HID Gamepad support

[a-chol]

Hello,
I am in the process of adding an analog stick to my keyboard (bminiex by winkeyless). I have researched current issues and the code repository and haven't found any previous work on the subject (apart from isolated work which doesn't seem to have been pushed into the main qmk repo)
I would like to expose it to the OS through an HID gamepad interface. The code works so far for this part, but it relies on modifications outside of my user keyboard configuration to add the gamepad HID interface to v-usb. Moreover, the key matrix could support sending HID gamepad button signals instead of keyboard events, which also requires changes inside qmk/tmk.
In this context, my issue is : does the qmk maintainers approve the integration of this feature?

• i should keep my modifications to myself, the qmk project is not about gamepads
  or
• this feature would benefit the whole project and a pull request adding it would be well received.

To advocate for the latter, there is already support for mouse and MIDI interfaces, so gamepad doesn't seem out of place. Moreover, I have seen people asking for advice to create their own gamepad controller (for homemade arcades) and qmk could be a good approach on multiple counts.
What do you guys think about this?

[yiancar]

I think anything that can be turn on/off using a DEFINE is welcome on qmk:) would have to be tested on all platforms, ARM and AVR, but we can help with that!:)

[jackhumbert]

@a-chol just to echo what @yiancar said - absolutely :) with the addition of the Qwiic Joystiic, this would be a great feature to build-out! Do you have your modifications on a branch somewhere?

[a-chol]

So here is where I'm at right now : a-chol@50a059c
I have chosen to rename the feature as Joystick instead of gamepad, which seems to fit the definition of the HID specification better.
Everything is protected by #ifdef ENABLE_JOYSTICK.
I support up to 6 axes and 32 buttons ; it is tested and working (I haven't pushed my personal keymap yet though)
I implemented hid report sending for joystick interface only for the v-usb driver for now. I don't have a keyboard that uses LUFA, but I can give a shot at an initial implementation.
I haven't built the electronics side of the project yet, but I'm planning on doing this in the coming weeks. I salvaged a dualaxis analogic joystick and the schematic of my kb (ps2avrgb) seems to indicate that I have a "hole in the matrix" available to connect two analogic inputs.
With the current implementation, one is able to control a virtual joystick from the keyboard, which is already a useful feature.
I'll create a pull request when the analog read is tested to be working, or sooner if you find the current state usefull.

[yiancar]

this would be good to get working on arm as well! I imagine u need to get the adc working for that?

[a-chol]

Yeah, I haven't checked what is available in the drivers/arm folder. The only function I use from the avr folder is analogRead.

[yiancar]

yea we will need to set it up but its quite easy. have a google regarding chibios adc. there should be a hal library for it. I mean that after you get it working on avr. I can have a look and help you as well when the time comes

[twschum]

Just like to voice my support for this! I am looking at building out a macropad + analog inputs device and would love to be able to use QMK for it.

[a-chol]

So this feature is materializing into PR #4226 .
I am in the process of doing the electrical wiring for the analog joystick I got. I thought everything everything was going the work out until the following problem : when one potentiometer is connected, all keys from the row get triggered.
I'm a bit of a noob in electronics, so my understanding might be off. Here is what I have done :

• I connected one pin of the 3-pin potentiometer to ground
• I connected one pin (middle pin) to a col (col01 in this schematic : https://raw.githubusercontent.com/showjean/ps2avrU/master/firmware/ps2avrGB/Circuit/ps2avrGB/20140829_140848_001.png)
• I connected one pin to a row read by one of the 8 ADC channel (row5 in the previous schematic).
  There is no key in my keymap at the intersection of col 1 and row 5.

My expectation was that the col was driven high when selected for reading, and when a switch is depressed, would drive the corresponding row input to high.

After reading the code, I'm not sure I understand how the matrix scanning works anymore : columns are set to low when selected (?), the others are never set to high : what does it matter to have column if they are never high? does setting columns as "input high" output current on the pin?
rows (inputs) are configured with pull-up resistor : I get from that that the default reading would be 1, except when connected to the ground? Does a switch depression short the pin towards ground?

I must be missing some basic knowledge of how MCU IO pins work, so a short explanation of what I am missing would be great.

[drashna]

You may want to check this out: https://docs.qmk.fm/#/hand_wire

[a-chol]

Thank you for the link. As I had understood,

send a logical 1 to the columns, one at a time, and read from the rows,

Then I don't get why each of the columns in qmk's scan_matrix are set to low successively, which should correspond to writing a logical 0.

[drashna]

IIRC, AVR doesn't support pulling high. But I'm not an expert with the hardware interface stuff, at all.
@jackhumbert, @skullydazed, or @fredizzimo would be more qualified to answer this. (and to which I'd likely open a PR to update docs)

[a-chol]

So I figured some things out regarding the circuitry that is a bit "wrong" in the handwiring guide.
Basically, input pins for the matrix are set with pull-up resistor. It means that the "default" tension to the input pin is 5v (Vcc) as the input reads high. As soon as "some" current (more than the current that flows through the pullup resistor, if I got that well) is drawn towards the ground, the input pin reads low. This is required to properly detect circuit closure when contacts of a switch might be a bit oxydized and conduct current with a bit of resistance.
The consequence of this is that when a switch is pressed, current actually flows from the input pin (at Vcc) towards output pin. During matrix scan, all outputs are set high (at Vcc) except the current column, which is set low and consequently acts as a ground. That's why, when configuring the qmk matrix, you have to define the DIODE_DIRECTION to go from inputs to outputs.
The good news is that, since this is opposite to the current flow we need to set up to read an analog device on the ADC, we can seggregate the keyboard matrix circuit from the analog device circuit using diodes, meaning that probably anyone can add an analog device to his keyboard.
For my joystick, the working circuit is set up with diodes at the ADC pin, and at the ground pin, which is a pin turned to output low during analogRead.

[pahlers]

@a-chol Can you update the status of this project. I would like to use something like the Qwiic Joystick in a project. Thanks for you good work 👍

edit: Strike through the question. The pull request
#4226 is almost merged! Well done!

[amrsoll]

@a-chol I presume this issue could be closed?

[drashna]

the HID joystick feature should support this.