First of all: Many thanks for the excellent photos. I also enjoyed the cat in the background.
I wanted to go over some of the similarities and differences to other older Topre keyboards that you may find interesting.
Firstly, this is very similar in operation to their other keyboards. There is a driver IC that applies voltage to the rows, and there are 16 columns which are sensed to detect if a key is pressed or not. This is then sent to the green "TP1680" device, which is commonly called a ceramic Hybrid, usually it has a mixture of integrated circuits and precision components such as laser-trimmed resistors. Do not attempt to open it. This device takes the output of the column and compares the voltage to a set point and outputs to the microcontroller if it thinks the key has been pressed. I can go into more detail here, but it's the same as other old Topre devices.
Fortunately for me, I had the chance to take apart and analyze a much smaller Topre keyboard from 1996, which has a very similar PCB and layout. Many of the overall standards are the same, from the PCB construction, hand-routed traces. Microcontroller, buzzer, PCB labeling, etc. For example, the buzzer is driven by the 555 timer (Z8) nearby, which generates the frequency. This is then passed through the nearby transistor Q1, which has the Emitter, Collector, and Base labeled.
Going from left to right, Z1 is a 74LS145 BCD to decimal converter. This takes a 3-digital pin input from the MCU and outputs to one of ten outputs. This is what drives the (up to) 10 rows in the matrix.
Next to that is Z2-4051. the first of two 4051. This takes 8 columns from the key matrix and selects one of them to pass to the keypress detector. Fun fact: these are still used in most Topre boards nowadays and almost all MX Hall Effect ones too. Below that is an arrow of pulldown resistors that actually terminate the lines. The big IC after that, Z3-CPU is the Microcontroller unit that runs the keyboard. This is a NEC version of an Intel 8048 chip that has been used in keyboards for many years. It has a dielectric resonator (MC1) instead of a crystal oscillator for the main clock, so the phase noise will be worse, but it is sufficient for a keyboard. The next is the special custom sense IC, seems to be called Z4-HIC perhaps.
After that is Z5-06. This is a HD7416P, a hex inverter. Note that the 7406 and 7416 have the same function. This most likely handles communications between the MCU and the cable connector, they are commonly used for that purpose. Next is Z6-4051. This is the second of two 4051, and handles another 8 columns. Just like the previous one it has pulldown resistors to terminate the capacitive lines.
After that is CN1, the cable connector. The puniout is
1: unknown possibly ground
2,3: VCC,
4, data IO
5: unknown, possibly clock
6: data IO
7: keypress detection? This is very interesting as it goes to R10 / C21 and into the TP1680 and it is sensitive enough to need guard traces.
8,9 GND or earth.
The next is Z7-06 which has another HD7416P, which is connected to the four LEDs below it (0-3) as well as the unpopulated one in the key matrix area (LED 4). After that is the 555 timer.
All the ICs were made by Hitachi except the MCU which is NEC. This is common for Topre to use exclusively Japanese components. Unfortunately date codes are difficult here since they used a single number, we don't know the decade just the year and week. Most are 5 and 6 indicating 1986 or 1996 construction.
The MCU is a NEC D8749H, which is an 11 MHz MCU, pin compatible with the uPD8048 (the Intel 8048). It has UV-erasable EPROM, so please don't remove the plastic label.
Moving on to the key matrix, this is an 10-row (at most) and 16 column design, with terminating (pull up) resistors on the rows (MR3, between the numpad and the nav area). Since this is an 8-pin resistor network, it's possible this only has 8 rows out of the 10 it could have, this would be a 128-key matrix which is compatible with the layout. The highest number I saw was 111. The very important earth connector (the ring terminal) connects from the cable and is electrically connected to the metal plate. This is typical. (+) is carried to the key matrix through some traces visible through the PCB, this is a wide traces that goes to C3 (+) and the 74-HIC custom chip. I suspect this is to discharge the lines between strobes? I'm not as familiar with this chip than I am with other things. Also visible on the back of the PCB is the ground (or earth) connections between the rows of the keyboard, between the alphas and the nav area. This is very important to the capacitive keypress detection, it will help to eliminate coupling between the pads when the key is unpressed, as well as isolate them from electrical noise. Finally, of interest is LED4, which is not populated. The key it goes with is not populated, and the plate does not have a cutout here. So they could have used this PCB for multiple layouts. I would suspect this is for caps lock based on the location. The IBM XT keyboard had it here. Of course the other LEDs (0, 1, 2, 3) are at the top.
This style of strobe / scan, with the rows (8 or so) being strobed and 16 columns being scanned is the same for many other Topre keyboards and is still used today. There are some exceptions, such as the NEC N8337 which has too many keys for such a matrix and implements a third 4051 for up to 24 columns, though not all are used. Topre was very reliable here, using this overall design and overall layout for many other boards.
If you have questions about Topre-style capacitive sensing or analog keyboard strobe/scan dynamics I'm happy to go into more detail. It's hard for me to know if other people are interested in this or not, I often go into much more detail than there is interest so I hold myself back.
That does it for PCB analysis. Very interesting kb for sure, and please do reach out if you have other questions. I no longer have the old Tope keyboard to analyze, but I'm going from my notes and what I see on the current PCB.
edit:
Yeah it looks like mDIN-8.
Sadly I don't think the protocol for these mDIN-8 Topre boards has been converted yet.
My keyboard was fortunately labeled on the PCB with the pinout of the connector. The communications pins were:
Data Out
KB Busy
Data In
CPU Busy
I had the same 8-pin connector on mine. Unfortunately I didn't put any time into decoding the protocol, but I would power the board, set the "CPU Busy" to tell the KB to operate and monitor the pins that output info from the keyboard (Data out, KB Busy) and press some keys. If nothing works, then attempt to put data in to the "Data In" pin and see what happens. I wonder if it's based on the NEC PC98 standard? It's also 8 pins and has a CPU busy pin (called RDY in the TMK firmware converter for PC98)
Edit: turns out I did fully decode the protocol on mine. It has a start bit and then 8 bit keycodes, looks like it sends codes immediately (held down keys send repeat codes) and there's no upcode / downcode so you can't use it as a normal KB. I wrote everything down and then stopped development on it since I really wanted a keyboard that sends make / break codes. I guess I can check what I got against the PC98 standard to see if there's overlap. I also didn't record anything about the timing.
Recent Posts
