Victor Series 1400 Model 321 Electronic Calculator

The Victor 1400 series of calculators (which included two machines, the exhibited machine, and the slightly more-capable 14-322) were Victor's second venture into the electronic calculator marketplace. Victor's first attempt, the Victor 3900, introduced in October of 1965, while a technological tour-de-force, had difficulties in the marketplace because it stretched the state of the art just a bit too far and suffered reliability problems as a result. The 3900's difficulties somewhat hurt Victor's market reputation as far as electronic calculators went for a time during the heyday of the emergence of electronic calculators. Along with this, Victor's management was concerned over the huge investment it had made in the Victor 3900's development, only to have it deliver disappointing results. The story behind the development of Victor's entry into the electronic calculator industry is quite interesting, and is told in in the detailed essay entitled "The Victor 3900 - History's Forgotten Miracle".

After Victor's problems with its first electronic calculator, the company waited nearly two years before venturing back into the electronic calculator market on its own. During the interim, it marketed electronic calculators made by other manufacturers (Nixdorf/Wanderer Werke in W. Germany was one) under the Victor brand name through OEM relationships with the manufacturers. This provided Victor a less-risky presence in the electronic calculator marketplace while management was working to decide whether to try again at developing its own electronic calculators, or continue marketing other manufacturer's machines under the Victor brand name. Given that all mechanical calculator manufacturers were seeing sagging sales due to the emergence of electronic calculators, either Victor had to strengthen its OEM relationships and market a wider range of electronic calculators made by other manufacturers, or it had to get busy and come up with its own electronic calculators that would be competitive and compelling to customers, as well as delivering Victor's reputation for quality and reliability, along with providing a solid profit margin.

It was finally decided that Victor would engage in a crash program to develop its own electronic calculators, with a conservative design utilizing tried and true off-the-shelf technology that would be simple, reliable, and relatively inexpensive to manufacture, but employing high quality parts and design to assure that the resulting machines would meet the high expectations that management held for them. After all, the future of the company literally depended on the success of these machines.

Development of the new calculator design began in late 1967. It was decided that there would be two calculators in the series; one machine offering a few additional capabilities over the other, in order provide the prospective buyer different options based on their need. The development of the machines proceeded at a rapid pace, as all aspects of the design were chosen to minimize risk, and increase the odds of success. In early 1969, two calculators were introduced, the entry-level Model 14-321 as exhibited here, and its slightly more capable big-brother, the Model 14-322.

The Victor 14-321 exhibited here appears to have been manufactured sometime in the latter part of 1970, based on date codes on integrated circuits and other parts in the calculator. The main logic board is populated with parts from the mid- to late part of 1969, but the MOS LSI shift register chips on the display generator board are dated in the late-1970 time frame. Either the display drive board is a later replacement (though there isn't obvious evidence that it has been replaced), or the main logic board was perhaps was from a batch production run done in early '70 that simply had not yet been used up by production. Because of this discrepancy, it is difficult to pin down with certainty when this machine was actually built.

CRT-Display of the Victor 14-321 in operation

The display utilizes a 5-inch RCA 4499 electrostatically-deflected CRT (Cathode Ray Tube) with Green P1 Phosphor for presenting output to the user, showing the content of three registers with digits drawn in a slightly modified version of the familiar seven-segment form. The modifications include adding a tail to the top segment of the "7" and "2", and slightly reducing the height of the "1", which also affects the "4". The digits are presented subtly slanted to the right, giving a less blocky look to the digits. The display provides leading zero suppression, and automatically groups displayed numbers into three digit blocks for easier reading, before and after the decimal point. Negative numbers are displayed with a "-" directly to the right of the number. The result is an easy-to-read display with a pleasant look to it. The 14-321 performs the four basic math functions with a capacity of fourteen digits.

Both of the 1400-series calculators share a great many common components, including the power supply circuitry, CRT tube, display generation and driving circuitry, keyboard design, chassis, and cabinetry (although the keyboard bezel for the 14-322 has more space for the additional keyboard keys). The stable-mate to the 14-321 offers has an additional store/recall memory register as well as providing the ability to add/subtract results of multiplication and division operations to/from the accumulating memory register.

The electronic architecture of the two machines is identical, with with only a number of logic changes in the 14-322 to implement the additional capabilities of the machine. The architecture utilizes a (mostly) bit-serial data flow as with is very common in electronic calculators in general. At idle, the operational speed of the machine is slowed down to allow the display to be generated, as the display process operates at a speed significantly slower than the capabilities of the rest of the logic. The content of the working registers circulates through MOS shift registers a bit at a time, through the arithmetic unit (which essentially operates in "add zero to everything" during the display operation), through the display generation logic, and back into the shift registers. This process continues indefinitely until the machine is asked to perform an operation by the operator pressing a key on the keyboard. Once a key is pressed, the calculator shifts into "high-speed" mode, turning off the display generation part of the data flow, performing the requested operation, then resuming the comparatively low-speed idle mode. This two speed mode of operation was implemented in the earlier Victor 3900 calculator, and was patented as US Patent 3,453,601.

The arithmetic unit (AU) of the 1400-series machines is slightly different than most traditional bit-serial designs, in that it operates on numbers a digit (four bits) at a time rather than operating on data a single bit at a time. There are two four-bit shift registers that serve as the two inputs for a 4-bit parallel arithmetic unit. As bits for each digit of a math operation stream out of the MOS shift register chain, logic gating directs the appropriate bits into the two arithmetic unit input shift registers a bit at a time. Once both shift registers contain a digit consisting of four bits, the arithmetic unit performs the requested math operation on the two digits in parallel, putting the result into another parallel-input, serial-output shift register whose 4 bits are shifted into the serial bit steam at the appropriate place for the result to reside, as well as making note (by setting a flip-flop) if the math operation resulted in a carry or borrow. This is a rather unusual arrangement, and it's not clear why this scheme was used instead of the usual single bit serial adder with a carry delay. The arrangement used in the 1400-series calculators seems to add no real benefit in terms of performance, and definitely adds cost in terms of requiring more chips to implement. The only reasons that the author can think of is that this design may have been used to avoid concerns relating to patent infringement, or perhaps to make the design unique for patent purposes, though no patents from Victor Comptometer have been found that reference this arithmetic unit design.

A close-up view of one of the MOS LSI Shift Register Integrated Circuits
Note Date code of 7042 (42nd week of 1970)

The 1400-series calculators were originally designed using a magnetostrictive delay line, with the delay time of 412 microseconds(0.000412 seconds). Clocked at 1MHz (one cycle every 0.000001 second), there was sufficient capacity in the delay line to store approximately 412 binary bits, one bit for every clock cycle. Due to manufacturing tolerances in the delay time through the delay line, as well as changes to the delay period due to thermal effects, the timing logic was designed to provide for 400 bit-times, with the remaining 12 bit-times (or roughly .000012 seconds), give or take, used as slack to provide for such timing variances.

An astute reader may realize that 400 bits seems like a bit more than would be needed to represent the four workng registers of the 1400-series. Each register holds 16 digits, and each digit takes four bits (in binary-coded decimal form) to represent it. That yields 64 bits per working register, and with four working registers, the total comes up to only 256 bits. Why, then, would the delay line hold 400 bits, when only 256 are theoretically needed? The reason comes down to cost. The 412 µS delay line was a delay period that was a standard item produced by the manufacturer. Specifying a standard delay line, even if it had more capacity than actually needed, was substantially less-expensive than having the manufacturer produce a custom delay line with a delay of something on the order of 265 to 270 µS, enough to comfortably hold 256 bit periods at 1MHz, along with some slack. Even though the full capacity of the stock 412µS delay line could hold well over the 256 bits needed, its availability as a standard stock itemp made it a wise choice, with the logic of the calculator designed to utilize only the 256 bits needed, with the rest of the bit times ignored on each cycle of the registers through the delay line.

Based on vintage documentation produced by Victor on the Victor 1400-series calculators, sometime during the latter part of 1969, a design change was made to the calculators whereby the delay line was replaced by two large-scale (for the time) MOS shift register ICs. Each shift register IC contains two 100-bit shift registers, which can be connected together externally (by connecting the output of one of the 100-bit shift registers to the input of the other shift register, with both shift registers connected to the same clock source) to form a two-hundred bit shift register. With two of these chips tied together, 400 bits of storage were, virtually the same as the delay line. Since each shift register IC held 200 bits, two of the devices were required to hold the working registers of the calculators, with 144 bits left-over No "slack" was needed with the shift registers, because of their solid-state digital nature. Delay lines are electromechanical devices subject to various conditions that can affect their delay period. Not too long after the 1400-series calculators had debuted on the market, the cost of large-scale MOS integrated circuit devices had come down to the point where it was more cost-effective to utilize the solid-state MOS shift register ICs as a replacment the delay line, even given the extra engineering cost to modify the design, as well as the changes required to circuit boards. Use of the shift register ICs also eliminated some of the problems that could affect delay lines, such as timing skew due to changes in ambient temperature, which had to be compensated for in the circuitry. Delay lines were also sensitive to physical shock, such that lifting one edge the calculator's cabinet an inch or so off the desktop and dropping it (not recommended, as such activity could also damage the fragile CRT), the shock could cause the bits in the delay line to be scrambled. while it wasn't expected that the calculator would endure such shocks, the delay line was a rather delicate assembly, and if the calculator was subjected to excessive shock during shipping and handling, or an accident when being moved from place to place, the delay line could be damaged, resulting in a calculator that required service to replace the delay line. The solid-state shift registers had no such sensitivities to temperature (within reason) and shock, meaning that overall reliability of the calculators was improved. The shift registers also required considerably less-complex circuitry to interface them to the DTL IC logic of the calculators.

The extra 144 bits of the storage in the shift register are simply skipped over, being ignored by the calculator logic during each cycle of the shift register. The reason that 400 bits of storage was used is that the delay line originally used had a delay time of 412 µS. At a bit-clock rate of 1 MHz, that would yield something on the order of 412 bits of storage as the capacity for the delay line. The logic related to the timing of the delay line-equipped 1400-series calculators was designed to skip over 156 bit times (give or take), much like the shift-register equipped machines would skip over 144 bit times. The delay line calculators used the slack time as a means to help keep the calculator logic synchronized with the delay line.

The reason that the delay line was specified with a delay of 412 µS is that it was a standard delay period provided by the delay line manufacturer. Using a delay line that was a standard stock item from the manufacturer would made the delay line considerably less expensive than specifying a custom delay line. When the switch was made to the MOS shift registers, the timing logic was modified slightly to adjust to 400 bit times versus 412, with 144 bits ignored. The logic relating to synchronizing the delay line with the calculator logic could be eliminated. Two shift reqister ICs were required because the total of 256 bits required to represent the working registers of the calculators was 56 bits more than a single IC could contain, thus requring the second shift register IC.

The exhibited 14-321 calculator was a benefactor of this update. The change involved a redesign of the display generator/drive circuit board to remove the delay line and replace it with the two MOS shift register integrated circuits. The fairly complex delay line input driver and output amplifier were replaced with simple level shifters to convert the DTL logic levels from the main logic into signals that the shift register IC can accept, as well as level shifters to convert the MOS shift register output levels to DTL logic levels. of the delay line. Also included was special clock driver circuitry for shift register ICs, and as mentioned above, slight modifications to the timing logic to adjust for 400 bits versus 412. The shift registers behave very much like the previously-used delay line; bits go into one end of the shift register a bit at a time, are delayed a specific amount of time as they shift through the 400 stages of the shift register, and come out the other end in the same order, but delayed a specific period of time. The fact that the data is held in the storage elements of the shift register during this time provides the memory needed for the working registers of the calculator. The shift register ICs use dynamic data storage cells that will forget their content if the shift register is not operated at a minimum shifting rate specified by the manufacturer. For this reason, the registers of the calculator are maintained by continually circulating the data through the shift register at a moderate speed, then through the arithmetic unit, to the display generator circuitry (when the calculator is idle), then and back to the shift register all the time the calculator is powered up. This continual activity of the shift registers assures that the data is retained.

All of the digital logic of the 14-321 (with the exception of the shift-register ICs) is contained on one large circuit board mounted in the base of the machine. Two square-pin connectors provide the connections between the calculator logic and the power supply, keyboard, decimal point selection switch, register storage, and display generation circuitry.

According to old Victor documentation, the 1400-series calculators were declared obsolete on January, 20 1971. The fact that the machines were obsoleted after just two years exemplifies the frenetic pace of electronic technology development during this period of time. Production of the machines likely ceased around or shortly after that date, with sales continuing until inventory was depleted.

Top view of 14-321 chassis

The display shows the content of three registers. The bottom line of the display shows the content of the entry/result register, into which numeric entry occurs and final result of math operations are placed. The middle line of the display shows the multiplicand/dividend register, which holds the first number of multiplication or division operation, which retains its content across calculations to serve as a constant register. Not displayed is a third working register used in addition and subtraction operations. The content of the 14-321's accumulator-style memory register is displayed on the top line of the display. All of the working registers have a capacity of 16 digits, with two non-displayed digit positions used for housekeeping purposes, including keeping track of the sign of the number in the register and storing the decimal point location, leaving the other 14 positions available for digit storage.

Decimal point settings are available for 0, 4 or 6 digits behind the decimal point. Oddly, there is no setting for two digits behind the decimal, which would be useful for financial calculations. The 14-321 does not have any round-off functionality. For example, 2 ÷ 3 results in 0.6666 with the decimal point selection thumb-wheel at the 4 digits behind the decimal position.

The main logic board for the Victor 14-321

Both of the machines in the 1400-series have three circuit boards. The rather large main logic board (measuring 18"x13") occupies the entire bottom of the chassis. This single board contains all of the digital logic of the calculator except for the register storage. This board in the 14-321 is populated with 114 small-scale DTL (Diode-Transistor Logic) devices made primarily by Fairchild, but with the few Texas Instruments parts sprinkled about. The devices are all in 14-pin dual-inline plastic packages. Unfortunately, Victor saw fit to have the IC manufacturers place Victor-specific part numbers on all of the IC's (e.g., 50210-x), so it's a bit of a challenge to figure out exactly what function of each particular part number provides. It is most likely that the ICs are from Fairchild's popular DTµL line, with the few TI parts being functional equivalents.

The Fairchild (Marked with f Symbol) and Texas Instruments (Marked with Texas State Outline) Diode-Transistor Logic (DTL) Chips that make up the logic of the 14-321.
Note Victor Custom Part numbers that take form of 50210-X, where X indicates type of IC.
Also note date codes ranging from 6933 (33rd Week of 1969) through 7004 (4th week of 1970)

Two groups of square pin connectors on the main logic board provide connections to the rest of the machine. One connector is for the keyboard, and the other connector provides power supply, connections for the storage element (either the delay line or a pair of MOS shift register ICs) located on a separate circuit board, and logic signals that direct the operation of the CRT display. A second board, approximately 11"x4", oriented vertically on the left side of the chassis, contains mostly discrete components, with one single linear integrated circuit. This board creates the low voltage power supplies (for the logic and deflection amplifiers for the CRT); provides Digital to Analog Converters that are used to generate the vectors that draw the figures on the CRT display; the CRT deflection amplifiers; the storage element, as well as signal conditioning for the storage element.

The keyboard in both machines utilize very high-quality modular key-switch units, with two sets of dual gold plated contacts for reliability and minimization of contact bounce. Even so, there is a special timer circuit that delays the sampling of key closures by a few milliseconds to allow any contact bounce to settle out. The surprising part about these keyboard switches is that they are an open-frame switch, with the contacts exposed to the environment inside the cabinet. This makes them susceptible to dust and atmospheric contaminants that could potentially cause problems with reliable key switch operation, but even after 30-plus years of service without any cleaning, the keyboard in the 14-321 exhibited here works very smoothly, with no glitches. The key caps are made of plastic, with molded in color and nomenclature. The power switch is made up of a slide switch located to the left of the keyboard panel.

The 14-321 is built upon an aluminum chassis that is quite nicely fashioned, and very study. The circuit boards are mounted to the chassis with machine screws that hold them in place. The wiring harness consists mainly of individual wires that interconnect the circuit boards, using hard-wired connections, or plug-type connections. The cabinet is made of plastic, with two parts; the base, and the upper cabinet. The upper cabinet is made of a few parts, including the main cabinet section, and two pieces that form the "bubble" that makes up the hood and viewing screen for the display. The base part of the cabinet is a darker, forest green color, and the upper part an almost turquoise-green color mix. These colors are consistent with Victor's corporate coloring scheme, although these colors are not what the author would consider stylish in an office environment. With age and exposure to atmospheric contaminants, the color of the upper cabinet oxidizes to a rather sickly looking brownish-green. The keyboard bezel is painted a medium-gray. Cooling is by convection only, with cooling grates in the base, and at the back of the upper cabinet to allow air to flow through the cabinet to cool the components. The machine does generate a sizable amount of heat, and the cabinet gets clearly warm to the touch after operating for 15 minutes or so.

Keyboard of the Victor 14-321

Both the 14-321 and 14-321 operate conventionally, with arithmetic addition and subtraction, and algebraic multiplication and division. Addition/subtraction are performed by entering a number, followed by the [+] or [-] key, which immediately adds or subtracts the entered number from the amount in the hidden working register, then copies the working register in to the entry/result register. Multiplication and division are entered as they would be written, by entering the first number, pressing the [X] or [÷] key, which moves the entered number into the multiplicand/dividend register, entering the second number, and pressing the [=] key, which calculates the product or quotient, and places the result in the entry/result register, leaving the multiplicand/dividend register untouched, allowing for easy multiplication and division by a constant. Two clearance keys are provided; [C], which clears the entry/result register (used mainly for correcting entry errors), and the [C ALL] key, which clears all of the working registers, along with the accumulating memory register. The memory register in the 14-321 is a full accumulator-style memory register, capable of accumulating sums and differences. The content of the memory register is displayed on the top line of the CRT display at all times. A total of three keyboard keys control the operation of this register. The [M+] key adds the content of the entry/result register to the memory register. The [M-] key subtracts the entry/result register from the memory register. In both cases, the number in the entry/result register remains unchanged. The [MR] key copies the content of the accumulator memory into the entry/result register. It is interesting that Victor opted not to include a function to clear the memory register. The only way to clear the register is to subtract (or add, if the number in the memory register is negative) the value currently in the memory register to/from itself, which results in the memory register being set to zero. The easiest way to do this is to recall the memory register using the [MR] key, then press [M+] if the number is negative, and [M-] if the number is positive.

Overflow indication on CRT display of the 14-321

The 1400-series machines have very few quirks, clearly a result of careful logic design. The machine is thorough in detecting error conditions, such as calculation overflow, entry overflow, and division by zero. The full 14-digit capacity of the machine is available for all operations. When an overflow/error condition occurs, the machine displays 16 digit positions filled with a rather odd character that consists of all segments and decimal point (including segments normally blanked) on at once, and ignores the depression of all keyboard keys except the [C] and [C ALL] keys. Pressing the [C] key will clear the overflow condition, restoring the display to normal, with the memory and multiplicand/dividend registers as they were before the error condition, and the entry/result register cleared. Pressing the [C ALL] key clears the error condition, as well all working registers and the accumulating memory register. If an overflow is caused by the accumulating memory register exceeding capacity, unusual results can be left in the memory register if the [C] key is used to clear the overflow condition. The machine does not have a power-on reset circuit for the registers, so typically when the machine is powered on, it comes up in overflow state, requiring a press of the [C ALL] key to clear things before calculations can begin.

The machine operates at a master clock frequency of approximately 2 MHz (Two million clock cycles per second). This master clock is divided in half to generate the calculator's main timing signal of 1 MHz, which is used to clock data through the shift register storage. The 1 MHz signal is further divided down to yield sixteen operational time-slices that are used to step the machine through its various states. With this rather fast clock rate, the calculator is quite quick. I have not been able to find published calculating speed figures for these machines, so the estimates of calculation times are only by observation. Addition and subtraction occur with no apparent delay, with only a barely-noticeable flicker of the display as the calculation is performed. A guess would be that these operations complete in somewhere between 30 (0.03 second) to 50 milliseconds (0.05 second). Multiplication and division are also quite quick, with a brief, but more noticeable flicker of the display noted for all but the most complex operations. An estimated average time for multiplication and division would be somewhere between 100 to 150 milliseconds. During calculation, the display is blanked, so on more complex multiplication and division operations, there is a quick, but noticeable dropout in the display. The Old Calculator Museum benchmark calculation of all 9's (in this case, fourteen 9's) divided by 1 completes in approximately 300 milliseconds (roughly 1/3rd of a second). Multiplication of 9,999,999 by itself takes a little over 250 milliseconds (1/4 second). Both calculations yield the correct result, which is somewhat uncommon for calculators of this vintage.