> From: Pauric Hennessy <Pauric.Hennessy@xxxxxxxxxxx> > Subject: [SI-LIST] LVDS vs RS422 > Date: Fri, 9 Aug 2002 08:08:49 +0100 > Hi all, have just subscribed and would greatly appreciate > your advice on > the following. > We are currently utilising a linescan camera to upload data > in RS422 @ > 20MHz to a host PC. I haven't alot of information on the > interconnect cable > except that its 32 pairs of twisted pair in a Shielded cable > running a > length of 15 metres. We use Dalsa TDI (glorified linescan) cameras here, and we used to run 4 meter cable runs back to commercial frame grabbers. (These days we convert to multi-gigabit fiber at the camera end.) At the time (~1997), the Dalsa cameras were using 26LS31/32 line drivers/receivers (this was with 15MHz pixel clocks). Shortly thereafter (around 98/99), Dalsa started using LVDS (90C031/32) drivers/receivers instead. Although our group had already gone to fiber by then, another group was still making long cable runs and found that they simply had "no signal left" (i.e., no signal to noise ratio) by the time they received the signal at the far end of the cable. Since the 90C031/32 only have ~200mV swing (compared to about 2V for the 26LS31/32), somewhere along the way they were picking up at least many tens of mV of noise. That group went back to the 26LS31/32 transceivers and fixed their problem. In your case, since you're already having problems with the slow stuff... I doubt going to LVDS transmission would improve matters. My understanding is that the faster edges may create more crosstalk than the lower amplitudes reduce it. I'd suggest the following: --> Take a look at the pixel clock as well as the pixel data. Is the noise you're encountering severely affecting the (differential) amplitude levels? Or is it more "smearing" the edges around and therefore violating setup and hold time requirements of an input buffer somewhere? In the later case, you may well be able to get away with simply rebuffering the signal on the far end -- using faster latches with shorter setup and hold times -- before presenting the signal to your frame buffer. In the former case, you might try using LVDS _receivers_ while still using RS-422 _transmitters_. (But be careful to use, e.g., a 90C032 -- 5V -- with a 26LS31 and not a 90LV032 -- 3.3V -- with a 26LS31... we had a camera that did that, and found out the hard way that a 90LV032 WILL eventually die when subjected to the output of a 26LS31. :-) ) --> Make sure you're clocking on the correct edge. This sounds silly, but at least with our Dalsa cameras at 15MHz, everything was so slow that -- over short cable runs -- you could get away with actually clocking data in on the wrong edge and it would still work just fine! As the cable was lengthened and setup/hold times eroded, you'd start to see "salt and pepper" noise in the image. > We are having ongoing problems with > signal integrity, > manifesting itself as crosstalk, resulting in noisy images being > transferred. If worse comes to worst, you might use one of the off-the-shelf SerDes'es and package everything up into one (or a small number) of much higher speed LVDS signals and ship it over, e.g., coax to significantly reduce crosstalk problems. (We went to fiber at the times for a variety of reasons, included perceived coolness on the part of customers as well as due to out not having any test equipment that would actually view the multi-gigabit signals themselves in copper.) National and Cypress (their HotLink transceivers) make lots of nice "provide parallel data here, comes out as high speed LVDS data there" chips. ---Joel Kolstad Electroglas ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu