PCB DESIGNED AND VERIFIED USING KICAD TOOLS
After about 22 hours of work I have a PCB layout that reflects my design for the replacement that will plug into the IBM 1130 and function as 8K of core memory. Assuming the final design is close to this, the bill of materials cost including the PCB fabrication plus all tax and shipping for parts is about $120.
CHECKING CHIP BY CHIP FOR DATASHEET AND ACTUAL PCB CONNECTIONS
For each chip on the board, I will check the pins against the data sheet and the connections from those pins to other chips on the board. A particular kind of failure in projects like this comes from unseen shorts on the schematic in KiCAD that bridge wires that should not connect. The graphical editor sometimes sticks a dot over crossing wires, converting them into a connection. If I don't spot these subtle changes then the board reflects the faulty schematic and I will have wasted a round of PCB fabrication and soldering.
I will check the package type for each chip as the pin numbering can vary. I want to be sure that the functional pin I wanted to connect is the one that has a trace on the board, including the other ends of the trace.
SANITY CHECKING THE LOGIC
It is all too easy to send a signal that should be high but is inverted because of the driving circuitry. Also, complex chips that have many modes of operation depend on carefully understanding the datasheet and properly setting up the input pins to achieve the mode you intended.
I want to model the flow of the external signals into the board and check that my thinking was sound. All logic conditions should work as I expected. No unexpected activations of signals should be possible.
STUDIED VARIATION IF ONE OF THESE WOULD GO ON A DIFFERENT 1130 MODEL
The original 1130 models had 3.6 microsecond core memory, with one 4K or 8K memory module seated in gate B compartment C1 of the machine . Over time IBM expanded the product to larger configurations - up to 32K words by use of an expansion frame containing gates D and E to hold as many 8K modules as were needed to reach the capacity ordered.
To support the larger memory, the machine has an Expanded Storage (ES) feature installed which re-uses gate B compartment C1 for additional logic cards to support the larger memory. With ES, all memory must reside in gates D and E since gate B compartment C1 is now unavailable.
Machines without ES have three cables running from the rest of the CPU to the top of gate B compartment C1, placed in cable slots T1, T3 and T4. When ES is on the machine a fourth cable is also connected to T2.
The machines with ES have three cables attached to gate B compartment C1 at slots A2, A3 and A4 which are connected to the installed core memory modules which would sit in gate D compartment A1, gate D compartment B1, gate E compartment A1 and gate E compartment B1. They hook directly to gate D compartment A1 and other cables daisy-chain to the remaining compartments based on the size of the total memory.
IBM also changed the 1130 later in its life to offer a faster 2.2 microsecond core memory technology. This faster memory also requires the ES feature even for small configurations. A cable runs from the CPU to slot A5 of gate B compartment C1 to support the special needs of the faster memory. Therefore, any 2.2 uS machine has all its core memory in the expansion frame in gates D and E.
My design would need to be modified to work with ES machines. Rather than plugging in the cables that normally hook to gate B compartment C1 connectors T1, T3 and T4, I would need to connect my board to gate B compartment C1 slots A2, A3 and A4. I would have to study how the signals are arranged on those cable connectors to determine any wiring changes on my board.
In addition, the faster memory would require slightly different component values on my board to align with the 2.2 uS core behavior, but this is relatively trivial.
I will look into this during the downtime once the PCB is ordered and I am waiting for the board to arrive.
PARTS ARE ON ORDER FROM DIGIKEY ALREADY
I have ordered all the parts for the board and expect to have them all before the PCB comes back from the fab. Until I am satisfied with the desk checking I won't pull the trigger on the board so that starts the clock.
No comments:
Post a Comment