A firmware mystery: What is the "autoload filament sensor?"  

A firmware mystery: What is the "autoload filament sensor?"

I'm preparing to submit a git commit to the Buddy firmware with a request to fix a bug, but I encountered something that I don't understand and I want to check with the community first.

The specific question I have is: Does the MK4/4S have an "autoload" filament sensor that is separate from the normal extruder filament sensor?

The problem that I'm trying to address is described in this GitHub bug report:

https://github.com/prusa3d/Prusa-Firmware-Buddy/issues/4474

...namely: when the printer runs out of filament mid-print, the UI can get stuck on a particular screen with no user input and no way to progress, forcing the user to power-cycle the printer and wreck the print. More specifically: When the printer detects filament runout, it automatically ejects the filament and displays the dialog: "Was the filament ejected?" If the user answers Yes, the printer sometimes switches to a screen reading: "Please remove filament from filament sensor" with no user interface options, and no response to fiddling with the filament or extruder - literally the only option is to reset the printer.

Initially, I chalked this up to my filament sensor going bad. The weird part is that after resetting the printer, when I instruct the printer to load filament, the filament sensor is perfectly fine: the printer asks me to insert filament, and when I do, it detects the filament and loads it. So it can't be a case where the filament sensor is flaking out during filament runout / reload and then works perfectly fine a moment later after resetting. I figured it must be a logic bug.

Since Prusa's team hasn't addressed this issue (and it kept wrecking my prints), I decided to roll up my sleeves and dig into the firmware. I tracked down the logic to a function in pause.cpp (excerpted here):

void Pause::filament_not_in_fs_process([[maybe_unused]] Response response) {
    ...
    if (!FSensors_instance().has_filament_surely(LogicalFilamentSensor::autoload)) {
    ...

In essence, this function goes like this: If the user responded "Yes" to "has the filament been ejected," check with the filament sensor to make sure that it isn't detecting filament. Wait for the filament sensor to respond "no filament" for a full second before continuing with the process.

So what's the problem? It's this part:

has_filament_surely(LogicalFilamentSensor::autoload)

Looking through the entire firmware codebase, every other time that the printer needs to check the filament sensor, it does this:

has_filament_surely(LogicalFilamentSensor::extruder)

The only place in the entire Buddy firmware that checks the "autoload" LogicalFilamentSensor instead of the "extruder" LogicalFilamentSensor is the filament_not_in_fs_process function. That's why it behaves so oddly.

So what is "LogicalFilamentSensor:autoload?" This is where things get weird. Looking in the file filament_sensor_types.hpp, I find this:

enum class LogicalFilamentSensor : uint8_t {
    /// Filament sensor on the current extruder
    extruder,
    /// Side sensor for the current extruder
    /// MK4+MMU: MMU sensor | XL: current side sensor | OTHER: none
    side,
    /// The first runout filament sensor - the one further from the extruder
    /// XL: side sensor | MK4+MMU: MMU sensor | OTHER: extruder sensor
    primary_runout,
    /// The second runout filament sensor - the one closer to the extruder
    /// XL,MK4+MMU: extruder sensor | OTHER: none
    secondary_runout,
    /// Filament sensor that is used to detect autoload
    autoload,
};

static constexpr size_t logical_filament_sensor_count = 5;

Okay, there's an "extruder" sensor, a few other sensors for the XL and MMU, and then... this "autoload" sensor, with no meaningful description. The rest of the firmware treats this sensor just like the extruder sensor, by polling it to detect its state and raise events and such.

The crux of the issue, I think, is that I don't believe that the MK4/4s *has* an "autoload" sensor separate from the main extruder filament sensor. If this sensor doesn't exist, then obviously its state never changes, it never gets any events, and the filament_not_in_fs_process function that is awaiting a "no filament detected" event on this sensor will wait forever - exactly what I'm encountering.

The solution here is very simple: just change has_filament_surely(LogicalFilamentSensor::autoload) in that function to has_filament_surely(LogicalFilamentSensor::extruder) to be consistent with the rest of the code. But before I submit that change to the GitHub repository, I want to see what people know about an "autoload" filament sensor in the context of the MK4/4S.