2005-03-29
22:13:10

Macrobanding on a Epson R200

While printing some negatives for my Cyanotypes (for some work see my gallery) I again had to experience some macro banding. As this occured on a panorama I first thought this might be the seams but in the original I can't see anything which resembles this banding. So I did some tests.

But before testing I did realign the head of my printer (unfortunately this isn't possible right now using Linux but I could borrow a laptop with Windows XP from a friend ;-). The head wasn't aligned perfectly, but realigning didn't solve my problems.

So here are the basic conditions of my tests (unchanged throughout the tests):

• printer: Epson R200 with epson inks
• printer driver: gutenprint 5.0.0beta3 (CVS version of 2005-01-29; I didn't see any important changes this then which I'd expect to have any impact on printing quality on a Epson R200)
• print quality: manual control
• medium type: Premium Semigloss Photo Paper (acutally I printed on a glossy medium for the test, but my final prints for the negative are done on a medium for which semigloss is a good choice)
• ink set: EPSON Standard Inks
• resolution: 2880x1440 dpi (former tests showed, that this really leads to the best results)
• printing direction: unidirectional
• interleave method: see below
• ink type: 6 inks (foto)

First I did test different raster algorithms with a interleave method: nearest neighbour avoidance. Here are the results (for some comments about the scanning parameters and the results itself see at the end of the page; the images are showed scaled down to a width of 400px, if you save them (or show just one of them), they'll be about 1 megapixels large).

raster algorithm	result
Default
Adaptive Hybrid (looks almost perfect)
Hybrid EvenTone

As you can see above the Adaptive Hybrid version does look best but in the original you can see some of the macrobanding that is easily visible with the other raster algorithms too. Aditionally if you look at the printout the Hybrid EvenTone looks better (smoother), but the macrobanding.

So I decided to test what a variation of the interleave method leads to with a raster algorithm of Hybrid EvenTone. And here are the results:

interleave method	result
Standard
Alternate Fill
Ascending Fill
Descending Fill
Ascending Double
Nearest Neighbour Avoidance (same as the last one above)

The original

If you want to test yourself or analyze the original, here is it (it a high quality jpeg with a resolution of 360 dpi, but if requested I could put a png on this site too):

Theres something to be noted for this image: it's not a grayscale image but a grayscale one overlayed with a layer filled with (R,G,B)=(10,0,0), mode "screen" in Gimp. With this I get the best results for my final process... But I did see similar banding with grayscale pictures before.

What should you see?

Fortunatly (at least to show it) the scans show the problem pretty clearly ;-). So the macrobanding is clearly visible (even more pronounced compared to the printouts) but because of the scanning the colours are not to be taken very seriously. So in the printout it always seems to me, that the darker/denser stripes are more yellow. The frequency seems to be similar but the position is depending on the printer settings (esp. on the interleave method)!

Because of the variation of the position of the stripes I draw two conclusions:

1. it's not directly in the source of the image but some kind of a moire effect caused by the raster algorithm/interleave method,
2. it's not a problem of the printer itself (but here I'm not sure)

Another observation I made throughout printing is that this problem needs some special features of the image; I just can't grab which features trigger this macrobanding. I get this macrobanding mostly for quite uniform pictures and if the image is a bit lighter, the banding might vanish...

About the scanning

I didn't take care very much about the scanning, so you can see many dust spots. The scanner is a IT8 calibrated Epson 2580 and I used VueScan 8.1.43. I didn't optimize the colour (manual control, black and white point to 0%). The scans were taken at 300dpi and the original printouts have a height of 10.1 cm.

What do I do this for?

All my printouts are negatives for Cyanotypes. To see some pretty results go to my homepage. There's a link to wikipedia too, for a short introduction to the process.