Why are GPIOs used?

Question

I have been searching around [in vain] for some good links/sources to help understand GPIOs and why they are used in embedded systems. Can anyone please point me to some ?

Answer 1

I think you have it backwards. GPIO is the default in electronics. It's a pin, a signal, that can be programmed. Everything is made up of these. For a processor, dedicated peripherals are a special case, they're extras for when you know you want a more limited function.

From a chip manufacturers perspective, you often don't know exactly what the user needs so you can't make the exact peripherals on your chip. You make generic ones instead. Many applications are so rare that there's no market for a specific chip. Only thing you can do is use GPIO or make specific hardware yourself. Also, all (unused or potentially unused) pins are worth turning into GPIO because that makes the part even more generic and reusable. Generic and reusable is very nearly the whole point of programmable chips, otherwise you would just make ASICs.

Some particularly suitable applications:

• Reset parts (chips) in a system
• Interface to switches, keypads, lights (all they have is one pin/signal!)
• Controlling loads with relays or semicondctor switches (on-off)
  • Solenoid, motor, heater, valve...
• Get interrupts from single signals
  • Thermostats, limit switches, level detectors, alarm devices...

BTW, the Parallax Propeller has practically nothing but GPIO pins. Peripherals are made in software. It works very well for many uses.

Answer 2

In most embedded systems, a processor will be ultimately responsible for sensing the state of various devices which translate external stimuli to digital-level logic voltages (e.g. when a button is pushed, a pin will go low; otherwise it will sit high), and controlling devices which translate logic-level voltages directly into action (e.g. when a pin is high, a light will go on; when low, it will go off). It used to be that processors did not have general-purpose I/O, but would instead have to use a shared bus communicate with devices that could process I/O requests and set or report the state of the external circuits. Although this approach was not entirely without advantages (one processor could monitor or control thousands of circuits on a shared bus) it was inconvenient in many real-world applications.

While it is possible for a processor to control any number of inputs and outputs using a four-wire SPI bus or even a two-wire I2C bus, in many cases the number of signals a processor will need to monitor or control is sufficiently small that it's easier to simply include the circuitry to monitor or control some signals directly on the chip itself. Although dedicated interfacing hardware will frequently have output-only or input-only pins (the person choosing the hardware interface chips will know how many signals need to be monitored, and how many need to be controlled), a particular family of processor may be used in some applications that require e.g. 4 inputs and 28 outputs, and other applications that require 28 inputs and 4 outputs. Instead of requiring that different parts be used in applications with different balances between inputs and outputs, it's simpler to just have one part with inputs that can be configured as inputs or outputs, as needed.

Answer 3

The reason you cannot find an answer is probably because if you know what an embedded system is and does, or indeed anything about digital electronic systems, then the answer is rather too obvious to write down! That is to say that if you get as far a s actually implementing a working embedded system, you should already know what they are.

GPIO pins are as a minimum, two state digital logic I/O. In most cases some or all of them may also be interrupt sources. These interrupts may have options for be rising, falling, dual edge, or level triggering.

On some targets GPIO pins may have configurable output circuitry to allow, for example, external pull-ups to be omitted, or to allow connection to devices that require open-collector outputs, and in some cases even to provide filtering of high frequency noise and glitches.

Answer 4

In any useful system, the CPU has to have some way to interact with the outside world - be it lights or sounds presented to the user or electrical signals used to communicate with other parts of the system. A GPIO (general purpose input/output) pin lets you either get input for your program from outside the CPU or to provide output to the user.

Some uses for GPIOs as inputs:

• detect button presses
• receive interrupt requests from external devices

Some uses for GPIOs as outputs:

• blink an LED
• sound a buzzer
• control power for external devices

A good case for a bidirectional GPIO or a set of GPIOs can be to "bit-bang" a protocol that your SoC doesn't provide natively. You could roll your own SPI or I²C interface, for example.