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

Re: Power savings in 1802

Expand Messages
  • wa9hsl
    If one tries too hard he may set up power-saving-control-logic that consumes more power than it saves! ... al ... any ... still ... oscillator ... the ...
    Message 1 of 23 , Jun 9, 2004
    • 0 Attachment
      If one tries too hard he may set up power-saving-control-logic that
      consumes more power than it saves!

      :o)

      al
      ....

      --- In cosmacelf@yahoogroups.com, "sbirdasn" <sbirdasn@y...> wrote:
      > Binh,
      >
      > --- In cosmacelf@yahoogroups.com, "Binh Dao" <ltt_doc@y...> wrote:
      > > Thanks Dave -
      > >
      > > The two ways of entering the PAUSE mode were what I was looking
      > for.
      > > It seems that you can enter PAUSE mode either through an
      > instruction
      > > (IDLE) or via hardware (dropping the WAIT pin low).
      > >
      > > I looked in the data sheet and could not find out if there are
      any
      > > special conditions that you need to meet prior to bringing the
      > > processour out of PAUSE mode. For example, I am concerned about
      > > ensuring that the clock signal is stable before the processor
      > starts
      > > running again.
      >
      > But the clock DOESN'T stop in either of those two modes. It's
      still
      > running. Full blast. Even an 1804/5/6 will leave the main
      oscillator
      > going when IDL mode is entered.
      >
      > The only requirement is that you meet setup/hold conditions for
      the
      > WAIT pin if you want the action to take place during a particular
      > clock cycle. If you don't meet the setup time for a given clock
      > cycle, then the state change will occur on the next appropriate
      clock
      > edge.
      >
      > A properly designed clock oscillator will be a small part of the
      > total system power consumption when the processor is active in
      most
      > cases.
      >
      > Now if your power budget is really tight or you really want to
      > maximize the power use, then you have to look at every current
      draw
      > in the system, and that will include an analysis of the clock
      circuit
      > and anything that it drives directly or is based on.
      >
      > All for now.
      >
      > Tony.
    • jplagasse@jparchives.com
      I don t know if any of this applies in this case, but here are some comments and low power tricks i used in the 1010/303 1802 DAQ system. The entire system was
      Message 2 of 23 , Jun 9, 2004
      • 0 Attachment
        I don't know if any of this applies in this case, but
        here are some comments and low power tricks i used in
        the 1010/303 1802 DAQ system. The entire system was
        powered through an MTL zener barrier, allowing only
        about 30ma max. for the entire system.

        This is entirely from (my own) memory, so if there are
        any errors etc. & anyone is interested, i could do a
        more formal check on this stuff.

        1823 ram chips provided the lowest power consumption
        out of everything i looked at. Typical CMOS ram chips
        used far more power.
        Smaller capacity National CMOS eproms (low speed
        versions) used small amounts of power compared to any
        other CMOS eprom mnf's i looked at.

        The 12 bit teledyne 8702 D/A converter (if i remember
        correctly) was very stable and noise free at low
        temperatures, yet used miniscule amounts of power
        compared to others.
        This chip is now obsolete, of course.

        (The original RCA development system we got, used a
        cheepie RC oscillator to power the CPU test board for
        it's emulator (micromonitor).)

        1800 series and 4000 cmos series was used almost
        exclusively, far less power consumption, wider supply
        voltage range etc.

        Spiking was a problem during switching transients...
        especially for the 7660 voltage inverter required for
        the D/A converter. Used a 10 ohm resistor in series
        with the flying capacitor, to give a longer
        rising/falling edge. This "spread out" the current
        formerly required during switching, & did not interfere
        with the negative voltage obtained.

        Voltage regulation was a problem, as regulator
        quiescent power draw vs ability to handle transients
        needed to be balanced. Couldn't use large capacitors,
        hazardous location stuff etc. I forget which regulator
        i used, but it's another old one with a good balance
        between these two specs.
        .01 uf caps were spread out throughout the board.

        It strikes me, that the 1010/303 system could NOT be
        reproduced today, with the chips now available.
        In some ways, the "old stuff" was better than what we
        have now...

        In case interesting or useful,
        Regards,
        JP
      • jplagasse@jparchives.com
        oops... Should be 8702 A/D converter, instead of D/A. These used polypropylene capacitors for their ultra low leakage spec. minimum wave distortion etc. All
        Message 3 of 23 , Jun 9, 2004
        • 0 Attachment
          oops...

          Should be 8702 A/D converter, instead of D/A.
          These used polypropylene capacitors for their ultra low
          leakage spec. minimum wave distortion etc.

          All CMOS is NOT equal, just cause it says CMOS doesn't
          mean it's lowest power especially when powered up but
          not active.

          Oh yes... interestingly, RCA 4000 cmos was the absolute
          best, had the best noise immunity across the power
          rails.
          We wound up removing & replacing national & motorola
          4000 stuff with RCA after a few production runs. A few
          stories here...

          Motorola zeners provided the best (lowest) leakage spec
          below their rated voltages. Here again, these were way
          below their rated spec. Other mnf's didn't even come
          close... and caused input voltages to droop. Published
          Specifications didn't help here, trial & error (&
          experience) was needed.
          These were used to protect all analog inputs against
          static discharge, & for the hazardous location
          protection etc.

          Where i work, they are trying to sell off the entire
          series of 1010/303 1802 systems. If any of these become
          available (if they garbage these), I'll try to make
          some of these boards available to forum participants.
          Can provide schematics & parts lists etc. also... But
          I'll have to see what happens.

          Regards,
          JP
        • Lee Hart
          ... Yes, you could. But the designer would have to be pretty dumb not to notice. What usually happens is that power consumption isn t known for sure until the
          Message 4 of 23 , Jun 9, 2004
          • 0 Attachment
            wa9hsl wrote:
            > If one tries too hard he may set up power-saving-control-logic that
            > consumes more power than it saves!

            Yes, you could. But the designer would have to be pretty dumb not to
            notice.

            What usually happens is that power consumption isn't known for sure
            until the product design is almost done. Then it's too late to make
            radical changes. The design engineer sometimes has to do some rather odd
            things to save power. Here are some that I've done.

            1. Change memory chips.

            Very few memory chips are true CMOS devices (whose power consumption is
            essentially zero when none of the inputs are changing). Most are
            partially CMOS, and partially NMOS. The NMOS parts of the chip draw
            power continuously when turned on. For example, a RAM might draw <1
            microamp when chip enable is high (only the CMOS parts of the chip are
            active), but 50ma when chip enable is held low (the NMOS parts are
            active).

            On one 1802 design, we used two RCA MWS5114 1Kx4 CMOS RAMs. They are
            spec'd at VCC=0.1ma when CE=high, 5ma when CE is cycling at 1 MHz. In
            our circuit with a 2 MHz clock (250 KHz cycle time), we expected they
            would draw 1/4th the power (since a true CMOS chip's power consumption
            is proportional to frequency). But the 5114 actually drew 4ma, putting
            us over our power budget.

            The short-term fix was to replace it with a Harris HM6514, which was
            pin-for-pin compatible and had the same power specs (0.1ma with CE=high,
            5ma with CE cycling at 1 MHz). But, the 6514 was a true CMOS part, and
            its power consumption *did* go down 4:1 at 1/4th the cycle time.

            Later, we found that the 5114 draws power continuously when its CE pin
            is low, so it apparently has some NMOS circuitry inside. Minimize the CE
            low time, and you minimize the power consumption. So, we redesigned the
            circuit to only pull CE low during the TPB pulse (1/8th the time). This
            cut the 5114's power consumption so it could be used.

            2. Design tricks

            When I was using the 1802 in the late 1970's, there were no CMOS EPROMs.
            You either had to give up ever changing the software and order a
            mask-programmed ROM (thousands of dollars in mask fees), or put up with
            the much higher power consumption of the NMOS EPROMs. This latter option
            was not feasible for battery-operated products.

            So, on one product we used an NMOS EPROM, but with an I/O bit to turn
            off its power. On power-up, the EPROM data was copied to RAM, the EPROM
            was turned off, and the program subsequently ran from RAM. This was
            messy because buffers were needed for every address, data, and chip
            select line to the EPROM (the EPROM shorted all its pins to ground when
            it was not powered). Later, we found that TTL fusible-link PROMs could
            be tied directly to the address and data buses, and when their power was
            turned off, all their I/O pins became open circuits. So this was the
            final design; a micropower 1802 product with a TTL PROM that drew 100ma
            when active! (but it was only active for <1 second on first powerup :-)

            3. Oscillator tricks

            The 1802 is very forgiving on its clock circuit -- far more so than any
            contemporary microcomputer. You can start and stop the clock any time,
            rise and fall times aren't critical, etc. So, many of my 1802 circuits
            would change the clock frequency; low when nothing was going on, fast
            when full speed was needed.

            Crystal oscillators normally take a long time to start oscillating. So,
            one easy way to add a low-power mode is to have an output bit that
            momentarily stops the oscillator. A rising edge on an output is
            capacitively coupled to a transistor, which shorts the oscillator. The
            capacitor finishes charging, the transistor turns back off, and the
            oscillator restarts. The oscillator thus "putts" along at a low
            effective clock rate for low power, though while it is running, it runs
            at full speed so your software timing loops etc. work correctly.
            --
            "Never doubt that the work of a small group of thoughtful, committed
            citizens can change the world. Indeed, it's the only thing that ever
            has!" -- Margaret Mead
            --
            Lee A. Hart 814 8th Ave N Sartell MN 56377 leeahart_at_earthlink.net
          • wa9hsl
            ... oscillating. So, ... The ... runs ... That s funny. Sort of like a rabbit in a clover field I suppose... :o) Speaking of non-synchronized switching of the
            Message 5 of 23 , Jun 9, 2004
            • 0 Attachment
              > Crystal oscillators normally take a long time to start
              oscillating. So,
              > one easy way to add a low-power mode is to have an output bit that
              > momentarily stops the oscillator. A rising edge on an output is
              > capacitively coupled to a transistor, which shorts the oscillator.
              The
              > capacitor finishes charging, the transistor turns back off, and the
              > oscillator restarts. The oscillator thus "putts" along at a low
              > effective clock rate for low power, though while it is running, it
              runs
              > at full speed so your software timing loops etc. work correctly.

              That's funny. Sort of like a rabbit in a clover field I
              suppose... :o) Speaking of non-synchronized switching of the
              oscillator, one of my 1802s has a dual speed clock with a toggle
              switch to go from a very low (1-20hz) to a higher 3579khz
              frequency. I've always been very meticulous in switching the speeds
              with the CPU in a non-running mode for fear of glitching the running
              program. From what you say that precaution isn't needed. Anyway I
              enjoyed your writeup and you made several good points.

              al
              ....
            • Mark Graybill
              ... I have used everything from conditioned manual switches to pulling wires out of one hole in a solderless breadboard and pushing it into another to change
              Message 6 of 23 , Jun 9, 2004
              • 0 Attachment
                On Jun 9, 2004, at 11:09 AM, wa9hsl wrote:

                > I've always been very meticulous in switching the speeds
                > with the CPU in a non-running mode for fear of glitching the running
                > program. From what you say that precaution isn't needed.

                I have used everything from conditioned manual switches to pulling
                wires out of one hole in a solderless breadboard and pushing it into
                another to change clock signals on an 1802 without causing software
                glitches. It's not exactly definitive, but I have to admit I've gotten
                pretty lacksadaisical about how clean the clock signal to the 1802 is.

                Noise immunity is not the name for it.

                Someone here said the 1802 would execute a brick if it had an even
                number of bits, as I recall. I think you could clock it off one, too,
                so long as you could tolerate the energy use caused by the slow
                transitions between rails. ;)

                -Mark G.
              • Lee Hart
                ... Well, times change... There are lots of circuits we can build easier today than years ago. But, there are also lots of circuits that we could build 10-20
                Message 7 of 23 , Jun 9, 2004
                • 0 Attachment
                  jplagasse@... wrote:
                  >
                  > I don't know if any of this applies in this case, but
                  > here are some comments and low power tricks i used in
                  > the 1010/303 1802 DAQ system. The entire system was
                  > powered through an MTL zener barrier, allowing only
                  > about 30ma max. for the entire system.
                  >
                  > This is entirely from (my own) memory, so if there are
                  > any errors etc. & anyone is interested, i could do a
                  > more formal check on this stuff.
                  >
                  > 1823 ram chips provided the lowest power consumption
                  > out of everything i looked at. Typical CMOS ram chips
                  > used far more power.
                  > Smaller capacity National CMOS eproms (low speed
                  > versions) used small amounts of power compared to any
                  > other CMOS eprom mnf's i looked at.
                  >
                  > The 12 bit teledyne 8702 D/A converter (if i remember
                  > correctly) was very stable and noise free at low
                  > temperatures, yet used miniscule amounts of power
                  > compared to others.
                  > This chip is now obsolete, of course.
                  >
                  > (The original RCA development system we got, used a
                  > cheepie RC oscillator to power the CPU test board for
                  > it's emulator (micromonitor).)
                  >
                  > 1800 series and 4000 cmos series was used almost
                  > exclusively, far less power consumption, wider supply
                  > voltage range etc.
                  >
                  > Spiking was a problem during switching transients...
                  > especially for the 7660 voltage inverter required for
                  > the D/A converter. Used a 10 ohm resistor in series
                  > with the flying capacitor, to give a longer
                  > rising/falling edge. This "spread out" the current
                  > formerly required during switching, & did not interfere
                  > with the negative voltage obtained.
                  >
                  > Voltage regulation was a problem, as regulator
                  > quiescent power draw vs ability to handle transients
                  > needed to be balanced. Couldn't use large capacitors,
                  > hazardous location stuff etc. I forget which regulator
                  > i used, but it's another old one with a good balance
                  > between these two specs.
                  > .01 uf caps were spread out throughout the board.
                  >
                  > It strikes me, that the 1010/303 system could NOT be
                  > reproduced today, with the chips now available.
                  > In some ways, the "old stuff" was better than what we
                  > have now...

                  Well, times change... There are lots of circuits we can build easier
                  today than years ago. But, there are also lots of circuits that we could
                  build 10-20 years ago that can't be duplicated with parts available
                  today.

                  People tend to think that "progress" moves in a straight line, but it
                  doesn't. There are MANY directions it can take. By pushing technology in
                  one direction, it gets better at some things but WORSE at others.

                  Our present semiconductor industry has decided to move single-mindedly
                  toward sub-micron geometries and GHz speeds. This is what you need to
                  build fast PCs, but it is the *wrong* direction if you want high
                  reliability, high noise immunity, long life, high breakdown voltages, or
                  high accuracy analog parts. Some examples:

                  1. The 1802 is naturally radiation-resistant; thousands of times
                  better than modern parts. That allowed its use in spacecraft and
                  other high-radiation environments. Modern parts require heroic
                  measures to make them survive.

                  2. The 1802 has phenomenal noise immunity; far higher than any modern
                  microcomputers. Modern parts need far more shielding and noise
                  filtering.

                  3. Germanium devices have half the voltage drop of silicon devices.
                  They are essentially out of production, so circuits that run on a
                  single solar cell or the 200mv thermopiles are no longer buildable.

                  4. Single-junction diodes, transistors, and SCRs were available past
                  5000 volts; but with today's processes, 1200 volts is about the
                  best we can do. When higher voltages are needed, multiple devices
                  must be stacked in series.

                  5. Analog circuitry has suffered the most severe limitations. The
                  limits that we can measure to with simple parts (low voltage, low
                  current, low temperature, etc. are worse with modern parts. For
                  example, I designed a picoammeter in 1980 using an off-the-shelf
                  Analog Devices part that could measure down to 1 femtoamp (10
                  electrons per second!) Analog Devices discontinued it in 1984
                  because they could no longer get the tunnel diodes it used. The
                  replacement part was 10 times worse.

                  We are fast approaching the point where simple devices like 1N400x
                  rectifiers, 1N52xx zener diodes, 2N4401 transistors, and 555 timers will
                  go out of production. No major manufacturers make them any more; they
                  are only made by a few third-world outfits that produce them on ancient
                  semiconductor equipment cast off by the big semiconductor houses.
                  --
                  "Never doubt that the work of a small group of thoughtful, committed
                  citizens can change the world. Indeed, it's the only thing that ever
                  has!" -- Margaret Mead
                  --
                  Lee A. Hart 814 8th Ave N Sartell MN 56377 leeahart_at_earthlink.net
                • Steve Ross
                  Okay, I ve got to ask: Why? What was the application? Steve [snip] The silliest 1802 clock I ve used is a phototransistor configured with the internal
                  Message 8 of 23 , Jun 9, 2004
                  • 0 Attachment
                    Okay, I've got to ask: Why? What was the application?
                    Steve

                    [snip]
                    The silliest 1802 clock I've used is a phototransistor configured with
                    the internal inverter as an RC oscillator whose frequency depends on
                    light intensity. In the dark it got so slow that it eventually stopped.
                    I didn't have any problems with program execution no matter what the
                    clock speed.
                    [snip]
                  • Steve Ross
                    I had to ask, and now I know. 8-) thanks Steve ... From: Lee Hart [mailto:leeahart@earthlink.net] Sent: Wednesday, June 09, 2004 3:03 PM To:
                    Message 9 of 23 , Jun 9, 2004
                    • 0 Attachment
                      I had to ask, and now I know. 8-) thanks
                      Steve

                      -----Original Message-----
                      From: Lee Hart [mailto:leeahart@...]
                      Sent: Wednesday, June 09, 2004 3:03 PM
                      To: cosmacelf@yahoogroups.com
                      Subject: Re: [cosmacelf] Re: Power savings in 1802


                      > The silliest 1802 clock I've used is a phototransistor configured with
                      > the internal inverter as an RC oscillator whose frequency depends on
                      > light intensity. In the dark it got so slow that it eventually stopped.

                      Steve Ross wrote:
                      > Okay, I've got to ask: Why? What was the application?

                      (You just had to ask, didn't you? :-)

                      It was a name tag, with nothing but the 1802 and that phototransistor as
                      the only active devices. All it did was blink a bunch of LEDs connected
                      between the various address bus pins. The "program" was just to wire the
                      data bus to the address bus, with a couple pins tied high to avoid it
                      ever hitting an IDL instruction.




                      [Non-text portions of this message have been removed]
                    • Lee Hart
                      ... The silliest 1802 clock I ve used is a phototransistor configured with the internal inverter as an RC oscillator whose frequency depends on light
                      Message 10 of 23 , Jun 9, 2004
                      • 0 Attachment
                        Mark Graybill wrote:
                        > I have used everything from conditioned manual switches to pulling
                        > wires out of one hole in a solderless breadboard and pushing it into
                        > another to change clock signals on an 1802 without causing software
                        > glitches. It's not exactly definitive, but I have to admit I've gotten
                        > pretty lacksadaisical about how clean the clock signal to the 1802 is.

                        The silliest 1802 clock I've used is a phototransistor configured with
                        the internal inverter as an RC oscillator whose frequency depends on
                        light intensity. In the dark it got so slow that it eventually stopped.
                        I didn't have any problems with program execution no matter what the
                        clock speed.
                        --
                        "Never doubt that the work of a small group of thoughtful, committed
                        citizens can change the world. Indeed, it's the only thing that ever
                        has!" -- Margaret Mead
                        --
                        Lee A. Hart 814 8th Ave N Sartell MN 56377 leeahart_at_earthlink.net
                      • Lee Hart
                        ... (You just had to ask, didn t you? :-) It was a name tag, with nothing but the 1802 and that phototransistor as the only active devices. All it did was
                        Message 11 of 23 , Jun 9, 2004
                        • 0 Attachment
                          > The silliest 1802 clock I've used is a phototransistor configured with
                          > the internal inverter as an RC oscillator whose frequency depends on
                          > light intensity. In the dark it got so slow that it eventually stopped.

                          Steve Ross wrote:
                          > Okay, I've got to ask: Why? What was the application?

                          (You just had to ask, didn't you? :-)

                          It was a name tag, with nothing but the 1802 and that phototransistor as
                          the only active devices. All it did was blink a bunch of LEDs connected
                          between the various address bus pins. The "program" was just to wire the
                          data bus to the address bus, with a couple pins tied high to avoid it
                          ever hitting an IDL instruction.
                          --
                          "Never doubt that the work of a small group of thoughtful, committed
                          citizens can change the world. Indeed, it's the only thing that ever
                          has!" -- Margaret Mead
                          --
                          Lee A. Hart 814 8th Ave N Sartell MN 56377 leeahart_at_earthlink.net
                        Your message has been successfully submitted and would be delivered to recipients shortly.