Might want to consider what you have done. The pixel dimensions of a 20d image would be something like 3504 X 2336. That would give you a little over 8,000,000 pixels. In order to get a print 40 X 25 at 350 ppi, you would need an image 40X350 X 25X350. That's 122,500,000 pixels, that is roughly a 15x increase. In other words, only one pixel in 15 is from your original image. Everything else was "made up" by your software. Assuming you had to crop the file to get those dimensions, it would be even worse. Now consider whether your eyes can distinguish the difference in size between a pixel measuring 1/180" and one measuring 1/350" from several feet away. If you re-sampled to 40X25 at 180 ppi, then the resulting file would be around 32,400,000 pixels, only around a 4x increase. It's your call, but I know which way I would go. I used 180 ppi as an example, but anything between 150 and 200 is probably going to give you a better print that "made up" pixels at the higher resolution.