Raspberry Pi HAT identity EEPROMs, a simple guide


I’ve been working on a RFID scanner than can best be described as an overly large Raspberry Pi HAT recently. One of the things I am grappling with as I get closer to production boards is that I need to be able to identify what version of the HAT is currently installed — the software can then tweak its behaviour based on the hardware present.

I had toyed with using some spare GPIO lines and “hard coded” links on the HAT to identify board versions to the Raspberry Pi, but it turns out others have been here before and there’s a much better way. The Raspberry Pi folks have defined something called the “Hardware On Top” (HAT) specification which defines an i2c EEPROM which can be used to identify a HAT to the Raspberry Pi.

There are a couple of good resources I’ve found that help you do this thing — sparkfun have a tutorial which covers it, and there is an interesting forum post. However, I couldn’t find a simple tutorial for HAT designers that just covered exactly what they need to know and nothing else. There were also some gaps in those documents compared with my experiences, and I knew I’d need to look this stuff up again in the future. So I wrote this page.

Initial setup

First off, let’s talk about the hardware. I used an 24LC256P DIL i2c EEPROM — these are $2 on ebay, or $6 from Jaycar. The pins need to be wired like this:

24LC256P Pin Raspberry Pi Pin Notes
1 (AO) GND (pins 6, 9, 14, 20, 25, 30, 34, 39) All address pins tied to ground will place the EEPROM at address 50. This is the required address in the specification
2 (A1) GND
3 (A2) GND
5 SDA 27

You should also add a 3.9K pullup resistor from EEPROM pin 5 to 3.3V.

You must use this pin for the Raspberry Pi to detect the EEPROM on startup!
6 SCL 28

You should also add a 3.9K pullup resistor from EEPROM pin 6 to 3.3V.

You must use this pin for the Raspberry Pi to detect the EEPROM on startup!
7 WP Not connected Write protect. I don’t need this.
8 VCC 3.3V (pins 1 or 17) The EEPROM is capable of being run at 5 volts, but must be run at 3.3 volts to work as a HAT identification EEPROM.

The specification requires that the data pin be on pin 27, the clock pin be on pin 28, and that the EEPROM be at address 50 on the i2c bus as described in the table above. There is also some mention of pullup resistors in both the data sheet and the HAT specification, but not in a lot of detail. The best I could find was a circuit diagram for a different EEPROM with the pullup resistors shown.

My test EEPROM wired up on a little breadboard looks like this:

My prototype i2c EEPROM circuit

And has a circuit diagram like this:

An ID EEPROM circuit

Next enable i2c on your raspberry pi. You also need to hand edit /boot/config.txt and then reboot. The relevant line of my config.txt look like this:


After reboot you should have an entry at /dev/i2c-0.

GOTCHA: you can’t probe the i2c bus that the HAT standard uses, and I couldn’t get flashing the EEPROM to work on that bus either.

Now time for our first gotcha — the version detection i2c bus is only enabled during boot and then turned off. An i2cdetect on bus zero wont show the device post boot for this reason. This caused an initial panic attack because I thought my EEPROM was dead, but that was just my twitchy nature showing through.

You can verify your EEPROM works by enabling bus one. To do this, add these lines to /boot/config.txt:


After a reboot you should have /dev/i2c-0 and /dev/i2c-1. You also need to move the EEPROM to bus 1 in order for it to be detected:

24LC256P Pin Raspberry Pi Pin Notes
5 SDA 3
6 SCL 5

You’ll need to move the EEPROM back before you can use it for HAT detection.

Programming the EEPROM

You program the EEPROM with a set of tools provided by the raspberry pi folks. Check those out and compile them, they’re not packaged for raspbian that I can find:

pi@raspberrypi:~ $ git clone https://github.com/raspberrypi/hats
Cloning into 'hats'...
remote: Enumerating objects: 464, done.
remote: Total 464 (delta 0), reused 0 (delta 0), pack-reused 464
Receiving objects: 100% (464/464), 271.80 KiB | 119.00 KiB/s, done.
Resolving deltas: 100% (261/261), done.
pi@raspberrypi:~ $ cd hats/eepromutils/
pi@raspberrypi:~/hats/eepromutils $ ls
eepdump.c    eepmake.c            eeptypes.h  README.txt
eepflash.sh  eeprom_settings.txt  Makefile
pi@raspberrypi:~/hats/eepromutils $ make
cc eepmake.c -o eepmake -Wno-format
cc eepdump.c -o eepdump -Wno-format

The file named eeprom_settings.txt is a sample of the settings for your HAT. Fiddle with that until it makes you happy, and then compile it:

$ eepmake eeprom_settings.txt eeprom_settings.eep
Opening file eeprom_settings.txt for read
Done reading
Writing out...

And then we can flash our EEPROM, remembering that I’ve only managed to get flashing to work while the EEPROM is on bus 1 (pins 2 and 5):

$ sudo sh eepflash.sh -w -f=eeprom_settings.eep -t=24c256 -d=1
This will attempt to talk to an eeprom at i2c address 0xNOT_SET on bus 1. Make sure there is an eeprom at this address.
This script comes with ABSOLUTELY no warranty. Continue only if you know what you are doing.
Do you wish to continue? (yes/no): yes
0+1 records in
0+1 records out
107 bytes copied, 0.595252 s, 0.2 kB/s
Closing EEPROM Device.

Now move the EEPROM back to bus 0 (pins 27 and 28) and reboot. You should end up with entries in the device tree for the HAT. I get:

$ cd /proc/device-tree/hat/
$ for item in *
> do
>   echo "$item: "`cat $item`
>   echo
> done
name: hat

product: GangScan

product_id: 0x0001

product_ver: 0x0008

uuid: b9e3b4e9-e04f-4759-81aa-8334277204eb

vendor: madebymikal.com

Now I can have my code detect if the HAT is present, and if so what version. Comments welcome!


Trail run: the base of Urambi Hill


This one has been on my list for a little while — a nice 10km loop around the bottom of Urambi Hill. I did it as an out and back, although there is a loop option if you cross the bridge that was my turn around point. For the loop option cross the bridge, run a couple of hundred meters to the left and then cross the river again at the ford. Expect to get your feet wet if you choose that option!

Not particularly shady, but nice terrain. There is more vertical ascent than I expected, but it wasn’t crazy. I haven’t posted pictures of this run because it was super foggy when I did it so the pictures are just of white mist.


What is Gang Scan?


Gang Scan is an open source (and free) attendance tracking system based on custom RFID reader boards that communicate back to a server over wifi. The boards are capable of queueing scan events in the case of intermittent network connectivity, and the server provides simple reporting.


1-Wire home automation tutorial from linux.conf.au 2019, part 3


This is the third in a set of posts about the home automation tutorial from linux.conf.au 2019. You should probably read part 1 and part 2 before this post.

In the end Alistair decided that my home automation shield was defective, which is the cause of the errors from the past post. So I am instead running with the prototype shield that he handed me when I started helping with the tutorial preparation. That shield has some other bugs (misalignments of holes mainly), but is functional apart from that.

I have also decided that I’m not super excited by hassos, and just want to run the orangepi with the OWFS to MQTT gateway into my existing home assistant setup if possible, so I am going to focus on getting that bare component working for now.

To that end, the gateway can be found at https://github.com/InfernoEmbedded/OWFS-MQTT-Bridge, and is a perl script named ha-daemon.pl. I needed to install some dependancies, which in my case were for armbian:

$ apt-get install perl libanyevent-perl cpanminus libdist-zilla-perl libfile-slurp-perl libdatetime-format-strptime-perl
$ dzil listdeps | cpanm --sudo

Then I needed to write a configuration file and put it at ha.toml in the same directory as the daemon. Mine looks like this:


	timeout=5 # seconds, will reconnect after this if no response
	sensor_period=30 # seconds
	switch_period=10 # seconds


Now run the gateway like this:

$ perl ha-daemon.pl

I see messages on MQTT that a temperature sensor is being published to home assistant:

homeassistant/sensor/1067C6697351FF_temperature/config {
	"name": "10.67C6697351FF_temperature",
	"current_temperature_topic": "temperature/10.67C6697351FF/state",
	"unit_of_measurement": "°C"

However, I do not see temperature readings being published. Having added some debug code to OWFS-MQTT, this appears to be because no temperature is being returned from the read operation:

2019-05-27 17:28:14.833: lib/Daemon/OneWire.pm:73:Daemon::OneWire::readTemperatureDevices(): Reading temperature for device '10.67C6697351FF'
2019-05-27 17:28:14.867: /usr/local/share/perl/5.24.1/AnyEvent/OWNet.pm:117:Daemon::OneWire::__ANON__(): Read data: $VAR1 = bless( {
                 'payload' => 0,
                 'size' => 0,
                 'version' => 0,
                 'offset' => 0,
                 'ret' => 4294967295,
                 'sg' => 270
               }, 'AnyEvent::OWNet::Response' );

I continue to debug.


Mount Bimberi on a Scout Bushwalking Course


Julian Yates kindly ran a bushwalking course for Scouts Australia over the last five days, which covered walking in Uncontrolled Terrain (the definition in the Australian VET scheme for the most difficult bushwalking — significant off track navigation in areas where emergency response will be hard to get). I helped with some of the instruction, but was also there working on my own bushwalking qualifications.

The walk was to Mount Bimberi, which is the highest point in the ACT. We started with a short night walk into Oldfield’s Hut on Friday night after a day of classroom work. The advantage of this was that we started Saturday at Oldfield’s Hut, which offered morning views which did not suck.

On Saturday morning we walked up to Mount Bimberi via Murray’s Gap. This involved following the ACT / NSW border up the hillside, which was reasonably well marked with tape and cairns.

Our route on the way to Bimberi:

And the way back:

On Sunday we walked back out to the cars and did the three hour drive back to Canberra. I’ll include the walk out here for completeness:


A nerd snipe, in which I learn to read gerber files


So, I had the realisation last night that the biggest sunk cost with getting a PCB made in China is the shipping. The boards are about 50 cents each, and then its $25 for shipping (US dollars of course). I should therefore be packing as many boards into a single order as possible to reduce the shipping cost per board.

I have a couple of boards on the trot at the moment, my RFID attendance tracker project (called GangScan), and I’ve just decided to actually get my numitrons working and whipped up a quick break out board for those. You’ll see more about that one later I’m sure.

I decided to ask my friends in Canberra if they needed any boards made, and one friend presented with a set of Gerber CAM files and nothing else. That’s a pain because I need to know the dimensions of the board for the quoting system. Of course, I couldn’t find a tool to do extract that for me with a couple of minutes of Googling, so… I decided to just learn to read the file format.

Gerber is well specified, with a quite nice specification available online. So it wasn’t too hard to dig out the dimensions layer from the zipped gerber files and then do this:

Contents of file Meaning Dimensional impact
G04 DipTrace* Comment
G04 BoardOutline.gbr* Comment
%MOIN*% File is in inch units
G04 #@! TF.FileFunction,Profile* Comment
G04 #@! TF.Part,Single* Comment
%ADD11C,0.005512*% Defines an apperture. D11 is a circle with diameter 0.005512 inches
%FSLAX26Y26*% Resolution is 2.6, i.e. there are 2 integer places and 6 decimal places
G04* Comment
G70* Historic way of setting units to inches
G90* Historic way of setting coordinates to absolute notation
G75* Sets quadrant mode graphics state parameter to ‘multi quadrant’
G01* Sets interpolation mode graphics state parameter to ‘linear interpolation’
G04 BoardOutline* Comment
%LPD*% Sets the object polarity to dark
X394016Y394016D2* Set current point to 0.394016, 0.394016 (in inches) Top left is 0.394016, 0.394016 inches
D11* Draw the previously defined tiny circle
Y1194016D1* Draw a vertical line to 1.194016 inches Board is 1.194016 inches tall
X1931366Y1194358D1* Draw a line to 1.931366, 1.194358 inches
Board is 1.931366 inches wide (and not totally square)
Y394358D1* Draw a vertical line to 0.394358 inches
X394016Y394016D1* Draw a line to 0.394016, 0.394016 inches
M02* End of file

So this board is effectively 3cm by 5cm.

A nice little nerd snipe to get the morning going.


Gangscan 0.6 boards


So I’ve been pottering away for a while working on getting the next version of the gang scan boards working. These ones are much nicer: thicker tracks for signals, better labelling, support for a lipo battery charge circuit, a prototype audio circuit, and some LEDs to indicate status. I had them fabbed at the same place as last time, although the service was much faster this time around.

A gang scan 0.6 board

I haven’t got as far as assembling a board yet — I need to get some wire thin enough for the vias before I can do that. I’ll let you know how I go though.


Trail run: Tuggeranong Stone Wall loop


The Tuggeranong Stone wall is a 140 year old boundary between to former stations. Its also a nice downhill start to a trail run. This loop involves starting at the Hyperdome, following the wall down, and the continuing along to Pine Island before returning. Partially shaded, and with facilities at the Hyperdome and Pine Island. 6km, and 68m vertically.


Trail run: Lake Tuggeranong to Kambah Pool (return)


This wasn’t the run I’d planned for this day, but here we are. This runs along the Centenary Trail between Kambah Pool and Lake Tuggeranong. Partially shaded, but also on the quite side of the ridge line where you can’t tell that you’re near the city. Don’t take the tempting river ford, there is a bridge a little further downstream! 14.11km and 296 vertical ascent.

Be careful of mountain bikers on this popular piece of single track. You’re allowed to run here, but some cyclists don’t leave much time to notice other track users.