Printable View
Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
buffers instead of bidirectional? Seems wasteful as every board
I've seen only converts them back to bidir.
I've been trying to figure out possible reasons for days now, and
found no explanations in Bursky's or Garetz's books. My guess is
there's either no reason, or something to do with front panel.
~ J
MdntTrain wrote:[color=blue]
> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
> buffers instead of bidirectional? Seems wasteful as every board
> I've seen only converts them back to bidir.[/color]
My understanding is that MITS got a good deal on 100 pin connectors.
My guess, then, is that they tried to find a use for those pins.
Much of S-100 seems to be an accident.
Note also that man y RAM chips at the time had separate data
in and out pins.
-- glen
glen herrmannsfeldt wrote:[color=blue]
> MdntTrain wrote:[color=green]
>> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>> buffers instead of bidirectional? Seems wasteful as every board
>> I've seen only converts them back to bidir.[/color]
>
> My understanding is that MITS got a good deal on 100 pin connectors.
>
> My guess, then, is that they tried to find a use for those pins.
> Much of S-100 seems to be an accident.
>
> Note also that man y RAM chips at the time had separate data
> in and out pins.[/color]
A couple of the minicomputer busses of the time also had separate data
lines (DEC Unibus comes to mind).
Bob
On 2007-08-30, BobH <WanderingMetalHead@DUMP.SPAM.yahoo.com> wrote:[color=blue]
> glen herrmannsfeldt wrote:[color=green]
>> Note also that man y RAM chips at the time had separate data
>> in and out pins.[/color]
>
> A couple of the minicomputer busses of the time also had separate data
> lines (DEC Unibus comes to mind).[/color]
UNIBUS did not have separate in and out data lines. Neither did
OMNIBUS, for that matter, although it did have separate data busses for
memory and peripherals.
--
roger ivie
[email]rivie@ridgenet.net[/email]
On Wed, 29 Aug 2007 13:08:52 -0700, MdntTrain <js@cimmeri.com> wrote:
[color=blue]
>
>Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>buffers instead of bidirectional? Seems wasteful as every board
>I've seen only converts them back to bidir.
>
>I've been trying to figure out possible reasons for days now, and
>found no explanations in Bursky's or Garetz's books. My guess is
>there's either no reason, or something to do with front panel.
>
>~ J[/color]
Made it easier to do simple output or input devices also simplified
the data hold time for read or write. The origin was likely the
basic MDS8 and later the 8080 version that used a bus that was
mostly CPU signals buffered minimally. It was a side effect of
both CPUs having control signals and the like multiplexed on the
CPU data bus. You need to buffer the CPU databus for output data
as it has two paths, one is a latch to catch the CPU state and cotrol
signals and the other is DATA out to the bus on writes.
The front pannel logic was not really simplified any by it though it
didn't hurt.
The amount of waste is similar to bidirectional busses you have
to buffer them and it takes about the same number of parts.
The down side was there were designs that didn't buffer the input
making for noise and capacitive loading problems (a bad thing).
Usually it was early 2102 ram cards that did that but there were a few
IO that weren't pretty.
Mostly that was how people did it then. Later when the 8228/8238
appeared and the few 8080 designs that used it tended to have
bidirectional data busses... I suspect so intel could sell the 8216
(4bit bidirectional buffer). ;)
Allison
On Aug 30, 7:29 am, no.s...@no.uce.bellatlantic.net wrote:[color=blue]
> On Wed, 29 Aug 2007 13:08:52 -0700, MdntTrain <j...@cimmeri.com> wrote:
>[color=green]
> >Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
> >buffers instead of bidirectional? Seems wasteful as every board
> >I've seen only converts them back to bidir.[/color]
>[color=green]
> >I've been trying to figure out possible reasons for days now, and
> >found no explanations in Bursky's or Garetz's books. My guess is
> >there's either no reason, or something to do with front panel.[/color]
>[color=green]
> >~ J[/color]
>
> Made it easier to do simple output or input devices also simplified
> the data hold time for read or write. The origin was likely the
> basic MDS8 and later the 8080 version that used a bus that was
> mostly CPU signals buffered minimally. It was a side effect of
> both CPUs having control signals and the like multiplexed on the
> CPU data bus. You need to buffer the CPU databus for output data
> as it has two paths, one is a latch to catch the CPU state and cotrol
> signals and the other is DATA out to the bus on writes.
>
> The front pannel logic was not really simplified any by it though it
> didn't hurt.
>
> The amount of waste is similar to bidirectional busses you have
> to buffer them and it takes about the same number of parts.
>
> The down side was there were designs that didn't buffer the input
> making for noise and capacitive loading problems (a bad thing).
> Usually it was early 2102 ram cards that did that but there were a few
> IO that weren't pretty.
>
> Mostly that was how people did it then. Later when the8228/8238
> appeared and the few 8080 designs that used it tended to have
> bidirectional data busses... I suspect so intel could sell the 8216
> (4bit bidirectional buffer). ;)
>
> Allison[/color]
I tend to agree with this assessment, and there were some cards like
the Northstar MDS disk controller that really overloaded the use of
the data in bus, by using "reads" to actually write to the hardware by
using the lower 8 bits of the address as the write data. Of course
this was st the expense of wasting precious space in the address map,
but it did simplify the hardware, and cut costs. I guess this was a
reasonable approach when 64k was a lot of memory.
-Howard
On Thu, 30 Aug 2007 13:21:20 -0700, "hharte@hartetec.com"
<hharte@hartetec.com> wrote:
[color=blue]
>On Aug 30, 7:29 am, no.s...@no.uce.bellatlantic.net wrote:[color=green]
>> On Wed, 29 Aug 2007 13:08:52 -0700, MdntTrain <j...@cimmeri.com> wrote:
>>[color=darkred]
>> >Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>> >buffers instead of bidirectional? Seems wasteful as every board
>> >I've seen only converts them back to bidir.[/color]
>>[color=darkred]
>> >I've been trying to figure out possible reasons for days now, and
>> >found no explanations in Bursky's or Garetz's books. My guess is
>> >there's either no reason, or something to do with front panel.[/color]
>>[color=darkred]
>> >~ J[/color]
>>
>> Made it easier to do simple output or input devices also simplified
>> the data hold time for read or write. The origin was likely the
>> basic MDS8 and later the 8080 version that used a bus that was
>> mostly CPU signals buffered minimally. It was a side effect of
>> both CPUs having control signals and the like multiplexed on the
>> CPU data bus. You need to buffer the CPU databus for output data
>> as it has two paths, one is a latch to catch the CPU state and cotrol
>> signals and the other is DATA out to the bus on writes.
>>
>> The front pannel logic was not really simplified any by it though it
>> didn't hurt.
>>
>> The amount of waste is similar to bidirectional busses you have
>> to buffer them and it takes about the same number of parts.
>>
>> The down side was there were designs that didn't buffer the input
>> making for noise and capacitive loading problems (a bad thing).
>> Usually it was early 2102 ram cards that did that but there were a few
>> IO that weren't pretty.
>>
>> Mostly that was how people did it then. Later when the8228/8238
>> appeared and the few 8080 designs that used it tended to have
>> bidirectional data busses... I suspect so intel could sell the 8216
>> (4bit bidirectional buffer). ;)
>>
>> Allison[/color]
>
>I tend to agree with this assessment, and there were some cards like
>the Northstar MDS disk controller that really overloaded the use of
>the data in bus, by using "reads" to actually write to the hardware by
>using the lower 8 bits of the address as the write data. Of course
>this was st the expense of wasting precious space in the address map,
>but it did simplify the hardware, and cut costs. I guess this was a
>reasonable approach when 64k was a lot of memory.[/color]
Yes, but it was properly buffered and interfaced to the bus.
The fact that it chewed up E800h though EFFFh is only 2k and
in the early years most people would have killed for even
56k of ram. Also that 2k included the Boot roms. It was resonable
when SA400s were still $399 each and a controller was another $500.
FYI: it was possible to relocate the MDS to F800h ( I have a set of
proms around that did that.). It was very easy to do. Also NS* for a
price would deliever a custom prom set and matching DOS for any
address combination.
When I refered to data bus loading I was talking in DC and AC terms
as in 74367s could only drive so much and 8 ram card amounting
to 64 rams per data bit line (64k of 8kx8 2102s) if they were not
buffered but a huge AC load on the bus. Some of the early ram cards
were not so good or at least tolerable if you only had one or two.
Allison[color=blue]
>-Howard[/color]
On Aug 29, 5:27 pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote
On Aug 29, 7:32 pm, BobH <WanderingMetalH...@DUMP.SPAM.yahoo.com>
wrote
On Aug 29, 8:56 pm, Roger Ivie <ri...@ridgenet.net> wrote
On Aug 30, 3:36 pm, no.s...@no.uce.bellatlantic.net wrote
On Aug 30, 3:21 pm, "hha...@hartetec.com" <hha...@hartetec.com> wrote
-------------------------------------
Thank you all for your cool and informative replies!
~ J
"MdntTrain" <js@cimmeri.com> wrote in message
news:1188418132.136470.243670@50g2000hsm.googlegroups.com...[color=blue]
>
> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
> buffers instead of bidirectional? Seems wasteful as every board
> I've seen only converts them back to bidir.
>
> I've been trying to figure out possible reasons for days now, and
> found no explanations in Bursky's or Garetz's books. My guess is
> there's either no reason, or something to do with front panel.
>
> ~ J[/color]
I may be stepping into a pile of something stinky, but it seems it was as
logical "back in the day" as it is now to use the manufacturer's application
notes and example architecture to implement new technologies, and I believe
such was the case with almost all early implementers of the Intel
microprocessors. I'll have to go back to my early data books to confirm,
but I seem to recall that the Intel MCS system and variants laid the
groundwork that was loosely followed by the esteemed and enshrined Mr.'s
Titus and Roberts and numerous others who chose the 8080 platform to
exploit.
On Thu, 13 Sep 2007 12:05:37 -0700, "Thomas \"Todd\" Fischer"
<trf@imsai.net> wrote:
[color=blue]
>
>"MdntTrain" <js@cimmeri.com> wrote in message
>news:1188418132.136470.243670@50g2000hsm.googlegroups.com...[color=green]
>>
>> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>> buffers instead of bidirectional? Seems wasteful as every board
>> I've seen only converts them back to bidir.
>>
>> I've been trying to figure out possible reasons for days now, and
>> found no explanations in Bursky's or Garetz's books. My guess is
>> there's either no reason, or something to do with front panel.
>>
>> ~ J[/color]
>I may be stepping into a pile of something stinky, but it seems it was as
>logical "back in the day" as it is now to use the manufacturer's application
>notes and example architecture to implement new technologies, and I believe
>such was the case with almost all early implementers of the Intel
>microprocessors. I'll have to go back to my early data books to confirm,
>but I seem to recall that the Intel MCS system and variants laid the
>groundwork that was loosely followed by the esteemed and enshrined Mr.'s
>Titus and Roberts and numerous others who chose the 8080 platform to
>exploit.
>[/color]
You are correct! One only has to look at the MCS80 manual to see
that.
Allison
<no.spam@no.uce.bellatlantic.net> wrote in message
news:96kje39dtqsvg8brbamqc5u34lj4n3r60c@4ax.com...[color=blue]
> On Thu, 13 Sep 2007 12:05:37 -0700, "Thomas \"Todd\" Fischer"
> <trf@imsai.net> wrote:
>[color=green]
>>
>>"MdntTrain" <js@cimmeri.com> wrote in message
>>news:1188418132.136470.243670@50g2000hsm.googlegroups.com...[color=darkred]
>>>
>>> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>>> buffers instead of bidirectional? Seems wasteful as every board
>>> I've seen only converts them back to bidir.
>>>
>>> I've been trying to figure out possible reasons for days now, and
>>> found no explanations in Bursky's or Garetz's books. My guess is
>>> there's either no reason, or something to do with front panel.
>>>
>>> ~ J[/color]
>>I may be stepping into a pile of something stinky, but it seems it was as
>>logical "back in the day" as it is now to use the manufacturer's
>>application
>>notes and example architecture to implement new technologies, and I
>>believe
>>such was the case with almost all early implementers of the Intel
>>microprocessors. I'll have to go back to my early data books to confirm,
>>but I seem to recall that the Intel MCS system and variants laid the
>>groundwork that was loosely followed by the esteemed and enshrined Mr.'s
>>Titus and Roberts and numerous others who chose the 8080 platform to
>>exploit.
>>[/color]
>
> You are correct! One only has to look at the MCS80 manual to see
> that.
>
> Allison[/color]
Correct that I stepped into something stinky, or that adoption of new
architecture generally followed preliminary guidelines?
As always, Allison, your insight and citations are a treasure!
On Thu, 13 Sep 2007 17:31:18 -0700, "Thomas \"Todd\" Fischer"
<trf@imsai.net> wrote:
[color=blue]
>
><no.spam@no.uce.bellatlantic.net> wrote in message
>news:96kje39dtqsvg8brbamqc5u34lj4n3r60c@4ax.com...[color=green]
>> On Thu, 13 Sep 2007 12:05:37 -0700, "Thomas \"Todd\" Fischer"
>> <trf@imsai.net> wrote:
>>[color=darkred]
>>>
>>>"MdntTrain" <js@cimmeri.com> wrote in message
>>>news:1188418132.136470.243670@50g2000hsm.googlegroups.com...
>>>>
>>>> Does anyone know why the S-100 bus used 2 x 8 bit unidirectional
>>>> buffers instead of bidirectional? Seems wasteful as every board
>>>> I've seen only converts them back to bidir.
>>>>
>>>> I've been trying to figure out possible reasons for days now, and
>>>> found no explanations in Bursky's or Garetz's books. My guess is
>>>> there's either no reason, or something to do with front panel.
>>>>
>>>> ~ J
>>>I may be stepping into a pile of something stinky, but it seems it was as
>>>logical "back in the day" as it is now to use the manufacturer's
>>>application
>>>notes and example architecture to implement new technologies, and I
>>>believe
>>>such was the case with almost all early implementers of the Intel
>>>microprocessors. I'll have to go back to my early data books to confirm,
>>>but I seem to recall that the Intel MCS system and variants laid the
>>>groundwork that was loosely followed by the esteemed and enshrined Mr.'s
>>>Titus and Roberts and numerous others who chose the 8080 platform to
>>>exploit.
>>>[/color]
>>
>> You are correct! One only has to look at the MCS80 manual to see
>> that.
>>
>> Allison[/color]
>
>Correct that I stepped into something stinky, or that adoption of new
>architecture generally followed preliminary guidelines?
>
>As always, Allison, your insight and citations are a treasure!
>[/color]
I was among the lucky when the 8008 hit and again when the 8080 hit
So being there and paying attention I saw. Didn't hurt to save all
preserve my docs.
Intel drove the unified bus (bidirectional) with cleaner controls when
the Bipolar part 8228/8238 became available. Those ideas were
encapsulated in multibus and later the very similar PC xt.
Allison