide the fine detail of the black and white dots. When using the grayscale in practice, insure the averages are correct by taking an image that is slightly defocused so that the fine black-white detail does not show, thus averaging the intensities. The defocus should be applied when taking the image, not afterwards.
Privacy Policy and E-Mail
Go to C F Systems Home Page