modelrail.otenko Trains, computers, vintage, retro, model railways… the lot…

IBM 45U0018 POS Keyboard

I saw this unit on eBay and couldn't resist. It really did remind me of the department store POS keyboards and I remember them being clicky. I checked the photos on the auction and saw that it had a non-standard socket and no cable included. You can download the drivers/configuration-utility here.

The main keys are fixed, but the surrounding keys have optional translucent key 'covers' where inserts have been placed for special functions. Some of them even cover standard keyboard keys.

But yeah, back to the socket, and do excuse the photo...

It's very similar to a standard IDC Header, but in-set which makes it really difficult to use anything other than IBM's proprietary cables. After a lot of googling, I found that there's both a USB and PS/2 cable version available.

Unfortunately, these weren't actually available anywhere... and if they were, the cost was out of control. It seems that these keyboards could also just dock onto the POS system above the till, so it's possible that the cable was only required if you wanted to mount the keyboard elsewhere. Anyway, a cable would be too easy... I'm sure we can work this out!

What does it look like on the inside?

With the hunch that the port supported PS/2 and USB, the keyboard was opened to work out the pinout. Opening it was actually very easy with just 6 small screws on the back. A large plate comes off and then there's just a bit of foil-cardboard shielding over the main circuit board.

Inside, of note, were a TUSB2046B IC and a STM32F103VBT6 Microcontroller with V0326_PF.bin programmed into it. The first IC happens to be the brains for a USB 4-port hub and provided a huge hint for determining the unit's wiring. I started tracing along the pins and the table below is the result.

Note that, when looking at the back of the keyboard into the socket when the keyboard is UPSIDE-DOWN on a flat surface, I've started Pin 1 from the top-right, with pin 6 then being top-left, pin 7 bottom-right and pin 12 bottom-left. This turned out to be silly as all the wires are in vertical pairs. Note this has been updated from the great resource at sharktastica.

Ok, so that's easy for the USB connection.. just wire the 'upstream' pins, +5v and GND into a spare USB cable. The other 2-pair of wires which run into the Microcontroller must be for the PS/2 mouse and keyboard? I wonder which is clock and which is data? There are plenty of images online showing microcontroller-to-PS/2-port hookups, so I'll try that after USB.

Hooking Up USB

USB cables contain 4 wires. First you need two wires for a very standard 5v DC rail and then you need two more wires to transmit and receive data. As you can see from the table above, we have all of these already mapped out. This was all very straight-forward thanks to the documentation for the TUSB2046B USB Root Hub IC. I grabbed a bunch of spare parts from the box'o'junk and started wiring something together. The USB cables I first tried to use happened to be power-only? No data lines! No wonder someone else threw them out.

Too short anyway! So instead, I remembered I had a mini USB socket in another tray'o'junk and set to wiring that up...

I plugged it into my RasPi 400 and it evicted all USB devices...

Sep  2 17:04:30 raspberrypi kernel: [63187.047153] usb usb2-port1: over-current change #1
Sep  2 17:04:30 raspberrypi kernel: [63187.202027] usb 1-1-port1: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.278524] usb usb2-port2: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.438880] usb 1-1-port2: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.518607] usb usb2-port3: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.678902] usb 1-1-port3: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.758546] usb usb2-port4: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.918864] usb 1-1-port4: over-current change #1
Sep  2 17:04:31 raspberrypi kernel: [63187.998562] usb usb2-port1: over-current change #2
Sep  2 17:04:32 raspberrypi kernel: [63188.238478] usb usb2-port2: over-current change #2
Sep  2 17:04:32 raspberrypi kernel: [63188.478480] usb usb2-port3: over-current change #2
Sep  2 17:04:32 raspberrypi kernel: [63188.718503] usb usb2-port4: over-current change #2
Sep  2 17:04:32 raspberrypi kernel: [63188.890074] usb 1-1-port1: over-current change #2

Ooops! Why am I chewing too much power? I google'd around and couldn't find an easy solution to inject my own power, as all answers indicated VCC needed to stay connected. Looking at my wiring again, I realised that this error was bound to happen as I'd wired the polarity around the wrong way! So, after correcting that...

Sep  2 18:44:47 raspberrypi kernel: [69203.309095] usb 1-1.3: new full-speed USB device number 9 using xhci_hcd
Sep  2 18:44:47 raspberrypi kernel: [69203.443182] usb 1-1.3: New USB device found, idVendor=0451, idProduct=2046, bcdDevice= 1.25
Sep  2 18:44:47 raspberrypi kernel: [69203.443193] usb 1-1.3: New USB device strings: Mfr=0, Product=0, SerialNumber=0
Sep  2 18:44:47 raspberrypi kernel: [69203.449815] hub 1-1.3:1.0: USB hub found
Sep  2 18:44:47 raspberrypi kernel: [69203.452677] hub 1-1.3:1.0: 4 ports detected
Sep  2 18:44:47 raspberrypi kernel: [69203.779047] usb 1-1.3.4: new full-speed USB device number 10 using xhci_hcd
Sep  2 18:44:47 raspberrypi kernel: [69203.922725] usb 1-1.3.4: New USB device found, idVendor=04b3, idProduct=4604, bcdDevice= 3.26
Sep  2 18:44:47 raspberrypi kernel: [69203.922734] usb 1-1.3.4: New USB device strings: Mfr=1, Product=2, SerialNumber=3
Sep  2 18:44:47 raspberrypi kernel: [69203.922741] usb 1-1.3.4: Product: IBM Retail USB Alphanumeric POS Keyboard                                     
Sep  2 18:44:47 raspberrypi kernel: [69203.922747] usb 1-1.3.4: Manufacturer: (c) Copyright IBM Corp. 2008
Sep  2 18:44:47 raspberrypi kernel: [69203.922752] usb 1-1.3.4: SerialNumber: 44T4011-200811231006420977000036
Sep  2 18:44:47 raspberrypi kernel: [69203.935428] input: (c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard                                      as /devices/platform/scb/fd500000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0/usb1/1-1/1-1.3/1-1.3.4/1-1.3.4:1.0/0003:04B3:4604.000D/input/input18
Sep  2 18:44:47 raspberrypi kernel: [69204.000545] hid-generic 0003:04B3:4604.000D: input,hidraw4: USB HID v1.00 Keyboard [(c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard] on usb-0000:01:00.0-1.3.4/input0
Sep  2 18:44:47 raspberrypi kernel: [69204.008558] hid-generic 0003:04B3:4604.000E: hiddev97,hidraw5: USB HID v1.00 Device [(c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard] on usb-0000:01:00.0-1.3.4/input1
Sep  2 18:44:47 raspberrypi kernel: [69204.016919] hid-generic 0003:04B3:4604.000F: hiddev98,hidraw6: USB HID v1.00 Device [(c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard] on usb-0000:01:00.0-1.3.4/input2
Sep  2 18:44:47 raspberrypi kernel: [69204.035363] input: (c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard                                      as /devices/platform/scb/fd500000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0/usb1/1-1/1-1.3/1-1.3.4/1-1.3.4:1.5/0003:04B3:4604.0010/input/input19
Sep  2 18:44:47 raspberrypi kernel: [69204.036075] hid-generic 0003:04B3:4604.0010: input,hidraw7: USB HID v1.00 Mouse [(c) Copyright IBM Corp. 2008 IBM Retail USB Alphanumeric POS Keyboard] on usb-0000:01:00.0-1.3.4/input5
Sep  2 18:44:47 raspberrypi mtp-probe: checking bus 1, device 10: "/sys/devices/platform/scb/fd500000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0/usb1/1-1/1-1.3/1-1.3.4"
Sep  2 18:44:47 raspberrypi mtp-probe: bus: 1, device: 10 was not an MTP device
Sep  2 18:44:48 raspberrypi mtp-probe: checking bus 1, device 10: "/sys/devices/platform/scb/fd500000.pcie/pci0000:00/0000:00:00.0/0000:01:00.0/usb1/1-1/1-1.3/1-1.3.4"
Sep  2 18:44:48 raspberrypi mtp-probe: bus: 1, device: 10 was not an MTP device

HAHA. It just worked. It's very loosely-clicky, but not bad at all. Hilariously sounds like the registers at David Jones. The trackpad and touchpoint actually work beautifully as well. Very impressive. I assume there's further commands that could be sent to change LEDs, use the peripherals, but I don't need it.

PS/2 Wiring

I didn't feel like plugging this into a PC and restarting it 17 times to work out the wiring, so I used an Arduino Uno instead! Unfortunately, no combination of pairs would get the unit to produce any output. I'll have to do some more digging at some point in the future... a cop-out, I know, but I'm happy with USB for now!

UPDATE: There's a pinout here at sharktastica.

Printing a proper plug

I searched Thingiverse and found a very nice set of IDC Plug Frames. I then dragged that frame into Tinkercad and stretched it to make it fit better. The result is here.

From there, I used standard 'single' jumper wires to build the 6x2 inner plug and glued them in place. I then soldered up a spare USB cable.

Perfect! Note there's still plenty of these keyboards on eBay from a WA seller. Highly recommended!

Dual Technologies SKD-4000

This beauty had been doing laps on eBay for a while. I had tried to low-ball it early on, only to receive a swift declination. A week or two later, I offered a little more and it was accepted! The seller took no time to send it to me, perfectly packed. It was listed as untested and with a cracked front-right case. Indeed it was, with one of the plastic clips actually hanging out the front of the opposing top-half, looking slightly like a buck-tooth.

After a little googlin', I came across a site talking about a slightly different version to this laptop. Turns out Dual Group were a Taiwanese manufacturer and built laptops that were then re-branded and on-sold. They therefore didn't really have their own site with any relevant info on this laptop. If anyone manages to find any information on this laptop, then please get in touch in the comments section! I therefore kept inspecting the unit, disassembling what I could prior to needing a screwdriver.

This unit didn't have a power supply, but thankfully the link above provided a pinout for the power supply their SKD-4000 needed. Under this unit, it stated it needed DC 20v @ 1.0A, so I tinned up a set of power wires and dialed in the voltage.

Jamming the wires in the power socket, after continuity-testing for GND, I flicked the power switch... lo'an'behold: the full orchestra of a vintage HDD and a POST beep!

The joy was very short-lived though... the BIOS quickly reported a low battery and incorrect date/time. Hardly surprising! I went into the BIOS anyway and did a HDD auto-detect.

116mb, nice! I saved the config and rebooted, but nothing seemed to have stuck. Seems I'll need to replace the battery to get further. Before that though, does VGA out work?

Nice! Ok, back to the CMOS battery. The link above provided a tear-down guide to help me, but in the end I didn't really need it as this laptop was surprising easy to open and work with! Even the CPU had a flap for easy access? It also has DIP switches for setting the CPU and they seem to configure the CPU pins directly? I'll try and decode them shortly.

The CMOS battery is a piece of work. It's hard-soldered in and it happens to be a rechargeable coin-cell. These are pretty rare nowadays... and... stupid me... I just tried to solder in a coin-cell holder and use a CR2032. DO NOT TRY AND USE NON-RECHARGEABLE BATTERIES IN-PLACE OF RECHARGEABLE BATTERIES!

Turning the unit on saw the coin-cell heat up rapidly, but fortunately I had my finger on the power switch, ready to prevent any spontaneous combustion. I then tried a 3.7v cell I had lying around, but the laptop started reporting 10 error beeps indicating that no battery was available at all. I really do hope I haven't trashed the charging circuit. At this point, I ordered an exact replacement Panasonic VL2330 battery and then looked for other things to fix up.

Power Plug

Whilst waiting for a battery to arrive, I realised I should solidify the power supply. Seems this type of plug is called a SnapNLock 4-Pin Mini DIN and I was pleased to see it available at Jaycar. I purchased a few (luckily) and attempted to wire one up. Soldering it was easy enough, but assembling it was next-level.

I mean, just look at the datasheet (which, actually, wasn't even east to find!), and you'll see what I mean... 8 components to jam together in a not-so-obvious order. I trashed the first plug and sorta-but-not-really succeeded with the second plug... it was good enough to test with.

Dip-Switch Dissection...

There are two blocks of four switches and, with no documentation, I set about trying to determine what these might control. Looking at the set closest to the CPU first, one can find the traces run off CPU pins.

Thanks to the glory of the internet, one can find pin maps and descriptions of pins readily available. The latter doesn't go into much detail, so actual CPU datasheets can also come in handy. After a little digging, it seems this block of switches controls maths error handling and interrupt line access to the CPU. I might just leave them as-is.

The second block's traces disappear into the board's layers and are impossible to trace and understand.

The BIOS

Just for fun, I popped out the PLCC32 and read it in my Willem Programmer.

You can download the BIN file here.

Back to that Battery

A proper Panasonic VL2330 arrived and I soldered it in place. Applying power heard the same beeping. Ok, let's follow the traces to the left and see what's going on... two diodes, both seem to conduct power... then the voltage goes into pin 20 of that IC... with no marking? Why no marking?

Under the right light, it turned out to be an Epson RTC-6593 Real-time Clock. It even has integrated battery-backed RAM! Of course, this unit stores my BIOS settings. Why has the label disappeared? It's COOKED! It tried to burn me when I touched it. We can only assume it's due to my testing of other batteries... if only I'd bought the correct replacement at the start.

Replacing the RTC

So, I'm game to try this... what I'm not game for is fake chips! Searching for this IC has only come up with seemingly-dodgy sellers on Alibaba and eBay. I've just thrown in a few orders and we'll see what comes back. In the meantime I've removed the existing IC and cleaned up the area. Interesting to note that the silk-screen shows the IC should be an RTC-6583, whereas a 6593 was installed. The datasheet seems to indicate that 659x has "Extended Alarms" whereas 658x doesn't.

During removal of the chip, I just happened to remove a pad... I blame the fact that the IC was actually glued down to the board! I couldn't work out why the IC wouldn't move, regardless of the pressure I applied. Furtunately, it was a quick fix with winding wire and it's ready to solder up.

I'll post a Part II once the ICs have arrived.

Mega Drive Circuit Board Repairs

A bunch of Sega Mega Drives came via the workbench recently, all looking for a 50/60hz mod and region unlocking. Two older Mega Drive 1s were in the batch and the first modded with zero concern. Unfortunately, the second wouldn't even boot... a quick opening of the case saw the reason:

This area of the circuit board is easily accessible from the cartridge slot and I can only assume that a screwdriver (or other sharp object) has been jammed in at some point, destroying a few traces. You'll also notice solder to the left, but that was me in a previous attempt to repair this. I tried again as I now have a USB Microscope and it really does work well for reviewing damage and making intricate repairs!

How not to repair it

I first-up thought I'd be able to use copper winding wire and simply solder to each pad where the IC contacts the circuit board.

From the shots above, you can see how quickly that process failed. The combination of using a soldering iron under a microscope, a large soldering tip and shakey hands meant that solder spread everywhere. Fortunately, solder wick cleaned all that up in no time.

Also how not to repair it

The next method was a little dodgy, but proved successful. From above, I knew i needed more room to solder, so the goal was to create this room by lifting and spreading the pins and running wires to the other end of the traces.

A multimeter found the other end of the traces and winding wire was attached.

Thanks to the tiny tweezers in the iFixit kit, I was able to bend, lift and then straighten the leftmost leg.

From there, the other legs were lifted easily.

Finally, they were all spread out both horizontally and vertically to make soldering as easy as possible. I then connected the right-most as it was only going as far as the local via.

With the wires hideously attached, I considered bending the legs back down... But a blind person could see my dodgy soldering job would make contact.

Instead? Glue gun to the rescue! Finally? Test. It worked!