[GreenKeys] Early early Model 28 catalog - many prototypes

[Jim Haynes]

On Thu, 3 Dec 2020, Keelan Lightfoot wrote:

> 
> stuff that an engineer would lay in bed until 3AM dreaming about. The one
> thing I always found disappointing is that they couldn't come up with a way
> to make the enclosures modular too, so that a KSR could be upgraded to an
> ASR by just bolting together the appropriate cabinets. I guess the sheet

See IBM 1050 for a terminal with that kind of concept.  One of the 
managers at Teletype asked me about doing something similar with our
product line.  I gave it some thought, but with the way that Model 32/33
worked electrically I couldn't see any neat way of doing it.  Many
years later I remembered that I had the answer several years earlier
when I was working on the low-voltage selector magnet driver.

The answer I had at that early time was that every selector magnet should
have a low voltage driver and the input to the driver should be something
like +/- 6 volts rather than a current loop.  And every signal generator
should produce a similar voltage.  Build the interior of a set that way
and you don't have the difficulty of inserting selector magnets into 
loops, and keeping current through them so they don't run open when they
are not in use.  change everything to keyed voltages rather than currents,
and use the current loop only when required to interface to the external
signal line.  But I failed to sell that concept at the time.  I was just
a summer student technician.

And there still would have been a small problem in cases where you had
parallel data to input, or wanted to look at parallel data.  Such as
the LBXD can do with its separate reader and distributor shafts.

I don't think the IBM 1050 was a highly successful product - probably was
too costly.  The 2741 was much more successful as a product, I think,
and it was simply a KSR with no options.  In fact you could turn a switch
and use it as an office typewriter.

Many years later, the Model 40 generation, Teletype came up with what
was called the Standard Serial Interface, and that made it possible to
connect things together in a more modular way.  But it was awfully
complicated, used a high-speed serial data word of something like 18 bits
per character.

Another factor in the whole situation is you have to remember Teletype
had a strong culture of designing and manufacturing complicated
mechanical stuff.  Making things modular is a lot easier if you
use electronics.  The IBM 1050 had a central electronics box that
was required regardless of which other components were to be used.
For all I know they may have done everything bit parallel outside
that box.  Making things modular is a lot more difficult when you
have to deal with motors and shafts and gears and connecting things
together with mechanical linkages.

I often point to the Model 37 selector as an illustration of the
power of mechanical fixation at Teletype.  In 1925 the single magnet
selector (as used in Model 14 and 15) was a matter of great pride -
at last they had got rid of the electromechanical selector and the
individual electromagnets per bit and all that troublesome wiring.
But then when they got to 110 baud the mechanical selector was a
bit dicey, and for 150 baud it was Really Hard.  They could have
made it electrically parallel input with an electronic selector and
saved a lot of design time.  But those mechanical designers were
determined to do it their way.  And at the time a handful of mechanical
parts made on a punch press seemed vastly cheaper than a PC board with
transistors and a power supply.

Another illustration is the Model 32/33 keyboard.  I was told once that
it cost something like $50 to manufacture.  If someone had told us that
in the future we would be able to buy keyboards with 104 keys for $10
at Walmart, how could we have believed it?