Andrew

Further to my earlier reply, I find that bitmap resolution on the page should equal or be an integer division of the Print Options > Output > Transparency resolution > Pixels per inch value to avoid low quality resampling. The Transparency resolution should be set to an integer division of the printer's maximum dpi, as stated before, although this may be less important due to the blurring caused by the printer's dithering pattern. In fact, on a 600dpi HP printer, I found that a Transparency resolution of 300dpi gave the best results for both 300 and 600 dpi bitmaps. This is probably highly model dependent and some experimentation will be necessary.

Anybody who's eyes glazed over before they finished the first paragraph should at least note the following: Xara's Print Options > Output > Automatic setting appears to be anything but automatic, remaining at 150dpi regardless of the printer selected. As a first approximation, click Manual and enter a value of half your printer's maximum dpi. I assumed that Xara interrogated the printer driver and made an intelligent choice of value ... silly me ;-) Also note that this setting is local to the drawing, so it has to be saved with the default template to become the default.

My attached diagram shows how five pixel values (red dots) might be used by linear (blue lines) and cubic (grey line) interpolation to produce new pixels at points indicated by vertical dashed red lines.

Linear interpolation works by taking the two pixels either side of where the new, interpolated pixel is going to be and draws a straight line between them. The value of the new pixel lies somewhere along that line and is a weighted average of the original two pixels. Averaging is synonymous with blurring and this technique tends to blur the edges of images.

Cubic interpolation uses four adjacent pixels to construct a cubic curve passing through all of them. It therefore takes account of how the pixels are changing beyond the immediate vicinity, rather like the behaviour of Bezier lines which also use cubic interpolation. This allows the new pixel to lie outside the range defined by the original pixels or to be disproportionately close to one or other of its neighbours, as shown in the diagram. The end result is that reduced bitmaps are less blurred and better retain their edge definition.

When enlarging bitmaps, however, it's usually not desirable to emphasise pixel boundaries, so the smoothing effect of linear interpolation is preferable. This is particularly important when enlarging JPEGs because their compression uses a sort of dithering that introduces shadows along edges which would be magnified by cubic interpolation. The Bi of Bicubic and Bilinear refers to the fact that the pixels are interpolated in both horizontal and vertical directions. I think that Xara uses Bilinear but just calls it Linear.

Regards - Sean