The Serial Peripheral Interface Bus or SPI bus is a very loose standard for controlling almost any digital electronics that accepts a clocked serial stream of bits. A nearly identical standard called "Microwire" is a restricted subset of SPI.

SPI is cheap, in that it does not take up much space on an integrated circuit, and effectively multiplies the pins, the expensive part of the IC. It can also be implemented in software with a few standard IO pins of a microcontroller.

Many real digital systems have peripherals that need to exist, but need not be fast. The advantage of a serial bus is that it minimizes the number of conductors, pins, and the size of the package of an integrated circuit. This reduces the cost of making, assembling and testing the electronics.

A serial peripheral bus is the most flexible choice when many different types of serial peripherals must be present, and there is a single controller. It operates in full duplex (sending and receiving at the same time), making it an excellent choice for some data transmission systems.

In operation, there is a clock, a "data in", a "data out", and a "chip select" for each integrated circuit that is to be controlled. Almost any serial digital device can be controlled with this combination of signals.

SPI signals are named as follows:

Most often, data goes into an SPI peripheral when the clock goes low, and comes out when the clock goes high. Usually, a peripheral is selected when chip select is low. Most devices have outputs that become high impedance (switched-off) when the device is not selected. This arrangement permits several devices to talk to a single input. Clock speeds range from several thousand clocks per second (usually for software-based implementations), to several million per second.

Most SPI implementations clock data out of the device as data is clocked in. Some devices use that trait to implement an efficient, high-speed full-duplex data stream for applications such as digital audio, digital signal processing, or full-duplex telecommunications channels.

On many devices, the "clocked-out" data is the data last used to program the device. Read-back is a helpful built-in-self-test, often used for high-reliability systems such as avionics or medical systems.

In practice, many devices have exceptions. Some read data as the clock goes up (leading edge), others read as it goes down (falling edge). Writing is almost always on clock movement that goes the opposite direction of reading. Some devices have two clocks, one to "capture" or "display" data, and another to clock it into the device. In practice, many of these "capture clocks" can be run from the chip select. Chip selects can be either selected high, or selected low. Many devices are designed to be daisy-chained into long chains of identical devices.

SPI looks at first like a non-standard. However, many programmers that develop embedded systems have a software module somewhere in their past that drives such a bus from a few general-purpose I/O pins, often with the ability to run different clock polarities, select polarities and clock edges for different devices.

The interface is also easy to implement for bench test equipment. For example, the classic way to implement an SPI interface from a personal computer to custom electronics is via a custom cable to the PC's parallel printer port. The parallel port generates and reads standard TTL logic voltages; +5V is high, ground is low. A number of helpful people have developed drivers to give access to this port in the most restrictive operating systems, such as Windows NT (see below), from the least likely environments, such as Visual Basic.