CDP18S030: A Microcomputer in a Briefcase

The original plan for the minimal target board was just the CPU, oscillator, and TC5517A with address lines A8-10 grounded for 256 bytes of memory. It was a small change to add TPA clocked capture FFs for those address lines for full 2 kB. 

Initially the target board turned out to have a serious flaw, as it was not possible to handler the ZIF socket lever. 

Other components than the memory are SMDs and thus on the copper side of the board.

This is the RC oscillator. Same circuit as in the self test card.

The address FFs are also singles:

The target board after fixing the ZIF lever issue:

To test the memory board, I first checked with memory reads and writes that some addresses work correctly, writing a few locations with some random bytes and then checking that they held the values.

The next test was to write at the beginning of each 256-byte page a distinct byte, values were 0x11, 0x22, etc. Those also worked fine.

The next test was to enter a small program that fills one page with bytes 0x00...0xFF. The program was as follows:

After executing this program I checked the address range 0x0040...0x013F for the values 0x00...0xFF. All was OK.

So, with this card I have a primitive computer with a CPU and 2 kB of memory.

The next step is to check the serial port and see if the debugger can be used from the external console.

After studying some alternatives for making the 7-segment hex displays one feasible may be to use a simple programmable logic device like Renesas SLG46824 to convert the nybble input to 7-segment display hex digit patterns. A quick check with AI provided a solution which fits into the limited resources of this chip. The logic patterns require 16 of the 19 logic elements of the chip leaving 3 for DFFs, and thus one external DFF like SN74LVC1G79 is needed. The SLG devices have relatively high drive current specifications so they can typically drive the LEDs directly making external driver transistors unnecessary.

The converter schema from the GreenPAK designer:

The LUT contents are:

P1: 0x04
P2: 0x9E
P3: 0x04
P4: 0xD0
P5: 0xE0
P6: 0x8E
P7: 0x20
P8: 0xE
P9: 0x18

Ma: 0x8D
Mb: 0x7B
Mc: 0x27
Md: 0x7B6D
Me: 0x8D
Mf: 0x27
Mg:0x72

The simulation with the SPICE in the GreenPAK designer confirms the correct patterns:

Even though feasible, for the moment I won't include the display to the target circuit, but only the memory and the logic it requires. That will be sufficient to further test the debugger.

Tinkering the self-test card took some prestidigitation with sharp knife, LT1SC soldering tip, and Leica A60.

The oscillator was made as described in the previous log. The inverter for the EF1 was made of a '1G57 multi-function gate.

After plugging in the card and the processor nothing remarkable appeared to happen. No eight dots on the display, suggesting the Micromonitor could be faulty. But not really. The manual tells to use known good processor as the target, and untested recycled processor is not such a thing. So after inspecting the device with oscilloscope, it seemed almost to work, sometimes.

After some closer study, it became obvious that the pins of the target processor were so dirty and oxidized that even the first time to get to the 8-dots prompt was a miracle. So cleaning the pins by removing the old solder and cladding with fresh tin, the processor began to respond correctly to the Micromonitor. No issues with the self test circuit so far.

After a little playing with the debugger it seems to be alive. 

The keypad has some contact issues as well as some bouncing, but nothing really unexpected so far.

So, now that I have and in-circuit debugger, I need a circuit to debug in. And what would be more apt to the theme than make something like Cosmac ELF? The schematic can be found in the net. However, some components may be hard to come by, so have to improvise.

For the RAM I found a 2 kB TC5517AP. Simply tying the high address lines low I can start with 256 bytes aliased across the whole memory space, like the original one. A bit more difficult to find are the two hex displays with internal latch and binary to hex chars converter. That calls for a hack... 

Instead of the original self test board which, according to the description in the manual, is a board that is plugged into the external memory board edge socket and connected with the DIL40 connectors of the target cable to the target connector socket, this self test board will be a DIL40 hack plugged directly into the target socket and connected with a single cable to the memory socket for the 5V supply (pin B of the external memory connector P1).

The 100 kHz oscillator will be a 200 kHz RC oscillator made with dual unbuffered 7404 gates (NL27WZU04) and a divider made of single D-flop with inverted output (SN74AUP1G80). The concept is like shown in this simulation.

Lacking the models of actual components the actual frequency will be seen by measuring the live circuit.

After building the circuit C=470pF and R=8.2 kΩ resulted in 104 kHz.

After checking the board, mostly visually, all appeared to be OK. The first power-on was to be tested with the logic only, so the only elcos in the serial line drivers don't matter at this time.

Raising the current carefully and observing the behavior of the device, all appeared to be electrically OK. At 5V the current needed was 180 mA, something expected of that many CMOS packages on the board.

Some status leds turned on but nothing on the display.

That, however, is expected, as the 8 dots in the display only appear after the debugger has detected the processor. For that a live processor is needed in the target processor socket. Also some target system should be there, but lacking that the self-test card that came with the system can be used. Except that no self-test card came with it.

However, the manual has the circuit for the self test card, so the next step is to make one.

However (again), some checks can be made right away, the interesting one is to see if the system's own processor is alive. Quick check with oscilloscope showed that the crystal oscillator is running. The next check was to see the timing pulses TPA and TPB.

Those appear to be OK, too.

So the next step is to make the self test card and check the system further.

To be able to verify and potentially repair a power supply sometimes requires figuring out the schematic in detail. For simple linear and switching regulator designs that follow the example circuits given in datasheets and application notes, that may not be necessary. This power supply appeared to contain some circuitry that I'm not familiar with, so the next step was to draw out the schema, with KiCad.

Before plugging in the AC it is advisable to verify that each part of the circuit is in working order. Especially the electrolytic capacitors of this age (date code 1977 week 21 for C9, for example) are know to be often, though not always, severely degraded. In this case I used multiple adjustable lab power supplies to feed each of the circuits at the output side of the rectifiers.

With some luck, all could have been just fine, but as often, some are not. The first step in the testing is to limit the current to something like 20-50 mA and slowly rise the voltage to the expected maximum, which can be determined from the specifications of the components, electrolytics and semiconductors. 

In this case the first ramping of the capacitor C9 to 9V went fine, but on the next ramp it shortcircuited. Without current limiting much more damage would have occurred. Others seemed to be OK on that firs testing, but because of the one major failure I decided to desolder all e-caps and test them individually. 

Each cap was first slowly ramped to the maximum working voltage without exceeding the charging current on 20 mA and held there until the leakage current went below the usual maximum leakage current CV/4000s (for 5300µF/25V that is 33 µA). The voltage was then mainatained for 10 min. Then the procedure was repeated.

All remaining caps passed that test.

After the leakage test each cap was measured with LCR meter (100Hz) for capacitance and Q. Requirements set was that capacitance was at least 80% of the nominal and Q above 10 (my arbitrary choice for this kind of a quality check). One capacitor, C7, had less than 40% left of the nominal capacitance. That had to be changed. C4, the other large cap, had 87% of the capacitance but Q was only 2. That indicates large ESR and that will need to be changed soon, too. Others were withing the requirements.

Lacking an axial cap for replacing C9, two 4700µF/25V radial caps were soldered to a trip boar in parallel and the combination was soldered to the place of C9; attached with two-sided tape and a zip tie.

After these fixes the PS appears to work correctly. The tracking section was tested by taking the tracking input from another channel of a power supply and the output voltage followed the input as expected. To understand the circuit correctly it was simulated with LTSpice (simulation files are in the attached zip file tracking-ps.zip).

The next step is to begin studying the main board. At first by powering also it from external current-limited lab supplies.

The restoration of a computer often begins with the power supply. It is critical not to turn on the potentially degraded and fragile device until the safety of the power supply and the other circuitry has been verified to not contain shorted components, especially the electrolytic and other capacitors. There is also a chance there are degraded solderings which may trigger a major destruction rare or unobtainable components.

The first step in this project was to take out the power supply and verify that it is wired for the correct voltage (220V for Europe). 

Though a bit unorthodox, also the PH-163 power cord receptacle was changed to a modern IEC C8 one for which cables are available. 

The original socket will be saved for a chance that the fitting plug becomes available.

The next step is to get familiar with the power supply. Though the CDP18S030 manual has the schematics and component layout of the logic board, the only thing about the power supply is the instructions for the voltage selector wiring.

Therefore the next step is to draw the circuit diagram of the power supply.