Communication buses

Presentation

Standard communication buses emerged from the need of getting MCU and IC communicating together. We will here present some of them.

Lines

Before discussing about buses, let's discuss how physical electronic lines work. Contrary to the "computer scientist" vision, an electronic line is not only "0" or "1", it is at a given voltage level, but also a certain impedance.

A line can be driven at low impedance using push/pull transistors:

Symbolically, this can be represented with a triangle, and called line driver or buffer.

Here, the input "B" is equivalent to pinMode in Arduino, or writing to DDRx register on the ATmega328P, it allows us to either drive the line or let it "floating".

Another possible configuration is "open-drain", that is like push-pull but without the upper transistor, allowing to pull the line to the ground, but letting it floating else:

When a line is input high impedance and nothing is plugged (the line is floating), any voltage can be read due to any kind of interference. If one wants to impose a "default" logic level, it should use the so-called pull-up/down resistors:

In the case of ATmega328P, each input/output can actually be set in 4 different modes:

  • Low impedance (output)
    • Output high level (ex: 5V)
    • Output low level (0V)
  • High impedance (input)
    • Input floating
    • Input with pull-up

Open drain mode can be implemented by writing 0 in PORTx and using DDRx to either set the pin as output or input.

More complex microcontrollers typically also offer pull-down resistor input possibility

Generalities

  • Voltage level?
  • Analog / Digital?
  • Synchronous / Asynchronous ?
  • Full duplex / Half duplex ?
  • Bandwidth ?
  • Robustness ?
  • Protocol ?
  • Efficiency ?

UART

UART, also sometime simply called "Serial", is one of the most famous buses. It is an asynchronous bus (without clock). This implies an agreement on the communication frequency, called baud rate.

Bytes are transmitted bit per bit (serial).

Bytes starts with a start bit and are followed by a stop bit allowing to keep synchronisation among lines.

TTL

Typically, UART uses two wires TX and RX. Each controller drives its TX pin and listens on its RX.

In the most standard form, UART bus is full-duplex.

RS-232

RS-232 is another UART standard using more robust voltage levels, typically -12V and 12V.

It is likely you already saw the historical connector on some old machines.

RS-485

RS-485 is another standard that is half-duplex and use 2 symetrical wires, the difference should be -200mV or +200mV to specify a 0 or a 1.

There are historical typical baud rates, most famouses are 9600, 57600 or 115200. UART can reach only some Mbit/s

I2C/TWI

I2C (Inter-Integrated Circuit), or TWI (Two Wire Interface) is a bus type that uses two wires, it is synchronous and half-duplex.

Those lines are pulled to high voltage level with pull-ups and driven by open drains

I2C is based on a protocol with address, arbitration and acknowledgements.

If multiple masters start a frame simultaneously, the arbitration is implemented by monitoring the line level (if LOW is read when the open drain is open, frame is stopped because of a collision).

Thus, addresses with the more 0s in the high part have more priority in the arbitration.

This protocol is very widespread, but is limited to 400kbit/s. There is however "fast" variants of I2C.

SPI

SPI (Serial Peripheral Interface) is a full-duplex and synchronous bus. On a SPI bus, there is one master and several slaves.

It is using 4 lines:

  • SCK: The clock indicating when bits are transmitted (synchronous)
  • MOSI and MISO: respectively Master Output/Slave Input and Master Input/Slave Output
  • CS: For Chip Slect, indicating to a slave if it is activated (this allows multiple slave to share MOSI and MISO)

Master is always ticking the clock, meaning that when a slave is responding, it is still ticked by the master.

SPI is widespread, and allows for high bandwidth (more than 10 Mbit/s)