Loading ...
Sorry, an error occurred while loading the content.
 

Re: 1802 Archetecture Critique

Expand Messages
  • ajparent1
    ... I think you meant 64pin. The 68K series were all aircraft carriers (large surfaces). The 008 in the 48pin was far later than the 64pin! ... RCA did not
    Message 1 of 21 , Nov 1, 2012
      --- In cosmacelf@yahoogroups.com, Lee Hart <leeahart@...> wrote:
      >
      > On 10/30/2012 10:42 PM, Rick Cortese wrote:
      > >> There are no 48 pin IC packages; 40 was the biggest you could get.
      >
      > > 48 pin 68008s not that much later.
      > > http://www.cpu-world.com/CPUs/68008/index.html

      I think you meant 64pin. The 68K series were all aircraft carriers
      (large surfaces). The 008 in the 48pin was far later than the
      64pin!

      > >
      > > Intel offering
      > > http://en.wikipedia.org/wiki/File:C8095-90_Intel.jpg
      >
      > 48-pin ICs came out much later. I don't think any 48-pin packages
      > existed until at least 5 years after the 1802 was introduced. Certainly
      > RCA had no such packages.
      >
      > But RCA *was* selling 40-pin parts at the time. My guess is that they
      > didn't want to design or tool up a new 48-pin package for a new part
      > that they expected to sell in low quantities.
      >

      RCA did not do 40 pin however the industry was doing that and larger
      for a while.

      Also someone commented on the 48 pin 68008 which is actually a
      64pin (have one) as was the TI9900 cpu.

      The NEC D372 FDC and a USART (D361) were in a 42 pin package and it predates the 1802.

      I'd add some of the early calculator chips that predated CPU on a
      chip were larger than 48pins. Back then pins were cheaper than multiplexers in the logic (on silicon).

      40 pin was something that sorta evolved and was a good balance
      for the time between needed IO and cost of pins to do it.


      > Also consider that CMOS was considered a novelty, and a detriment at the
      > time. "Everybody knew" that CMOS cost more, and was slower than NMOS or TTL.

      It was not novelty and was an established logic family from about
      1970 maybe earlier. It was widely used where power or noise
      immunity was an important factor and was not that much slower
      then TTL of the day (1974ish). CMOS counters then were good
      to about 5mhz and TTL on a good day could hit 35mhz, some
      never more than 25hz. By 1974 National Semi and Motorola
      were making CMOS parts in the RCA pattern.

      The basic logic functions were cheap. They "cost" more in silicon
      area though, so MSI and later LSI functions were slower to come
      until demand plus finer lithography came about.

      I tended to view RCA back then as cutting edge.


      Allison


      > --
      > Don't worry about people stealing your ideas. If your ideas are any
      > good, you'll have to ram them down people's throats. -- Howard Aiken
      > --
      > Lee A. Hart http://www.sunrise-ev.com/LeesEVs leeahart@...
      >
    • ajparent1
      ... They did but when you look at instruction execution rate it was slower than many. ... It s a good part but the CMOS 6502 was as common or more so as well
      Message 2 of 21 , Nov 1, 2012
        --- In cosmacelf@yahoogroups.com, Andrew Wasson <awasson@...> wrote:
        >
        > The reason I built an 1802 elf and didn't attempt a Dyna Micro or some other small hobby computer of the time was because my electronics teacher pointed out that the 1802 was a single supply chip. Give it a power source around 5V and it works. All the other processors I was looking at needed several power supplies. That it didn't need a bootloader was sort of secondary because I didn't realize it was the only chip that didn't need one. The coolest feature for me once i started figuring out how to write code on it was that I could load an address in one of the registers and then make that register the program counter. I thought that was pure genius.
        >
        > The only thing I wished the 1802 had was a faster clock. I know that the 1802 was plenty fast when its architecture was used to its advantage but I recall the pundits of the day looking at clock speed rather than what could be done within each tick of the clock.

        They did but when you look at instruction execution rate it was
        slower than many.


        > I don't think RCA cared about the 1802 as a consumer product though. They had lots of use in cell phones, engine management, aerospace, inventory management, satellite communications, etc... It's too bad because I think that the 1802 was a competent as the 6502 and Z80's of the day but without industrial level backup, it was destined to become sidelined. Then again, my 1802 computers power up and run just as well as they did when they were new.

        It's a good part but the CMOS 6502 was as common or more so as well as the CMOS z80 (used in many places including the Epson PX8 portable).

        The 6502 in the fast flavors were as fast or faster than 4mhz z80s
        with average instruction excution in the 800,000 to 1,000,000 instructions per second. Even the lowly 8080 was averaging a instruction execution rate of 400,000. Put that against the
        1802 running with a 3.2mhz clock (about as fast as they go at 5V)
        of maxing at 400,000 single byte instructions per second. The big difference is all the other got faster, much faster (z180s run up
        to 33mhz). An 1802 that ran at 10mhz or even faster would indeed
        be something.

        The industry was a crazy place.


        Allison

        >
        > Andrew
        >
        >
        >
        >
        >
        > On 2012-10-29, at 9:40 AM, Douglas wrote:
        >
        > > OK, this might be fun.
        > > Based on experience we have so many years later,
        > > what do you think the 1802 design got really right and
        > > what do you think they missed or could have done better that
        > > time around? (pulling the pin and lob it in... taking cover)
        > >
        > >
        >
        > Andrew Wasson | Luna Design
        > 1651 West 15th Street,
        > North Vancouver BC., V7P 1N3
        > 604 987-2850
        >
        >
        >
        >
        >
        >
        >
        > [Non-text portions of this message have been removed]
        >
      • Rick Cortese
        ________________________________ From: ajparent1 To: cosmacelf@yahoogroups.com Sent: Thursday, November 1, 2012 5:10 PM Subject: [cosmacelf]
        Message 3 of 21 , Nov 2, 2012
          ________________________________
          From: ajparent1 <kb1gmx@...>
          To: cosmacelf@yahoogroups.com
          Sent: Thursday, November 1, 2012 5:10 PM
          Subject: [cosmacelf] Re: 1802 Archetecture Critique


           


          --- In cosmacelf@yahoogroups.com, Lee Hart <leeahart@...> wrote:
          >
          > On 10/30/2012 10:42 PM, Rick Cortese wrote:
          > >> There are no 48 pin IC packages; 40 was the biggest you could get.
          >
          > > 48 pin 68008s not that much later.
          > > http://www.cpu-world.com/CPUs/68008/index.html

          <big snips that may lose some context>
          -Also someone commented on the 48 pin 68008 which is actually a
          -64pin (have one) as was the TI9900 cpu.

          Most references I've found refer to it and show pictures as a 48 pin device. I agree the 68000 was a 64 pin device.

          It was probably a bad example for me to use anyway since it went the other way with respect to pin count. Only thing it shows is that pins and packages were a bit more flexible back then. I'll defer to Lee's point that RCA didn't have the capability of more then 40 pins. Been there, done that: There may have been people, marketing or bean counters, that overruled or dictated what was going to happen. Frequently engineers say "We can do this" manufacturing people say "We can make this" and the marketing people say "We don't want to sell this" or bean counters say "We will never recoup our capital investment" and it dies. It was a 'wishful thinking' type question anyway.<grin>

          A question if you have the time? When I was commuting to my job back ~1980s there was a billboard on the side of 101 that announced 'RCA Worlds first 32 bit CMOS processor.' I never saw an actual product or datasheet. I chalked it up to vaporware. Do you know anything about what this was going to be?

          Rick

          [Non-text portions of this message have been removed]
        • Lee Hart
          ... It seems like RCA used 40-pin ICs before the 1802 -- the 1801 is the first example that comes to mind. ... Yes; there were indeed many special packages
          Message 4 of 21 , Nov 2, 2012
            ajparent1 wrote:
            > RCA did not do 40 pin however the industry was doing that and larger
            > for a while.

            It seems like RCA used 40-pin ICs before the 1802 -- the 1801 is the
            first example that comes to mind.

            > I'd add some of the early calculator chips that predated CPU on a
            > chip were larger than 48 pins.

            Yes; there were indeed many special packages with large numbers of pins.
            They were often ceramic, and intended for high-cost applications. They
            also had large areas inside, and were used for hybrid ICs where there
            were several chips and even discretes inside.

            > Back then pins were cheaper than multiplexers in the logic (on silicon).

            Sounds about right. It may also be that there simply weren't many MSI
            multiplexer chips to choose from.

            > 40 pin was something that sorta evolved and was a good balance
            > for the time between needed IO and cost of pins to do it.

            Yes; it became the defacto standard for big ICs. You could get
            inexpensive plastic 40-pin DIPs, but for more than this they usually
            went to much more expensive packaging.

            >> Also consider that CMOS was considered a novelty, and a detriment at the
            >> time. "Everybody knew" that CMOS cost more, and was slower than NMOS or TTL.

            > It was not novelty and was an established logic family from about
            > 1970 maybe earlier. It was widely used where power or noise
            > immunity was an important factor and was not that much slower
            > then TTL of the day (1974ish). CMOS counters then were good
            > to about 5mhz and TTL on a good day could hit 35mhz, some
            > never more than 25hz. By 1974 National Semi and Motorola
            > were making CMOS parts in the RCA pattern.

            I suppose it depends on where you worked at the time. Were they a
            "bleeding edge" outfit, or a more conservative company? Here's my view
            of the history:

            In the mid-1970's (say 1975), TTL was the reigning king. It had largely
            replaced all the earlier bipolar digital technologies (RTL, DTL, HTL,
            etc.). TTL had enough variants (74, 74L, 74H, 74S, 74LS) that designers
            figured they could get "all" the speed and power they needed within the
            family.

            RCA came late to the game, and had no chance of garnering any of the TTL
            market. So they chose to promote CMOS (called COSMOS at first). It was
            much slower, but also much lower power and had much higher noise
            immunity. It found niche markets, but by 1975 I doubt it had even 1% of
            the market of TTL chips.

            Other manufacturers (notably Motorola and National) felt that it looked
            like a big enough niche market, so they second sourced RCA's CMOS chips.
            National tried to make 74-family compatible CMOS chips (their 74C
            series), but it was a flop in the market. (It was another 2 decades
            before 74HC could match the speed of 74LS chips).

            TTL's high power consumption turned out to be its Achilles heel. If you
            put enough bipolar transistors on a chip to make a decent microcomputer
            or memory chip, it burned itself up! Or, you had to slow it down so much
            that it was nearly useless. Designers first turned to PMOS, then NMOS.
            All the popular micros and memory chips of the 1970's were NMOS.

            I think RCA "missed the boat" on the early microcomputer market as well.
            By the time RCA came in, NMOS micros already ruled the market. So they
            chose to use their innovative CMOS technology to make a micro. CMOS
            "doomed" them to need twice as many transistors for the same
            functionality as NMOS, so the 1802 was more expensive. But they could
            make up for that with CMOS's traditional advantages; lower power and
            higher noise immunity. But it did relegate them to a niche market.

            Ironically, CMOS was the right choice in the long run. Today,
            *everything* is CMOS, because they are putting so many transistors on a
            chip that power consumption would be unmanageable if they still used
            NMOS! Plus, various technological improvements have eliminated the speed
            penalty of CMOS.

            In my case: I worked for Eastman Kodak throughout the 1970s. Purchasing
            ruled; they chose what brands we were allowed to buy. RCA was verboten,
            and Motorola was their favorite nephew. We couldn't use CMOS until
            Motorola started selling it. Before that, my CMOS work was done at home,
            or as secret "closet" projects. :-)

            I'm sure that each of us will have a different view of the situation,
            depending on where we were at the time. We're all "blind men describing
            the elephant". This is just the situation as I saw it! I'm sure others
            will have other perspectives. We'd have to collect them all and somehow
            assemble them to produce a more complete and accurate picture.

            --
            If you would not be forgotten
            When your body's dead and rotten
            Then write of great deeds worth the reading
            Or do these great deeds, worth repeating.
            -- Ben Franklin, from Poor Richard's Almanac
            --
            Lee A. Hart http://www.sunrise-ev.com/LeesEVs leeahart@...
          • Douglas
            ... I d say that was $.25! I only typed in very short little ditties on my Super Elf, so I am learning here from these posts. It seems to me a reassignable
            Message 5 of 21 , Nov 2, 2012
              --- In cosmacelf@yahoogroups.com, "urrossum@..." <mark@...> wrote:

              >
              > I really like the 16 x 16 register array, and apparently I'm not the only one that thinks it's a good thing - the Texas Instruments MSP430 series has the same array, too. (Actually, theirs isn't quite as flexible, since the functions of the first four registers are permanently assigned, rather than the great flexibility that the 1802 provides.)
              >
              > However, a real stack, and built-in CALL, RETURN, PUSH, and POP instructions would have been a really nice thing to have on the 1802. In my mind, that's the most glaring thing I wish was different.
              >
              > The Q bit and EF flags bits, while handy for a hobbyist computer or a very minimal system, seem sort tacked on; the I/O ports should generally be sufficient. And what's with the processor waking up from reset with the interrupt system turned on?
              >
              > Just my $0.02...
              > ~~
              > Mark Moulding
              >

              I'd say that was $.25! I only typed in very short little ditties on my Super Elf, so I am learning here from these posts.
              It seems to me a reassignable program counter and no stack instructions would make this a tough processor to generate code for
              from a high level language. Sounds like the architect had state machines in mind, not procedural logic. Who and where is that architect, anyway!
            • Dave Ruske
              ... http://www.cosmacelf.com/weisbecker.html Dave [Non-text portions of this message have been removed]
              Message 6 of 21 , Nov 2, 2012
                On Nov 2, 2012, at 10:04 PM, Douglas <yahoo@...> wrote:
                > Sounds like the architect had state machines in mind, not procedural logic. Who and where is that architect, anyway!

                http://www.cosmacelf.com/weisbecker.html

                Dave

                [Non-text portions of this message have been removed]
              • Andrew Wasson
                Actually, I believe those who were writing programs for the 1802 at that period were quite happy with the way it was designed. You can read a number of
                Message 7 of 21 , Nov 2, 2012
                  Actually, I believe those who were writing programs for the 1802 at that period were quite happy with the way it was designed.

                  You can read a number of articles in the Quest and ipsofacto newsletters where software writers of the day offered their 2-bits about what they liked/disliked about the 1802. For example, I recall an article back in the day where the author of Tiny Basic mentioned that because of the way the 1802 was designed that he was able to build Tiny Basic for the ELF using less memory (for the program) than other CPU's of the day.

                  In machine language, I've never written anything larger than a monitor program or a simple motor controller but it's been my opinion that the instructions work quite well for procedural code and the ability to swap your program counter with any general purpose register is just a bonus because you can jump to anywhere in memory at runtime, execute and instruction and then jump back in a couple of cycles of the clock.

                  Maybe someone else who has experience writing real code on the 1802 can chime in with a more informed opinion than mine.

                  Cheers,
                  Andrew


                  On 2012-11-02, at 8:04 PM, Douglas wrote:

                  >
                  >
                  > I'd say that was $.25! I only typed in very short little ditties on my Super Elf, so I am learning here from these posts.
                  > It seems to me a reassignable program counter and no stack instructions would make this a tough processor to generate code for
                  > from a high level language. Sounds like the architect had state machines in mind, not procedural logic. Who and where is that architect, anyway!
                  >
                  >

                  Andrew Wasson | Luna Design
                  1651 West 15th Street,
                  North Vancouver BC., V7P 1N3
                  604 987-2850







                  [Non-text portions of this message have been removed]
                • William Donnelly
                  The 1802 is PERFECT in EVERY way. Don t try to convince me otherwise. Shut Up! La-La-La-La-La-La-La-La-La-La-La -- bIll _________________ ( __/) This is Bunny.
                  Message 8 of 21 , Nov 2, 2012
                    The 1802 is PERFECT in EVERY way.

                    Don't try to convince me otherwise.

                    Shut Up!
                    La-La-La-La-La-La-La-La-La-La-La

                    -- bIll
                    _________________
                    (\__/) This is Bunny.
                    (='.'=) Copy and paste Bunny into your signature.
                    (")_(") Help Bunny gain World Domination.

                    .¸¸.·´¯`·.¸¸.·´¯`·.¸.::::? www.ChicoSkyWatch.org
                    <http://www.ChicoSkyWatch.org/>
                    GeoEngineering . Aerosol Spraying . Toxic Aluminum in Rainwater

                    ?__? ? ? ? www.AE911Truth.org <http://www.AE911Truth.org/>



                    [Non-text portions of this message have been removed]
                  • Douglas
                    ... Very cool. Thanks. I ve been on that site and missed that page.
                    Message 9 of 21 , Nov 2, 2012
                      --- In cosmacelf@yahoogroups.com, Dave Ruske <dave@...> wrote:
                      >
                      > On Nov 2, 2012, at 10:04 PM, Douglas <yahoo@...> wrote:
                      > > Sounds like the architect had state machines in mind, not procedural logic. Who and where is that architect, anyway!
                      >
                      > http://www.cosmacelf.com/weisbecker.html
                      >
                      > Dave
                      >
                      > [Non-text portions of this message have been removed]
                      >

                      Very cool. Thanks. I've been on that site and missed that page.
                    • Douglas
                      ... I will look those up. I also am going to read Tom Pittman s book. I suspect it will wise me up to how the unique features could be leveraged. Would have
                      Message 10 of 21 , Nov 2, 2012
                        --- In cosmacelf@yahoogroups.com, Andrew Wasson <awasson@...> wrote:
                        >
                        > Actually, I believe those who were writing programs for the 1802 at that period were quite happy with the way it was designed.
                        >
                        > You can read a number of articles in the Quest and ipsofacto newsletters where software writers of the day offered their 2-bits about what they liked/disliked about the 1802. For example, I recall an article back in the day where the author of Tiny Basic mentioned that because of the way the 1802 was designed that he was able to build Tiny Basic for the ELF using less memory (for the program) than other CPU's of the day.
                        >
                        > In machine language, I've never written anything larger than a monitor program or a simple motor controller but it's been my opinion that the instructions work quite well for procedural code and the ability to swap your program counter with any general purpose register is just a bonus because you can jump to anywhere in memory at runtime, execute and instruction and then jump back in a couple of cycles of the clock.
                        >
                        > Maybe someone else who has experience writing real code on the 1802 can chime in with a more informed opinion than mine.
                        >
                        > Cheers,
                        > Andrew
                        >
                        >
                        > On 2012-11-02, at 8:04 PM, Douglas wrote:
                        >
                        > >
                        > >
                        > > I'd say that was $.25! I only typed in very short little ditties on my Super Elf, so I am learning here from these posts.
                        > > It seems to me a reassignable program counter and no stack instructions would make this a tough processor to generate code for
                        > > from a high level language. Sounds like the architect had state machines in mind, not procedural logic. Who and where is that architect, anyway!
                        > >
                        > >
                        >
                        > Andrew Wasson | Luna Design
                        > 1651 West 15th Street,
                        > North Vancouver BC., V7P 1N3
                        > 604 987-2850
                        >
                        >
                        >
                        >
                        >
                        >
                        >
                        > [Non-text portions of this message have been removed]
                        >

                        I will look those up. I also am going to read Tom Pittman's book.
                        I suspect it will wise me up to how the unique features could
                        be leveraged. Would have loved to have had such a book back in... 1978... ?... when I built the Super Elf. It likely would have made me a programmer years earlier.
                      • David W. Schultz
                        ... There are stack instructions. STXD works as a push but to pop you need to use IRX and LDX. This is because the LDXA instruction increments the address
                        Message 11 of 21 , Nov 3, 2012
                          On 11/02/2012 10:04 PM, Douglas wrote:
                          > I'd say that was $.25! I only typed in very short little ditties on
                          > my Super Elf, so I am learning here from these posts. It seems to me
                          > a reassignable program counter and no stack instructions would make
                          > this a tough processor to generate code for from a high level
                          > language. Sounds like the architect had state machines in mind, not
                          > procedural logic. Who and where is that architect, anyway!
                          >
                          >

                          There are stack instructions.

                          STXD works as a push but to pop you need to use IRX and LDX. This is
                          because the LDXA instruction increments the address after the load
                          rather than before. It can't do it before because there is all of one
                          half clock cycle between the time the instruction is latched and the
                          address for the next machine cycle has to be on the address bus.



                          --
                          David W. Schultz
                          http://home.earthlink.net/~david.schultz
                          Returned for Regrooving
                        Your message has been successfully submitted and would be delivered to recipients shortly.