PDF Import

[tim_newsome]

PDF Import works very well on some PDFs. A few different PDFs I am converting still turn in to thousands of segments.

Any ideas of what causes that?

[cadtag]

at a guess, the software that created them used spines, instead of arcs or circle.

[rkmcswain]

There are multiple factors at play.

First, what did the original PDF driver do when it converted the original vectors to PDF linework?
Secondly, what did the PDF>DWG application do with the geometry it encountered?

As cadtag said, perhaps an original entity was a spline and the PDF driver constructed the PDF out of multiple, tiny segments to represent the curvature.
Another case might be a dashed line. Does the PDF driver know it's really one line with gaps in it, or will it create a series of individual lines?
Will the PDF>DWG application know to "re-connect" a bunch of co-linear dashes back into a line?

All sorts of variables like this, which is why trying to make a CAD drawing from a PDF is generally going to give poor results.

[MMccall]

It could be the difference between a PDF that was created by software and a PDF that was created by a scanner. The software created one will have perfect line work. The scanned one ... not so much.

[rkmcswain]

The software created one will have perfect line work.

Probably not perfect. Vector will almost certainly be better than scan, but there will always be some degree of error.

Using AutoCAD, I constructed a line with a length of 500 units, an arc with a radius of 176 units and another line with a length of 100 units.
I printed to PDF at a scale of 1"=40' on a D sized sheet, the PDF driver settings are set to 600dpi for vector.
I then used PDFIMPORT to import this geometry back into a new AutoCAD drawing.

The imported geometry now measures:
The first line is now 500.0016 units long
The arc's radius is now 175.9744 units.
The last line is now 99.9415 units long

[remi678731]

Probably not perfect. Vector will almost certainly be better than scan, but there will always be some degree of error.

Using AutoCAD, I constructed a line with a length of 500 units, an arc with a radius of 176 units and another line with a length of 100 units.
I printed to PDF at a scale of 1"=40' on a D sized sheet, the PDF driver settings are set to 600dpi for vector.
I then used PDFIMPORT to import this geometry back into a new AutoCAD drawing.

The imported geometry now measures:
The first line is now 500.0016 units long
The arc's radius is now 175.9744 units.
The last line is now 99.9415 units long

This a perfect example of almost all pdf and/or images loaded there always seems to be some degree of skew, some more than others but its there. I have heard numerous critiques from those who don't understand this I can't explain it other than experience has taught to be very careful relying on a scan image for exact linework.

[rkmcswain]

This a perfect example of almost all pdf and/or images loaded there always seems to be some degree of skew, some more than others but its there. I have heard numerous critiques from those who don't understand this I can't explain it other than experience has taught to be very careful relying on a scan image for exact linework.

Yes. When you "print" to PDF, even though you are printing vectors, there is a limit to the accuracy based on the DPI and other factors. There is simply no reason for the PDF output to be held to the same accuracy as the CAD program, and if the source (the PDF file) is not accurate, then the result (PDF back to CAD) is not going to be either.

The last post in this thread does a good job at explaining further.

[dotson]

Most PDF content streams I've looked at round to 2 or sometimes 1 decimal place. Never forget, PDF is a popular plot file format, nothing more.