- hp-2621+2.jpg (339.5 KiB) Viewed 1220 times
As an aside, I find the suggestion on that HP Computer Museum page that orange oil goo remover (active ingredient D-Limonene) can reverse ABS bromination like Hydrogen Peroxide rather implausible. Sure it'll get rid of decades of skin oil, but so will dish washing liquid. Maybe it was just a plug for an Aussie solution.
The switches are Hi-Tek linear grids. All this is shown in the Wiki, so I am not bothering with more detailed pictures here.
All these terminals use the same protocol, as explained in the Keyboard Module manuals.
The HP engineers had the same idea as the DEC engineers who did the VT52: have the terminal do the scanning, but abstract the key switch matrix by having the keyboard implement a device that returns the key state of a given keycode. Split this code into column and row using a decimal decoder and demux. Because of how the schematic was drawn, HP used "row" and "column" with row being the three bit LSB and column the four bit MSB.
The keys that are the same between the standard and extended keyboards have the same keycodes. But somehow they use slightly different numbering schemes in the respective manuals' diagrams. I am not sure why. The extended keyboard scheme, which is just the hexadecimal value output on the address lines, works for any arrangement. And the standard keyboard scheme, which has one digit each for column and row, in that order, extends straightforwardly to have a hex digit for the MSB, rather than packed bits, which end up with one in the LSB. The manual does note that the keycodes are the same when appropriate but makes no mention of this discrepancy.
The two digital elements of the keyboard are CMOS. Which means, in the words of the same Compatibility issues section,
In fact, both chips have a supply voltage range of 3V to 18V. So, in theory, it might be possible to use a modern 3.3V MCU. I tried with a Teensy LC, but couldn't get it to work, so I guess it was just too close once the passives are accounted for. So Teensy 2.0 it is.The keyboard can run on +5 volts or +12 volts. The CMOS inside really wouldn't care.
The entire Keyboard Module manual, and that section in particular, seems to me to have clearly been written by an engineer addressing other engineers. The timing section's explanation of the slowness from the Q1 transistor inverter is either a veiled complaint about budgets or meant to anticipate criticism from others who didn't understand the design trade-offs they were making.
Somewhat unusually for the time, the keyboards have diodes for every switch, so N-key rollover works reliably.
There is a speaker inside the keyboard. It is wired between a signal line and +5V. So, rather than driving it directly it is meant to be switched by an NPN transistor. On the original terminal, this was done by dividing one of the video trace signals. They used a current limiting resistor of 47Ω, allowing the 8Ω speaker to reach its full half-watt rating. That seemed like a lot of additional current, so I used 470Ω, for only around 10mA and it's still plenty loud.
Commendably, the J1 connector on the PCB uses the same numbering as the DA-15 connector on the other end. Too often working out a keyboard means keeping a pin mapping for the its cable.
Here is the converter with original cable:
- hp-262x-converter.jpg (287.61 KiB) Viewed 1220 times
In terms of the QMK mapping, I think the standard keyboard might actually be more pleasant to try to use with modern systems. It has fewer keys, but critically a pair either side of the space bar for modern shifts. And the blank adobe brown key could be a layer shift for a "65%."
