Is there a standard that propose voltage/current thresholds for GPIO in automotive ECUs. Something that would define ViH, ViL, Iil, IiH, etc for 12V or 24V networks.
I have good intuition of what is needed by working with different types of controllers, but I'd like to find something more rigorous to back up a design. The best I found is IEC-61131-2 which defines these levels for PLC.
There are no GPIO standards in the automotive world to speak of - they are not necessary. Even if I gave you one, I am absolutely sure that you wouldn't be able to meaningfully use it.
That's because discrete signals aren't designed in the way you seem to think they are: they are all purpose-designed inputs and outputs, there's nothing general-purpose about them. Anything general-purpose would use a standardized digital bus! That's what those are for!
Just because an input or output may have two logic states doesn't mean that there's any other commonality between them - because it'd be pointless. The remote dumb nodes (sensors and actuators) are the cost drivers, so they are designed to be cheap, and the control module has to deal with it - because it costs comparatively nothing to adapt a control module input or output to interface with those. Any sort of premature standardization would only drive up the costs of the nodes, while not appreciably lowering the cost of the control module (ECU, etc.).
With PLC I/O it's a different world, because each PLC application is bespoke, with low volumes, and nobody really cranks out a million identical machines with a PLC in them (unlike cars!). To do so would be a waste of money: it'd be cheaper with bespoke I/O, tailored to the sensors and loads. PLCs and the standards they mandate make sense in custom low-volume integrations, and solutions that evolve over time, like on a factory floor.
In automotive world, digital outputs are used to drive actual loads, and current capacity is the most important spec, while voltage drop is usually secondary, i.e. as long as the driver won't self-immolate while delivering rated load power, it's if not OK then at least functional. The outputs are designed to drive particular loads, and aren't treated like some general-purpose GPIO.
Digital inputs are mostly captured within a module, e.g. on my 15-year-old car there's exactly zero GPIO lines running on accessible wires - by GPIO I mean I/O that was designed to be "general purpose" and potentically could serve various application-specific uses. It just doesn't work that way. E.g. all buttons interface directly to network nodes.
Sensors with discrete outputs (such as an oil pressure switch) are subject to their own specifications, and it's up to the consumer of the signal to deal with it. Typically, the levels are determined at least partially by the signal sink. E.g. a short-to-chassis pressure sensor or overtemperature sensor's "idle high" voltage level depends on what the ECU feeds on that line - it may even be a voltage present only sporadically! - whereas the "active low" level depends on the current source setpoint in the ECU, coupled with various parasitic resistances along the circuit. The sensor itself is just a switch, not some GPIO signal source.
Same goes for discrete signals that feed, say, ignition coils: they are not universal GPIO outputs, but are designed specifically to drive a particular ignition module, or a specific range of them, and are designed to the spec of those modules.
The SAE SENT protocol defines an interface standard for automotive sensors.
More here: https://www.sae.org/standards/content/j2716_201001/
