Category Archives: Uncategorized

Sharp PC-G850 Serial over WiFi Module

YAEP (Yet Another ESP Powered Project): An ESP8266-12 powered WiFi module for the Sharp Pocket Computer via the 11-pin connector connects wireless to PCs for transfer of programs and data.

The ESP8266 connects to your home WiFi or acts as an AP for initial configuration. The Serial port of the G850 becomes accessible via TCP on port 23. You can use telnet or netcat (nc) for simple direct transfers to/from the PC or use socat to create a virtual com port. The module supports 9600baud and shortens CTS/RTS and pulls these signals up to +5V, so you need to use XON for flow control.

Please note:

  • The G850 uses inverted serial protocol logic levels (i.e. logical “high” is represented by a “low” (0V) TTL level, logical “low” is represented by a 5V TTL level. The ESP8266 uses the SoftSerial library on GPIO 4 and 5, which supports inverted logic levels
  • Raw TCP is implemented without encryption on port 23. You can connect via telnet to receive or send data or programs but everyone on the same network can read the transferred data in clear text
  • The module does only support serial port communication (i.e. it does not emulate the CE-126 synchronous communication for print and cassette tape commands)
  • In the TEXT/Sio/Format menu enable 9600 baud, 8N1 and no flow-control. End-of-file “1A” allows the G850 to stop listening when receiving a file. You can still send files without the end-marker, but need to interrupt the “load” command by pressing “ON/BREAK”. The received text will be in the editor.
  • Sleep timeout can be set. The module starts blinking befor it goes to sleep and pressing the PRG button resets the sleep timout.
  • Some basic “AT” commands allow configuration changes without having to re-flash the module
Tex/Sio/Format menu

Why not Bluetooth? I started out using a HC-06 BT module, but learned along the way that this module is no longer supported by Windows 10’s BT stack. After several hours of fruitless tests, I gave up and resorted first to an ESP32. However, the BT stack that implements a serial port pushes even the most bare-bone program beyond 1MB in size (thank you Espressif), so I opted for raw TCP instead.

I have not noticed that the WiFi module decreases battery life dramatically, especially since I implemented deep-sleep.

WiFI Serial Module for Sharp PC-G850

The module is a double sided PCB. I isolation-routed it with FlatCAM. Top-side is very much limited to traces and the two buttons, all vias can be manufactured with just copper wire (i.e. no through-hole rivets required) and all pins only need soldering on the bottom-side. You can find the EAGLE files at the end of this post.

The module consists of a double sided 1.5mm PCB, an ESP8266-12, a 3.3V voltage regulator, a 6x 1K, 2 x 330Ohms, 1 x 1.5K, 1 x 3K3 SMD resistors (0805 type), a 100uF capacitor, 1x LED (0805) and a 90 degree 11-pin header. The G850 connector and the case are 3D printed.

Disclaimer: I am not an electrical engineer. These projects are a way for me to learn and figure out ways to overcome the problems I encounter. By all means, if you have suggestions for doing things better, please let me know.

You can reach me on twitter at @ChrisHerman if you need help.

ESP8266 module PCB and 3D printed 11-pin connector (bottom view, version 3)
Schematic, version 4 (.sch file available in resource section)

11-pin connector: A 3D part for the 11-pin connector which others may be able to re-use in their designs. The 2mm drill holes need to be positioned 200mil (5.08mm) above and below the connector and 100mil (2.54mm) in the orthogonal direction, away from the axis formed by the 11 pins (see picture below)

Sending and receiving files

To receive a file, plug-in the module (hostname of my module is G850V.local), type on the PC (OSX/Linux):

nc G850V.local 23>test.c

And on the G850 press TEXT, then S (for Sio) and S again (for Save).
Terminate nc with ^c on the PC

To send a file, first press TEXT, S (for Sio) and L (for load) on thje G850.
Then type on the PC:

nc G850V.local 23 <test.c

terminate nc with ^c on the PC

AT commands

Baudrate (1200..9600), WiFi settings and sleep timeout can be set via a small set of “+++AT+” commands. These commands can be sent from the G850 or via netcat ot telnet from your PC/Mac (of course, this will require your WiFi settings work). There’s also a “failsafe” configuration that can be selected when pressing the PRG button of the module for 5 sec:




this sets the SSID to “GUEST” and the wifi access point password to “pw”.

numeric configuration items:
rev: integer, referencing the version of the configuration. I recommend leaving at 1
sleep: seconds until unit goes into deep sleep
baud: baudrate for connection with G850 (600…9600)
port: TCP/IP port (use 23 for telnet compatibility)

string configuration items:
ssid: the ssid name of your wifi
wifipw: password of your wifi network
host: hostname for use when requesting an IP address
otapw: password for protrcting your ota passwords (use espota protocol)

Returns: OK
Saves current configuration as failsafe.ini to LittleFS Flash file system. Pressing the PRG button for longer than 5 sec will load failsafe.ini configuration and reboot. This allows you to recover from a messed-up config.ini (e.g. wrong wifi credentials).

Returns: OK
Puts the adapter immediately into sleep. Pressing the RES button on the module will wake it up.
There is an issue with the ESP8266’s deep sleep behaviour, because regardless of all interrupt and wake-up sources being disabled, the

Resources (all in my github repository)

Vertical Pinball Machine

Vertical Pinball Machine


  • Windows 10
  • Pinball FX3 (Steam)
  • 34” QHD Ultra Wide 1440p Monitor (Samsung LS34J550WQUXEN)
  • 7”Screen Display 1024*600 LCD Monitor
  • Radeon RX590
  • Dell Optiplex 3020 (8GB RAM, 3.2GHz i5-4570
  • Inductive Plunger (allows trick shots)
  • 2x45W Stereo, 68W Subwoofer
  • Oak case

The vertical Pinball machine was build on a Windows 10, Optiplex 3020 (i5) with 8GB RAM and a 16GB Radeon RX 590 (8 GB GDDR5).

The Dell Optiplex shows a BIOS error when the front cover is not connected (and I threw it away before realizing I should have kept the cable). So I added an additional Arduino that emulates a keyboard and sends an “F2” keystroke 5 seconds after power-up.

I am particular proud of the inductive plunger, which allows skill shots without use of a linear potentiometer (no wear!). You can Google “Arduino Inductance LM339” and will find good description of how to do it. Schematics and board can be found in the resource section.

One problem remains unsolved: I need an external keyboard to shutdown windows. I did not want to mess with any software that maps keystrokes, so at the end of the gaming night, I pull out a little wireless keyboard and shutdown Windows. No enough of an inconvenience to warrant drilling another hole in the cabinet.

I ordered a kitchen worktop (2m x 0.9m x 2cm), cut it into 12cm wide boards of 2cm thicknes and various lengths, halved the thickness to about 8mm (thereby doubling the number of boards), glued the boards together to form 8mm thick sheets and cut the sides of the cabinet. Easy.
Could have ordered oak boards right away.

For the subwoofer and speaker cut-outs I used a laminate cutter (also known as a hand router), with a ball bearing at the tip. I printed jigs with my 3D printer that I screwed to the bottom of the board to guide the router.
This produced super neat and perfectly round cutouts.

I used a biscuit jointer when making the large boards. That was a stupid idea because you need to know where you make the cuts and ensure there’s no biscuit where you cut the board: it looks ugly when the biscuit shines through at an edge, like the diagonal cut that I had to make for the front of the cabinet.

Windows PC
The Dell Optiplex 3020 was gutted and basically only the mainboard survived, mounted on an open frame at a 20 degree angle. The Dell PSU was replaced with a Corsair VX550 (the Optiplex mainboard requires a PSU adapter cable which can be ordered from Amazon for £10. A cheap 256GB SSD stores the data.

Beware that the Dell motherboard will report an error and not booting unless the frontpanel is attached. I learned the hard way AFTER I had thrown away the Dell case and NOT retained the front panel or cable. So I programmed an Arduino to simulate an “F2” keypress shortly after boot to get past the error message.

The mainboard is mounted at an angle on a simple frame.

A Samsung LS34J550WQUXEN 34″ Ultra Wide LED Monitor (WQHD 3440×1440, Freesync, 2 x HDMI, DisplayPort) was dissected (use a credit card to buy it and then later to pry the case open. Comes apart very neatly.

I build a pine frame that the monitor rests on, with an 5mm tempered glass pane on top. Black spray paint from the back creates a nice bezel that covers the edge of the display and the wooden frame. Everything is held in place by 20x30mm aluminium profiles (I screwed up and cut one side 1mm too short, so had to patch it. not my proudest moment, but I ran out of profile and then lockdown hit and I couldn’t get to the hardware store).

The small top screen is a super cheap AliExpress 1024×600 screen. A small OLED display would have been better I suspect, because I went a bit too cheap on the smalls screen and as a result it’s not quite as bright as I would have liked it to be.

Buttons and Plunger/Inductance Meter
Figuring out how to measure the position of the plunger with a coil took the longest. Usually you either trigger a switch or pull a linear potentiometer back and forth. The latter allows trick shots.

I did not like the idea to attach a potentiometer to the plunger and having to deal with the mechanical hassle, potentially incurring problems after hours of use (the plunger is difficult to reach and not easily removable.

The idea was to wind a coil around the plunger and as the plunger retracts, it pulls out of the coil which should change the magnetic flow and the inductance of the coil.

There are great guides out there that describe how to use an Arduino to measure inductance. So I thought how hard can it be?

The circuit works as follows:

  • The Arduino triggers a pulse that “rings the bell”. The bell in this case is the coil and a capacitor.
  • Coil and capacitor form an LC oscillator. The frequency depends on the capacitance and the inductance. If you keep the capacitor (1uF) fixed, then the change in frequency is an indication for the change in inductance, which in turn is an indication of how far the plunger is pulled out of the coil.

    Mujahid has written a nice piece about it here:
  • Using an LM339 helps converting the analog signal into a TTL signal that the Arduino can measure. The length of a pulse is a measure for the position of the plunger.
    See it in action here:


  • The Arduino simulates a joystick and sends the position of the plunger as z-axis values to the PC. It also takes care of the buttons that control the flippers, the gamepad, Select, Start and Back. The Atmega 32U4 has a build-in USB controller (you cannot use an Arduino Uno because it relies on an external USB-Serial controller).

Was it worth it?
Was it worth it? Many people on the Internet will tell you that virtual pinball machines are inferior, but I disagree. While virtual tables can’t match the tactile feedback and nostalgia of a real pinball machine, they should be viewed as a unique category. Is it the same as a “real” pinball machine? Absolutely not—it’s essentially a vertical screen. However, the experience is highly engaging, and the FX3 software is impressively realistic. The reduced space requirements, combined with the flexibility to play any pinball machine I want and the ability to own dozens of different tables in one package, make it worthwhile.

I enjoyed the building process and figuring out how to make the inductive plunger work. Even after about eight months, playing my vertical pinball machine remains fun, and it’s exciting to see new tables being released each month.