TI Silent 700 print mechanism mechanical issues, and Model 763/765 bubble memory notes

[Eric Smith]

I've now got a TI Silent 700 Model 763, and it is partially working.
The previous owner said that the line advance wasn't working reliably,
but that otherwise it was working. As I received it, the thermal print
head is only darkening the top scan line of characters, and only
partially. The intensity pot is already at maximum. The line advance
is working, but carriage return is unreliable, which results in the
carriage banging against the right side of the mechanism and getting
stuck there. The horizontal carriage positioning is done by stepper
motor with an encoder for feedback, so I'm surprised that it isn't
smart enough to recognize that it can't home the carriage, and avoid
ramming it into the right end of the mechanism.

I was worried that the thermal print head might be damaged. I opened
it up and found that it uses the solenoid line advance mechansim,
probably the same as is documented for the early production of Model
743/745. (Later 743/745 use a stepper motor for line advance.) With
power off, the carriage can be manually returned to the left side.

There were two loose parts inside, a spring and a small knob. These
apparently belong to the solenoid line advance mechanism, belonging on
the shaft at the opposite end of the solenoid from the link to the
pivot.

There are screws on the left and right end of the mechanism, which
when loosened, allow the printhead to be moved up and down relative to
the mechanism. The Model 743/745 maintenance manual suggests adjusting
that if characters aren't fully visible. I tried that, and could see a
little bit more of the characters, but couldn't adjust it to get the
characters entirely visible.

I accidentally discovered that with a small amount of additional
pressure against the carriage assembly, toward the paper path, the
characters are printed fully formed, and (quite suprisingly to me) the
carriage return works properly as well.

If I can't get this working reliably, I may start searching for a 743
or 745 with the solenoid line advance mechanism, to try a mechanism
transplant.  (The mechanism with stepper motor line advance is not
interchangeable, and requires a different circuit board. I'm not sure
whether that mechanism was used in any model 763/765 units.)

The Operating Instructions manual for the Model 763 and 765 gives
information on 13 different commands which are accepted in command
mode. They mention the TEST command, which performs a self-test, and
the TEST INIT subcommand which resets everything to factory defaults.

There's a TEST MASK subcommand not documented in the manual which
allows examining or altering the bubble memory minor loop masks. When
installing or replacing a bubble memory module, it may be necessary to
enter this information for the new module, from the mask data printed
on the label of the bubble device.

Unlike the later Intel bubble memory, the TI parts (at least of the
92Kbit devices) don't have a specified dedicated "boot loop" to store
the mask, nor do they have a defined synchronization pattern to
provide a detectable home position, so the TMS5502/TMS9916 bubble
controller chip has to ensure that the device is rotated to the home
position on power down or power fail. The terminal firmware may be
using a specific normal minor loop, probably the first or last loop,
to store the mask and a sync pattern, but the TMS5502/TMS9916 don't
provide any automation for that.  The TMS5502/TMS9916 also require
that the mask bits be provided in a bit-serial fashion at the precise
times needed during data transfers; TI app notes show the mask stored
in PROM with a counter for addressing, but the 763/765 terminal
doesn't do that. I'm not sure whether they have dedicated logic for
the mask, or whether it's being done by firmware.

The TEST MASK subcommand is probably documented in the Maintenance
Manual, which I don't have; the only reason I know about it is that a
technician left a printout showing the TEST MASK output for this
terminal between two pages of the Operating Instructions manual.

The terminal uses a TMS9980 microprocessor, which is an 8-bit-bus
version of the TMS9900, which can only address 16KB of memory. The
terminal has five TMS4732 4KB ROMS, as well as some RAM, so there must
be some bank-switching going on.

The bubble memory modules for the terminal came in two types:

1) A large "discrete" bubble module with one 92Kbit bubble device and
a whole lot of non-bubble-specific chips (presumably because the
SN753xx bubble memory support chips were not yet available for
production). Two large modules are fitted, to provide the terminal's
basic rated storage capacity of 20K characters (actually up to 22,860
characters if 18, 36, or 72 character record lengths are used). This
leaves no room for additional bubble modules. The modules show up to
the TEST MASK command as modules 2 and 6.

2) A small bubble module with two 92Kbit bubble devices and the
SN753xx bubble memory support chips. A single module provides the 20K
(22860 character) basic rated storage. Up to four of these modules can
be fitted, for up to 80K (91,440) characters.

My terminal had two large bubble modules installed. A friend gave me
three more large modules, which have been treated roughly, so they
have bent pins and possibly damaged components. I have one small
module that came from eBay some years ago. I haven't tried any of them
yet.