Re: Design flaw in the SCSI spec?
On Wed, Jan 8, 2020 at 4:57 PM Bill Gunshannon via cctalk < cctalk@classiccmp.org> wrote: > And then you have DEC. I have often heard DSSI described as > SCSI done right That's reportedly what DEC thought of it, but in practice it was SCSI done differently so as to be incompatible. The only technical "advantage" of DSSI was that it better matched some of DEC's other proprietary storage standards, thus enhancing vendor lock-in and profitability. If you were another company trying to decide whether to use SCSI or DSSI in your products, even if licensing and related cost weren't at issue, the _only_ reason to use DSSI would be to interoperate with DEC. I was a fan of DEC, but they did plenty of things for vendor lock-in rather than technical reasons.
Re: Design flaw in the SCSI spec?
On Wed, Jan 8, 2020 at 4:44 PM jim stephens via cctalk < cctalk@classiccmp.org> wrote: > The electronics of the SCSI interface were designed initially by the > precursor to Adaptec. At the time that the SCSI committee was formed, > there was work by a number of companies. Among them were HP, and the like. > There was no one common electronic design for SCSI. The electronic designs were unique to many companies. The earliest implementations were board-level interfaces consisting mostly of TTL. Early single-chip implementations included those from NCR and Adaptec, though many others came later. While there were many companies involved in the standardization of SCSI, the original SCSI 1 specification, ANSI X3.131-1986, was a relatively minor elaboration on SASI, which had been developed exclusively by Shugart Associates, hence the name SASI being an acronym of Shugart Associates System Interface. Shugart Associates was an early maker of 8 inch floppy drives and invented the 5 1/4 inch floppy.
Re: Design flaw in the SCSI spec?
On 1/8/20 6:44 PM, jim stephens via cctalk wrote: > > > On 1/8/2020 9:54 AM, Liam Proven via cctalk wrote: >> With his express permission, I'm forwarding a mail from a public list. >> I am interested in Gene's comments about the design of SCSI, but I >> don't know enough electronics to judge. >> >> I thought others here might. >> >> I have trimmed the mail a little to the relevant parts. >> > The electronics of the SCSI interface were designed initially by the > precursor to Adaptec. At the time that the SCSI committee was formed, > there was work by a number of companies. Among them were HP, and the like. > And then you have DEC. I have often heard DSSI described as SCSI done right. bill
Re: Design flaw in the SCSI spec?
On 1/8/2020 9:54 AM, Liam Proven via cctalk wrote: With his express permission, I'm forwarding a mail from a public list. I am interested in Gene's comments about the design of SCSI, but I don't know enough electronics to judge. I thought others here might. I have trimmed the mail a little to the relevant parts. The electronics of the SCSI interface were designed initially by the precursor to Adaptec. At the time that the SCSI committee was formed, there was work by a number of companies. Among them were HP, and the like. I know the design for the 40mhz flavor was done by Emulex with Bill Roberts, one of the founders involved in the design work at Emulex. I'm not sure that bean counters factored into anything there, but business did, as Emulex saw dominating the parallel bus market emerging as SCSI as a critical goal for the corporation. I went to an ANSI meeting with Bill Roberts when he presented work that Emulex did that defined the 20mhz flavor. Thru the process of the definition of SCSI-I there was a continuous evolution of the performance of the bus, and of the protocol to best exploit the additional performance. The one thing that I do recall was that AMP would always be present with a new and wonderful connector tweak or design if needed. They got to supply the bulk of the product as a result. thanks Jim
Re: Design flaw in the SCSI spec?
Al, Yep!! GOD Bless and Thanks, rich! On 1/8/2020 1:15 PM, Al Kossow via cctalk wrote: On 1/8/20 10:58 AM, Richard Pope via cctalk wrote: The A3000 did have a problem with the hardware. If you upgraded the SCSI controller chip it fixed the A3000 problem. I guessed it was a WD http://eab.abime.net/showthread.php?t=73936 yup.. Amigas weren't the only systems that had problems with WD33C93s
Re: Design flaw in the SCSI spec?
On 08/01/2020 18:56, Pete Turnbull via cctalk wrote: I recall the Christmas when our accountant insisted on an inventory of the electronics repair workshop. We kept our small components in a bank of drawers some 1.5m (around 5 feet) wide by about 1m (3 feet or so) high. Each drawer is just over 50mm (2") by 50mm. A drawer would hold several tens of 1/10W resistors, or some loose transistors or capacitors, or half a dozen ICs, or... And he wanted each one counted. And a cost given for each single item. And insisted that estimates were unacceptable for either count or cost. A few years later, when the company was about to be wound up, a colleague and I bought the bank of drawers and several other items for £100. I had a similar experience. "We need you to list each item, with item. We'll check a sample later". There were many hundreds of boards to list (as well as many systems, where the requirement was more reasonable). So I listed each board and when they came to check I picked (at random, of course) a bag of M9047 cards. "As you can see, they have an ID on the handle, but no serial number". They checked, they agreed with my analysis and were perfectly satisfied. Just simple people skills required :-) Antonio -- Antonio Carlini anto...@acarlini.com
Re: Design flaw in the SCSI spec?
On 1/8/20 10:58 AM, Richard Pope via cctalk wrote: > The A3000 did have > a problem with the hardware. If you upgraded the SCSI controller chip it > fixed the A3000 problem. I guessed it was a WD http://eab.abime.net/showthread.php?t=73936 yup.. Amigas weren't the only systems that had problems with WD33C93s
Re: Design flaw in the SCSI spec?
On 08/01/2020 18:06, Paul Koning via cctalk wrote: Allowing accountants to do electrical engineering makes just as much sense as allowing sociologists to do brain surgery. Off topic, but allowing accountants to do accountancy isn't always much better. I recall the Christmas when our accountant insisted on an inventory of the electronics repair workshop. We kept our small components in a bank of drawers some 1.5m (around 5 feet) wide by about 1m (3 feet or so) high. Each drawer is just over 50mm (2") by 50mm. A drawer would hold several tens of 1/10W resistors, or some loose transistors or capacitors, or half a dozen ICs, or... And he wanted each one counted. And a cost given for each single item. And insisted that estimates were unacceptable for either count or cost. A few years later, when the company was about to be wound up, a colleague and I bought the bank of drawers and several other items for £100. -- Pete Pete Turnbull
Re: Design flaw in the SCSI spec?
Hello all, I used Amigas for several decades and I really liked SCSI. The problems with the Amiga and SCSI were related to firmware and not the hardware. It was important to use the latest SCSI Roms on the 2091 and the 590. The A3000 did have a problem with the hardware. If you upgraded the SCSI controller chip it fixed the A3000 problem. I still use SCSI with my windoze machine. GOD Bless and Thanks, rich! On 1/8/2020 12:49 PM, Eric Smith via cctalk wrote: On Wed, Jan 8, 2020 at 10:55 AM Liam Proven via cctalk < cctalk@classiccmp.org> wrote: With his express permission, I'm forwarding a mail from a public list. I am interested in Gene's comments about the design of SCSI, but I don't know enough electronics to judge. And all that pickityness can be laid at the feet of a bean counter between the interface card designer, who specified a $2.00 schotkey diode for buss isolation, which had a maximum voltage drop across it of perhaps .1 volts, and changed to have an 8 cent Si diode with .666 volts drop across it, thereby lowering the logic one voltage by .45 volts. The complaint is with the design of a specific host adapter, not of SCSI itself, and specifically with the diode that host adapter uses to provide termination power. The complaint is only relevant to the use of passive termination, though the specific host adapter might include passive termination on board. The problem can be solved by removing the on-board passive termination and using an active terminator. A schottky diode will have a minimum drop around 0.3V (not 0.1V!), and typical with an actual termination load the drop will be closer to 0.4V to 0.5V. Better than a normal silicon rectifier, but not hugely better. Using real-world values of 5V termination power, 220/330 ohm passive terminator, 0.45V for the schottky diode drop, and 0.7V for a silicon diode drop, the resulting termination voltages are 2.73V with a schottky diode ((5.0 - 0.45) * 0.6) and 2.58V with a silicon diode ((5.0 - 0.7) * 0.6). While 2.73V is clearly better than 2.58V, the difference isn't as huge as Gene suggests. The SCSI data lines are normally actively driven, so during data transfer the termination voltage is not the only thing pulling the line up to a logic high. Even in the 1980s, a schottky diode cost under ten cents, not $2. It was maybe a few pennies more than a silicon diode. It is highly unlikely that a "bean counter" was involved in this decision at all. An engineer made the decision, presumably believing that the slightly higher voltage drop of the silicon diode didn't make a big difference. Is a schottky diode better for powering a bus terminator? Certainly. Is it a disaster to use a silicon diode in stead? In my opinion, no. I suspect that the Amiga SCSI host adapter had other issues that are much more significant with regard to system reliability. Eric
Re: Design flaw in the SCSI spec?
On Wed, Jan 8, 2020 at 10:55 AM Liam Proven via cctalk < cctalk@classiccmp.org> wrote: > With his express permission, I'm forwarding a mail from a public list. > I am interested in Gene's comments about the design of SCSI, but I > don't know enough electronics to judge. > > And all that pickityness can be laid at the feet of a bean counter >> between the interface card designer, who specified a $2.00 schotkey >> diode for buss isolation, which had a maximum voltage drop across it of >> perhaps .1 volts, and changed to have an 8 cent Si diode with .666 volts >> drop across it, thereby lowering the logic one voltage by .45 volts. >> > The complaint is with the design of a specific host adapter, not of SCSI itself, and specifically with the diode that host adapter uses to provide termination power. The complaint is only relevant to the use of passive termination, though the specific host adapter might include passive termination on board. The problem can be solved by removing the on-board passive termination and using an active terminator. A schottky diode will have a minimum drop around 0.3V (not 0.1V!), and typical with an actual termination load the drop will be closer to 0.4V to 0.5V. Better than a normal silicon rectifier, but not hugely better. Using real-world values of 5V termination power, 220/330 ohm passive terminator, 0.45V for the schottky diode drop, and 0.7V for a silicon diode drop, the resulting termination voltages are 2.73V with a schottky diode ((5.0 - 0.45) * 0.6) and 2.58V with a silicon diode ((5.0 - 0.7) * 0.6). While 2.73V is clearly better than 2.58V, the difference isn't as huge as Gene suggests. The SCSI data lines are normally actively driven, so during data transfer the termination voltage is not the only thing pulling the line up to a logic high. Even in the 1980s, a schottky diode cost under ten cents, not $2. It was maybe a few pennies more than a silicon diode. It is highly unlikely that a "bean counter" was involved in this decision at all. An engineer made the decision, presumably believing that the slightly higher voltage drop of the silicon diode didn't make a big difference. Is a schottky diode better for powering a bus terminator? Certainly. Is it a disaster to use a silicon diode in stead? In my opinion, no. I suspect that the Amiga SCSI host adapter had other issues that are much more significant with regard to system reliability. Eric
Re: Design flaw in the SCSI spec?
On 1/8/20 10:32 AM, Peter Corlett via cctalk wrote: > On Wed, Jan 08, 2020 at 10:17:29AM -0800, Chuck Guzis via cctalk wrote: >> Before I go delving into my pile of SCSI X3T10 documentation and interface >> chip datasheets, exactly *which* flavor of SCSI are we talking about here? > > Given the reference to the Amiga, almost certainly SCSI-1, i.e. 8 bit wide > single-ended HVD, clocked asynchronously at low single-digit MHz. > > The A2091 (and thus presumably the A590) was a pretty hateful controller, but > the main sources of pain were its shoddy firmware and the limited Zorro-II bus > rather than the SCSI interface. Perhaps the third-party GVP controllers > swapped > it around so the firmware was great but the SCSI side sucked. SCSI-1 was pretty chancy to begin, but even SCSI-2 had a bit of voodoo associated with it. --Chuck
Re: Design flaw in the SCSI spec?
On Wed, Jan 08, 2020 at 10:17:29AM -0800, Chuck Guzis via cctalk wrote: > Before I go delving into my pile of SCSI X3T10 documentation and interface > chip datasheets, exactly *which* flavor of SCSI are we talking about here? Given the reference to the Amiga, almost certainly SCSI-1, i.e. 8 bit wide single-ended HVD, clocked asynchronously at low single-digit MHz. The A2091 (and thus presumably the A590) was a pretty hateful controller, but the main sources of pain were its shoddy firmware and the limited Zorro-II bus rather than the SCSI interface. Perhaps the third-party GVP controllers swapped it around so the firmware was great but the SCSI side sucked.
Re: Design flaw in the SCSI spec?
Before I go delving into my pile of SCSI X3T10 documentation and interface chip datasheets, exactly *which* flavor of SCSI are we talking about here? --Chuck
Re: Design flaw in the SCSI spec?
> On Jan 8, 2020, at 12:54 PM, Liam Proven via cctalk > wrote: > > With his express permission, I'm forwarding a mail from a public list. > I am interested in Gene's comments about the design of SCSI, but I > don't know enough electronics to judge. > > I thought others here might. > > I have trimmed the mail a little to the relevant parts. > > -- > Liam Proven - Profile: https://about.me/liamproven > ... > And all that pickityness can be laid at the feet of a bean counter > between the interface card designer, who specified a $2.00 schotkey > diode for buss isolation, which had a maximum voltage drop across it of > perhaps .1 volts, and changed to have an 8 cent Si diode with .666 volts > drop across it, thereby lowering the logic one voltage by .45 volts. Allowing accountants to do electrical engineering makes just as much sense as allowing sociologists to do brain surgery. That's true whether you're building a disk interface or an airplane. But why is such a corporate organizational screwup any reflection on the SCSI standard? I suppose it depends on the spec, and in particular the conformance requirements. A properly constructed spec is written such that two conforming implementations will interoperate. Ethernet and the DECnet specs are examples of this, and in fact that property was an explicitly stated requirement in the DECnet architecture team. Unfortunately "conformance implies interoperability" is not nowadays all that common a standards property. I've even found myself involved in a standard development effort where the standard document editor explicitly told me that "conformance implies interoperability" was NOT a goal for that standard. I never quite understood why he thought that was reasonable, or why he was permitted to "hold the pen" on a major technical spec. paul