Subscribe via RSS

Building an Arduino from scratch

OK, OpenLCB is the new cool... to make 'nodes' for it, we need a skeleton Arduino to build from. This article will define how to go about this. The goal in the end is to have a prototype that can, via the Arduino IDE, have a sketch uploaded to do as you wish.


Firstly, a list of random links for random information on building your own Arduino:

Serial or USB?

I know the DB-9 Serial Port is archaic... needs an Interrupt Request (IRQ) assigned to it etc, etc... but there's something nostalgic about plugging in a chunky cable instead of a new USB-B connector. It's also cheaper and easier to use parts on hand and I had a few MAX232s sitting around.

To use USB you'll need an FTDI chip to convert the USB signal from your PC to the TTL protocol of which the Arduino understands. Note that there are also FTDI chips which convert USB to RS232, the serial protocol that allows devices (specifically your old dial-up modems) to communicate over serial cables. You don't want one of these as you'd then have to convert the RS232 again to TTL.

Which Microcontroller?

Good question! Your choice of microcontroller comes down to size and quantity of digital/analogue inputs and outputs. I decided on the Atmega328P as it has enough IO lines for what I need on each OpenLCB node; I'll produce more node types and have them specialise rather than one node that does everything from one board. They're cheap on eBay and they are very similar to the MCU used in the Arduino Duemilanove (5 analogue and ~13 digital IOs.)


You'll find schematics all over the web, although none seem to agree on what components are the exact bare-minimum for an Arduino... it actually makes it quite difficult to know exactly what you need. Below is a solid basis to start your Arduino-based invention, containing everything required to program over the serial port.


Building the prototype

There was not much out of the ordinary on the breadboarding of this project. It's actually quite simple and the standard rules of check, check and double-check your wire placement, IC pin numbers, etc... as you go is imperative. The serial cable connection on this project, due to the cable being rigid, can cause issues of dragging the board around. I'd recommend to seat connection and cable first and secure everything to your workbench... You'll then prevent your breadboard from upending when the serial cable chooses to move around.

Programming from the IDE

The MCU chosen is the Atmega328P and, as mentioned, it's closest match is the "Arduino Duemilanove w/ Atmega328". Once everything was connected I attempted to flash using this board, selected in the Arduino GUI, but it didn't work straight away. The initial error was that the chip device ID didn't match. It turns out that the Atmega328 has as a different code to the Atmega328P (something to do with Pico Power.) I had to hack the 'avrdude.conf' file buried in the arduino folder and then it just worked.

Before I knew it I had the servo library in and a mini servo from eBay controlling a Peco point on my layout.

Re-programming an Atmega MCU

I accidentally learnt how to do this... At one point I'd put the MCU in the wrong way around and expected that I'd killed it. The circuit then failed to work after numerous attempts with the TX and/or RX lights constantly lit. It wouldn't accept sketches and I therefore thought the best bet was to re-flash the bootloader + program.

It turns out that my Arduino Mega can be used as an In-Serial Programmer and so I set this up. This can be selected via the "Arduino ISP" option in the "Programmer" menu under the new Arduino 1.0 software.


(Note that the capacitor from +5v to RESET is a 100n non-polarised.)

After the re-flash and a reconfigure of the MAX232 in/outs (the schematic above is the final version that now works perfectly) it all just worked again... for one of the MCUs anyway. The second simply refused to receive sketches; the RX light would not turn off. I took this as a fundamental failure in the chip (I blame my bread-boarding skills) and therefore tried something sneaky: I copied the entire code off the MCU that worked and flashed it to the 'broken' one with avrdude and it's associated avrdude-GUI [direct download link]. It turns out this worked... the chip now did what the previous one did... but in the MAX232 breadboard it still would not accept a new sketch from the Arduino GUI. I attempted to find other methods to 'completely erase/reset' an Atmega, but I couldn't get it to work and so just accepted it's fate of having to copy it's brother.