Integrated circuits:

• Hitachi HD 3103 (7x): quintuple P-channel MOSFET transistors
• Hitachi HD 3104 (2x): dual 4-in OR gates, plus dual inverters
• Hitachi HD 3106 (6x): dual 2-in OR gates and dual 3-in OR gates
• Hitachi HD 3112 (1x): serial BCD adder, performing serial addition, bit carry, tens carry, correction for digit sums greater than 9, negation, etc. Presumably it contains two serial adders, a 4-bit shift register and additional logic for negation and ten's carry.
• Hitachi HD 3117 (1x): dual shift register (4 and 44 bits per register). Serial input / output

• NEC uPD 101C (2x): quadruple 2-in OR gates
• NEC uPD 116C (1x): BCD to decimal decoder (serial BCD-input)

• Toshiba TM 4351P (8x): quadruple D-type flipflop + 1 inverter
• Toshiba T1191 (1x): keyboard decimal to BCD-encoder (parallel BCD-output)
• Toshiba T4097 (3x): nixie tube driver device

• SSR0109 / H7444A (1x) : dual 48-bit shift register. Serial input / output (manufacturer unknown)

• black: GND
• grey: -24 V
• yellow: +180 V
• brown: +90 V

The Casio model 121-B dates from the late 1960s. The one I got piece is probably built in 1970, which is derived from the date codes that are printed on the integrated circuits (00). The calculator provides the basic four arithmetic functions and has a separate shift register for the total memory. There is no decimal point key on the keyboard. Negative numbers can neither be entered.

All components of the calculator are situated on two separate pcb’s which are mounted parallel to each other and are connected by a number of short soldered wires. The keyboard and the power supply unit, which is located on a small pcb, are each connected to the upper pcb by their own connector.

The upper pcb contains the display driver and the keyboard encoder. Also the clock is found here. The shift registers and the serial adder are situated on the lower pcb.

The display consists of 12 small nixie tubes (Hitachi CD-71).

The calculator uses negative logic. A logic "1" is represented by -24 volt, while 0 volt represents a logic "0".

The clock pulse generator consists of a multivibrator which runs at a frequency of around 100 kHz (PG). Clock pulses are distracted from this multivibrator and are called Ø1 and Ø2. These are 180 degrees shifted with respect to each other and have a width of about 5 microseconds (us) at a period of 20 us. The pulses that are fed to the anodes of the 12 nixie tubes, are generated by a circuit consisting of 13 D-type flipflops. These pulses have a width of about 80 us. Although there are 12 nixie tubes, a total amount of 13 pulses is generated. During the 13^th pulse, none of the nixie tubes lights up, but BCD-coded numerals can enter the shift register (HD3117). The contents of the HD3117 shift register is displayed by the nixies. After pressing a function key, the contents of the HD3117 is copied to a second shift register (SSR0109 / H7444A). When the first numeral of the second operand is entered, the contents of the HD3117 is cleared while at the same time the first numeral of the second operand is stored in this ic. After pressing the += or = - key, the result of the calculation is generated by the HD3112 on pin 15 (SUM out) and fed to the HD3117 shift register and displayed. As the HD3117 is a 48 bit register, the remaining 4 bits are found in the HD3112, giving a total of 52 bits, which is equal to 13 (pulses) of 4 bits (BCD). Multiplications and divisions are basically carried out in the same way, although pin 9 (SUB in) receives a certain pulse train during these operations, probably in order to let the HD3112 perform a number of additions and subtractions successively. Pin 9 is not used for simple additions.

The calculator was bought on eBay and shipped to Europe. It was sold as not functional (item description: powers up but doesn’t work). The calculator was tested with a mains transformer (220 to 117 Volt) and did indeed not react to any keypress. All nixies were displaying a zero, and all decimal points were also visible. Dismantling and cleaning the machine and its connectors didn’t change this state of non-functionality.

A search on the www brought me the contents of some of the integrated circuits that were situated in the machine (HD3104, HD3106, HD3112, uPD101C, uPD116C and T1191). After having read the way Brent Hilpert solved the problems on his calculators, I decided to reverse engineer the schematic of the 121-B. By completing the schematic, I also discovered the contents of the remaining ic’s (TM4351, HD3103, HD3117, and a small metal can with ten connections with the legend SSR0109 / H7444A — no manufacturer). Although it has been quite lot of work, it was really fun to discover the architecture of the machine!

I discovered that the HD3117 was not functional. After having replaced this ic (I was lucky to find some pieces at a German firm), the machine was able to display the numbers that were entered by the keyboard. The second problem were the decimal points that were visible all together at the same time. I found out that this was due to a bad connection on the lower pcb. Applying heat to some of the printed wires resolved this problem. By now, the calculator was able to perform additions and subtractions, but stuck in the middle of multiplications and divisions. This was also due to a bad connection of one of the pins of a HD3103. Therefore, the SUB-D pin (#9) of the HD3112 did not get any signal pulses. Apparently this does not matter for additions and subtractions. After having resoldered the connections, the calculator works perfectly well. Because the power transformer has a number of (unused) connections for different input voltages, I was able to adapt the calculator to the European mains voltage (220V).