Early
Electronic
Calculator
Monroe 925 Calculator

Manufacturer: Monroe
Model: 925
OEM: Canon
Year: 1970
Form: Desktop
Functions: Basic four, accumulator, constant
Number of Digits: 13
Display Type: NIX
Display:JRC B-5755
Logic Technology: SSI
Memory Technology: DELAY-MS
Diodes: 153
Transistors: 42
Principal ICs: SN3900/4500 Series
IC Complement:Texas Instruments SN3920 (7)
Texas Instruments SN3925 (26)
Texas Instruments SN3931 (11)
Texas Instruments SN4553 (4)
Texas Instruments SN4554 (21)
SN7400 (1)
(70 total)
Tech. Data Level: 2
Tech. Data Source: RE
Tech. Data Pages: 14
Tech. Data:Schematic (pdf)

This model is actually OEM'd by Canon. It shares some implementation aspects with the Canon 163, such as the IC-series used, the use of a magnetostrictive acoustic delay-line for the register memory, and two write transducers on the delay-line.

The 163 uses an interesting register arrangement in the delay line with a short loop for one register, a long loop for all the other registers, and empty bit slots to allow interleaving the registers. This model takes the complexity a step further. A bit-slippage is introduced in the long loop so there is a 'rotation within the rotation'. The resultant variation in the bit interleave allows register selection to be done in time as with the 163, but eliminates almost all of the wasted empty bit-slots of the 163, thus shortening the required length of the delay line. (See schematics for more detailed explanation).

Notes:


Case opened.

Keyboard from below. The key-slides with magnets attached, and magnetic-reed switches, are mounted in a frame of high-density plastic.

Power supply.

Upper board, component side.
Notice the crystal in the upper-right corner of the photo. Generally calculators didn't require the timebase precision provided by crystal oscillators and instead got away with simple RC astable flip-flops for the master clock. However for calculators using acoustic delay-line memories, master clock precision was needed to keep the clock rate synchronous with the delay-line pulse velocity.

Another method of maintaining sync was to use a bit pattern in a channel-slice of the delay-line to set up a phase-locked-loop that tweaks the master clock rate with drift in the delay-line.


Upper board, solder side.

Upper board from the rear.
The pulse transformers are for high voltage isolation of the Nixie anode drivers.

Lower board, component side.

Lower board, solder side.

The magnetostrictive delay-line memory.

The delay-line from below.
Opening the delay-line requires drilling out the brass rivet flanges on the 4 mounting spacers.

Inside the delay-line.
This is a torsion-mode delay-line. The transducer solenoids are in the blue housings. The flat metal tapes going through the solenoids in those housings are the actual magnetostrictive material. The delay-line proper is the coil of wire mounted in small rubber bushings for acoustic isolation from the frame.

The tapes are spot-welded in tangent to the circumference of the delay-line wire. The magnetostrictive contraction and expansion of the tapes, by the solenoids, thus torques the delay-line wire.

The read transducer is on the right. The first write transducer is on the left. The transducer in the middle is the second write transducer, used to achieve the 'register slippage'.

The coil is 6 loops of ~ 91mm diameter. With the ends and the tapes, it's length comes to:
6*3.14*91 + 2*65mm + 30mm + 20mm = 1895mm.
The delay is 635µS, thus a pulse velocity of ~ 10,740 Km/h.



- Unit Log -

Serial Number: E017715
Year of Manufacture: 1970 (Acoustic Delay Line unit stamped with 08 Oct 1970)
Date of Receipt: 1970s
Source: Sears Service Department.
State upon Receipt: Good physical condition.
Current State: Fully functional (Dec 1997).


Date: Dec 1997
Analysis: Grounding of IC112 pin 7 intermittent due to poor solder on associated feedthru connecting the two sides of the board.
Solution: Resoldered.


Date: 31 Jan 2005
Symptoms: No display. Overflow lamp will light on powerup sometimes and can be cleared by C key but nixies are dark.
Analysis: Clock signals disappear at slice counter. Feedthru stub connecting 110-6 to trigger caps for 110-10,13 have bad solder connection on solder side.
Solution: Resoldered.


Date: 04 Feb 2005
Symptoms: Arithmetic operations intermittently produce incorrect results. Key sequence "1+" results in display of "1.88".
Analysis: When "1" is pressed, equivalent bits show up in YABC registers. Speculate that M pulses for X are being injected into delay line at wrong slice times, perhaps due to problem around diode gate feeding W1 during slices 123. Scope indicates intermittent problem with not-phase-S4 feeding diode gate. Confirmed on slice level simulation: forcing input TRUE (open) produces identical results.
Solution: Probably bad feedthru, not confirmed yet.



  Monroe 925
Calculators | Integrated Circuits | Displays | Simulations
EEC
bhilpert