dyer@spdcc.COM (Steve Dyer) (11/08/89)
In article <469@siswat.UUCP> buck@siswat.UUCP (A. Lester Buck) writes: >All I know is that I was configuring some RTs for resale by a client. I had >written SCSI drivers for various tape devices, so having the extra slot >filled by the IBM streamer seemed like a waste. I called the local Cipher >distributor asking for a SCSI streamer, but unfortunately no one (at that >point) had a 1/4" drive with a differential SCSI interface (thanks IBM!), Can someone here tell me what "differential SCSI" is, and how it's different from the SCSI I'm used to on Macs, {DEC,VAX}station 3100s, Adaptec 1542s, and Suns? We've got lots of IBM Athens 400mb drives, some of which would be nice to attach to our {DEC,VAX}station 3100s, and some people are asking about attaching DEC RZ53s to the IBM SCSI controller (yes, I know that the AOS 4.3 SCSI driver seems wedded to the weirdness of the Athens drive-- I'm concerned with things I have no control over like hardware incompatibility, and not with presumably tractable driver problems.) Is there really an electrical problem here? Can someone accurately characterize the issue as a compatibility problem or not? -- Steve Dyer dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer dyer@arktouros.mit.edu, dyer@hstbme.mit.edu
artk@congrunt.uunet.uu.net (Art Kreitman @ Congruent) (11/10/89)
> Path: uunet!snorkelwacker!spdcc!dyer > From: dyer@spdcc.COM (Steve Dyer) > Newsgroups: comp.sys.ibm.pc.rt > Reply-To: dyer@ursa-major.spdcc.COM (Steve Dyer) > >filled by the IBM streamer seemed like a waste. I called the local Cipher > >distributor asking for a SCSI streamer, but unfortunately no one (at that > >point) had a 1/4" drive with a differential SCSI interface (thanks IBM!), > > Can someone here tell me what "differential SCSI" is, and how it's > different from the SCSI I'm used to on Macs, {DEC,VAX}station 3100s, > Adaptec 1542s, and Suns? The SCSI spec (ANSI X3.131-1986) allow for two types of electrical SCSI bus connections, single ended and differential. For single ended, a bus signal is asserted on negated based on the voltage on a single wire relative to ground. 0 to .4 volts== asserted; 2.5 to 5.25 volts==negated. For differential, there is a signal line for assertation and a line for negation. Assertion is when "+SIGNAL" is more positive then "-SIGNAL", etc. Its all in the spec, if you're claiming to be writing SCSI Device Drivers, you can't live without it. Art Kreitman uunet!congrunt!artk Congruent Corp. congrunt!artk@uunet.uu.net 110 Greene Street New York, NY 10012 212-431-5100
buck@siswat.UUCP (A. Lester Buck) (11/10/89)
In article <509@ursa-major.SPDCC.COM>, dyer@spdcc.COM (Steve Dyer) writes: > Can someone here tell me what "differential SCSI" is, and how it's > different from the SCSI I'm used to on Macs, {DEC,VAX}station 3100s, > Adaptec 1542s, and Suns? The physical SCSI bus comes in two flavors - single-ended and differential. Single-ended is much more common. Single ended has each signal on a single wire, with all the extra lines as grounds. Differential uses two lines per signal, with a balanced plus and minus signal. Differential interfaces have much better noise immunity due to elimination of common mode noise. According to the SCSI standard, differential cables can be up to 25 meters, while single-ended cables are limited to 6 meters. This allows significant distances between the drives and the adapter, as in a computer room environment. The disadvantage of differential SCSI is that the driver chips are quite large, relatively expensive, and take a lot of power, so they have not been integrated onto the SCSI protocol chips common today. Differential and single-ended are not compatible on the same bus, though most differential implementations have circuitry to detect single-ended devices and shut down the differential drivers before they are damaged. Rancho Technology sells a bi-directional converter box for about $150 that can connect the two flavors, if you must mix interfaces. IBM was somewhat odd in choosing a differential SCSI implementation for their adapter. The 9332 SCSI disks (are these called Athens?) are also differential drives. > We've got lots of IBM Athens 400mb drives, some of which would be nice to > attach to our {DEC,VAX}station 3100s, and some people are asking about > attaching DEC RZ53s to the IBM SCSI controller (yes, I know that the > AOS 4.3 SCSI driver seems wedded to the weirdness of the Athens drive-- > I'm concerned with things I have no control over like hardware > incompatibility, and not with presumably tractable driver problems.) I would suggest just the opposite. The whole SCSI architecture has removed virtually all of your hardware incompatibility problems, but software can be a nightmare. As I have said before in this newsgroup, I have experience with a number of SCSI implementations, and the IBM RT version is some of the worst hardware and the worst software (AIX/VRM) that I have ever run across. Maybe things are a little better when you have full source for the drivers. I hope IBM gets it together with the MicroChannel SCSI adapter that they will soon announce. -- A. Lester Buck ...!texbell!moray!siswat!buck
griefer@ibmarc.uucp (Allan D. Griefer) (11/13/89)
There are two types of SCSI interfaces, differential and single-ended. Very simply single ended uses signals that are either positive or ground and differential uses signals that are expressed between two wires in the same manner that phone signals are expressed. The single-ended is simpler, but has some very short cable length limitations when compared to differential SCSI. There are a number of companies that produce SCSI convertors to switch between the two levels. Opinions are strictly my own, Allan D. Griefer, IBM Almaden Research Center, San Jose, CA BITNET: GRIEFER at ALMADEN Internet: griefer@ibm.com UUCP: ...!uunet!ibmarc!griefer mcimail: 398-8024