Microcontroller options

[Thymoleptic]

Hello, Im hoping to get some assistance with speccing a controller base for a project. I work for a Aviation training school and want to simulate the running of a large engine for the students with the option of simulating defects later on with software. The controller and system will be mounted to the engine stand.

I made a mock up on arduino that allowed me to simulate and display a full start up, idle and shut down sequence (with leds, screen, fuel pump, ignitors etc.). I'm now looking to scale up the project to be more robust so have a list of stuff it needs to run and then also a list of nice to haves. Must: 12v fuel pump Large screen or 2x smaller ones Fuel pressure sensor a number or switches (both toggle and momentary) Ignition relays 12v linear actuator proximity switches (for actuator) speaker 28V starter motor

Nice to have (ability to read/monitor): speed sensor oil temperature sensor thermocouple Oil pressure sensor Air temperature sensor

I thought I might run out of ports trying to run on an arduino and then not sure how I could load in and out faults easily. So im up for trying out something else. Any knowledge would be greatly appreciated.

[voltsandjolts]

None of that needs micro-second or even milli-second response times, so using an arduino with port expanders would be fine. Make a list of exactly what you need (digital inputs, outputs and any analogue inputs, outputs) and add some spare. Here's an old video that talks about port expanders, no doubt there are more modern options;
www.youtube.com/watch?v=sKdUgGVf0jc

Remember that some arduino boards have 5V IO and others are 3V, so be careful if switching to a different board.

[brucehoult]

How many IOs do you need?

The ATmega2560 already has 86 GPIOs of which the standard Arduino board exposes 54.

It's not like an AVR is going to be lacking in speed for anything you'll want to do.

The Arduino GIGA R1 WiFi has 76 GPIOs.

Teensy 4/4.1 has 55 GPIOs but a vastly faster CPU.

Some STM32 chips have up to 168 GPIOs. I don't know what pre-made boards would expose that many.

WCH MCUs top out at 80 GPIOs for the CH32V307 or 95 on the CH32H417.

If you need more than the above kind of numbers then you can start chaining 595 shift registers and push bits into them at 20 MHz so putting one '595 on each real GPIO would allow you to update 8 output bits in under 1 µs (or over 10,000 output IOs in under a ms from a single real GPIO pin).

[ledtester]

What kind of screens do you mean by "Large screen or 2x smaller ones"?

What Arduino board are you using in your mockup?

[Doctorandus_P]

There are lots of DIY projects of flight simulators who have built whole cockpits around the PC hardware. That goes from simple I/O such as switches, potentiometers and LEDS, to reused old aviation equipment.

The amount of I/O is never a significant problem. There are simple ways to expand, such as with 74hc595 shift registers, or you can dedicate an 8 or 16 bit port as a databus, another port for an address bus and some more I/O pins for read, write and chip select signals. This way it's easy to get to several hundred or so I/O pins. More too, but latency or bus buffering becomes more of an issue. And as BruceHolt already mentioned, microcontrollers with a lot of I/O pins are readily available, even up to 160 I/O pins or there abouts. Working with a data / address bus does make the system more flexible (when using connectors). You can for example add modules later.

An FPGA can have more I/O pins, but it's probably overkill for this, and too complicated to set up.

Another approach is to use an industrial bus bases system. With for example modbus, you just need to add an RS485 driver (costs around EUR 2) and a bit of software to your uC, and then you can put a whole lot of uC's on that bus, and control them all from a single location. Main bottleneck here is throughput. With older 8-bitters, 115k2 is around the top for a "sensible" baudrate (the uC's also need time to do something with that data). With more modern uC's, 3MBaud is around the top what still makes sense. Such a bus does have quite a lot of overhead. Nearly half the data sent is for addressing and checksums, and you normally need two packets (command and response) on the bus to do anything. CAN bus is also an option.

If you want more throughput, microcontrollers that support ethernet are also available, and there are boards such as the beagle bone black. It runs linux, (including ethernet stack and more) and has 70+ I/O pins natively. The  Olimex A20-OLinuXino-LIME2 has 160 GPIO pins. This does necessitate using connectors with a smaller pitch and that does make it more fiddly to work with. This board also has Gbit Ethernet, but due to some hardware bugs, the throughput on for Ethernet is a bit disappointing. And there are many other SBC's too. Main thing here is that you to realize that using an SBC as "just an I/O expander" is a viable and affordable option. These linux boards can easily run without a monitor or keyboard attached and then be dedicated to a single (or multiple) user applications.

[voltsandjolts]

Just throwing in an off-the-wall option here, in case it's of interest;

If you want bigger displays like a computer screen, you can interface custom hardware via a MCU to USB. Then your custom hardware can trigger actions on a computer which runs your custom display software. For example, FreeJoy, FreeJoy Configurator and JoyToKey are used together by DIY hardware enthusiasts to map custom-built peripherals to keyboard and mouse commands. Model train enthusiasts love this kind of thing.

[Kilrah]

With this vague request and no info about how much development effort is possible pretty much anything can do.

Plenty of boards with displays e.g. https://de.aliexpress.com/item/1005006331465838.html
Then if you need lots of gpios since the timing doesn't seem critical you connect one or more boards to it, either dumb port expanders or other MCUs and communicate with them.