back from the service

I left Seeburg in the late spring of 1966, went to Navy boot camp at Great Lakes, IL, followed by Electronics Technician (ET)‘A’ school there, Nuclear Power school in Vallejo, CA, and prototype nuclear reactor plant training at the S1W facility outside of Idaho Falls, ID. I enlisted for six years, to learn how to become a Nuclear Reactor Operator. I qualified, but did not want to go on submarines, so I ended up as the leading ET on a brand-new Attack Cargo Ship (LKA) out of Norfolk, VA for the rest of my tour. You might ask what in the world an LKA is. It is an amphibious cargo ship, used to transport dry stores and vehicles to support a marine invasion force, hence the name ‘attack’. When asked what the ‘Attack’ meant, we used to joke that we would pull up alongside the enemy and throw landing craft at them. Six years after joining, I was again a civilian, but now I was a qualified Electronics Technician and a qualified Nuclear Reactor Operator, having qualified on the prototype reactor plant for the USS Nautilus, the first nuclear-powered submarine.

The first three months after getting out of the Navy were spent partying and just plain goofing-off. After that amount of time doing nothing, I got bored, so I went out looking for a job. Just for fun, I interviewed at Seeburg, looking for a technician job. I was accepted, and started working in the Quality Assurance Lab. From the time I had left Seeburg till now, the VP of Quality Assurance had changed, and my mother no longer worked there, having retired.

Seeburg had just started building the SPS160 model jukebox at this time, and had been experiencing intermittent problems with credits on the black box, which was starting its fourth model year in production. The credit problem was traced to intermittent shorted or open diodes, used on the Pricing Programmer Board in the black box. For credit addition, the diodes determine how many credits are added for each coin switch closure. For credit subtraction, the presence and absence of diodes in the subtract path determines how many credits are subtracted. So an intermittently shorted or open diode could exhibit some very strange symptoms. My first task was to put together a test rig that would permit these diodes to be quickly checked. Since the diodes came on large reels with thousands of parts on each reel, I set up a rig which was based on a couple of mechanism magazine end castings, used as pivots to let a feed reel and a take-up reel rotate. I then put together a ‘device under test’ fixture, which consisted of a couple of spring steel contacts which in turn were connected to a curve tracer. We would run a reel of diodes through the test three times. First, we would heat shock the diodes using a heat blower. Next, we would cold shock them using freeze spray. Finally, we would physically shock them by tapping with a pencil eraser. All this, while watching the curve tracer for any sign of an open or shorted diode.

The curve tracer would be set up to sweep the diode out to its reverse breakdown voltage, with the current limited to a value which would not destroy the diode. The diodes used (type 1N4148, Seeburg part number 61-309481) had a reverse breakdown voltage of 40 Volts. When properly set up, the curve tracer display looks like a backwards capital ‘L’, with the horizontal part of the ‘L’ depicting the reverse biasing voltage and the vertical part depicting the reverse current. Reverse Breakdown is exhibited when the horizontal trace snaps vertical. For good diodes, it is a nice sharp snap upwards. Questionable diodes have a soft knee going into breakdown. If, while applying the temperature and physical stresses to the diode under test, the curve tracer display went either totally horizontal (open) or totally vertical (shorted), we knew we had a bad diode and would cut it out of the reel. We usually found problems with the small DO35 packaged diodes. The diode used came in this or the DO41 package, which is bigger, or the DO7 package, which is larger still. The DO35 package has a couple of metal ‘buttons’ which connect to the leads on one side and the diode die on the other. Looking at a bad diode through a microscope usually revealed a die that was not properly mounted between the buttons. It was usually crooked, or you might find a cracked die or maybe even two in the package. The DO41 package used a whisker to make contact between the ‘button’ and one side of the die, which would relieve any stress applied to the die. The smaller DO35 package had no way of relieving stress on the die, since it was pressed between the two ‘buttons’, and held in place by the glass package. We found several thousand defective diodes this way, which helped improve the reliability of the black box tremendously.

About this time, we started getting many defective SHP1 amplifiers back from the field. They all had the same symptom: burned driver emitter resistors, burned or shorted driver transistors, and shorted output transistors. Many of the driver PC boards were also scorched. Yet all of these amplifiers had been 100% tested on the amplifier production line and worked fine. We spent several weeks diagnosing this problem. The difficulty in finding the cause was the fact that so many of the components had been destroyed. It was hard to determine which one went first, taking the rest with it. I walked into the lab one day to find an SPS-160 jukebox (the SHP1 had been introduced with the SPS-160) sitting there, with everything above the titlestrip holder burned. When that amplifier failed, it took much of the juke (and, from what I was told, part of the building it was in) with it. I remember spending many hours removing parts from the amps and testing them, but I never came up with the answer. We started testing amplifiers from the line, doing everything we could think of to them, trying to get them to fail. No luck.

It was my supervisor who did come up with an explanation, and a solution. Solid-state amplifiers using the quasi-complementary output structure (where the output transistors are of the same type, connected in series across the power supply and with the output taken at the junction of the two output transistors, the emitter of the upper transistor and the collector of the lower for NPN output transistors), need to have a biasing current flowing through the transistors when no signal is applied. The reason for this is that the bipolar-construction output transistors used are non-linear when very little current flows through them, exhibiting what is called ‘cross-over distortion’. A ‘bump’ can be seen in the output waveform, just as the output goes from negative to positive, and vice-versa. This is just as one transistor turns off, and the other on. The solution is to not fully turn the transistor off, to avoid this non-linear region. The amount of bias current through the output transistors should change as a result of the temperature of the output transistor. As temperature increases (due to the ambient, or a record being played at high volume) the bias current should decrease, and vice-versa. To make the bias current correctly track the output transistor temperature, a sensing transistor (later replaced by a thermister and dual diode) were mounted up under the heatsink. The earlier TSA series amplifiers featured a bias current adjustment right on the driver board to control the amount of quiescent current flowing through the output transistors, and no remote temperature sensing of the output transistors. The SHP1 was the first amplifier to mount the bias adjustment pots on the chassis, instead of on the driver board like the TSAs. All this required wiring between the heatsink, pots, and the driver PC board, and a connector for the PC board. To improve connection reliability, Seeburg used what was called a ‘bifurcated’ contact on their PC board edge connector cable assemblies. This was a split contact, using the theory that if one of the two segments lost connectivity, the other wouldn’t. Unfortunately, there were times when both contacts were open, which would cause one or both drivers to turn fully on, turning the outputs fully on, resulting in the driver emitter resistors burning up. Since the driver and output stages of the amplifier are DC coupled (there are no coupling capacitors between stages) a catastrophic problem in one stage takes everything else with it.

As a fix, Quality Assurance immediately required the production line to hard-wire connections in parallel with the four (two per channel) heatsink sensor and bias pot connections to the driver PC board. Small holes were drilled into the PC board, into which wires were inserted and soldered from the opposite (foil) side. These wires were then crimped into the contacts making up the edge connector. This solved the problem, so if you have an SHP amplifier without these connections, you should seriously think about making them. Later on, Seeburg Engineering modified the PC board to include little solder cups and made the wiring harness change official to deal with the problem. At the same time, they changed the bias sense circuit to delete the transistor sensor, and replace it with the dual diode circuit, which is the ‘classical’ design approach. I was told that the transistor circuit was originally used to get around a patent that someone had on that circuit arrangement.

One of the guys brought in an amplifier that had just come off the line. He was going to check it out. He connected it to the dummy loads, and plugged it in to the wall outlet. He heard a hissing noise, and put his head down next to the amplifier to identify the source of the noise. Just then, an electrolytic capacitor, which had been installed backwards, exploded. Sure woke him up!

We started getting some video game boards in for repair, and I was assigned to work on them. This was my first introduction to video games, and I was very interested in how they worked. Williams Electronics, Inc., was part of Seeburg at the time, and they evidently jumped on the video game bandwagon along with several other game manufacturers. This was the Pong era, with lots of knock-off games around. Williams’ version was called Pro-Tennis, and used a large PC board filled with about 80 or so TTL chips to generate the video and beep-bop-boop audio. Williams did not assemble these boards, since they were not set up to do any of this at the time (their microprocessor-based pinball machine was still in the future). So it follows that they were not set up to trouble shoot non-working boards. Later on, a consultant came in and modified the design somewhat, resulting in the Pro-Hockey game. This game used four smaller PC boards which plugged into an interconnect board, and used ROMs to generate the goal patterns on the screen. Otherwise, the game was identical to Pro-Tennis. Seeburg assembled these boards for Williams, so it made sense for us to repair the ones that didn’t work. Along with this being my first introduction to video games, it was my first opportunity to work with TTL logic chips. Quite interesting, I thought.

I always use an oscilloscope while troubleshooting. It’s easier for me to see what’s going on. I don’t like to guess what the waveform looks like at any specific point in the circuit, I’d rather know for sure. But after trouble shooting a fair number of boards, I got to know how they worked very well. The most difficult circuit to analyze was the one responsible for puck movement. It was pretty tricky, using several counters loading to different values at different times, but I eventually figured out how it worked. After awhile, I got so good at trouble shooting that there were times when I didn’t even use a scope or voltmeter, just looked at the video screen, looked at the schematic, and decided what chip must be bad. This worked many times, but there were also times when I had to really dig to find the problem.

Working on the Williams video games didn’t last too long. I think the market was saturated with Pong knockoffs and most game makers got out of it. So I started working on black and gray boxes. Seeburg had introduced this new credit and selection system several years earlier, with the LS3 jukebox. The black box took care of credit and selection entry, encoding, and transmission, and used a pair of custom chips, developed by Seeburg, to perform this function. The gray box took care of selection decoding, writing the selection into the Tormat memory, and starting the mechanism. It had one custom chip. The idea behind the selection encoding and decoding process was that a common system could be used in both the jukebox and consolettes, and that the data could be sent from the consolette to the jukebox very quickly to avoid one consolette transmission beating over another, a situation which happened from time to time with the earlier, single-wire stepper system. Each jukebox and each consolette had a black box. Only the jukebox or hideaway had a gray box.

Each box was a sealed unit, and carried a three-year warranty. Service people were expected to carry spares with them, and repair jukeboxes in the field by replacing whichever box didn’t work with their spare, followed by sending the bad box back to the factory for repair. Several of the larger distributors were provided with field testers, to verify that the box was bad before returning it. Besides the custom chips, each box had a board full of interface circuitry. For the black box, this circuitry mainly consisted of pulse-stretchers (i.e., one-shot multivibrators) and lamp drivers. For the gray box, there were a series of SCSs (Silicon Controller Switches, the name used by Seeburg for an SCR) used to pass current through the appropriate Tormat write-in loops. Precisely what the three custom chips did was a well-kept secret. This was evidently done so that none of Seeburg’s competitors could get any inside information on what they did. Very few people within Seeburg were permitted to have copies of the logic diagrams for these chips. People high enough on the ‘food chain’ included the Chief Engineer, and the Purchasing and Legal Departments. The Section Engineer who had design and support responsibility was noticeably absent from this list, even though copies were submitted to him for approval of the logic before the chips were released for production. When he pointed out the fact that he could not do his job if he could not have copies of the logic diagrams in his possession, the ‘powers that be’ relented and added his name to the list. It was the responsibility of the Quality Assurance department to repair all boxes returned from the field. We would fix, test, burn in and return them to the various customers, and keep a record of each one, so that we could spot any failure trends. But, we were not allowed to have copies of the logic diagrams.

There were three of us working on the boxes. Two of us fixed black boxes, since their usage and failures were more numerous. The other person worked on gray boxes. We were each issued a set of the pertinent schematics, but no logic diagrams for the custom chips. Each of us had a field tester (Seeburg type DSST1), which was built by the Test Equipment Engineering Department. It was a chassis which included a control center, a set of switches for coins, etc., lights for Single Credit, Album Credit, etc., and connectors for a selector, a black box, and a gray box. You would plug a known good box of the type you weren’t working on into its connector, and use the system to troubleshoot the other box. There was a set of lamps running across the middle of the tester, which came from a now playing indicator. These lights would display which selection you made, so they basically replaced the Tormat and mechanism. We worked up a test routine to exercise each box to verify proper operation of the credit circuits, selections, etc.

Most of the problems with the black box had to do with shorted lamp driver transistors. If, for some reason, one of the lamps in the jukebox instruction window shorted, it would usually take the transistor inside the black box with it. The usual reason was that the repairman replaced the bulb with one requiring too much current. If the repairman didn’t check the bulb before replacing the black box, the replacement black box would have the same transistor blown out. Incorrect credit or selections was generally caused by a problem in one of the one-shots on the interface board. I got pretty good at fixing black boxes, so I went moved over to the gray box, learned it, and then alternated from one to the other.

The gray box was more difficult to trouble shoot, since it configures many parallel paths through the Tormat. Only one of the eight or ten SCSs in each parallel path is supposed to fire. Sometimes none of them would fire. Other times, more than one would fire. Sometimes you couldn’t find anything wrong, and had to make an educated guess based on the symptoms given in the write-up that came with the box. Intermittent incorrect selections were especially hard to trouble shoot. If we could find nothing wrong, we would tap on all the SCSs and diodes, along with heating and cooling them and the Decoder custom chip while making selections. In other words, we really beat on the box trying to get something to fail. If it still would not do anything wrong, we usually replaced the Main Trigger SCS, since this part was generally the cause of some of the stranger problems. The custom chips themselves were seldom replaced. Once a chip was soldered onto a board and made it through all the testing, it usually lasted.

Every custom chip received at Seeburg was tested in Incoming Inspection using a purpose-built, specialized tester. After being soldered onto a PC board, the assembly with the custom chip was tested again. The completed box was tested yet again. The latter two tests used a huge tester at the end of the sub-assembly line. This machine was easily 10 feet on a side, and about 6 feet tall, based on some super-duper mini-computer, which was state-of-the-art for the time. It always ran, even at night and on weekends. There was one engineer whose sole responsibility was keeping this ‘monster tester’ as everyone called it, running. After we completed the repair of a box, it was tested on the ‘monster’, after which it was returned to us. We would then put a new warranty seal on it, with the letters SRT (Seeburg Reliability Tested) stamped on the seal. Any box having SRT stamped on the warranty seal went through Quality Assurance at least once.

Seeburg used two suppliers for the custom chips, first General Instruments (GI) and later American MicroSystems, Inc. (AMI). The GI chips were in square ceramic packages, while early AMI chips came in 40 pin ceramic (white) DIPs (Dual Inline Packages). Later AMI chips came in black plastic DIPs. Whenever we came across a board mounting a GI chip, we would automatically replace it with a board having a newer AMI chip. There were lots of reliability problems with the GI chips. Also, due to some confusion on the part of Engineering, the first few boards built had the GI chips actually mounted on the bottom of the PC board. The drawing provided by GI evidently showed the bottom-side view of the packaged chip, and the board was laid out to accept that pinout. And, there was a flip-flop deleted from the first few lots of Pricing chips from GI, so an external flip-flop in a small metal can was added to the early Pricing Logic PC boards. Due to this replacement policy, there are very few boxes with square-chip GI chips out there today.

It was right about this time I started going to college, and brought down my hours at Seeburg until I was only working there part-time. I would arrange my course schedule so that I could spend the late afternoon and part of the evening working at Seeburg. When my class schedule permitted, I did take the occasional full day off from school to work full time. This would typically happen when I was running low on money, since my technician job at Seeburg still paid hourly. On one of these days, I walked into Seeburg and decided that I would spend the day knocking out as many black boxes as I could. I believe I still hold the world’s record of 75 black boxes fixed in one day. Obviously, most of the boxes had problems that were easy to find.

I also spent some time repairing TSA9 amplifiers after the day shift went home. That way, I could use the test rig out on the production line to analyze and repair the amplifiers. I was asked to help out, since they were getting a fair number of amplifier rejects and the daytime repairman was having problems keeping up, probably due to the many SHP amplifier defects. The same line built SHP1s and TSA9s until the TSA9 was discontinued when the 100-77D (Topaz) jukebox came out. The TSA9 was replaced by the SHP2 in this model. I remember the daytime repairman being mad at me because I was fixing more amplifiers than he was.

It was also during this time I decided that I should draw up my own logic diagram for each of the custom chips used in the black and gray boxes. This would be useful as a diagnosis aid for me, and for the other technicians working on the boxes. When the drawings were complete, my supervisor passed on copies to Seeburg’s Chief Engineer, who passed them on to the Section Engineer responsible for the black and gray box design. I believe that the correctness of my logic reconstruction made it possible for me to later move on to become a junior engineer, working in the Engineering Department, even before I finished college.

My supervisor and I had spent some time musing about how much better the FC1 Regency jukebox would look with a color organ. This jukebox was a low production console-type machine, finished in black, with a pair of lighted panels, one running across the top and the other across the bottom. Each panel had many small 2182 ‘grain of wheat’ lamps behind a black-painted perforated panel. This gave the lamp a halo-effect, which apparently moved as you walked by. We got permission from the Vice President of Quality Assurance to modify one of the machines, since they were not selling too well in the market place. We came up with a circuit, which I built a prototype of. We divided the bulbs into four groups. The first group was left with the original clear color. This group numbered just over ¼ of the bulb total, and would be on when there was no music, retaining the motion effect of the original box. We applied red, green and blue gel filters to equal numbers of the remaining bulbs. The control circuit input audio from the top of the amplifier volume control, since the volume level at that point is pretty constant, due to the AGC circuit in the amplifier. The controller sensed the presence of audio, and if detected (i.e., a record is playing), turned off the white lights. Next, an active filter took over to split the audio into bass, midrange and treble frequency ranges which would light the red, green, and blue bulbs, respectively. Since an auxiliary AC transformer powered the bulbs, we used the same triac as was used in the Income Totalizer as the switching element. This proved to be a marginal part to use, since the amount of current required for the lamps was close to the maximum rating for the part.