Our pdf output:
(1"= 500') = 225 MB for 12 sheets or 19 MB per sheet - approximately 2 hrs to generate
(1"= 1600')=375 MB for 6 sheets or 63 MB per sheet
I am wondering why, if all the settings are the same, the files are so much different in size. I assume it has something to do with the quality factor being a percentage of the overall raster that is being displayed. So when the scale of the sheet is changed, more raster is displayed. Even though the raster takes up no more room on the page (still 11x17), the file size has increased drastically. Something that further complicates the lowering of the quality factor is if we have a rasterized logo on the page it prints horribly because it's quality setting has also been applied to it.
I would assume that a practical way to deal with this is that the rasterized object should be created based on the dpi capability of the printer not the quality of the image. So if I create a pdf, I would designate it to be 600 dpi rather that adjusting the raster quality. This should mean that the pdf or plot will not contain more information than is required to plot a 600 dpi print. This would also help reduce the time required to generate the plots.
Is it possible to set an "absolute raster value (dots per inch)" to pdf's and other plots generated?
You might be interested in this blog article about rasterized printing.
I don't have a ready explanation for your results. I agree -- the final print size matters, the print scale should not. There must be some other difference, perhaps in the type, color depth, or density of the data being printed. Or perhaps you are not printing consistently in rasterized mode, and thus would be comparing apples and oranges.
When printing in rasterized mode, then the rasterized quality and the print size are the important factors in the PDF file size. By default, the rasterized quality is 100% of the device resolution (600 DPI), and there is no way to change that from the Print dialog. The rasterized quality may only be adjusted in the printer driver configuration file.
The raster quality factor (which can be adjusted from the Print dialog) really makes sense only in the case of non-rasterized printing. The feature is designed to allow MicroStation to embed a smaller raster image in the plot file, and allow the plotter firmware (or Acrobat) to scale that raster up at display / render time. When printing in rasterized mode, all the plotter firmware or Acrobat receives is one big raster image. There is no method by which to selectively rescale portions of a rasterized plot file.
The raster quality factor is still used during a rasterized print, however. Say the rasterized quality is 100% and the raster quality is 50% (the default for pdf.pltcfg). You have an aerial photograph covering the entire print area. Instead of producing print tiles at 600 DPI, Raster Manager produces print tiles at 300 DPI. Those print tiles are fed through the graphics hardware, where they are scaled up to rasterized resolution (600 DPI) and written into the PDF file. As they are written into the PDF file, they are compressed (by default using LZW compression, but may be configured for JPEG).
I would expect 300 DPI print tiles scaled up to 600 DPI to compress somewhat better than print tiles originally generated at 600 DPI and not rescaled by the graphics hardware. But I wouldn't anticipate that difference to produce large differences in PDF file size.
Performance is another issue. I'd expect Raster Manager to take longer, and consume more memory, to produce 600 DPI print tiles instead of 300 DPI tiles. It's a judgement call as to whether that performance improvement is worth the quality degredation that occurs when the print tiles are then scaled up to rasterized resolution.
There is a way to specify both the rasterized and raster resolutions in DPI, rather than a percentage of the device resolution. The hpglrtl.pltcfg printer driver configuration is set up to do that, since the device resolution is an odd value. However, that capability isn't exposed for PDF in the printer driver configuration editor. If you are comfortable working in XML, you can remove the Quality records and replace them with the DPI records as found in hpglrtl.pltcfg. But I don't think there is any need to do so, and don't recommend attempting it.
If you want to make a big difference in the size of your rasterized PDFs, set the rasterized quality to 50%. That in effect will yield 300 DPI PDFs. Raster Manager will produce print tiles at 300 DPI (via the unchanged 50% raster quality of 600 DPI), producing no rescaling by the graphics hardware. The downside is that when the PDF is printed from Acrobat or send directly to your plotter, the entire rasterized image (text, linework, and raster data altogether) will be scaled to plotter resolution, Text embedded in raster images doesn't scale very well. For best results, you want to set your rasterized resolution to the same resolution as the plotter you'll eventually print to.
You should note that PDF supports transparency on vector data. Depending on what you are doing, you may be able to print in non-rasterized mode.
.
Thanks for your thoughts on the issue. I have read through your blog and it was very helpful.
I believe that we have been trying to compare apples to apples. We have been plotting with the same pdf.pltcfg file with the setting not changed from one plot to another. The same aerial is being used just the scale at which it is presented has changed.
I am familiar with the two different areas within the .pltcfg file that can be edited that have something to do with the raster quality.
To sum up, I don't believe that the Raster Quality Factor (#1) functions the way it should (as you described). If it did, I would easily be able to achieve the results I am looking for. At 30%, the image quality should be reduced to a dpi rating of 180. This should be very obvious when printed. As I said previously, we can see no difference.
Thanks for the insite on how to modify the .pltcfg file. I may give it a shot because when trying to create these 30-40 exhibits it literally takes more than a day to print.
Thanks again for your help.
Regarding the raster quality factor, I can definitely say that it is not a percentage of the original image quality. It is a percentage of the device resolution. It is the equivalent of the old PIXEL_RESOLUTION record in the .plt file (specified in DPI). The raster quality was added to the Print dialog user interface as a percentage primarily due to the Windows printer driver. Users wanted to specify the raster resolution as a fraction of the device resolution, and hold that fraction constant as Windows printers with different resolutions were selected.
That said, you are absolutely correct in that the image resolution plays an important role in print quality. Say you are printing to 17x11" PDF with 30% raster quality. Your image size exactly matches the paper size. Raster Manager will need to scale the image to print size at 180 DPI, or 3060 x 1980 pixels. If your original image is already 3060 x 1980 pixels, then RM doesn't have to scale up or decimate the image at all. Your raster quality will be 30% of the device resolution, but 100% of the original image resolution.
If your image resolution is larger than 3060 x 1980 pixels, than RM will decimate to print resolution. That's not typical -- usually RM has to scale original images up to print resolution. However, a 30% raster quality factor goes a long way towards bringing print resolution down to the size of many raster files.
If you set the raster quality factor to 50%, then RM will need to produce a 300 DPI 17x11" image, or 5100 x 3300 pixels. If your original image resolution is 3060 x 1980, then some upscaling is required. In this case, 50% raster quality doesn't mean the original image resolution is reduced by half. Just the opposite.
Say your original image file is larger: 6800 x 4400 pixels. When printing with 30% raster quality, RM produces a 3060 x 1980 print image. The print image is 45% of the original image, not 30%.
When printing the same with 100% raster quality, RM produces a 10200 x 6600 print image. The print image is 150% of the original image.
When printing in rasterized mode, the images produced by RM are then sent through the graphics hardware where they are scaled up to rasterized resolution (600 DPI), regardless of the size they were produced at.
Aerial photographs rescale well, so I could easily believe you might not notice a visual difference between the 30% and 100% raster quality cases. Especially if you are comparing results on paper after printing to a 300 DPI or 400 DPI plotter.
If you are not printing in rasterized mode, then RM never scales raster beyond the original image size and the graphics hardware doesn't get involved. In that case, the maximum print image size embedded in the PDF is the same as the original image size. Whatever scaling needs to be done to match the plotter resolution is performed by the plotter, or by Acrobat's printing code.
--
If your raster attachment is actually a PDF file or a multi-resolution raster format such as MrSID or ECW, then this gets a little more complicated. In that case, Raster Manager isn't limited by the original image resolution.
If you'll upload your data to Bentley's FTP site and send me the file name along with precise instructions on how to repeat your test conditions, I'd be happy to take a look at it when I can.
Andrew,
Thanks for all of your help. We are using .ecw's and the size of those files are between 2 & 3 GB. I don't believe that uploading those files is a practical idea.
You mentioned that if we were using PDF, MrSID or ECW the process gets a more complicated. Could you explain why this is? Most of the rasters we use are of these types and having a clear understanding of the issue would be of great help. Also, would using the method you mentioned of modifying the .pltcfg file to use an absolute dpi value help resolve this?
Thanks again, Your insight has been most beneficial.
Chris
With ECW rasters, most of what I wrote in my last post is irrelevant. For all practical concerns, an ECW raster attachment has an unlimited resolution. Since Raster Manager is not restricted by the size of the original raster data, it will always produce print images at the print raster resolution (set by the raster quality factor).
When printing in rasterized mode, the print images produced by Raster Manager get further scaled up to rasterized resolution (set by the rasterized quality factor) by the graphics hardware. So I would not expect to see much if any PDF file size impact from changing the raster quality factor. You would need to set the rasterized quality factor for that to happen.
I would definitely expect to see a file size impact by changing the raster quality factor when printing in non-rasterized mode. In that case, the rasterized quality factor is irrelevant.
When printing in either rasterized or non-rasterized mode, changing the raster quality factor should make a difference in the MicroStation process's memory usage. When printing, Raster Manager builds a cache of images at print resolution. The higher the print resolution, the larger that cache will be. So even though all the output will eventually be scaled up to rasterized resolution, lowering the raster quality factor can still improve performance when printing in rasterized mode.
Regarding my .pltcfg DPI comment, I only included that for accuracy. There is absolutely no functional difference between specifying either the raster quality factor or the rasterized quality factor as a percentage of the device resolution or as an absolute DPI value. The printer driver configuration editor is designed to show only the percentage in all but of couple of cases. Swapping out the percentage XML record with a DPI XML record is just a theoretical exercise; there's no reason to actually do it, and plenty of reasons not to.
I've got a few ECW images in my test suite. When I have a chance, I'll run some tests and see if my results match my expectations. However, I can't imagine any reason that print scale would play a role. I suggest that you reconfirm that all your tests were performed with consistent rasterized or non-rasterized mode.
Also, have you verified that rasterized mode is truly required for your workflow? You mentioned transparency, but transparency is supported in PDF for vector objects.