Archive for the ‘Technical’ Category

16-Mar-09: One Step Foward… [More Wang 370]

March 16, 2009

In the last post concerning the Wang 370, the 371 punched card reader was checked out and appears to be generally healthy. In spite of this, the 370 still has problems with hanging when simple program-related functions are attempted. After looking through the logic, the first place to check is the master clock and timing chain.

The bottom cover of the 370 was taken off, providing access to the backplane connectors. This would be the best place to connect an oscilloscope to monitor signals on the backplane.

One Logiblock (#571 – System Control) contains the master clock oscillator and the three timing chain flip flops connected as a simple three bit binary counter. A scope probe was placed on the backplane pin containing the master clock signal. The 370 was disconnected from the 360E electronics package, and the 370 powered up. Immediately, the oscilloscope showed a reasonably clean, symmetrical square wave at a frequency of about 36KHz. It was clear that the master oscillator was running just fine. Next, the probe was moved to the output of the first stage of the timing chain. Interestingly, the output was static, with only tiny (a few millivolts) transients occurring occasionally. If the flip flop was working properly, this output should be a square wave running at half the frequency of the master clock. Clearly, it wasn’t. This in itself was a big clue. If the timing chain isn’t running, the logic states that sequence the 370’s logic can’t operate properly. This could well be the reason why the 370 hangs when trying to perform a simple single-step operation.

Just out of curiosity, the outputs of the two other timing chain flip flops were checked, and they too were static, which would make sense, as these flip flops are clocked by the action of the first flip flop.

The outputs of the timing chain flip flops are passed across the backplane to another Logibloc (#570 – System Timer) that contains, among other things, combinatorial logic that decodes the outputs of the timing chain flip flops into a series of timing signals that choreograph the operations of the 370’s functions. To see if perhaps there might be a fault on this board, the board was removed from the backplane, and the 370 powered back up. This time, each of the three timing chain flip flop outputs (T1, T2, and T3) were wiggling the way they should, with the first stage output at half the speed of the master clock, the second stage at one quarter the rate of the clock, and the last stage at 1/8th the master clock frequency.

It seems that something in the System Timer Logibloc was interfering with the output of the first stage of the timing chain. The combinatorial logic on the System Timer board consists of diode-resistor networks configured as AND gates, with transistor stages providing inversion and/or buffering of the gate outputs. There are a number of these logic networks that decode various states of the timing chain flip flops. All of the diodes in the gating networks were tested in-circuit using the diode check mode of the DVM, and they all appeared to be OK. Resistors were measured for proper resistance values, and all were within spec. So, attention was turned to the transistors. A shorted transistor could easily drag down the output(s) of the timing chain flip flops.

Testing transistors in-circuit can be a hit or miss proposition. It is much better to remove the transistors from the circuit, and test them individually. The difficulty with this is that the Wang Logiblocs were hand-assembled. At the time this 370 was manufactured (1969), automated component insertion equipment was very rare. Thus, components were manually inserted into the circuit boards, with the leads bent over on the wire side of the circuit boards to temporarily secure the parts on the board. Once all of the components were installed, the boards were flow-soldered. This makes component removal a little difficult, because with the component leads bent over, the solder needs to be removed, then the leads unbent carefully, and then any residual solder removed while the component is very carefully extracted from the circuit board. A high quality temperature controlled soldering iron, combined with a vacuum solder-removal system makes it possible to remote components, but even with these tools, component extraction is still tedious and time-consuming.

It was decided to remove the seven transistors involved in the timing chain decoding logic and test each of the transistors using a transistor tester.

All seven transistors (RCA-made 2N404 Germanium PNP transistors) were carefully removed from the circuit board, and each was tested. As it turned out, one transistor was found to be bad. It had a short between collector and base that applied -11V to the output of the diode gating network. This served to drag the output of the T1 timing chain flip flop down to near -11V. This was certainly the reason why the timing chain flip flops weren’t running properly. The defective transistor was replaced with a known-good one (the museum has a stock of spare parts for the Wang 300-Series calculators) and all of the transistors re-soldered back in place on the 571 board.

The 571 board was replaced in the backplane, and the scope set up to monitor both the master clock signal, along with the T1, T2, and T3 timing chain flip flop outputs. Power was applied, and low and behold, all three flip flops were flipping and flopping exactly as they should be. To double check, the outputs of each of the logic networks that decode the various states of the timing chain flip flops into timing phases that sequence the operations of the 370 were tested, and each output was as-expected, depending on the state of the three timing chain flip flops at any given point in time.

Now that the timing chain was running properly, it was time to see if programming operations would function properly without hanging the 370.

The 371 Card Reader was connected up to the 370, and the 370 plugged into a Wang 360E electronics package. When powered up, the display showed “+0.000000000” as expected. A program card had been prepared, punched with codes to perform keypresses of [1], [2], [3], followed by a STOP instruction. This program, if it ran, should result in “+123.0000000” being shown on the display.

The card was loaded into the reader, the [PRIME] key pressed to initialize the calculator and the 370, and then the [CONTINUE] key was pressed, which should start the program executing. Nothing happened. The display still read “+0.000000000”. Pressing the [DISP PROG] key showed that the program counter was at “00”, and the step at that location was as it should be. It appeared that the program did not run, as if it had, the program counter should have been something other than “00”. The [DISP PROG] key was released, and the display reverted back to “+0.000000000” as expected. Then the [STEP] key was pressed (with no apparent response by the machine), followed by again holding the [DISP PROG] key. The machine still said that the program counter was still at “00”. If the STEP function worked properly, I would have expected the program counter to have incremented to step “01”. It didn’t.

It appeared that some problems still exist. Just out of curiosity, the [PRIME] key was pressed, followed by the [7] key, just to make sure that the 370 was still talking to the electronics package. Nothing happened. A number of different key-presses were attempted, and none of them had any effect – the display just kept reading “+0.000000000”. Pressing the [PRIME] key resulted only in a slight dimming of the Nixie tubes – not a good sign. I powered everything down and powered it back up, and tried using the 370 as a plain old calculator keyboard/display unit — functionality which was known to work in the past. Sadly, there was no response to any keypresses on the 370. It appears that in the process of trying to figure out what was up with the timing circuitry, some other problems have cropped up.

Just to make sure that the 360E electronics package didn’t have a problem, a 360K keyboard unit was connected up to it, and it worked fine. So, the problem is definitely something new that has developed with the 370. It seems that I had made one step forward then taken two steps back.

This isn’t the first time that such a situation has occurred when working with old electronics like this. With equipment that is nearing 40 years old, there is much potential for problems. Sometimes just the physical shock of unplugging circuit boards can cause components that were marginal to fail completely. Sometimes no matter how carefully old circuit boards are manipulated, just the act of moving them can create other types of non-component-related problems such as solder joints that were perhaps weak to begin with suddenly failing. Such things are a simple fact of life when it comes to working with electronics that are this old.

While this is a bit of a setback, it isn’t a disaster. It just means that it’s going to take a little more digging to get to the bottom of the problems that this Wang 370 has. Of course, I will make update postings here as time presents itself to do more troubleshooting.

06-Feb-09 – More Busicom 161

February 6, 2009

The Busicom 161 calculator chassis donated to the Old Calculator Museum arrived Tuesday (2/3) morning. The box it was packed in looked a slight bit worse for the wear due to handling during shipping, but structurally, the box looked OK.

The machine was double boxed, which is one of the best ways to ship old calculators (the best way being custom-made conformal foam packing). The machine was cushioned by styrofoam packing “peanuts”, another good thing, however, it’s usually a good idea to first seal the machine in some kind of bag (preferably an anti-static bag) to keep the packing material from working its way into every nook and cranny — especially in a machine like this that is a bare chassis, without the cabinet. It took about 1/2 hour with long handled tweezers to pick all of the packing material out from between circuit boards, under the keyboard, and in various other cavities in the machine.

The first order of business, after harvesting the peanurs, was to give the machine a complete visual inspection once all of the packing material was removed. It was noted that one of the discrete neon lamps that are positioned between the Nixie tubes as decimal point indicators was missing. The wires for the lamp were sticking out of the circuit board, but the lamp itself was missing. Referring to the photograph of the machine placed in the eBay auction showed that the lamp wasn’t there in the auction posting, so it wasn’t missing as a result of shipping damage. This isn’t a big deal, as it likely can be replaced. The chassis was turned upside-down, and the glass neon tube fell out from inside the machine. Apparently somewhere along the line, the indicator was broken, probably as a result of improper handling without the cabinet in place, and it fell down inside the machine.

The next step was to very carefully pull out each of the 41 logic circuit boards. There are 42 logic circuit boards in the machine, but one of them isn’t removable. It contains the Mitsubishi-made 16×16 magnetic core array, and it is hard-wired into the backplane. Why this board is hard-wired is beyond me, – many machines of similar design use plug-in core arrays. It is interesting to note that the IME 84, upon which it appears the Busicon 161 was patterned, also has a fixed circuit board for the core memory array. Anyway, all of the logic boards were individually carefully withdrawn from their backplane connectors.

As discussed in the original posting about the Busicom 161, the museum’s existing machine has severe problems with the edge connector sockets losing their structural integrity. It was strongly hoped that this chassis didn’t share the same problem. Unfortunately, as the circuit boards were removed, it was found that of the 41 sockets, six of them suffer the same kind of problem.

The problem manifests itself by cracks developing in the plastic-like material that the socket contacts are encased in. The socket contacts are gold-plated, and look somewhat like a “Y”, with the open part of the “Y” being where the card edge connector is inserted. Each connector has 22 of these contacts. Once the plastic material cracks sufficiently, the tension that holds the contacts in place is released causing the contacts to become mis-positioned, and also to lose the spring tension that keeps them in contact with the gold-plated fingers on the wire side of the logic circuit board.

Given the problems wtih the backplane, it was not safe to try to power up the machine with the circuit boards in place. But, it was possible to check out the power supply. The damaged backplane connectors were carefully inspected to make sure that none of the contacts were shorted, so that when the machine is powered up without the circuit boards, there’s no chance of short circuits. The Nixie tube display subassembly, and the keyboard subassembly were removed. Both subassemblies connect into the backplane wiring with high-quality connectors, allowing these units to be easily removed and serviced. Unfortunately, the power supply is hard-wired into the backplane, meaning that some disassembly is required in order to look at the power supply circuit board. Once things were taken apart enough, the power supply board was found to conveniently have nomenclature on it identifying the various voltages. The power supply makes +5, -12, and -5 volts for logic supplies, and +85 and -85 Volts for powering the Nixie tube displays. The power switch for the machine is located in the keyboard assembly, and thus the wiring for the power switch had to be traced and a jumper fashioned to plug into the socket that the keyboard plugs into, in order to simulate the power switch being “on”. Three digital voltmeters were connected up to the power supply, and the power cord was plugged into a Variac, and the voltage slowly ramped up. There were no signs of any trauma as the mains voltage was ramped up to 50%, then slowly up to 75%. The digital voltmeters started to register voltages. Once the mains voltage was at 100%, the +5V supply read +5.31V, the -12V supply was at -13.21V, and the -5V supply was at -5.89V. The voltages were all a little higher than expected, but this is liekly because there was no load on any of the power supply voltages. The Variac was turned off, and two of the DVM’s connected up to the +85 and -85V Nixie tube power supplies, and the machine powered back up again. The +85V supply was running at almost 89V, and the -85V supply was running at just over -90V. These voltages aren’t as critical as the logic supplies, and aren’t actively regulated. Transformer windings are simply rectified and filtered, so some variance in the output voltages is to be expected. Lastly, the machine was again powered off, and the oscilloscope connected up to the logic supplies to measure ripple, to make sure the power supply filter capacitors were OK. All of the logic supply voltages had only slight (1-2 millivolts) of ripple, all within safe boundaries. With all of these measurements performed, the power supply of the machine seems to be in good health after all these years.

Speaking of years, this 161 appears to have been made at very close to the same time as the original 161 in the museum. Transistors are all coded with dates in the range of the 2nd week in 1969 to the 5th week in 1969, while the original 161 has date codes that are just a few weeks later, with the latest date codes listed being in the 6th week in 1969. In a rather silly move, Busicom put the serial number tag on the back surface of the upper part of the cabinet. The upper part of the cabinet on the recently donated machine was missing, thus, there’s no way to identify the serial number on this machine. While disassembling the machine, I looked high and low for any hints of a serial number stamped or written in other locations within the chassis, but none was found. This is rather unusual given the price of machines of this vintage (frequently $1000 or more, which was a lot of money in 1969), many manufacturers put the serial number on a fixed part of the machine that was not easily removable, and also had markings inside the chassis that provided secondary identification of the serial number to aid in identifying a machine if it was lost/stolen.

It is going to take some time to figure out a stragegy to replace the edge connector sockets that are bad. Experience trying this with the other Busicom 161 proved to be futile. That time, the connectors were replaced with the backplane in place. This involved very carefully separating the dense wiring to clear it away from around a failed edge connector socket (horribly difficult because the backplane wiring is very tightly laced into a big bundle of wire), carefully desoldering the wires to the bad connector, putting in a new connector, and then soldering the wires back to the new connector. This proved to be nearly impossible due to the density of the backplane wiring. A different approach is going to be needed to replace the faulty connectors on this machine. Fortunately, the connectors used are standard 0.156-spacing 22-pin sockets, easily available through any electronics supply outlet. The edge connector sockets in the 161 are held in position by rectangular holes in the bottom of the circuit card guides that are screwed into the chassis. There are bosses on each end of the edge connector sockets with a hole through them that would allow the connectors to be screwed into a circuit board or chassis to hold them in place. The screw holes are not used in the 161, but these bosses fit into the rectangular holes on the circut board edge guides on each end of the connectors to hold each connector in place. If the card guides are removed from the chassis, this would free up the edge connectors to “float”, allowing the backplane to be worked on in a much easier fashion. The only support for the edge connector sockets at this point would be the backplane wiring itself. This is likely that strategy that will be used to try to safely replace the damaged sockets. The real difficulty with any means to try to replace the connectors is that the backplane wiring uses only minimally color-coded wire, and due to age, heat cycling, and other factors, the coloring on the wire insulation is sometimes very difficult to make out. There’s no schematic or service documentation that I’ve run across for the 161 (if anyone knows of any such documentation, I’d love to hear from you), so it means that very careful observation and documentation of the wiring will have to be done. This will rquire a lot of digital photography at close-quarters, a lot of time with bright lighting and a magnifying glass, and meticulous note-taking. Keeping all of the wiring sorted out to assure that it all goes back together again with the correct connections is going to be absolutely critical. No room for any errors here.

I need to get hands on replacement connectors, which is something I hope to do sometime soon. Once I have them in hand, and have been able to take things apart enough that I can test my method for repair, I’ll make a new post here letting everyone know how its going. In the meantime, I plan on taking the original machine and taking photos of it (though it’s certainly not operational), and begin the process of creating a museum exhibit for the Busicom 161. The goal of the Old Calculator Museum is to have fully operational machines, but for the time being, the Busicom 161 exhibit, when I put it online, will be one of the few exhibits where the machine shown isn’t operational. But, over time, I hope to be able to get a fully-operational machine. Betweent he two machines, there are plenty of spare parts that an operational machine can be made, assuming I can get through the backplane repairs without causing more problems or, worse yet, irreparable damage.

Thanks for reading.

29-Jan-09: Wang 360SE

January 29, 2009

Today I had some time to look into the Wang 360SE that was recently donated to the museum.

Visual inspection was the first order of business.  First, the cover over the backplane was removed so that a complete visual inspection of the backplane and power supply circuit board could be performed.  The backplane in the 360SE is a hard-wired affair.  Wang used a unique method for wiring the backplanes on these machines that used special edge connector sockets with long rectangular tail pins on the backside that special clips attach to.  The clips are soldered to the individual wires in the backplane wiring harness.  The clips visciosly grab the tails on the edge connector sockets to provide connectivity between the various edge connector sockets.  Intuitively, this method doesn’t seem like it’d be very good in terms of long-term reliability, but amazingly, it seems to work very well.  Even though the edge connector tails and the clips are made from what appears to be a tin alloy, the connectivity this system provides seems to be very relaible, even after 40 years.

Looking over the backplane carefully showed no signs of any loose connections or broken wires.  The power supply circuit board looked as if some of the components (mainly rectifier diodes) had been replaced at some time during the machine’s life.  The machine’s service tag did not indicate anything of these repairs, so it is likely that these repairs were not done by a Wang service technician.

The backplane cover was replaced, and the cover over the Logiblocs was then removed.  All of the Logiblocs were then carefully pulled from the sockets so that all of the edge connector socket fingers could be inspected, as well as checking the edge connector fingers on the Logiblocs for corrosion.  The edge connectors were all in good shape, and there was only light corrosion on the Logibloc fingers, which were all cleaned with a contact cleaning solution.  Each Logiblock was also inspected to make sure there were no sign of obvious problems, and all looked good.  The core memory board was inspected under a magnifying glass to check for any broken wires in the core array, and no problems were found.

The power supply transformer and large computer-grade capacitors were visually inspected, and looked good.  The line voltage fuse was checked, and it was found to be good.  Without power connected, an ohmmeter was placed across the power supply hot and neutral prongs on the plug, and the power switch turned on to check for too little resistance, which could indicate a shorted line filter device, or the mains winding in the transformer. The ommeter showed about the same resistance reading as the known good Wang 360SE already in the museum.

The power supply of the 360SE generates +11V and -11V, along with 230V for drive of the Nixie tubes on the keyboard/display units.  With all of the Logiblocs removed, it was possible to power up the machine and check the power supply voltages.  The 360SE electronics package was plugged into a Variac, and the voltage very slowly raised up to full line voltage, while monitoring the +11 and -11V supplies using two digital voltmeters.  It became apparent as the line voltage neared 50% that the power supply wasn’t doing anything.  The +11V and -11V DVM’s showed +0.000 on their displays.  As I ramped the voltage up to full line voltage, there still was no output at all on either of the logic power supplies.   Clearly, there’s something amiss with the power supply, that will require more digging to sort out.

There are two thermally activated "pop out" circuit breakers that provide protection for the +11V and -11V supplies.  The first thing to do was check them.  Visually, they weren’t popped.  They were both checked wtih an ohmmeter, and had full continuity.  Unfortuntately, the problem was not going to have a simple solution.

I ran out of time to work on the machine any further during this session.  The next step will be to check the transformer secondary voltages with power applied to see if the transformer is working properly.  If a problem is found with the transformer, finding a replacement is going to be difficult.  Any replacement would have to have the proper primary and secondary windings, as well as be very close to the size of the original, as the transformer is pretty tightly packed in the chassis.  If the transformer proves to be OK, then the problem is likely somewhere in the circuitry that takes the AC voltages out of the transformer, rectifies it to DC, filters it, and then regulates it to the proper supply voltages.

I’ll write about further digging into this in a future blog entry.

23-Dec-08: More Wang 370

December 23, 2008

In the last post, I talked about getting the recently-received Wang 370 keyboard unit running as a standard keyboard/display unit (equivalent to a Wang 360K).  The next step is to check out the programming functionality of the 370.

The 370’s programming capabilities are a function of local logic within the 370 that provides the interface to the 371 card reader that provides the source for the program steps; the test and branch logic; the interface between the 370 and the conencted electronics package; and interfacing of the 370 to peripheral devices.  The logic is contined on nine Logiblocs that are the same form-factor as those used in the 300-Series calculators.  The Logiblocs utilize the same Germanium transistor logic as the 300-Series calculators, primarily PNP 2N404 transistors made by RCA.

The 370 has a toggle switch that selects the mode of the display.  The two positions of this switch are DISP PROG, and DISP WREG.  When the switch is in the DISP WREG mode, the working register of the calculator is displayed.  With the switch in the DISP PROG position, the display is changed to show the current program step number (represented as two decimal digits 00 through 79), and the program code punched into the card at that step number (displayed as two octal digits (00 through 77).  Without the 371 card reader connected, I expected that the program code would display either as 00 or 77, if things were working properly.

The 370 was powered up connected to the 360E, and the display mode switch was set to “DISP PROG”.  The display immediately changed from the calculator-mode display of “+0.000000000″ to ”   00  00 “.  This indicates that the program counter (step number) was 00, and the program code at this step is 00.  This was a good sign that at least the logic that switches the display from the calculator’s working register to the internal program counter and code read from the (not-connected) 371 was operating properly.

A key on the 370 keyboard labeled [DISP PROG] allows the display to be momentarily (as long as the key is held down) switched from showing the calculator working register to the program mode.  This key’s function was tested, and it too seemed to function properly.

Another toggle switch on the 370 controls whether the programer automatically advances program steps when a program is run, or puts the machine in single-step mode.  This switch as two positions, “AUTO” and “STEP”. When this switch is in “STEP” mode, the [STEP] keyboard key is supposed to step through the program one instruction at a time with each depression.  With the display mode toggle switch in the “DISP PROG” position, I tried pressing the [STEP] button.  Nothing happened.  I tried a few more depressions, and the display still read ”   00  00 “.  I had expected for the program step counter to advance.  When I switched the display mode switch back to the “DISP WREG” position, the display showed “+0.000000000” asa expected.  But, when I tried entering some digits, there was no response. Pressing the [PRIME] key reset the 370’s logic, and then I could enter numbers and perform math just fine. 

To me, this indicated that either the 371 must be connected for the programming features to work properly, or that there was some kind of problem with the 370’s logic that caused the machine to hang when programming-related operations were attempted.

To answer this question, it would be necessary to get the 371 card reader tested out.  As with any vintage electronics, a thorough visual check is needed before attempting to power up the device.   The 371 was somewhat grubby, but looked to be in OK physical condition.  Mr. Allen (the donor of the 370/371 system) had indicated that there were some problems with the 371 based on his observations, so was going to make sure that a complete inspection was done before attempting to hook it up to the 370. I took the front cover off of the 371, and it immediately became obvious that some form of amateur repair had been attempted to the 371 (not by Mr. Allen, but by some previous owner).  The circuit board was not secured by mounting screws into the case, and the holes in the circuit board did not even match up with the mounting bosses cast into the case.  The only thing securing the circuit board inside the 371 was the connector that allows daisy-chaining of other 371s (the 370 can control up to four 371’s).  It appears that the original connector had been damaged at some point, and someone attempted to replace it with a new connector.  The problem was that when the new connector was soldered in place, it was mis-positioned such that the circuit board was not able to be aligned properly.  The connector is secured to the cabinet with two screws, which are all that was holding the circuit board in place.  While this is a problem that needs to be corrected, inspection revealed that the mis-positioning of the circuit board did not cause any potential short circuits or other problems. 

The next thing to look at is the connector and cable that connects the 371 to the 370.  The connector was inspected, and it looked to be in usable shape, but there were some signs of deterioration of the plastic material that the contacts in the connector are embedded in, but not to the extent where any contacts could short to others.  The cable goes through a grommet in the cabinet of the 371, and terminates in a bunch of individual wires soldered to points on the circuit board.  Each wire was checked, and one wire was found that was not connected to anything.  From there, an ohmmeter was used to check continuity from every contact in the connector to its mate on the circuit board.  It was found that this un-connected wire was simply an “extra” wire that isn’t used.  All of the other connections tested good.  One other thing noted on the 371 circuit board is that an incandescent lamp that lights (shining through a white plastic lens protruding through a hole in the cabinet) was missing.  This indicator lights to show that the 371 is actively being controlled by the 370.  Not knowing what kind of lamp is needed, it was decided to not worry about this problem at this point.

It was decided at this point to button the 371 up, and try connecting it to the 370.  The cover on the 371 was replaced, and the 371 plugged in the “READER” connector on the back panel of the 370.  A few of program cards were prepared with codes “77”, “52” and “25” codes (alternating patterns of 0’s and 1’s) punched in step number 00.

The system was powered up, and tested in calculator mode, with everything operating properly. The punched card with code “77” (all 1’s) was inserted into the card reader, and the reader clamshell closed. The display mode toggle switch was set to “DISP PROG”, and low and behold, the display read ”   00  77 “.  The “77” punched card was removed, and the “52” installed, with the display showing ”   00  52 “.  Then the “25” card was inserted in the 371, and likewise, the display showed that the code was being properly read.  The program run mode switch was set to “STEP”, and the [STEP] button was pressed.  The display didn’t budge, still showing ”   00  25 “.  The display mode switch was set back to “DISP W REG” to put the 370 back in calculator mode, and it was found that the 370 was again hung.  Pressing the [PRIME] key unfroze the 370, allowing it to operate as a calculator keyboard/display unit again.

The assumption at this stage was that the 371 seemed to be working properly, at least being able to read the card code punched into step 0.  It was also assumed that something was amiss with the programming logic of the 370, causing it to hang when operations relating to the programming functions were attempted.

Fortunately, the museum has an original copy of the Wang 300-Series Service Manual.  This document includes schematics of nealy all of the Logiblocs used in the 300-Series calculators, keyboard/display units, and peripheral devices, as well as the Model 370 and 380 programmers.   It was time to sit down and study the logic of the 370, and try to figure out what might be the cause of the system hanging when attempts were made to step the program counter.

In perusing the schematics, it was found that, like most sequential logic systems, the 370’s logic relies on a master clock circuit.  The clock generator, consisting of a transistorized oscillator circuit, drives a chain of three flip flops that are connected as a binary counter.  The clock steps the counter through eight different combinations of outputs.  The outputs of the counter are applied to diode gating networks that decode the counter outputs into various states that control the operation of the 370.

With this information in hand, the first place to check would be the clock generator and the timing counter flip flops.  That will be something for the next posting.

22-Dec-08: Wang 370

December 22, 2008

For a very long time, the museum has been looking for a Wang Laboratories 370/371 Programmer for the 300-Series calculators. The 370/371 is a system consisting of the 370 Programming Unit and the 371 Punched Card Reader. The 370 is a special keyboard/display unit that plugs into any Wang 300-Series calculator electronics package (although all features work only on the 360E, 360SE, or 362E electronics packages) that provides the standard function of a 360K/362K keyboard unit, but adds a bunch of programming-related functions. The 371 is a special punched card reader, similar in construction to the CP-1 and CP-2 punched card readers, but plugs directly into the 370 keyboard/display unit to provide the program code that the 370 interprets and passes on to the electronics package. The 370/371 system was developed to augment the very basic programming functions that the CP-1 and CP-2 Card Programmers offered by adding conditionals, branching, and looping capabilities, and the ability to control peripheral devices such as the Wang 372/373 data storage units.

Unfortunately, the 370/371 has proven to be extremely elusive. I have run across perhaps six or seven of these devices in over 15 years of looking, and of those that were available for sale (e.g. on eBay), all efforts to acquire them failed.

So, when I was contacted recently by Mr. Arnold Allen indicating that he had a Wang 360SE, Wang 370, and Wang 371 that had been sitting in storage for a very long time, I was intrigued. Mr. Allen graciously offered to donate the equipment to the Old Calculator Museum, which is greatly appreciated. 

Mr. Allen indicated that the machines had been in his possession for a long time, in excess of 20 years, and that they showed definite signs of age, and some signs of attempted repair.  The equipment was originally acquired by Mr. Allen as part of an auction lot that he purchased, with the Wang equipment being of secondary interest.  Fortunately, though, even though the primary reason the lot was acquired was for other items in the lot, Mr. Allen felt that this stuff was cool enough that he kept it around all these years.

The equipment was very well-packed and arrived at the museum without any problems.  Upon its arrival, the 370/371 was checked out.  The 370 is in darned good condition considering its age (the QA stickers indicate 1969).  There were indeed some signs that some non-professional repairs were attempted on the power supply circuitry of the 370 sometime during its life.  Also, one Nixie tube was missing.  The 371 was a little worse-off, but not too much so.  Someone at one point tried to replace the daisy-chain connector on the 371 card reader, and soldered it in with its placement just a bit off.  This resulted in the inability for the screws that retain the electronics circuit board in the card reader could not be installed, leaving the circuit board “loose” within the cabinet, retained only by the screws that hold the daisy-chain connector to the housing of the reader. 

The first order of business was to check out the power supply of the 370.  There are a total of N Logibloc circuit boards in the 370 that give it its local intelligence.  All of the boards were pulled, and inspected.  No signs of any catastrophic component failures, fortunately.  The was the usual layer of oxidation on the tin-plated edge connector fingers, which were cleaned using a contact cleaning brush and contact cleaner.  With all of the boards removed, some electrical tests of the power supply components were done.  The power supply transformed coils were tested for shorts and opens, with none found.  The rectifier diodes were also checked, with all of them testing good.  It appears that all of the diodes were replaced at some point, albeit a bit sloppily.  There is a 1000uf filter capacitor that was definitely a replacement also, and it looked good, and tested OK on the capacitor tester.  Two other can-type filter capacitors (3000uf) also looked like replacement units, and tested good.  The fuse was tested, and was good, so then AC power was applied via a Variac, with the line voltage slowly ramped up to 110V, while monitoring both the +11 and -11 volt DC supplies with digital volt meters.  The power supply voltages came up good, with the +11 reading +13.5V and the -11 at -11.9V.  Given that the supplies had no load on them, these voltages seemed reasonable. An oscilloscope was then used to view the power supply levels to check for excessive ripple, and, with no load, there was <1mv of ripple on either of the supplies, indicating that the rectifier diodes and filter capacitors were good.

After verifying that the power supply was working well, attention was turned to the backplane of the 370, to make sure that there were no bent pins or loose wires.  A detailed visual inspection showed that the backplane was in very good shape.  The cable/connector that goes from the 370 to the electronics package was ohmed out, and the connector inspected, and it looked to be in fine condition, with no broken or intermittent connections. 

The Logiblocs were re-installed in the 370, and power again applied (without the 370 being connected to an electronics package), and the power supply voltages checked under load.  The +11V supply was running at around +11.4V, and the -11V supply was running right on -11V.  Ripple was checked, and was only slightly more with the load in place, but still insignificant. 

Power was left on for a while, allowing the circuit boards to warm up, while the power supply voltages were monitored.  No signs of any problems occurred during this time. Without an electronics package connected, the 370 would not come up, as the high-voltage (180V) for the Nixie tube drive is supplied by the electronics package.The next step was to connect up an electronics package.  The museum’s trusty 360E electronics package was powered up with a 360K keyboard and tested to make sure it was still completely healthy, and it was found to be sound.    The 370 was then plugged into the 360E, and the 370 powered up, then the 360E powered up.  Some of the Nixie tubes started to light up – a good sign.  The PRIME key on the 370 was pressed to clear out everything, and zeroes started showing up on some of the Nixie tubes, and the + sign at the far left end of the display was also lit.  After letting the displays warm up for a while, there were two digit positions which refused to light up, but all of the other digits and the sign tube seemed to work.  Keyboard entry of digits gave the expected result, although after a little experimenting, it was found that the most-significant digit of the display had a problem. It would concurrently show the digit in that position, along with the digit in the 3rd digit from the left at the same time.  For example, if 1234567890 was in the display, the display would read +X2Y4567Y9Y, with the X being a “1 and 3” lit at the same time, and the Y’s being blank (with the least-significant digit having no Nixie tube in its socked). In spite of the display aberrations, performing math functions gave the expected results.  For the most part, the 370 was properly operating as a 360K keyboard/display unit just fine.

At this point, it was decided to focus on getting the display aberrations taken care of.  The two Nixies which didn’t light at all were removed from their sockets, and tested in space 320K keyboard that was known good.  They didn’t light up there, either, so the tubes were likely simply worn out.  These two tubes, and the missing tube, were replaced with spare, known-good tubes.  The 370/360E were powered up again, and this time, the display was completely lit, but the problem with the most-significant digit showing two digits at once still persisted.    Given the 300-series Wang machines’ propensity for edge connector finger and socket corrosion to cause problems, all of the cards were again removed from the backplane, and the edge connector sockets were cleaned carefully, but thoroughly.  After time for the contact cleaner to evaporate, the cards were reseated in their sockets, and the system was powered up again (again, without the 371).    This time, the most significant digit was displayed properly, with no “ghost” digits.  At this time, the full display was tested out, with all of the discrete neon decimal points checked out, and each digit ran through all of its combinations.  Everything worked perfectly.  The 370 is well on its way to recovery.

That’s all for this installment.  Check back for more on this story.