This project contains source code for:
- a Linux embedded system (in my case Olimex Lime2 A20 single board computer)
- digital radio transceivers (in my case a couple of Texas Instruments CC1110 boards)
The aim of the system is to provide via a web interface the possibility to open or close remote relays, which in my casea are attached to electrovalves that open or close some water flows (in my case to provide irrigation in my garden). However the target of the system is much more generic and you can attach to the "remote" node pretty much anything you like.
This project assumes that you have:
- an embedded Linux system, in particular Olimex Lime2 is assumed here. Moreover I tested this project only with a recent Debian-variant "armbian" installed, using DeviceTree overlays for accessing the SPI bus of the embedded system. See https://docs.armbian.com/User-Guide_Allwinner_overlays/. The DeviceTree config file used in my case is available here.
- two digital radios, based on Texas Instruments CC1110, operating in the 433 or 868/915 Mhz ISM bands. See e.g., http://www.ti.com/tool/CC1110EMK868-915 for the commercial boards used in this project. This choice is motivated by the fact that these frequencies provide high wall penetration and low battery consumption compared to other radio technologies like the well-known Wi-Fi.
The hardware design for the remote node is available as Cadsoft Eagle schematics (see https://www.autodesk.com/products/eagle/overview) in the hardware-remote folder. The design is based on 3 major parts:
- the CC1110 evaluation module which provides the antenna, the CC1110 radio+micro and its programming interface. See http://www.ti.com/tool/CC1110EMK868-915 ; this can be connected via SPI bus to an embedded SoC Linux system like a Raspberry Pi or similar boards. I used the Olimex Lime2. I documented the wiring between the CC1110 board and the Lime2.
- one or more commercial relay boards. These are usually unbranded chinese boards which you can find by googling for e.g. "DC 12V 2CH isolated high low level trigger relay module". Here's a picture of the one I used: The important aspect to keep in mind is that in my hardware design the CC1110 will drive the inputs of these relay modules directly (thus applying 3.3V as logic high signal) so that they must be both opto-isolated and sensitive enough (most modules out there expect 5V as logic high signal).
- a custom "glue" board to provide right power and cabling between the other 2 parts. I built this on a simple stripboard (https://en.wikipedia.org/wiki/Stripboard). This board connects the battery source (a 12V lead-acid battery in my case) to the radio module and relay module.
This is the overview of the custom glue board (extremely simple):
Finally a small caveat: typical electrovalves will require a positive pulse to move the internal valve to the OPEN position and a negative pulse to go in the CLOSE position. This requires the driving hardware to be able to invert the output polarity. This can be achieved using 2 channels of a relay module and wiring the electrovalve as shown in this picture:
Note that the normally-open (NO) contacts are attached to the 12V battery while the normally-closed (NC) contacts are attached to the ground. When no signal is applied to the relay module, the electrovalve has both its wires connected to the 12V and thus no current circulates. When one of the relay modules is triggered then the electrovalve will receive +12V or -12V. Thus the polarity applied to the electrovalve can be controlled by triggering just one of the 2 relay channels.
The current consumption budget of the remote node when the firmware puts the radio in sleep mode is:
- 100 uA for the 12V to 3V current regulator (in the hardware schematic shown above an ADP3333 low dropout 300mA-max regulator was chosen)
- 130 uA for the static resistor divider used for battery voltage probing
- 0.5 uA for the CC1110 (considering the power mode 2 used by the firmware of this project)
For a total consumption of about 240uA. The current consumption budget of the remote node when the firmware puts the radio in RX mode is dominated by the CC1110 and will be around 22mA. With current firmware settings the RX window lasts for about 3sec. Of course the "remote" node also needs to transmit an acknowledge to the "lime2" node (when a command is received in that RX window) raising current consumption up to 36mA but the TX time is so short that can be neglected in computations.
Assuming that a 7Ah lead-acid battery is used for powering the system, and that the sensor will wake up once every 6 seconds to check for commands over the radio channel, the battery life can be easily computed using e.g. https://oregonembedded.com/batterycalc.htm.
Data entered on that page is thus (considering current firmware version in this Github project):
- 7000mAh capacity rating
- 100uA + 130uA + 0.5uA ~0.240mA current consumption of device during sleep
- ~22mA current consumption of device during wake (radio in RX mode)
- 3600sec / (6sec + 3sec) = 400 number of wakeups per hour
- 3000 ms duration of wake time
The result is a battery duration of ~33 days.
The computations above do not include of course the current consumed to actually activate the relays and to move the electrovalve from the OPEN to the CLOSE position or viceversa. However such event is typically very rare (let's say once per day) and thus can be ignored to simplify computation.
Of course it's clear that the more the remote node remains in sleep mode the longer the battery will last. This resolves to a basic tradeoff in the design criteria: by increasing the sleep time the battery life is increased but the latency for delivering commands from the lime2 node to the remote node increases as well.
Firmware and Software Design
Both the remote node and the lime2 nodes have custom firmware written in C and based upon Texas Instruments BSP and MRFI packages for CC1110 (its datasheet is available on Texas Instruments website at http://www.ti.com/lit/ds/symlink/cc1110-cc1111.pdf) The directory with the source code is "firmware-cc1110-lime2-remote"; it contains a project for the IAR Embedded Workbench for 8051 IDE. Unfortunately if you need to change it (you will probably need unless you use exactly the same evaluation module from TI with the same pinout), you have to either buy the IAR software or use the 30days evaluation trial; indeed the 4-kb limited edition is not suitable for any project employing CC1110 radio.
The firmware can be built in 2 modes: REMOTE and LIME2 to produce the firmware for the 2 peers of this project. SPI protocol documentation is available and provides an overview of supported SPI commands. Radio protocol documentation is available as well.
The firmware logic of the LIME2 node can be summarized as follows:
- poll the SPI RX buffer looking for commands
- whenever an SPI TURNON/TURNOFF command is received from the Lime2 Linux system, then a radio packet is filled with that command and sent (there is no real addressing logic: we assume only 1 receiver will be present); for each transmitted command, the system waits for an over-radio acknowledge and, if that arrives, the data contained in the acknowledge packet is saved; if it does not arrive, the command is retransmitted until the acknowledge arrives;
- whenever an SPI STATUS command is received from the Lime2 Linux, then the data contained from the most recent acknowledge packet is provided to the Lime2 Linux.
In practice the LIME2 node acts as an SPI-to-radio bridge. The presence of the "transaction ID" in both command packets and acknowledge packets allows all parts of the system to discard duplicates (e.g. the remote node can discard multiple commands having the same transaction IDs, which may be transmitted by the lime2 node when the acknowledge packet gets lost) and the Lime2 Linux system to understand if a command has been successfully received by the remote node.
The firmware logic of the REMOTE node can be summarized as follows:
- sleep for a certain amount of time, using a low-power timer to wake up the MCU;
- when the MCU wakes up, the radio is turned on in RX mode and the remote node waits for a certain time window for incoming commands;
- whenever a radio TURNON/TURNOFF command is received from the Lime2 node, then the command data is saved and an acknowledge radio packet is generated and transmitted; the system then waits some time to allow the Lime2 node to receive the acknowledge and stop repeating the command. After such "cool down" time window, the remote node drives its GPIO pins to turn on/off the relay modules which in turn will open/close the electrovalves.
The presence of the ADC read of the battery in each acknowledge allows to exploit the uplink channel remote->lime2 node to also provide an indication of how much battery power is left.
Tree of contained source code is:
Main source files to understand the application:
TO BE WRITTEN
This how my final assembly of the remote node looks like:
These are instead a screenshot of the web interface:
It is very very simple and is currently integrated with my other project: http://github.com/f18m/light-media-center
As of Aug 2018 I put "on the field" (literally) my assembly with the software and firmare of this GIT repo. It works well so far. I will develop and maintain this project in my free time, though I don't plan to add any new feature. Eventually I will improve the look and feel of the web interface but that's pretty much it.
The most comprehensive project similar to this is MySensors: https://www.mysensors.org
Similar projects designed to drive relay boards that are directly connected to the Linux system (without the radio bridge):
- http://www.logicaprogrammabile.it/come-pilotare-elettrovalvola-bistabile-usando-2-rele/ (in Italian)
Similar projects designed to drive REMOTE relay boards (not BATTERY powered though):