pdftex: Embedded pdf graphics not printed or partially printed

My problem concerns PDF not being portable across platforms.

I use these apps: * MacOSX 10.6.8. * Texshop 2.47, right out of the box, nothing customized. Default command is LaTeX. Default script is pdftex. * Adobe Illustrator CS5.

I am writing a long book. Some of the graphics are bitmap images (e.g. TIFFs), that are imported to Adobe Illustrator using "Place." Then I add labels to the Illustrator document, save as EPS, and convert to PDF with Acrobat Distiller X 10.1.3. The bitmap images are not placed as links; they are actually in the Illustrator EPS document. Finally the LaTeX document includes the PDF via \includgraphics and graphicx.

It works perfectly -- for me. The problems are that

• My friends print the document, e.g. from Mac LionOS or from Windows, and these figures are sometimes blank (other, simpler figures print fine).

• My other friends view the document on iPad, and the same figures are again blank.

You can find one of the bad figures here:

Illustrator EPS version:

When this was embedded in my LaTeX document, processed by pdftex, and then printed from Mac Preview in OSX 10.7.?, two of the three bitmaps were blank. One bitmap was printed, along with the labels added in Illustrator.

It's hard to know where to begin here, so any general advice is welcomed. Thanks.

-
Can you make one of the eps files available for download? – Ian Thompson Sep 8 '12 at 13:21
I add the same request than above but with a PDF file converted from EPS. – Paul Gaborit Sep 8 '12 at 13:26
If you have CS5 you can run the Acrobat preflight tool on the PDF to warn you about potential problems. – Lev Bishop Sep 8 '12 at 17:35
Ian and Paul: Thanks. I edited the original post to give you an example of the files that generate the issue. – Phil Nelson Sep 8 '12 at 18:20
Here is another example: physics.upenn.edu/~pcn/BadFigs/g224hertzLABEL.eps physics.upenn.edu/~pcn/BadFigs/g224hertzLABEL.pdf In this case, there is only one placed bitmap, which came out blank when printed. Again, the labels added in Illustrator were OK. – Phil Nelson Sep 8 '12 at 18:21

Something seems to have gone wrong with the export to eps. According to Encapsulated Postscript files, the first line should just be a version comment such as

%!PS-Adobe-2.0 EPSF-2.0


but in both of your files there is some trash preceding this. Open the files in a text editor and delete the trash. There is also a very large amount of trash at the end of your files. Everything that follows

%%EOF


can (and should) be deleted. If this doesn't fix the problem, I suggest eliminating Adobe Illustrator from your workflow. It is the work of Satan. You could convert the bitmaps to jpg or png format, include them directly in your LaTeX file, and use the picture environment to add labels as in this answer. This method has the advantage that the font used for the labels will automatically be the same as the font used in the rest of your document.

-
While I agree that illustrator can be problematic (I still use it, because I haven't found something better), I don't think these issues with the EPS are what's causing the problem here, since the EPS successfully distilled to PDF and I can't see obvious problems with the PDF. – Lev Bishop Sep 9 '12 at 1:22
@LevBishop Uncompress the PDF file to see "garbage"... – Paul Gaborit Sep 9 '12 at 6:18
@PaulGaborit, what "garbage"? Do you mean the xmp metadata? That can't be the problem, because pdflatex \includegraphics will not copy that metadata to the resulting pdf and so the ipads, etc that are causing problems will not even see that metadata. – Lev Bishop Sep 9 '12 at 13:15

Your EPS file and your PDF file contain useless informations (thumbnail, preview, etc.). You can try to clean it with GhostScript.

From a PDF file, you can do:

pdf2ps original.pdf original.ps
ps2pdf original.ps original-clean.pdf


I applied this method to http://www.physics.upenn.edu/~pcn/BadFigs/g113pereiraABClabel.pdf and the result is http://perso.mines-albi.fr/~gaborit/g113pereiraABClabel-new.pdf.

Does this new version works better?

-
This lossily recodes images with the /DCTDecode filter (like JPEG), which may be unacceptable for scientific images (I assume Phil was using TIFF images rather than JPEG for a reason). To refry without lossy compression: gs -o refry.pdf -q -dNOPAUSE -sDEVICE=pdfwrite -dBATCH -dSAFER -dPDFSETTINGS=/prepress -dAutoRotatePages=/None -dDownsampleColorImages=false -dDownsampleGrayImages=false -dDownsampleMonoImages=false -dAutoFilterColorImages=false -dColorImageFilter=/FlateEncode -dAutoFilterGrayImages=false -dGrayImageFilter=/FlateEncode g113pereiraABClabel.pdf – Lev Bishop Sep 9 '12 at 13:26
@LevBishop When using an iPad or a conventional printer to view images (even scientifc), I think that jpeg compression is not a problem. – Paul Gaborit Sep 9 '12 at 13:35
I agree. But often people expect the same PDF to be useful for multiple purposes. It's a pain to have to produce, distribute (and explain to users) that one version of the PDF is intended for ipad viewing and another version is suitable for careful inspection on high-resolution screens with yet another version for book-quality printing, and keep all these versions properly synchronized. As Phil wrote the "problem concerns PDF not being portable across platforms", and achieving this portability can be quite desirable – Lev Bishop Sep 9 '12 at 13:46
Thanks Lev and Paul. I'll try this and the other suggestions. Lev is right that ideally I'd like to stick to lossless methods if possible, for the reason he cites. – Phil Nelson Sep 10 '12 at 11:59