SPI interface Tutorial.

The SPI Interface (Serial Peripheral Interface) bus is a high speed, 3-wire, serial communications protocol (4 if you include SSn - see below). Its primiary purpose is to reduce on-PCB wire routing by replacing the traditional parallel bus with a serial interface. (You can just about manage an 8 bit bus routing it through a several layer PCB but when you get to 16, 32 bits and more it gets far more difficult).

The connections are:

• MOSI (Master Out Slave In)  [SDO].
• MISO (Master In Slave Out)  [SDI].
• SCK  (Slave Clock) [SCK]. 
• SS_n (Slave Select).

            [] - denotes PIC nomenclature

Note: The last signal SS or slave select is separate from the protocol and is usually implemented as an enabling control pin from the microcontroller. It is included here for completeness.

The SPI PIC interface allows connection of peripherals using a high speed serial interface. SPI Flash Memory and SPI SRAM can easily be added to any system. Other types of device include:

• ADC.
• DAC.
• Flash.
• EEPROM.
• Accellerometers.
• LCDs.
• Temperature & Humidity sensors.
• ...and many more.

Its only other real competition is the I2C bus which is why you often see these interfaces both available on processors and microcontrollers. 

The SPI interface was designed in the 1970s by Motorola, who used it in their 68000 processor, and it was quickly adopted by many other manufacturers as a defacto standard. 

It is intended for transmission of data from a master device to one or more slave devices over short distances and at high speeds (MHz).

How it Works

It works by transferring data one bit at a time between two devices with the master device sending the clock signal (SCK). The clock controls the timing of the data transfer.

Data (MOSI [SDO] )is sent out of a shift register in the Master SPI device along with a clock signal (SCK) while at the same time another shift register receives data from the slave (MISO, [SDI]).

The Master is always in control and initiates data transfer using the clock signal. Slave devices are selected using a separate slave select signal that is software controlled i.e. those signals are separate from the SPI hardware module.

SPI interface Clock Definition

The SPI clock is not defined at all - meaning that different slave devices can assume different clock operation; One slave may require an idle clock that is high, while another may require an idle state of low. Others will react to the rising edge while others react to the falling edge!

The SPI interface protocl copes with this by letting you program these details to the SPI hardware module.

The following image shows the PIC interface for SPI (all SPI hardware modules in other microcontrollers will allow the same operation just using different registers).

SPI Interface PIC signals (extract from PIC datasheet DS39582B)

It is up to you to set the correct clock for you device when it is activated by a slave select signal (or in the case of a single slave it is always selected so there will be no slave select signal).

Arduino SPI Connection

The following diagram shows the interface to the SPI connections which also happens to be the programming connector (ICSP), on the Arduino board, that allows direct programming of the ATmega328p.

Location of the SPI connector on Arduino Uno R3

Photo Credit : kenming_wang (mod shows ICSP - modified with rectangle red highlite)

Different SPI Interface operational Modes

There are three ways of using the SPI interface

1. Single Master - Single Slave.
2. Single Master - Multiple Slaves - Chip selected.
3. Single Master - Multiple Slaves - Dasiy chained.

Single Master - Single Slave.

 Here is the setup for a single SPI device connection:

SIP Interface : Using chip selects

With this scheme you control each slave device using its chip select line (usually active low- red arrows show control lines).  When disabled the Data output from the slave goes into a high impedance state so it does not interfere with the currently selected slave and the slave's data input is ignored (check datasheet).

The advantage of this scheme is that you can consider (control) each device separately when you compare it with the daisy chain method - allowing connection of SPI devices that require different clock schemes.

If you have SPI slaves that operate using different clocks (edges/idle states) you can re-programme the master SPI hardware module before enabling a specific CS so each slave has the correct signals sent to it.

If you use the Daisy chain method then you need to make sure all the chips use the same clock edge and idle clock state.

SPI Interface : Daisy chaining

With this scheme all data sent by the master is shifted into all devices and all data sent from each device is shifted out to the next (shown by red dotted arrow).  For this scheme to work  you have to make sure that each slave uses the clock in the same way and you have to get the right number of bits, so there is more work to do in software.

SPI Interface Daisy Chain

SPI Daisy Chain Example

Use several HC595 chips, daisy-chain linked, giving an easy increase in the number of outputs available without using many microcontroller pins.

The limits of operation are the speed of HC, the speed of the SPI output. Also design consideration e.g. If driving a set of LEDs that must be updated every 20ms.

Parallel Versus Serial SPI

The trade off between using a parallel interface and the SPI interface is speed e.g. if you read a parallel 12bit ADC at 200ksps then you could read the device at a 200kHz rate but if you want to get the same data rate using SPI then you need a serial speed of 200kHz x 12 = 2.4MHz.  So the actual trade off is speed and the consequential noise introduced into the circuit.

Summary

The SPI interface defines a very popular protocol that works at high speed. The main problem is that slave devices are not bound by any particular clocking scheme so they may operate differently to each other and that can make controlling them difficult i.e. the clock idle and edge must be set correctly for each device.

Advantages:

• High speed.
• Exremely simple interface that is easy to replicate in software.

Disadvantages:

• Slave chips are free to adopt any clocking scheme they require.