On Tuesday 14 February 2006 21:59, Jeff Vian wrote: [...] >> Yeah, my post wasn't very clear. Two channels per controller, >> master/slave device per channel, master device should be on the >> connector closest to the controller because of noise issues and >> that's why some half-bright got the idea that it should just be left >> up to cable select anyway, ... > >NOTE: If you have a CS cable putting the master in the middle > probably will not work. The CS cable has the master connector at the > end and the slave in the middle. > >I have always been told (even in hardware schooling) that you should >always have a device at the cable end to prevent 'ringing' and noise >caused by signal reflection from an unterminated cable end. This is >mandatory on SCSI and highly recommended on IDE cabling. (it is >sometimes unnecessary, not because the problem does not exist, but >because the data transfer is so [relatively] slow that some > interference does not impact performance seem by the user). >I never install a single device on a cable unless it is connected to > the end of the cable. This by default makes the single device the > master, and thus implies a reason for putting the master connector at > the end of the cable. > >> And it's no surprise that it sometimes "just works" for some people, >> and sometimes just can't be made to do what you really wanted it to >> do. >> >> ATA confuses me. It irks me no end that ATA has become mainstream, >> while SCSI gets no respect. > >SCSI gets a lot of respect. It also costs a lot more $ than IDE/ATA > so it is harder to get the home user to justify the extra expense. That $ factor, and the reputation for needing a nearly endless supply of virgins to be sacrificed over it to make it work in the presense of hidiously designed, then further emasculated by the bean counters changing parts without telling engineering because this parts 5% of the cost of the part the engineers specified, terminations, or should I say miss-terminations. I scsi bus, built with the usual 220/330 ohm passive resistor packs for termination works great on paper, and still works pretty error free if the isolation diode the engineers wanted, either a power GE, better yet a schotkey was used, both of which will have less than say .2 volts of voltage drop across them. But some bean counter sees the cost and decides an SI diode, with .75 to .8 volts of drop is good enough and the board gets populated and shipped before the engineers that have a clue discover it. Whats the problem? Well, with no diode, you get a logic one voltage of about 3 volts, which is comfortably in excess of the standard ttl logic one voltage of 2.4 volts, a 600 millivolt noise and echo margin. Stick an SI diode with a voltage drop of .7 volts in series with the terminations power supply, and you've lost about .45 volts at the logic line itself, leaving only a 150 millivolt echo and noise margin. You need a chicken or goat occasionally to make that work. Its there to keep the power supply in the outboard box with the scsi stuff in it, from backfeeding its 5 volt line into the hosts 5 volt line, effectively preventing you from turning the host off! Then toss in that most computer psu's fade downward with age, and the 5.03 volts you had when it was fresh a year ago, is now about 4.88 volts. Now the logic one noise/echo margin is reduced to 30 millivolts, and you can sacrifice all the virgins on the planet without making it reliable enough to even boot up. The bottom line is that when active terminations are used, the terminator chip regulates this voltage internally and can tolerate a fadeing psu well enough that when the box goes down, it will probably be from some other low voltage induced failure, not a scsi bus error. Active terms also have the advantage of much less power dissipation, so the moral of this tome is to, when dealing with building scsi stuff, demand and get from the suppliers, devices that are actively terminated for both ends of the cable. If the one you are dealing with cannot supply such devices, then I'd suggest you look elsewhere for a similar device. >> To say what I meant to say, I removed the _second_ controller card >> (not slave controller or whatever it was I said) and have both >> drives on the mobo's controller's primary channel, with the CD drive >> all by itself on the secondary channel (pretending to the BIOS that >> it's SCSI no less). Grub is on the drive on the "master" connector >> (closest to the controller on the cable, strapped CS), along with >> FC3, and two BSDs. The drive connected to the slave controller >> (which is actually there to be a spare because I don't trust ATA >> drives) has another BSD and sometimes other Linux on it. Thats probably wrong, unless its actually being seen as /dev/hdb due to the CS working and its not the last drive. If the CS master, which does the terminating of the cable, then it must be on the END of the cable to do this correctly. Is this drive on the end of the cable then seen as /dev/hda? [...] -- Cheers, Gene People having trouble with vz bouncing email to me should add the word 'online' between the 'verizon', and the dot which bypasses vz's stupid bounce rules. I do use spamassassin too. :-) Yahoo.com and AOL/TW attorneys please note, additions to the above message by Gene Heskett are: Copyright 2006 by Maurice Eugene Heskett, all rights reserved.
