Group Actions
Engagement
  • Author Author: COMPACT
  • Date Created: Date Created
  • Views 1400 views
  • Likes 10 likes
  • Comments 6 comments
Related
Recommended

Holiday Special

COMPACT

I'm playing with my S100 Bus computers which I built over 40 years ago.

They still work!!

They played a major role with making computers mainstream.

Parents
  • in reply to DAB

    Excellent stuff.  Have a look at them and laugh at their PCB design.

     

    S100 boards are inherently of poor design because at the time EMC and Signal integrity were not considerations due to false assumptions.

    It was when Steve Dompier used an AM radio for sound effects for his "Target" game and nearby TVs were getting poor reception did the FCC step in and formulate the necessary safety regulations.

    Signal Integrity was addressed as a afterthought through the use of active termination boards whose technology was then inherited by the FDD, HDD and SCSI busses.

    Even with termination in place, many S100 boards had their buffers and transceivers too far way from the card edge connectors creating stubs that caused havoc to signal integrity.

     

    The PDN (Power distribution networks) were woeful with ground bounce being generated everywhere.  This is where the concept of installing zillions of so-called decoupling or bypass capacitors originated as an attempt to alleviate it.

    (COMPACT always refers to these caps more as Coulomb buckets where ICs grab their necessary power.)

    RAM cards are a great example of this with a Coulomb bucket installed for every RAM IC.

     

    The CPU's 8 bit bidirectional data bus was split into 2 singularly directional 8 bit data in and data out busses. This was to make front panel manipulation of I/O and memory easier.

    e.g. Using toggle switches and LEDs to read and manipulate data values in RAM.

    It was amazing that Steve Allen was able to input Microsoft BASIC into an S100 machine to demonstrate it.

    Front panels were were phased out when other forms of I/O such as serial terminals and video cards replaced them.

Comment
  • in reply to DAB

    Excellent stuff.  Have a look at them and laugh at their PCB design.

     

    S100 boards are inherently of poor design because at the time EMC and Signal integrity were not considerations due to false assumptions.

    It was when Steve Dompier used an AM radio for sound effects for his "Target" game and nearby TVs were getting poor reception did the FCC step in and formulate the necessary safety regulations.

    Signal Integrity was addressed as a afterthought through the use of active termination boards whose technology was then inherited by the FDD, HDD and SCSI busses.

    Even with termination in place, many S100 boards had their buffers and transceivers too far way from the card edge connectors creating stubs that caused havoc to signal integrity.

     

    The PDN (Power distribution networks) were woeful with ground bounce being generated everywhere.  This is where the concept of installing zillions of so-called decoupling or bypass capacitors originated as an attempt to alleviate it.

    (COMPACT always refers to these caps more as Coulomb buckets where ICs grab their necessary power.)

    RAM cards are a great example of this with a Coulomb bucket installed for every RAM IC.

     

    The CPU's 8 bit bidirectional data bus was split into 2 singularly directional 8 bit data in and data out busses. This was to make front panel manipulation of I/O and memory easier.

    e.g. Using toggle switches and LEDs to read and manipulate data values in RAM.

    It was amazing that Steve Allen was able to input Microsoft BASIC into an S100 machine to demonstrate it.

    Front panels were were phased out when other forms of I/O such as serial terminals and video cards replaced them.

Children
  • in reply to COMPACT

    For its day, it was fine.

    We did a lot of prototype work with them before committing to full PCB.

     

    DAB

  • in reply to DAB

    I did a lot of prototyping but I was aware of the caveats.

    When prototyping things are done on a small scale maybe 1, 2 or 3 boards or so.

    Back then 3 level wire wrap prototyping sockets were often used.

    When prototyping these boards are often placed close together and on small motherboards mitigating possible signal degradation.

     

    The problems usually start appearing on 11 slot systems and are nearly always present on 22 slot backplanes when the signal traces became substantially longer.

    Yep back then we had whole 5 x 10 inch cards providing a whopping 4K of RAM with each one consuming about 20 Watts or so.

    The good old 2101 1k x 1 static RAM reigned supreme.

     

    Wirewrapping DRAM was never a good idea because of the perpetual activity on the multiplexed address lines from the relentless refresh cycles.

    For this you'll often see 33 ohm series terminator resistors on the address lines. (A first sign of some signal integrity control.)

     

    Even without an oscilloscope it was easy to spot the signal degradation just by observing the behaviour of the data lines when observing unpopulated decoded memory space.

    What should be reading a steady and stable value would be observed with flickering values.

    These flickering values could be fixed by applying the appropriate pull up, pull down or termination resistors.

     

    A false premise was the assumption that low clock frequencies didn't require didn't need signal integrity measures when in reality it was the slew rates that there the critical factor.

    You can completely stuff up a perfectly working LSTTL design by substitution of a single part from a faster family such as an S, ALS or F TTL part.

     

    As Eric Bogatin nicely put it @22:44. "Many designs work despite these habits and not because of them"

     

    .