Raspberry Pi to Arduino SPI Communication
There are many reasons to include a dedicated microcontroller when developing hardware projects based on a Linux single board computer like the Raspberry Pi. Whatever the specifics, typically the end goal is to take advantage of the microcontroller's built-in peripherals such as analog inputs and PWM outputs. Additionally, microcontrollers provide multiple timers, interrupts and other features to make easy work of precision real time processing and control.
With its enduring popularity, it's no surprise that an Arduino variant is often the go-to solution for such projects. The platform provides a robust, general purpose microcontroller appropriate for many projects, with a rich ecosystem of libraries, tutorials, and other resources.
This tutorial presents a basic framework for Raspberry Pi to Arduino communication and control using SPI - the Serial Peripheral Interface bus. SPI represents a very well established chip-to-chip communication methodology that is implemented in hardware on both devices. Here we will explore SPI in some detail, discuss hardware and software considerations, and develop a working example of a bi-directional communications scheme that could be adapted for any number of command and control applications.
Arduino at 3.3 Volts
Choosing an Arduino that runs at 3.3v will simplify your Raspberry Pi Arduino projects by enabling a direct connection between the input and output pins on the two devices without the need for level-shifting to align voltages and protect the Raspi inputs.
Arduino is often described as a 5 volt platform, but this is not entirely true. The ATMEGA328P-PU microcontroller at the heart of many standard Arduinos is actually rated to operate from 1.8v to 5.5v. Although the flagship Arduino Uno and a number of other products are designed for 5v operation, there are also plenty of official and unofficial 3.3v Arduinos to choose from, including the Fio, Pro, Pro-Mini and Lilypad. (Pro and Pro-Mini have both 3.3v and 5v versions.)
The "Arduino" used in developing this tutorial is an ATMEGA328P-PU in a barebones configuration on a breadboard. The setup is truly barebones. There is no voltage regulator or external clock crystal, just the chip powered by the Raspi's 3.3v tap and a 10K resistor to pull the reset pin high. (Of course it wouldn't hurt to put a few decoupling capacitors on the power rails.) In addition to the enticing simplicity and frugality of the circuit, the barebones configuration is ideal for connecting to the Raspberry Pi. Not only is the 3.3v operation Raspi-friendly, but foregoing the voltage regulator and crystal significantly reduces overall power consumption which is especially important on older Pis which have a 50mA limit on the 3.3v line.
I routinely use this barebones setup on small, battery powered projects. When brewing your own, there are a few considerations to running the 328P at lower voltages. The most significant concern is the need to run at a lower clock speed - 8MHz is the fastest clock that can be reliably supported at 3.3v. Another consideration is that the 328's brownout detection must either be disabled or configured such that the dropout is less than 3.3v.
The official Arduino site has a board.txt file here: From Arduino to a Microcontroller on a Breadboard that will provide the proper settings to run a minimalist board and load an 8MHz compatible bootloader. If you're new to burning bootloaders and building your own boards, this page will also provide a nice jumping off point to gathering the necessary knowledge and skills.
A Few Notes on the Raspberry Pi Environment and C++ Compiler
This tutorial was developed and tested using both Raspberry Pi Model B and Raspberry Pi 2 running the standard Raspian Linux distribution. The examples should work without modification on any current Raspberry Pi model. I run my Raspi "headless" (without monitor or keyboard) connecting from a terminal on a Linux PC via SSH, so virtually all of my interaction is at the command line. The code for the examples is written in C++ and compiled using the GNU Compiler Collection (GCC). GCC is part of the Raspian distro and should therefore be available without the need for installation or configuration. All of the examples here can be compiled using a very simple command line syntax:
Super user permissions are required for access to the SPI peripheral, so compiled examples will need to be executed using sudo, as in:
The SPI Interface
SPI is one of several viable options for Raspberry Pi Arduino connectivity. A few advantages of the SPI interface are widespread adoption, signal integrity, flexibility in protocol design, and simplicity.
SPI employs a master-slave architecture. The master device initiates and controls all communication with the slave. SPI communication is synchronous, meaning that data transmission between the two devices is synchronized by a shared clock signal. SPI communication is also full-duplex, meaning that data can flow in both directions simultaneously.
The master controls the data transmission via the clock line. Every time the master pulses the clock line, one bit of data is pushed from the master to the slave. Simultaneously one bit of data is pushed from the slave to the master. As a practical matter, and for our purposes here, data can be thought of as moving at the byte level. The master and slave each load a byte into their respective SPI data registers, and the master pulses the clock line eight times to push one byte in each direction. Although there are various hardware flags that can be set and interrupts that can triggered, there is no SPI protocol. The simultaneous transfer of one byte of data between the two devices is the entirety of the process. If required, acknowledgement signals, error checking, packet definition and other common protocol features can all be handled in software.
The SPI bus itself is made up of four physical connections: two data lines, a clock line, and the slave select line. The master must have one slave select line for each slave, so the number of required slave select lines could theoretically grow large.
Nomenclature and labeling for the clock and data lines are identical for the Raspberry Pi and Arduino:
- SCLK : Serial Clock (output from master).
- MOSI : Master Output, Slave Input (output from master).
- MISO : Master Input, Slave Output (output from slave).
However on the slave select line, nomenclature is slightly different between the two platforms. On the Arduino the line is called Slave Select and is abbreviated SS. On the Raspberry Pi, the process is called Chip Select, abbreviated CS, and the CS lines are referred to as Chip Enable lines, abbreviated CE. The PI actually makes two slave select lines available identified as CE0 and CE1.
In contrast to some serial bus schemes where the transmission line (TX) from one device is connected to the receive line (RX) on the other, in SPI the four lines on one device are wired directly to their counterpart on the other as shown in the diagram below.
The physical location of the SPI bus on the ATMEGA328P-PU is straightforward as depicted in the pinout diagram below. SS is on physical pin 16, MOSI on pin 17, MISO on pin 18 and SCLK on pin 19. If you are using a manufactured board instead of a raw chip, these physical pins correspond to the Arduino digital pins 10 to 13.
A pinout diagram for the Raspberry Pi GPIO header is even more straightforward. Following the pinout diagram below (and consistent among version A, A+, B, B+, and 2B) MOSI is pin 19, MISO pin 21, and SCLK pin 23. In this tutorial we're using CS CE0 which is on pin 24.
Enabling SPI on the Raspberry Pi
SPI on the Pi is disabled by default. Before you can use SPI in your programs it must be enabled. The easiest way is via the raspi-config utility which you can launch by entering:
at the command line.
In the resulting screens select Advanced Options and then select SPI. You should answer yes to both questions to enable the SPI interface and load the SPI kernel module by default.
Reboot the Pi to load the SPI kernel module. You can verify that the change was effective by entering lsmod at the command line to generate a listing of loaded modules. The SPI module spi_bcm2708 should appear in the results.
This should all work fairly effortlessly, however, if the SPI module fails to load you may find yourself needing to troubleshoot the activation. It is possible to manually examine the Pi's configuration files to ensure the settings are correct using nano or the text editor of your choice. Note that if you need to make changes you will need to be running the text editor as sudo.
The Pi has a "blacklist" file which blocks modules from being loaded: /etc/modprobe.d/raspi-blacklist.conf. The line #blacklist spi-bcm2708 must begin with the hash symbol "#" which "comments out" the line. Without the hash, SPI is blacklisted and won't load.
Another potential culprit is the Device Tree parameter settings in the /boot/config.txt file. Look for a line in the file which reads dtparam=spi=on. If the line reads that spi is off, or the line is missing altogether, then edit the file accordingly. It is fine to simply make the line the last line in the file.
Remember that you will need to reboot the Pi before any kernel module changes will take effect.
Coding SPI on the Raspberry Pi
Raspian Linux provides several layers of abstraction to the SPI bus. To gain access to the bus, a virtual file is opened that points to the SPI device, and a file descriptor is assigned. Subsequent interactions with the SPI device are made by reading and writing to that file. The SPI virtual file will follow the naming convention of /dev/spidevB.C where B is the device number and C is the chip select. In our case, device 0 and chip select 0, we will open the file /dev/spidev0.0 for read/write access, setting the variable fd equal to the file descriptor:
With the file now open, communication with the SPI hardware is made via the ioctl (input/output control) system call. The spidev device driver provides an interface for passing information to and from the SPI hardware via ioctl. The next step before actually transmitting and receiving data is to set configuration options for the session. The spidev driver provides functionality to set/override various SPI options that include the SPI Mode for clock polarity and clock phase, the bit width of transmitted words, the bit order of transmission, and the maximum speed of transmission. For the most part, the Raspberry Pi Arduino communications project uses the Raspberry Pi's default settings: SPI Mode 0, 8 bits per word, and most significant bit first. (For clarity's sake, these are actually the default settings of the Broadcom SPI peripheral hardware itself.)
One setting worth considering is the maximum transmission speed. The Atmel reference for SPI states that the system clock of the slave must be at least 4 times the SPI transmission rate. Since the barebones ATMEGA328P we're using is configured for an 8MHz clock, transmission from the Pi master could be as high as 2MHz. Here we'll set the SPI clock speed to 1MHz with a system call to ioctl, using the spidev defined SPI_IOC_WR_MAX_SPEED_HZ and the value 1000000. Note that all values must be passed to ioctl by reference. First we assign the desired value to a variable (in this case speed and then rather than pass the actual variable, we pass a pointer to the variable using the syntax &speed.
The SPI device is now ready to transmit and receive data. The spidev driver establishes a data structure spi_ioc_transfer that is used to format both data and any temporary configuration data to be passed to the device via ioctl. Here we'll use that structure and create a variable named spi that can be passed by reference to ioctl. We'll also use memset to allocate memory for the struct:
Again, we're mostly content with the default settings for the transfer, so we'll only need to load the struct with three pieces of information:
- .tx_buf - a pointer to a variable holding the data to be transferred
- .rx_buf - a pointer to a variable to hold received data
- .len - length in bytes of tx and rx buffers
Two variables will need to be initiated for the transfer and receive buffers, which we'll name txDat and rxDat. Although spidev will enable multibyte transmissions, the Raspberry Pi Arduino project will only use single byte transfers, so the length value will be set at 1.
All that remains is to call ioctl and pass the spi struct to the virtual file:
The txDat byte will be transmitted to the Arduino and the returned byte will be in rxDat.
Putting the pieces together, the following function, spiTxRx, accepts a single byte, sends that byte by SPI, and returns the received byte:
Lastly, to enable the functionality described above, the following header files must be referenced for inclusion in your C++ source file:
SPI on the Arduino
The ATMEGA328P is equipped with a fully configurable high speed SPI peripheral that can function as either SPI master or slave. Interaction with the bus is conducted via three hardware registers - the SPI Control Register (SPCR), the SPI Status Register (SPSR) and the SPI Data Register (SPDR).
We will be configuring the device as a slave. As with the earlier configuration of the Raspberry Pi, we are able to rely mostly on the default settings of the ATMEGA328P, so there will be no need to make changes to settings for mode or bit order. There is also no need to configure clock speed on a slave device because the master drives the clock. While an Arduino SPI library does exist, it does not have support for SPI slave operation. Fortunately, the desired functionality can be easily established by direct register access. In fact, it only takes two lines of code to setup the SPI slave.
First the MISO pin has to be defined as an output pin. All other pins are configured automatically as input pins if the SPI is enabled:
Second the SPI enable bit needs to be set:
Reading and writing of SPI data is performed through SPDR. Programmatically you can treat SPDR as you would a variable. To read the contents of SDPR, it can either be accessed directly, or another variable can be set equal to it:
To load the data register with a value to transmit back to the master, the statement is reversed:
At the hardware level SPDR includes both an 8-bit shift register and an 8-bit receive buffer. When the slave is receiving data, that data is shifted into the shift register one bit at a time while the original 8-bits in the register are shifted back to the master. When a complete byte has been shifted into the register, that byte is then copied into the receive buffer. The receive buffer won't be updated again until the next complete byte is received.
There are several implications of this sequence of operations. A byte to be transmitted cannot be written to SPDR until the current transmission cycle is completed, but must be in SPDR before the next transmission cycle begins. The byte in the receive buffer must be read before the next transmission cycle is completed or it will be overwritten. However, reading SPDR is actually reading the receive buffer and writing to SPDR does not overwrite the receive buffer. Therefore it is possible to write the next byte prior to reading the received byte.
Because the master device will initiate and control communications by shifting a byte into the slave device, the slave requires a method for recognizing that a byte has been received. There are two possible methods for an Arduino/ATMEGA slave to monitor the SPI device - interrupt or polled. In an interrupt driven scheme, the interrupt enable bit in SPCR is set, and when a byte is received the Arduino immediately stops processing and executes a special function called an Interrupt Service Routine (ISR). In a polled scheme, the Arduino is coded to periodically check a flag in SPSR which will indicate that a byte has been received.
The choice to use interrupts or polling is based on the specifics of application being developed. Polling places a greater burden on software to continually monitor SPSR for a change. Interrupts create extra overhead to enter and exit the ISR, and extra care must be taken to avoid conflicts with other interrupt and timer dependent processes. In the code developed for this tutorial, polling is used.
The SPI End of Transmission Flag (SPIF) in SPSR is set when a byte has been received. Reading or writing SPDR clears SPIF. A quick method for checking the status of a single register bit is to mask off all the other bits in the register with an AND operation and then test whether or not the result is zero. For the tutorial, we'll use an IF-THEN construct to test for the flag and execute the code within the IF statement anytime the flag is set. If the executed code includes a read or write of SPDR the SPIF flag will automatically be cleared:
The SPI_Hello_Raspi and SPI_Hello_Arduino programs listed below, provide a demonstration of a minimal working Raspberry Pi Arduino communications system. In addition to putting some shape to the concepts introduced thus far, the demo can be invaluable in establishing that you have a working SPI connection, and that the correct kernel module and drivers are available and working on the Pi.
Otherwise, as with any "Hello World" program, the setup accomplishes little. On the Pi, a character array is established, and the ASCII characters for Hello Arduino are pushed through the SPI bus one byte at a time. On the Arduino the characters for Hello Raspi are stored in a similar array. As bytes are received on the Arduino, SPDR is loaded with the Hello Raspi characters and, driven by the clock from the Pi master, they are sent to the Pi one byte at a time. This sequence is in a loop that will repeat endlessly until the program on the Pi is forced to break. (Control Z will do the trick.)
The primary element to observe in the example is the coordination between the devices. The Raspberry Pi sends its byte and then pauses for 10 microseconds. The Arduino needs a small window of time in order to load SPDR with next the byte to send to the Raspi. The Arduino can't load the byte until the Raspi completes the current one-byte transmission, but the Arduino must have loaded the byte before the Raspi begins sending the next byte. On the Arduino side, the code loops endlessly until SPIF is set indicating the byte has been received and SPDR can be loaded.
Lastly, the astute will notice that the Arduino is not processing the incoming data. However, because of the master/slave architecture it is necessary to have the Raspi push one byte of data through for each byte it receives, so no harm was done saying hello to the Arduino. (In a less frivolous application, one might push through a series of zeros to accomplish the same result.)
Hello Arduino (Raspberry Pi Code)
Hello Raspi (Arduino Code)
A More Robust Example
To provide a more robust example of Raspberry Pi Arduino communication we'll now present a simple application in which the Pi sends a command to the Arduino to add or subtract two numbers and the Arduino returns the result. Creating an application that solves a math problem avoids the need to interface any additional hardware or other circuitry while providing a fairly straightforward troubleshooting path. Nonetheless, the application has many fundamental elements of a master slave system and it should be easy to appreciate how the techniques could be applied to a more realistic scenario such as reading an analog sensor.
As seen in the Hello Raspi application, accomplishing even the simplest of exchanges between devices requires establishing and adhering to a rigid set of expectations regarding the timing and sequence of communications and other events - in other words a communications protocol. We've actually already made quite a bit of progress toward developing a basic, general purpose protocol that can be expanded or adapted for specific purposes.
By accepting certain default settings of the Raspberry Pi and Arduino, we've established that the evolving protocol uses SPI Mode 0, transmission words are 8 bits, and the most significant bit is transmitted first. Additionally we've determined that transmission rate is 1MHz with a 10 microsecond delay between bytes.
To support our math application we'll be adding just a little more structure. Specifically, we'll add a handshaking step to the beginning of each transmission, and we'll also establish a fixed packet size for transmissions.
The handshaking step to be added is as simple as they come. The Raspberry Pi will send a one-byte start code ("c") which the Arduino will acknowledge with a one-byte acknowledge code ("a"). Both devices will loop until the sequence has been successfully executed. As simple as this technique may be, it effectively addresses two potential problems that are likely to crop up as soon as you begin incorporating SPI communications into actual projects.
The first issue being addressed is that using only the SPI bus, an SPI master has no way of confirming that the slave is actually online and receiving information. If the Arduino were offline or otherwise unavailable, the Raspberry Pi would cheerfully send a stream of data into the ether. The Pi will also quite happily read the incoming data as a string of zeros or whatever random noise may be present on the MISO line.
The other issue addressed by handshaking is the potential for the slave to either miss a byte somewhere along the way or start in the middle of a transmission. Should either scenario occur without a mechanism to realign the master and slave, transmission from that point onwards would be somewhat hopelessly out of synch. With the handshaking sequence, the two devices should realign on the very next transmission.
In most situations communication between the master and slave device will involve a multi-byte packet. The optimal size of the packet depends very much on the application being developed. If the application is sending large amounts of raw data, a larger packet size might be in order. If the application is sending short command sequences, then a much smaller packet might be called for. Likewise, the decision on whether to use fixed or variable packet length throughout the application will depend on the nature of the application.If you look to the Raspberry Pi source code below you'll see that the command to either add or subtract two numbers is called in the following format:
Where sendCommand is a function, command is a single byte instruction, and parameter1, parameter2, and results are all two-byte integers. When called, the function packages the command and parameters, and transmits them to the Arduino by SPI. The Arduino performs the calculation and transmits the results back to the Pi, where sendCommand will return the Arduino's response as the result of the function. The program uses the letter 'a' as the command for addition and the letter 's' as the command for subtraction. A command to add two numbers appears as:
Either command is the same in terms of length, therefore the optimal packet size is simply the number of bytes required to conduct the handshake, send the command and two integer parameters, and return the result: 8 bytes.
The graphic below depicts the packet and can serve as a guide to understanding the sequence of events:
Of Bytes and Integers
In transmitting data between the Raspberry Pi and Arduino we encounter a very common problem in serial communication. The SPI unit of transmission is a byte, but the parameters and result required for our application are two-byte integers. In order to work fluidly we must have a method of splitting the integers into two bytes for transmission and then reassembling the bytes into an integer once they are received. Bit shifting is one standard method of dealing with this conversion requirement. However, here we'll use a perhaps more obscure, but nonetheless elegant solution. C++ has a construct called union which allows two or more variables to occupy the same memory space. In the snippet below a union named p1Buffer contains both the integer variable p1Int and a two byte array variable p1Char:
Later in the program we are able to seamlessly refer to the contents of the union by either referring to the integer - p1Buffer.p1Int or either of the bytes - p1Buffer.p1Char or p1Buffer.p1Char.
With the above description, the code below should be fairly straightforward to navigate.
On the Raspberry Pi, the sendCommand function serves as a wrapper for the spiTxRx function we introduced in the Hello Arduino application and executes our protocol step by step.
On the Arduino, we've added a function called spiHandler This function is called every time the SPIF is set and uses a variable named marker to keep track of its position within the packet. The switch case construct relies on the value of marker to be able to execute the appropriate action at each step of the protocol, from responding to the handshake request, to interpreting the command and parameters, to return the result.
As before the program itself executes in an endless loop performing first an addition problem and then a subtraction problem, and then sending the results to the console.
SPI Raspi Arduino (Raspberry Pi Code)
SPI Raspi Arduino (Arduino Code)
About the Author: Ralph Heymsfeld is the founder and principal of Sully Station Solutions. His interests include artificial intelligence, machine learning, robotics and embedded systems. His writings on these on other diverse topics appear regularly here and across the Internet.
Buster - A Voice Controlled Raspberry Pi Robot Arm
Buster is a fully voice interactive robot arm built around the Raspberry Pi. He acts upon commands given in spoken English and answers questions too.
iCE40 and the IceStorm Open Source FPGA Workflow
Project IceStorm is the first, and currently only, fully open source workflow for FPGA programming. Here, the software and hardware are discussed and a small sample project implemented.
Adding a Character LCD to an FPGA Project
Adding a text LCD to an FPGA project is a simple and inexpensive way to get your project talking. This post discusses interfacing an FPGA with garden-variety generic 16 character by two-line LCD.
An Arduino Neural Network
An artificial neural network developed on an Arduino Uno. Includes tutorial and source code.
From Boole to Bits - Claude Shannon's Digital Revolution
Claude Shannon propelled computer engineering into the modern age in 1937 when he published a paper demonstrating that Boolean algebra can be applied to the design of electronic circuits to express any mathematical or logical function.
George Stibitz and the Bell Laboratories Relay Computers
In the 1940s, driven by the innovations of George Stibitz and a heritage in switching technology, Bell Telephone Laboratories produced a series of increasingly capable computers using electromagnetic relay logic circuits.
Grace Hopper - Matriarch of Programming
Grace Murray Hopper enjoyed one of the most storied careers in computing history. From the earliest days working on the Harvard Mark I, her insights and innovations helped establish the foundations for modern, user friendly computers and launched the information age.
SCELBI, Altair and the Journey to Home Computing
In the late 1960s and early 1970s a network of electronics enthusiasts galvanized around the idea of building their own computers. The hobby computers they brought to market were an entirely new class of machine that created fortunes and shaped the personal computer industry well into the twenty first century.
The Commodore VIC-20 - The Friendly Computer
Released in 1981, the Commodore VIC-20 was the best selling computer of its day. Inexpensive and user-friendly, the computer opened up the market to a new group of consumers. It was the first computer many families owned, providing the earliest exposure to computing for countless future programmers, engineers and entrepreneurs around the world.
Migrating to the 1284P
The ATMEGA1284P is one of the more capable microcontrollers available in the hobbyist and breadboard-friendly 40-pin PDIP package. Here I discuss migrating the neural network project to the 1284p to take advantage of its relatively generous 16K RAM.
Getting Up and Running With a Tamiya Twin-Motor Gearbox
Tamiya makes a full line of small gearbox kits for different applications that are capable for their size and an easy, economical way to get a small to medium size wheeled robot project up and running.
Flexinol and other Nitinol Muscle Wires
With its unique ability to contract on demand, Muscle Wire (or more generically, shape memory actuator wire) presents many intriguing possibilities for robotics. Nitinol actuator wires are able to contract with significant force, and can be useful in many applications where a servo motor or solenoid might be considered.
Precision Flexinol Position Control Using Arduino
An approach to precision control of Flexinol contraction based on controlling the voltage in the circuit. In addition, taking advantage of the fact that the resistance of Flexinol drops predictably as it contracts, the mechanism described here uses the wire itself as a sensor in a feedback control loop.