engineering

So, I transferred into the Engineering Department as a Junior Engineer. This was my first salaried position, and also my first experience wearing a tie to work each day. Nowadays, I only wear a tie for interviews and funerals.

The Production Color Organ FC1 kit

My first project upon joining Engineering was to prepare the color organ for production as a field update kit. This involved preparing drawings for all the new parts, cables, etc., and writing up a step-by-step procedure to be followed when installing the kit. We designed a PC board to mount all of the parts. The Color Organ Control Center or ‘COCC1’ as it was called used four of the same triacs used in the Income Totalizer to switch the lamps on and off. They were mounted to the PC board, and held in little heatsink cups, which screwed to a pair of Z-shaped brackets that were, in turn, screwed to the top cover of the unit. This was all mounted inside the same black plastic box that was used for the ‘quadraphonic decoder’ (which was really just a few wirewound pots interconnecting remote speakers for an SQS160 ‘Quadraphonic First Edition’ jukebox. That’s why the top of the ‘Quadraphonic Decoder’ was screwed on with tamperproof screws, to hide what was really going on in there.) The mounting arrangement for the color organ was an attempt to provide a modular heatsink and assembly, since the triacs ran pretty warm. This box was mounted on an internal metal wall of the FC1, which separated the mechanism from the bill validator chute going down into the cash box.

Both of the lighting panels on this jukebox consisted of a fibreboard panel, with two metal posts staked on either side of a hole, through which the 2182 grain-of-wheat lamps protruded. The bulb leads were soldered to the posts, and bare wire was used to connect one side of all the posts together, ending up in a two-wire cable that would plug into the power supply transformer. All this interconnect wiring was on the far side of the panel, which couldn’t be seen from outside the jukebox. The wiring posts were painted black from the visible side, so that only the bulbs could be seen. For the top panel, the cable was fed through the right-hand top panel support and out through the pivot around which the top panel would rotate to latch the panel down for shipping.

For the prototype, we had used double-stick foam tape, which was about ¼ inch thick to form a light-proof (only on the sides, not from the top) box around the area where the light bulb protruded through the hole in the lamp panel. One side of this foam ‘box’ stuck to the lamp panel, and a color filter was stuck to the other side. Unless you were real careful when assembling the foam ‘box’, you would get white-light leaks where the pieces of foam tape joined. For the production version, we used die-cut foam boxes, with the double-stick on both sides to solve this problem.

The conversion procedure required that you disassemble both panels from the jukebox, place the boxes and correct color filters over ¾ of the bulbs, remove one wire from all of the bulbs, and connect the correct bulbs together into four separate circuits. Then you had to connect and route a new five-wire cable through the upper panel pivot arm, connect everything up and reassemble the box. It easily took several hours to do this, which is why management canceled the project after I had spent several weeks getting all the drawings, bills-of-material, etc., together and wrote a twenty-some page procedure on how to perform the modification. I don’t believe more than about three were ever built, but they sure looked nice when playing a song. I managed to get my hands on one of them and had it at my house for several years. I sold it to an operator before moving out to California because I didn’t want to pay the shipping costs. Big mistake, I wish now I had it back. It was the FC1 prototype cabinet, the only one finished in gold trim. All production machines were finished in silver trim. If anyone knows where it is, I’d like to buy it back! I have since bought an FC1 of my very own, and have restored it to original operating condition, without the color organ. More about this in a later section.

Quadraphonic Sound

Quadraphonic sound seemed to be the next big consumer push (what would later be called a ‘killer app’) and Seeburg wanted to be on the bandwagon. That is how the ‘Quadraphonic First Edition’ version of the STD160 (renamed the SQS160) came to be, using the wirewound pots and weird speaker interconnect mentioned earlier. I have since learned that a few SPS2 Matadors were modified for Quad in the field. There was a kit offered, with a new upper glass, and the Quad ‘decoder’ was in a metal box instead of the more familiar plastic box. Since it seemed as though the consumer stereo industry was going to go Quad, Seeburg felt the coin-operated music industry had to go there too, lest they lose customers to the home, and later portable, stereo systems. I think that portable cassette and later, CD players is what sealed the fate of Seeburg and everybody else, except Rowe/AMI and RockOla. Someone had to survive and Rowe/AMI probably had a better management team than anyone else. RockOla was bought by a totally separate company. But CDs hadn’t yet been introduced. In the meantime, quadraphonic seemed the way to go. We got samples of several quadraphonic decoders and did some listening tests with them connected to both our jukebox and regular stereo systems. As I recall, the Sansui QS decoder did the best job of separating out the four channels, except for the CD4 system, which would require a major re-tooling effort for the mechanism pickup arm (to reduce tracking force, which would in turn give audio feedback problems with the speakers in the same cabinet as the mechanism) and Pickering, the pickup supplier. The CD4 system encoded the rear audio channels onto a subcarrier, at about 35 KHz. This required the major modifications in order to be able to track a stylus moving at a maximum of 50 KHz, rather than the normal audio maximum frequency of about 17-20 KHz. Then, all of a sudden, no one seemed to have any interest in four-channel sound, and we dropped the whole project after the SQS3 ‘Quadraphonic’ version of the Sunstar. I believe that conversion kits were made available for the FC1 and FC2 machines, but I have never seen one. For that matter, I have never seen an SQS3, except for photos in the brochure. The real reason for the death of Quad for consumers was that there were no standards established for it. There were simply too many competing systems out there. Multi-channel sound finally became standardized by Dolby Labs, Inc., with their AC-4 and 5.1 surround-sound systems. These started out in movie theaters and finally migrated into the home when the semiconductor industry had advanced to the point where decoding circuitry could be put onto a single inexpensive chip.

The semi-solid state Play Control

Well, since the color organ kit idea didn’t work out, my next assignment was to come up with a replacement for the Play Control Assembly and Scan Start board, used on the DCC (Digital Control Center) for the 160 selection jukebox. Seeburg had used the Play Control assembly for many years. It consisted of a pair of solenoids controlling a switch. Whenever a selection was made, the add solenoid energized, closing the switch (if it was not already closed from a previous selection) which in turn applied power to the mechanism motor. Each time the mechanism carriage reached the 179/279 side of the magazine, a switch would close, energizing the subtract solenoid. The mechanical gizmo which controlled the switch made it so the subtract solenoid had to energize twice to actually turn off the motor. The result: two complete scans before the mechanism shuts down. On the older boxes, a contact on the pricing unit closes to energize the add solenoid, but for the black & gray boxes, a small PC board containing a two transistor flip-flop, a Unijunction Transistor (UJT) oneshot, and a triac were used to perform the same function. I have no idea why they went to such extremes on the design of this board. An output of the gray box flips the flip-flop so that the triac turns on, which energizes the solenoid. At the same time, a capacitor charges to fire the UJT after a time delay, resetting the flip-flop and turning off the triac. Why not just use a oneshot to fire the triac?

Anyway, the idea for the replacement was to use a dual flip-flop chip, and a relay driven by a transistor controlled by the second stage of the flip-flop. The output of the gray box would set both of the flip-flops through a level-shifter, needed since the flip-flop ran off the positive supply and the gray box ran off the negative supply. The flip-flops were configured as a shift register, so that the transistor would turn off after the shift register was clocked twice. Each time the mechanism switch closed, the flip-flops would be clocked. I came up with a working circuit, built a prototype, and started testing it. Part of the testing included rattling the power switch to see what would happen, and injecting noise using a spark generator, which would discharge to the cabinet. Try as I might, I could not get that circuit to ignore the sparks, so I eventually gave up on the idea. I’ve learned a lot about noise reduction since then, but I don’t think I’ll be doing any jukebox control systems any time soon. The last digital Seeburg (the STD4 Mardi Gras) still used the original scan start board, with flip-flop, UJT, and triac.

The 5BS1 Five Bit Sequencer

One of the nifty things about working in the Engineering Department is that you get to see what the new jukebox will look like long before most people in the company. Seeburg used an outside consultant to style their new jukeboxes. Each year, in the early spring, the consultant would come in with about a half-dozen design sketches, which upper management would look at. They would usually pick two or three that they liked, and the consultant would go ‘develop’ those sketches, which meant build a foam-core (similar to sheet rock or plasterboard, except that the plaster is replaced with Styrofoam) prototype, about two feet tall. When finished, the prototypes would be put into one of the locked engineering demo rooms, and certain people were invited to look at them and give their opinion. My opinion certainly wasn’t asked, since I was about as low as you could get on the engineering ‘food chain’ and still be able to see the mockups. But it was still really interesting to see what the proposed machines looked like. Eventually, management would decide on one of the mockups, and the consultant went off to build a full-size mockup, still in foam-core. This is the time when the color scheme, materials to be used, etc., is decided.

Seeburg went through an annual model design cycle, just like cars coming out of Detroit. Every year had a new model for the 160-selection machine, but usually there were several years between major design cycles. The 100-selection machine usually ran for at least two years, sometimes more. For cars, there is a major change every few years, with minor styling changes for the in-between years. These minor styling changes are called ‘design refreshing’ in the automotive industry. Seeburg followed a similar practice. The basic cabinet style would remain unchanged for a few years, followed by a major change. Then, minor styling updates would follow for a few more years. For example, Seeburg used the same basic cabinet for the LS1, 2, and 3 jukeboxes. The USC cabinet design lasted for two model years, as did the SPS cabinet. The STD160 started a new cabinet design, which would last for four years. The trick was to make the box look significantly different for each model year, but change the fewest number of parts possible, since changing a part meant spending tooling money, which was very scarce at Seeburg during these times. On the other hand, if you didn’t change much, somebody would come out with an update kit so that the operators could make last year’s jukebox look like this year’s. This was obviously bad, because then the operator would not buy the new box. So, how do you make a box look completely different, even though it’s basically the same?

The STD160 was a completely new cabinet design for 1974. For 1975’s STD2 ‘design freshening’, the consultant came up with the idea of using a moving lighting display. He was the same guy (Bob O’Neil) who came up with the FC1 design, and evidently liked the look of the small lamps. This time, he put a row of lamps into a mirrored, U-shaped channel, and covered it with partially reflective glass. The result was the ‘infinity glass’ effect, as it was called. The glass and the mirrored channel reflected the lamps, so it appeared that there were at least 10 or so lamps for every real one in the box. And, as you moved, the display would appear to move, too. To make the box more attention getting, it was desired to add a lamp animation controller. I was assigned the task of coming up with this gadget.

It would cost way too much to build a sequencer that used a discrete lamp driver for every lamp (if I recall correctly, there were about 55 lamps in the entire string, with more on the lower panel than on the upper). So we decided to cut that down to a much more reasonable number (say, 5) and then repeat the pattern every five lamps. I found a digital serial-to-parallel shift register chip that would get the job done, but I had to come up the serial input to it. The circuit was actually pretty simple. Any time any of the first four outputs were at logical 1, I fed a logical zero into the serial input. Otherwise, the input was logical 1. The end result was the shift register walked a logical 1 through five stages. At any time, one and only one of the stages would contain a logical 1. This was used to turn on a driver transistor, which turned off one of the five output transistors. The end result was that there was an ‘off’ lamp traveling through a field of ‘on’ lamps, and the pattern repeated every five lamps. The circuit had an adjustable speed, but most machines I have seen set the speed at a constant rate. With four fifths of the lamps on at all times, the circuit ran pretty warm. For output transistors, I used the same part as was used for the +27 VDC regulator on the control center. In fact, the design used only two or three new electronic components, so it was not a lot of work assigning new part numbers for these parts.

My supervisor was the same Section Engineer responsible for the overall design of the black and gray boxes. One of the things he was upset about was the number of black boxes returned from the field having blown-out lamp driver transistors. He wanted to re-design the black box to take care of that problem, but that would make the new box incompatible with all those in the field, so he was prevented from fixing this problem. But he was determined that it would not happen on any new designs. So, part of my design requirements for the 5BS1 was that it must be able to withstand driving a shorted bulb for at least one hour. We figured that if the design survived an hour working into a short, it would survive until the next time the repairman showed up on location. When the design was complete, he took a wire jumper, shorted out one of the lamps, and went to lunch. When he came back, he removed the jumper, and the lamp started blinking again! You better believe I was happy when I saw the lamp blinking again! This was my first design ever to make it to production, and I was quite proud. I still am, whenever I see an STD2,3, or 4 with the lamps still working after all these years.

The STD2 was one of the more popular jukeboxes of the era, and its popularity seems to have held up well over the years. And, it fulfilled the requirement of having been put into production with minimal tooling money spent. Very little changed between the STD160 and STD2. Most, if not all of the front panel pieces were identical, except for paint. The mirror channels required tooling for the holes and forming. There was also a molded plastic piece used to mount the bulbs. This plastic piece gave some problems. The terminals for connecting the bulbs and wiring, which were staked in, had a tendency to work their way out of the holder. Also, the connections were made using insulation displacement terminals, which sometimes nicked the wire and would therefore fail over time. I recall that UL (Underwriters Laboratories) gave Seeburg some grief over this choice of connection and plastic holder. While Seeburg never submitted their jukeboxes for UL approval (at least during this era), they certainly made sure that the machines were built to UL standards. The reasoning behind this, from what I understand, is that it took too long for UL to do their tests on each new model. Evidently, Seeburg informally submitted machines to UL, and changed whatever UL insisted on, lest UL give them serious grief in legislative quarters. Because of this, the STD3 and 4 used PC boards to mount the bulbs, and regular crimped edge connectors instead of the insulation displacement terminals of the STD2. Replacing bulbs in an STD2 takes a lot of patience, since it is very easy to break the fine wires going to the bulb, or rip them out of the glass. As I mentioned above, the terminals have a tendency to work themselves out of the plastic holder, too.

When it came time to start working on the STD3, tooling money was again scarce. The cabinet design agreed upon (the half-circle upper and lower pieces mounting the flashing lamps) would be expensive to tool if those panels were made of metal. The engineers responsible for cabinet design started investigating alternatives, and came up with the structural foam material actually used in production. Tooling for this process was virtually non-existent. I believe the molds for the panels were made of wood, into which the foam mixture was poured. It then took a while for the foam to cure, after which it became a rigid piece. But, unforeseen during the design phase, there were some serious problems with these pieces warping with time and temperature, which resulted in several stiffeners being added later. This increased the amount of labor required to build a jukebox. Seeburg did not repeat the rigid foam panels for the STD4 jukebox.

Since the animated lighting was a success on the STD2, it was carried over to the STD3 and 4. I was asked if there was anything I could do (cheaply) to enhance the effect. I came up with a modification to the 5BS1 that varied the brightness of the bulbs, along with ‘moving’ them around the box. It was felt that this ‘breathing’ didn’t enhance the effect any, and the idea was dropped. I had left Seeburg by the time the STD4 was being worked on, so I don’t know if any additional changes to the lighting display were discussed.

LP Mechanism Band Select

Seeburg built a home stereo machine, which played 50 12-inch LP albums. They spent an awful lot of money tooling this mechanism, and did not sell very many machines at all. There were two different control systems offered for the machine. The first was based on the remote control stepper used for the wallboxes, and used a rotary dial telephone dial to enter selections. The second used a 12-button keypad, which along with the Tormat write-in circuitry, was the forerunner to the black and gray box jukebox selection system. While in-between ‘real’ projects, a mechanical designer and myself were assigned the task of investigating the possibility of developing a control system permitting band select on this mechanism. What I mean by the term ‘band select’ is this: the vast majority of LP albums have 3 to 8 ‘bands’ or songs on each side. Each of these is separated by a change in recording lathe pitch, which puts a noticeable band in between each selection. This makes it easier for persons using manual turntables to cue the song they want to play, without having to play them all. If we could come up with a reliable way of detecting these bands, we could then develop a coin-operated jukebox using this mechanism, offering many more selections with wider variety than the standard 45-RPM mechanism could offer. If each LP had, say, 10 selections, we could build a 500-selection jukebox, once again leaving the other companies in the dust. Also, there were tunes available on LPs during that time which never made it to 45s for one reason or another.

A British company by the name of BSR was selling a single-play turntable, which used an infrared LED to detect the changes in reflectivity between recorded grooves and the dead band in between selections. We bought one of these (called the ‘Accutrac 4000’), and spent some time figuring out how it worked. We quickly discovered the problems with using an infrared LED for this task. The reflectivity of the dead band varies a lot from record to record, especially when there is a mix of old and new records. Indeed, some records have no dead bands at all between selections. BSR handled the reflectivity problem by including a sensitivity adjustment knob. Clearly, we could not do this on a jukebox, so some automatic means would have to be used which would require some fairly sophisticated (for the time) analog circuitry. We discarded the LED idea. Besides, BSR probably had a patent on it.

I came up with the idea of detecting the amount of pickup arm rotation around its pivot point to determine where the dead band was. After all, to select one band out of all on a record side, you had to be able to rotate the pickup arm to some point, and then place it down on the record. You also had to determine where that selection ended, so that you could lift the pickup out of the groove and either move on to the another selection on the current record or reject it. So in any event, there needed to be some sort of detector, which would split the 25 or so degrees of arc the pickup traveled while ‘in the groove’ into small enough increments to catch each dead groove correctly. The detector also had to move freely with the pickup arm (which is what Seeburg called the tone arm) and offer no drag, since it would be active even while the record was playing. Also, the detector could not add any appreciable mass to the pickup arm. We felt the best implementation would be either a bar-coded panel attached to the pickup arm which would move past a stationary reader, or a panel with a pattern of holes which would interrupt a stationary optical emitter/sensor combination. The latter would hold up better under normal operating conditions.

We also had to solve the problem of getting the dead band position information into the system when the record was changed. We thought of applying a barcode to the record label, which could be read as the record rotated. We rejected that idea, since Seeburg didn’t have enough ‘clout’ with the record producers anymore to get them to print that information on the label, plus there would be problems with the contrast between the bars and the label background. We also figured that it would not be possible to get the operators to stick a preprinted label over a record before loading it. Then we hit upon the idea of using the sensor itself for programming the system for a new record. When a new record is loaded, the service man places the pickup on each dead band, and pushes a button. The system then reads the code for that position, and knows that this band is at that position. Now whenever that selection is made, the pickup arm travels to the start of the band, places the needle down, and monitors the detector until it matches the end of the band, and then lifts the needle out of the groove.

One problem solved, assuming we could get a microprocessor with enough horsepower and memory at a reasonable price to do all of this. But then, we came up with another problem.

When the 12-inch mechanism was first designed, LPs were made of a certain thickness of PVC (Poly Vinyl Chloride) plastic. As time progressed, the record producers started using thinner and thinner PVC blanks, evidently to save money. For normal record players, this is no problem since the record is placed on a horizontal flat turntable, which supports it everywhere. But the vertical play mechanism only supports the record with about a 4-inch diameter turntable, at the center of the record. The rest of the record (about 4 inches of the 6-inch radius) is totally unsupported, and tends to warp or wobble with no support, due to the tracking force applied by the pickup arm to keep the needle in the groove. To make matters worse, any heavy bass notes at high volume tend to move the air in the cabinet around and cause cabinet vibrations, which can feed back into the pickup arm, causing a high-pitched squeal from the speakers. The only way we could come up with to get around this problem was to completely change how the mechanism works. The records would be stored vertically as before, since that takes the least space and is best for storage to minimize warping due to gravity. But, the record would be placed on a normal turntable for playing, the same way everybody else’s jukebox worked. This also meant that the cabinet would have to be made wider, since the moving portion of the mechanism would have to stick out at least 6 inches (half the diameter of an album) past each end of the magazine. With no serious tooling money available, this was out of the question and therefore the project was dropped.

The CSP1 Red Box

My next ‘blue sky’ project was to develop a selection and credit system to replace the Tormat memory, which was a staple of all Seeburg jukeboxes from 1955 onwards. The Tormat is a solid-state, computer core-based memory, which replaced the lever memories used by Seeburg until then and by everybody else up until the early 1980s. The purpose of this memory is to remember which records the customer selected until the mechanism had a chance to play them, followed by ‘forgetting’ each selection as it was played. Building the Tormat was a labor-intensive operation. Each toroid core was individually tested on an automated test machine, which, as it got older, required almost constant maintenance. Each core was then hand-stitched into the Tormat assembly. The assembly line consisted of about 20 or so women armed with needles and thimbles, sewing the toroids into the assembly using fine gauge wire as thread. This was followed by an extensive automated test of the completed assembly. Replacing all of this with a single solid-state memory would save a lot of labor.

So, I started working on a digital credit and selection system, featuring a RAM (Random Access Memory) to replace the Tormat. At the same time, my supervisor started looking at microprocessors as a possible control system for a jukebox. Intel had recently introduced its first microprocessor, the 4004, followed shortly thereafter by an improved version, the 4040. He got a hold of a 4040 development kit and started playing around. This was before there were any development systems, PCs, etc., to permit you to rapidly develop an application. All that was provided was a board with the CPU on it, some RAM, control switches, etc. You were expected to develop your code on a mainframe computer using a dialup account with a modem and teletype machine. The mainframe would do your program assembly, and you could then download the assembled code over the modem and save it using a punched paper tape. You could load the memory from the paper tape using the teletype machine. At that time, the only computers at Seeburg were used to run the monster tester (which tested black & gray boxes), and the machine that tracked inventory, payables, receivables, payroll, etc. There was no way we were going to get access to this latter machine since it was jealously guarded by the bean-counters, so my supervisor built himself a little memory rig. This consisted of a wooden frame that would hold about 20 or so black box pricing programmer boards, each of which was wired to be at a different 4040 address. Each board had a different arrangement of diodes. Where a diode was present, the board returned a logical 1 on the data line when that board was addressed. Otherwise, the board returned a logical 0. Each board then contained a single 4040 instruction, and he built (short) programs by rearranging the boards in the wooden frame. As crude as this set up was, he was successful in teaching himself how the 4040 worked.

I took a purely hardware (i.e., no microprocessor) approach, and developed a system using digital logic chips. I decided to use CMOS chips, since they had significantly better noise immunity than the other available choice, TTL. Seeburg had some experience with TTL at that time, since they were building the SRT1 remote translator to permit the newer jukeboxes to work with the SC series Consolette. One of the advantages of the Tormat system was that it remembered what selections still queued (entered, but not yet played) when the power was turned off. Since CMOS RAMs used very little current, it would be possible to do the same thing with the new system.

I organized the CMOS RAM similar to the way the Tormat worked: one location for each record side. Therefore, I needed 160 bits for 160 selections, with a logical 1 meaning the record was selected; logical 0 meant that it wasn’t. There was a 256 bit CMOS RAM available, so I could use this part with memory left over. Then I came up with a bright idea. With a bigger memory (which didn’t cost a lot more once you were willing to take the leap) I could remember and play the records in the order they were selected. That way, if a person liked a song and played it twice, it would indeed play twice. I talked to the Chief Engineer about this idea and he responded “son, let me tell you about overplay and dead time”. Say two people make the same selection, unbeknownst to each other. With the current design, the record would play once, and both people would be happy, thinking that they got their selection. The operator and location owner would be very happy since they got paid for a selection that didn’t play (the extra one). They made an extra 12 ½ cents (two for a quarter), and at the same time minimized wear and tear on the record, needle, etc. So, if two people make the same selection within some reasonable time, it should only play once. This is what the industry calls ‘overplay’. Overplay is good. Selection order play is bad. A jukebox built many years before (the Packard Playmore, if I’m not mistaken) played in selection order. It was a bad idea then, and it certainly was a bad idea in the middle ‘70s. Dead time is the amount of time that nothing is playing, while the mechanism is searching out the next selection. Playing in selection order tends to maximize dead time, since the next selection is not necessarily close to the current selection. Nobody makes any money during dead time, so the idea is to minimize dead time by playing the records in the order that the mechanism gets to them. Dead time, like overplay, is bad. So, I forgot about that idea.

The biggest problem to be solved with this new implementation was how to detect the mechanism carriage position (which record the carriage is currently alongside). This is something inherent in the design of the Tormat, since there is a contact rivet for each record side. Each contact rivet is pulsed to interrogate the associated toroid as the mechanism carriage passes by while in scan. For the new system, there were a couple of ways to do it. The first was to use a counter that would count up in one direction of carriage travel, and count down in the other. I rejected that idea because the circuitry would lose track of the position if power was interrupted in the middle of a scan. If that occurred, the carriage would have to scan to one end of the magazine, where the correct count could be re-established. This exact same solution was used a few years later, when the SMC1 jukebox was introduced. The other approach, which I settled on, was to directly encode the carriage position into a 7-bit number (80 records is greater than 6bits {=64 max} but less than 7 bits {=128 max}). The direction of travel would be the eighth bit for a total of 160 record sides. There were rotary encoders available that could be screwed to the carriage, and geared to the rack position. But we would then have to route seven more wires through the mechanism cable, which was already getting too stiff. Instead, I came up with an encoding PC board that would be screwed to the magazine. A set of contacts would move with the carriage, applying a ground to the PC board contacts. The question now was how to encode the PC board. There were several options here. You could use a straight binary code, which would require seven contacts but no logic. Or, you could use two contacts for a decimal code, but that would require some logic to interface the memory. The binary code was attractive since it used no logic, but had the problem of mis-alignment causing incorrect codes and therefore wrong selections. Using a gray code (where only one bit changes from one position to the next) would solve this problem, but also required logic to convert the code. I finally settled on the decimal version, since only two contacts were needed. With some conversion logic, this could also be used for the now playing indicator. In fact, the encoding PC board was used (without the logic) for the now playing indicator on the later jukeboxes. This replaced the phenolic boards with staked contacts and wiring harness of the earlier black & gray box machines.

I had completed the design and was in the middle of building a prototype when fate intervened. From time to time, Seeburg would delay paying certain suppliers, when money was short. The middle ‘70s was no exception. I think most jukebox manufacturers were having tough times then. This occurred once too often with the supplier of the credit paddle switch used in the Single Pricing Unit (SPU5) for the 100-selection jukebox. They responded by refusing to ship any product until they were paid, meaning that we could not build any 100-selection machines. This switch was the heart of the SPU5. Whenever a coin switch closed, one of the paddles was flipped to the on position by the coin solenoid. This completed a circuit permitting the customer to make a selection. As the selection was written in, the switch rotated about 60 degrees by the cancel solenoid, driving the switches up a ramp to turn each off. When all switches were off, there were no more credits and no more selections could be made. The Chief Engineer asked me if I could get the credit portion of my digital design ready for production. I responded, “Sure, but why not go for the whole thing except RAM selection memory?” He gave his permission, and I had another real project to work on.

The CSP1 design was based on the black & gray box design, but greatly simplified. Since there were only 100 selections, we only needed two digits instead of the three used for the 160 machines. The black box would accumulate up to 31 credits. For the red box, we limited this to 15, to save on chips. The red box combined the functionality of the black and gray boxes into one and at the same time addressed some of the problems that had been found with several years of black and gray box production.

The first problem addressed was the blown-out lamp driver transistor problem that had plagued us in QA. I used the same technique as I had in the design of the 5BS1 lamp driver. As the current increases through the driver transistor, it tends to shut itself off. A short circuit results in increased current flow, but not a destroyed transistor.

The next problem was the intermittent diodes on the pricing programmer board. I addressed this problem by putting the diodes inside the box, so that we would use less of them and thus would not have to perform the specialized tests on so many parts. By that time, this problem had been pretty much solved by eliminating the DO35 package option from the part drawing. Instead of having diodes on the pricing board to control credit addition, there were only jumpers. Since the 100-selection machine only handled singles, we didn’t have to be able to handle subtracting more that one credit per selection, so this also simplified the circuitry.

Another desire was to eliminate as many components as possible from the write-in circuitry. The gray box uses three series-connected loops to write in a toroid. The first loop is the hundreds digit, which selects one of two paths through the units loops using a pair of SCSs (Seeburg’s name for an SCR). There are ten units loops, each of which has two hundreds paths through it. Each of the ten is controlled by a series-connected SCS. Next, there are eight tens loops, each of which is controlled by a series-connected SCS. Finally, there is the main trigger SCS, which acts as a switch connecting the write-in capacitor to the rest of the SCSs. So what the gray box does is to ‘program’ all of these various loops to provide a current path through the selected toroid to write it in. Since the loops are all in series, there has to be a level-shifter for each circuit, to shift the SCS trigger pulse from the level used by the custom decoder chip to the level required by the SCS. Each level shifter consists of a resistor, capacitor and diode, for a total of sixty-some parts. If I used parallel paths through the Tormat instead of series paths, I could eliminate all those parts and get it to fit in one box.

Finally, there were some problems identified in early gray boxes, which would cause an intermittent random selection to written in at random intervals. This proved very difficult to track down, since it only happened on rare occasion. Due to some diligent effort on the part of the field service group, QA and the Engineering Department, it was discovered that some SCSs used in the main trigger circuit exhibited a problem called ‘channeling’. This occurs sometimes when an SCS has a constant voltage applied to its anode. Since the main trigger SCS anode is connected directly to the write-in capacitor, this exact situation exists any time the jukebox is powered up and no selections have been made for a while. The easiest way around this problem was to eliminate the main trigger SCS altogether. I replaced it with a write-in relay. In the 160-selection machines, it was solved by performing specialized tests on the parts in Incoming Inspection, meaning that they fell under a different part number.

Since the CSP1 logic worked at +13 Volts, it made sense to use a 12 VDC relay for write-in. The design would thus use two relays, one for write-in and the other for trip. However, the trip circuit worked at +27 Volts, and I wanted to insure that both relays were operated from the same voltage so that they would be interchangeable. If not, I knew the chances of plugging the wrong relay into a socket were about 100%. Simply changing the trip circuit to work at +12 VDC would not work, since the gate trigger sensitivity of the trip SCS is partially determined by the anode voltage, which would have to change from +27 VDC to +12 VDC. My solution was to trick the trip SCS into thinking it was working at +27 VDC, but really run it off the +12 VDC supply. This was implemented simply using a resistor to the +27 VDC supply and an additional diode to isolate the +13 VDC supply from the +27VDC supply. The resistor connected the SCS anode to the +27 VDC supply. The additional diode was used to connect the SCS anode to the +13 VDC supply through the trip relay. This diode would be reverse-biased most of the time due to the +27 VDC supply applied to its cathode through the resistor. This set the SCS anode voltage to +27 VDC for sensing. Once the SCS fired, the resistor would de-couple the anode from +27 VDC, and the relay energized through the SCS. The blocking diode also served to isolate the SCS from the relay holding contacts, to insure that the SCS turned off at the correct time. The blocking diode was carried over from the 160-seleciton machines.

The red box was also designed to accept selections from a DEC-style Consolette, using the DCT1 Remote Translator. It made sense to design both the CSP1 and the DCT1 at the same time, since they were to work hand in hand. The salesman who sold us the black and gray box plastic came in one day with samples of the boxes molded in red and blue plastic. We immediately decided that the CSP1 would go in the red box. A later project (discussed below) was earmarked to go into the blue box, but when that was cancelled, the DCT1 (and later, the DMT1, translator for the SMC1 jukebox) went into the blue box.

It was right about this time I finished college, and started working again full time.

I spent the next couple of months getting the red box ready for production. This entailed the standard sort of thing. Work with the PC designer, get together a bill-of-materials, prepare new piece part drawings, test specifications, cable diagrams, and all the miscellaneous things that need to be done to get a project ready for production. I also had a chance to work with the cabinet guys on cable routing and the placement of the red box. Since the design included a new chassis (the PPC1) I had to get everything together for that, too, including drawings for the chassis, which located all the holes for the power transformer, relay sockets, terminal strips, etc. For this drawing, I worked with one of the mechanical guys and also with the guys from Production Engineering to develop the best parts placement. Production Engineering had the responsibility of actually getting a product into production; insuring that a product was as simple as possible to build, test, etc. They were also responsible for actually setting up the production line, and determined the order in which a product was built, how many people it took, etc. This meant that whenever the daily build rate changed, they had to add or delete people from the line, and figure out how many operations were required of each person to make the new quantity of product. This part of their job was called ‘time study’.

Since Production Engineering was responsible for everything that happened in production, their busiest time was during the line set up for a new machine, which happened every summer. Most if not all of the hourly employees were basically laid off for this period, which lasted between two weeks and a month. The longer time period tended to occur when sales were bad.

Yearly Inventory

During this same time period the majority of the salaried employees (and some senior hourly) were required to help with the plant inventory. The stock room occupied most of the second floor of the new building, and was pretty dark and dirty. The shelves usually were three or four high, and went from floor to ceiling. In order to get at the items on the higher shelves, there were wheeled ladders available, which would put your feet about 6 or 8 feet above floor level. Each group of shelves extended for about 30 or 40 feet. The engineers were assigned as group leaders for each counting group, which was generally about four or so people. The reason why the engineers were assigned as group managers was that they had the best chance of identifying a part when there was some question as to what the part number was. All of Seeburg’s parts were documented on microfiche (small pieces of film embedded within a cardboard holder on which was printed the part number, etc.), and there were several readers available so that you could look up a part. There were thousands of different parts, some of which had not been used for many years, like certain cabinet parts left over from a production run that ended years before. The number of different parts on a shelf was determined by their size, so bigger parts meant that the shelves held fewer different part numbers. Large, expensive parts were counted individually by one person and checked by another. Smaller parts in large quantity were counted by weight. You would dump most of the parts onto one side of a scale, and then start moving some of these to the other side, until you got the scale to balance. The count was then equal to the sample quantity multiplied by whatever ratio the scale was set for. When you started a new shelf, you were handed a deck of cards, which supposedly related to all parts on the shelf. If there was any plan to how the stock room was organized, I was never made aware of it. The only part of the organization that made any sense was that all of the semiconductors and pickups and any other small, expensive and easy to steal parts were kept in a caged, locked area off to one side of the main stockroom. We did this each summer, when it was hot and humid in Chicago. The stockroom was not air conditioned, so it was hot, dirty work.

Lunches

Three days a week, two of us would take lunch orders from the rest of us in Engineering, and bring back food. On Tuesday, we would go the Vienna Sausage factory, which was about two miles away from the office. We would get fresh Corned Beef or Hot Pastrami sandwiches. Thursday was gyros day. Seeburg was located about two miles from Chicago’s New Town area, which had several Greek carryout restaurants. Everyone except our secretary ordered gyros. The entire office reeked of gyros for the rest of the afternoon. Since our secretary couldn’t stand gyros, our normal procedure was to sneak the wrappers into her trash can when she left her desk. Her cussing could be heard for miles, but she usually took it with good humor. Friday was fried shrimp or fish & chips day. We used to get these from a small place down by the Chicago River. We never found out where the fish or shrimp came from. Hopefully, it was not from that river. The seafood was packed in brown paper bags for us to bring back. By the time we got back to the office, the bags were transparent. That’s how you could tell if the fish was good or not. If you couldn’t see the fish through the bag, it wasn’t any good. Later on, I used the same criteria to determine if a doughnut was good or not. The good ones made the bag transparent, just like the fish. We called these ‘belly-bombs’. Mondays and Wednesdays was either lunch in the cafeteria, or bring your own.

Williams Pinball

After getting the red box to the point where it was very close to being ready for release to production, I got pulled off to work on a Williams pinball machine project. By this time Bally had already released their microprocessor-based pinball system, so there was a lot of interest at Williams to develop a system of their own. Evidently, Williams decided to play all ends against the middle, because there were actually three different development programs set up by Williams/Seeburg management. The first, and ultimately successful, program was staffed by a new group of engineers. This group was actually housed at the Williams offices on California Avenue in Chicago and developed a system based on the Motorola 6800. Seeburg and Williams senior management were given a presentation by National Semiconductor, Inc., showing them how nifty their SC/MP (Single Chip Micro Processor or SCAMP) was. Management went for it, and decided that they would set up a group within Seeburg (of which I was made a member) to develop a system based on that part. National also developed a system internally, perhaps as a backup in case ours failed, or as a system they could offer to the other pinball companies without any strings attached to our development. Our management thought that it would be a good idea for us to develop (and of course build) a system that we could sell to Williams. We could use our electronic design and production expertise and get an additional product line out of it. This thinking evidently went back to the electronic video game experience between Seeburg and Williams.

I was assigned the task of designing the hardware controller, which would include the CPU, RAM, ROM/PROM, control interface to the lamps and solenoids, and displays. My supervisor took over the task of designing the actual lamp and solenoid driver circuitry. Since none of us had ever written any software, an outside consultant (recommended by National Semiconductor) was brought in for that part of the job. He lived in one of the suburbs of Milwaukee, WI, and was affiliated with Marquette University in that city. We built two prototypes, based on Williams’ Space Mission pinball machine, which was in production at that time. As the project progressed, we also built a prototype based on Williams’ Grand Prix. Both machines we designed by pinball legend Steve Kordek, whom I had a chance to work closely with during this period. He is a fantastic guy, and knows all there is to know about pinball.

Since we were on this colored box kick, we were directed to fit the hardware controller into a box of the same size as was used for the black, gray, and red boxes. This time it was blue. Fitting all this circuitry into the box was pretty difficult to do, since all the chips came in DIP (Dual Inline Package) packages at that time, and we also had to use ½ watt resistors everywhere, since this was the Seeburg standard. When the project was cancelled, there were a lot of blue boxes left in inventory, so they ended up being used for the red box (DCT1) and MCU (DMT1) translators.

For the control software, the idea was to put all of the control code into a single ROM, which would be contained within the blue box. This ROM would contain all the various subroutines needed to run the game, count coins, control displays, lamps, and solenoids, etc. Game-specific code (award 100 points for this spinner and move a lamp along a sequence as a result, etc., basically a pinball interpretive language or rules code) would be programmed into a smaller ‘game personality’ PROM, which would be mounted on a programmer board similar to (but significantly bigger than) the black box pricing programmer board, which would plug into the back of the blue box. The consultant wrote all of the ROM code, and he and I developed the interpreter codes. By using two different games, we thought we could cover most, if not all, possible situations that could occur in pinball rules (pretty naïve thinking, as I look back on it). Using this technique, Seeburg could continuously build the same box, which would be used for all games by merely changing the game personality board. That this approach does not work is evidenced by the fact that none of the major pinball manufacturers ever used it. These were the days prior to the introduction of the Personal Computer. But we needed computer to enter code into and have it do our software assemblies. There was no way the bean counters were going to give Engineering access to their mainframe computer, so we had to rent time on another computer, communicating via dial-up modem. This meant that whenever the consultant made a change to the code, he had to type it into his source code (stored on the Marquette University computer) using a telephone modem. The modems of today plugged into a PC were unknown. We had a teletype machine with a paper tape punch and reader, and an acoustic coupler. When you wanted to make a change, you dialed up the computer, put the telephone handset into the coupler, and started typing away on the teletype machine. After an hour or so, the computer finished the code assembly, after which we downloaded the assembled code, and saved it on punched paper tape. Luckily, I had become an expert at winding these tapes so that it would not snag, a trick I had learned while helping out in the radio shack during off hours on the ship on which I served while in the Navy. After all this, we could load it up into an Emulation platform we bought from National and run it. Once we were happy with the code, we would make a set of PROMs and play it.

My supervisor designed the actual lamp and solenoid driver circuitry. He hit upon the idea of designing a pair of small driver boards, one for each device type. The driver board would then be mounted under the playfield, close to the device it was driving. He also came up with variations having two or three of each driver type on a single PC board, used in areas of the playfield where there were a lot of lamps or solenoids in the same immediate area. This approach worked well, in that it was obvious which circuitry was driving a given lamp or solenoid, but greatly added to the wiring congestion under the playfield, since each board had to have a control signal, +5 volts power, Ground, and driven signal out connection. This was especially cumbersome for lamps; since we had not hit upon the idea of time-division multiplexing the lamps as was done in the Bally and (later) Williams’ hardware systems. This meant that there were a lot of wires going between the playfield and the blue box (one each per lamp and solenoid). The driver board scheme would probably have been popular with the game mechanics, since they could carry several replacement boards in their pocket.

By far the biggest problem we had with this system was the display drivers. The Bally system used gas-discharge digital displays to replace the old score reels. Their display boards used a lot of transistor level-shifters and drivers, since the displays needed a power supply of around +190 VDC. We thought we could do it cheaper by using a pair of chips recently introduced by Signetics (which would later become part of Philips Semiconductor). One chip was used as a digit driver, while the other was used to drive the segments. We had all kinds of problems with these chips, which were advertised as being a glue-less (i.e., no components in-between) connection to TTL logic. They were very temperamental, and would self-destruct if you looked at them sideways. We ended up re-designing their PC boards several times, adding isolation diodes, resistors, etc. Finally, as I recall, Signetics gave up and removed them from the market.

Eventually, the consultant and I had both machines working fairly well, except those troublesome display boards. About this time, a new vice-president was hired at Seeburg/Williams. The word we got was that he was to examine the corporation from a technology point of view, and make recommendations for making improvements. This proved to be a stepping-stone position for him, as he later became president of Williams. Prior to doing so, he discovered that I had never signed a patent agreement with Seeburg, and demanded I do so. I was a pretty headstrong engineer in those days, thinking that I was just a week or so away from an idea that would make me an instant millionaire. All these years later, I’m still just a week or so away. I looked at the standard agreement, and refused to sign it, since it was entirely one-sided in favor of Seeburg. Basically, by accepting continued employment at Seeburg I was willing to relinquish all rights I had to anything I came up with, whether or not it had anything to do with Seeburg’s line of business, and whether or not I came up with the idea on Seeburg’s time or my own. Nowadays, such an agreement is clearly un-enforceable, especially in the state of California (where I now live), but it was standard practice back then. I really enjoyed working at Seeburg, and didn’t want to leave, but I also didn’t want to give up any patent rights to anything I independently came up with, so I had a lawyer friend of mine start with their agreement, and come up with something a little more equitable for both sides. This was done at my expense. I had a lot of support from my boss and the Vice President of QA in this matter, who tried to negotiate a mutually-agreeable settlement between management and myself. Evidently, both thought I was a valuable employee and didn’t want to lose me. I submitted the revised agreement to the new ‘VP of Technology’ who basically said ‘sign the standard agreement or leave’. So I left Seeburg for the second time. The VP of QA and the entire Engineering Department came to my going-away luncheon, which was held at the famous Golden Ox restaurant, just around the corner from Seeburg’s main plant. I went to the last Seeburg reunion held, which was in 1996 in Chicago. While in town, I had a chance to visit the old plant and had lunch at the Golden Ox. I went down into the basement party room where they held my going-away party back in January of 1977. Off in the corner of the room was an STD2 Entertainer, with the lamps blinking merrily away. Click here to go to the next installment.

class="F4 Bold FontArial Black"