laser-lovers@uw-beaver (01/03/84)
From jacobson@LBL-CSAM Mon Jan 2 19:54:00 1984 We have just gotten TeX82 to drive our QMS 1200 laser printer and I thought people on the net might be interested. For those not familiar with it, the QMS uses the Xerox xp12 marking engine: a dry toner, plain paper, 300 DPI, 12 PPM device. The xp12 is also used in the DEC LN-01 and Xerox 2700 laser printers and "laser lovers" has had generally good reviews on its print quality. The QMS electronics include a 68000 with some special `rasterop' hardware, 512KB of font & buffer memory and a separate memory to bitmap a complete 8.5 x 14" page (1.4MB). There are RS-232, Centronics and IBM interfaces. We use the RS-232 interface to a Vax 780 running VMS & Eunice (a 4.1BSD Unix on top of VMS). The interface is supposed to run up to 34Kbaud but our Vax dies above 19.2Kbaud so we run it at 19.2. I second the raves about xp12 print quality. The blacks are BLACK and the toner transfer is uniform even on huge black areas (I've tried printing 6 x 8" black rectangles -- no problem). I haven't tried printing on a brown paper bag (yet) but I have run 16 through 22 weight paper, including our letterhead stock, and the results were uniformly good. Most of our printing was done over Christmas week & the heat was off in the building. I've run about 4000 sheets through the unit in temperatures varying from 45 deg. to 70 deg. and with the relative humidity up to 97%. There didn't seem to be any effect on print quality and only 2 paper jams. We got the QMS instead of the DEC or Xerox versions of the xp12 because QMS lets you do everything the printer can do. The DEC LN-01 can (currently) only be used as a line printer and, while the Xerox 2700 has the capability to download fonts, Xerox wouldn't tell God what the download commands are. QMS gives you line printer & Diablo daisy-wheel printer emulation, vector & raster graphics and multiple, downloadable fonts, all of which you can intermix arbitrarily, even on the same page. It took less than 3 hours, total, to get a Unix plot filter for the QMS running, most of which was spent in deciphering the lousy QMS manuals. We had some 256x256x4 digitized video images lying around (courtesy of Vortex Video Systems) so I also did a raster image filter. The QMS has some built in half-tone patterns & I'd already decoded the curious version of English they use in the manuals so this filter took about an hour. The results were beautiful -- not book quality but much better than newspaper quality. We're using the current version of the TeX AMR fonts (I think -- we have 3 different font tapes & PXL files don't seem to have dates in them). Anything above 7pt looks great (to me). I don't have Brian Reid's educated eye but I have been looking for the problems he mentioned. The glyphs aren't "blurry or hard to read"; the edges are sharp & smooth, even under a magnifying glass. The descenders don't stick to the letter bodies, the serifs are symmetric & the thickness of hairlines & stemlines doesn't seem to vary. At 5pt & below the toner transfer seems to get uneven and some characters have voids (although this may have been an artifact of the 50 deg. temp & high humidity -- I'll try again when our heater is back on). Glyphs smaller than 3pt turn into unreadable blotches but we have very few applications that need 3pt type. I've tried running the same document on both the QMS & an Imagen and the difference is easily visible. The large characters on the Imagen are grey & the edges wash out and the bodies of the small characters tend to run together (the only Imagen I have access to has been in continuous use for almost 2 years so this comparison isn't very fair). I think most of the difference is due to the dry vs. wet toner process since the 25% improvement in the QMS resolution shouldn't be very visible in 10pt type. We're downloading 12 fonts to the QMS, a mixture of 8, 10 & 12pt styles. It takes about 2 minutes to download the rasters for 128 10pt characters at 19.2Kbaud. The current version of the QMS DVI driver sends characters that haven't been downloaded as raster images. Most of the documents I've been printing are about 3000 char. per page. Pages containing only downloaded fonts print at the 5 sec. per page printer speed. Pages containing no downloaded characters are limitted by the communication line speed to about 1 minute per page. Most documents (e.g., the TeXBook or the LaTeX manual) average 10 sec. per page. Texas A&M is working on a new version of the driver that will download fonts on an LRU basis, similar to the way Richard Furuta's DVI2LGP driver works. That should keep things running at the printer speed, even on a 9600 baud line. We don't have enough experience with the QMS to have accumulated many complaints, but I have a few. As I mentioned before, the manuals are awful (though I hear that new versions are in the works). The QMS people seem to be unfamiliar with typesetting applications (when I called to try to find out how to download fonts, they said "Download what?" & put me on to a guy who explained why I couldn't download code to the 68000. Other than this, the QMS people have been very helpful and prompt in answering questions). The communications software interface is pathetic: The only characters that the QMS ever sends back to your computer are the ^S/^Q for flow control. It's impossible for the computer to find out if the paper is out or jammed or if you're in the middle of a power up cycle (the QMS does a good job of handling these things, though. I've yet to see it lose a page of output). What's worse, it's impossible to find out what fonts have been downloaded and how much free memory is available for fonts. (If you try to download more fonts than there's room for, the entire excess font(s) get discarded WITHOUT NOTIFICATION.) Since the QMS uses a bizarre fixed block allocation scheme for font memory, trying to figure out how much is left from knowing what you've already loaded looks like an NP-hard problem. The 200 page paper trays on the xp12 are the pits. They make unattended operation almost impossible. The paper registration also seems to go to hell when the trays are low so the operator needs to pay a lot of attention to how much paper is left. The two paper jams we've had both occurred with <10 pages in the tray. That's probably enough for now. Thanks again to the many people who helped us with answers to questions & shipments of software, particulary David Fuchs & Richard Furuta. Also, many, many thanks to Norm Naugle of Texas A&M who supplied us with the a first version of the QMS driver, dropped by to help us install it & even brought the latest PROMs from QMS to make sure everything would work! Cheers. -Van Jacobson, RTSG, Lawrence Berkeley Lab (jacobson@lbl-csam.arpa)
laser-lovers@uw-beaver (01/04/84)
From Nemnich@MIT-MULTICS.ARPA Tue Jan 3 15:16:01 1984 We (at Thinking Machines Corporation) have been running TeX82 on the QMS with the Texas A&M driver for about a month now. I agree with Jacobson's assessments of the printer (I think it is great), but I have not had the paper-jam problem. The only times I have ever jammed it were by opening the thing up while it was printing. One caution. The toner denisty knob and associated machanism is sensitive to movement along its axis of rotation; i.e., don't push or pull it, just turn it! It is a design bug. Pushing it even slightly can throw the toner distribution mechanism out of whack. On ours, the print kept getting lighter and lighter until it was hardly readable. They came out to fix it and told me what had happened. --bruce