Powering off via UI while using a Mausberry Circuit

cyperghost

ATTENTION! Definate correct GPIOPin!

EDIT: I exchanged one line... I would rather do the echo 1 first
After that es-shutdown can be removed by script...
You may extent sleep timer?

The script is located in /opt/retropie/supplementary/emulationstation/emulationstation.sh

#!/bin/sh

esdir="$(dirname $0)"
while true; do
    rm -f /tmp/es-restart /tmp/es-sysrestart /tmp/es-shutdown
    "$esdir/emulationstation" "$@"
    ret=$?
    [ -f /tmp/es-restart ] && continue
    if [ -f /tmp/es-sysrestart ]; then
        rm -f /tmp/es-sysrestart
        sudo reboot
        break
    fi
    if [ -f /tmp/es-shutdown ]; then
        sudo echo "1" > /sys/class/gpio/gpio$GPIOpin1/value && sleep 3 #<-- very likely here
        rm -f /tmp/es-shutdown
        sudo poweroff
        break
    fi
    break
done
exit $ret

caver01

@cyperghost
I am away from my pi at the moment so I cannot test, but I got to thinking about this. . . If we are just setting a variable to trigger the sleeping MB script, it will end up sending the same sudo poweroff command from the mausberry script . . so I am going to make a guess here that nothing will change. It will still keep the MB circuit itself in limbo and the Pi will be off/asleep, but not completely powered off. In other words, I am predicting no change.

Here's why: I suspect the MB circuit itself is monitoring the actual pin voltages, and all we are doing here is triggering the software shutdown. While that's all well and good, we may not be giving the MB hardware any indication that it should be preparing to cut power like it does when we press the button. This is my prediction anyway.

With that possibility in mind, and recognizing that I am a bit of a GPIO novice, is there a way to force the signal on the GPIO pin itself so that the MB circuit hardware really DOES detect something? This would get picked up by the script too.

cyperghost

@caver01

With that possibility in mind, and recognizing that I am a bit of a GPIO novice, is there a way to force the signal on the GPIO pin itself so that the MB circuit hardware really DOES detect something? This would get picked up by the script too.

You send a trigger to the MB script! Every second the MB checks the state of the GPIO! So I would extend the sleep timer in the script posted above to at least 2 seconds.

Type:
nano /opt/retropie/supplementary/emulationstation/emulationstation.sh
and just add this one line between the if-then-by es-shutdown clause

sudo echo "1" > /sys/class/gpio/gpio$GPIOpin1/value && sleep 3

Please definate $GPIOpin1 value! I can't see it as my glass globe is broken now. The modified MB-script checks existance of ES-running and some emulators and then exits and gives command back to the original ES-shutdown script. As there is no emulator running and ES is already shutting down the MB script will take command back to ES and ES ends as usual but the GPIOPin from MB is changed!

It will still keep the MB circuit itself in limbo and the Pi will be off/asleep, but not completely powered off. In other words, I am predicting no change.

No - I'm pretty sure it works - maybe not by this attempt but in general it should ;)

caver01

@cyperghost said in Powering off via UI while using a Mausberry Circuit:

Please definate $GPIOpin1 value!

Sorry, I don't know what this means.

Thanks for the added explanation. I understand how the updated Mausberry script works. It's so great that it looks for emulator PIDs now and does a graceful shutdown.

My point was that if all we are changing is the value of a variable defined by the script--and not an actual pin voltage-- I wonder how the outcome would be any different--as it still leads to a sudo poweroff.

But, if changing that variable also triggers a voltage change on the GPIO PIN itself, we are probably good to go.

What I suspect happens when the physical button is used is more than just falling out of the sleep loop. I may be mistaken, but the mausberry hardware is actually cutting power to the Pi and going into a dormant state itself--something it doesn't seem to do if I just trigger a software shutdown. That makes me think the circuit is doing something in hardware when it detects a voltage change on the pin it is monitoring IN ADDITION to our script getting triggered. Even the LED stays brightly lit from the Mausberry circuit. It "thinks" the Pi is still running. We need to send the signal to the circuit to trick it into thinking the button has been pressed.

In any case, I won't be able to test this until I get home in 5-6 hours.

cyperghost

@caver01 Well it could be that the switch triggers some hardware thing to the mausberry. But keep in mind

with echo 1 the GPIO sends 3,3V to the mausberry
with echo 0 the GPIO sends 0,0V to ther mausberry

You can connect a simple LED to a GPIO and ground directly and just send 1 to let the LED glow and send 0 to switch it off ;)
I'm pretty sure it works....

In theory there is no difference if you press the button to send the 1 to the Pie or do it via software and tell the GPIO - set to 1 now ;)
Maybe you're right and the hardware button does something with the mausberry but as far as I understood the mausberry works with permanent and momentary switches ;) So it just triggers the GPIO ....

caver01

@cyperghost said in Powering off via UI while using a Mausberry Circuit:

ATTENTION! Definate correct GPIOPin!

Are you saying we should adjust the line as it applies to our setup--so if we used pin 23 for the for GPIOpin1 as suggested by Mausberry, the line should be:

sudo echo "1" > /sys/class/gpio/gpio$GPIOpin23/value && sleep 3 #<-- very likely here

cyperghost

No!
sudo echo "1" > /sys/class/gpio/gpio23/value && sleep 3

caver01

@cyperghost Ok, so setting this value really does affect the pin voltage. That's was my hope, fearing that it might just be a script variable.

cyperghost

@caver01 You can just test without altering any script. Quit ES and type in terminal the command above. Maybe the sudo command is not needed :) (I doubt it but let us find out)

caver01

@cyperghost said in Powering off via UI while using a Mausberry Circuit:

No!
sudo echo "1" > /sys/class/gpio/gpio23/value && sleep 3

Hahaha. Gotcha. Ok. I will give this a try later. Thanks so much, and especially for your patience. You were way ahead of me, going straight for the pin itself.

If this works, as I now suspect it will, you just closed the last little loop as I can see with using this circuit on a RetroPie setup. Thanks in advance!

@madmodder123 I hope you are following along and can try this too. I think we were having the same issue (actually, everyone who uses a MB circuit can benefit).

caver01

@cyperghost said in Powering off via UI while using a Mausberry Circuit:

@caver01 You can just test without altering any script. Quit ES and type in terminal the command above. Maybe the sudo command is not needed :) (I doubt it but let us find out)

Good idea. Consider that my first test. I will report back later.

caver01

@cyperghost I tried the command at CLI. got an error:
echo: write error: Operation not permitted

caver01

@cyperghost after a little reading I tried something else:
First, I considered the fact that pin 23 is exported and set to input. That may have been preventing me from setting the value, so I tried changing it to output with:
sudo echo "out" > /sys/class/gpio/gpio23/direction
then I tried setting the value to 1 with
sudo echo "1" > /sys/class/gpio/gpio23/value
this immediately triggered the shutdown. However, the LED wired to the mausberry is still lit, the Pi is sleeping but power still on. And in this state, the power switch is non-functional, as if the MB circuit still thinks the Pi is on, like it has not seen a button press.

It strikes me that the power button is wired to the MB circuit, not to a GPIO pin. The MB circuit is wired to pins 23 INPUT and 24 output. So, the button tells the circuit to shutdown and the circuit triggers the input on pin 23 which kicks off the script. I honestly don't know what pin 24 is doing. I expect the MB is monitoring this pin, so when the power goes out, it sees that the pin is no longer set to "1", but it appears that the circuit does not bother to look at 24 unless it has received a button press from the wired button.

I am starting to think I need a transistor in this equation, or that the MB circuit is not as versatile as I wish it was. It should detect if the pi is shutdown, not just when the button gets used.

cyperghost

@caver01
I See .... That is because the switch is labeled as input.
So there is no chance to with an simple echo AS it expects an input. I will report in a few hours ...
I have a small test circuit available at home. One LED connected with a button.

cyperghost

@caver01 I was on the wrong track - sorry!
But there is an solution... you need a Solid State Relais like this and connect the output in parallel direction to your switch. So you can use the switch or the SSR...

The innput for the SSR is annother GPIO in ouput direction... if you set now echo 1 command in this GPIO direction.. LED inside the relais will glow and switch is closed.

So you have 2 switches ... the physical and one software remoted. I have some of these bought cost 1$ each and work fine.

@caver01 What happens if you change direction to the second GPIO?

caver01

@cyperghost

What happens if you change direction to the second GPIO?

If we change direction on pin 24, I can’t write to it. I tried change the value too and it does nothing.

I think I will look into the relay option. Do I need any other components such as resistors connected anywhere? I don’t think I have ever hooked up a relay before.

cyperghost

@caver01 I'll be back ;)

I tried change the value too and it does nothing.

I can't get this... Was there an error message? So can you please test if you change status uf $GPIOpin2 with sudo command
sudo echo "1" > /sys/class/gpio/gpio$GPIOpin2/value
It is setted to 1 what happens if it is setted to 0?

I don't get what the purpose. It seems like a check state... for, is the Pie is running!
As the mausberry contains a AVR µc I can't get it why the developer sends voltage output and does not wait for receiving a signal?

As you mentioned:

is there a way to force the signal on the GPIO pin itself so that the MB circuit hardware really DOES detect something? This would get picked up by the script too.

and I said:

Maybe you're right and the hardware button does something with the mausberry but as far as I understood the mausberry works with permanent and momentary switches

the mausberry sends a high signal to the Pie..... sadly!

I tested with a switch and LED and I was able to trigger the switch by change input to output direction. But as the MB sends the signal.... it won't work like that.

About the opto-mosfet... yes you will need a resistor to not damage the build in LED. For 3.3 V I used 330 Ohm. But those values are in the data sheet delivered with the IC. I think that will be the nicest way to get the mausberry full working and you need just to plug one more GPIO. And this device acts as a switch... I think the resistance is just 1 Ohm... that's a very bad value for a "switch" but it will surly work.

caver01

@cyperghost
If I set the gpio2 (actually, it's wired to 24) to "0" it does nothing.

That pin gets set to "1" in the script, so it must be for monitoring. When the Pi finishes the soft shutdown, the pin will drop, and the MB circuit should see that, right? But my thinking is that the MB circuit doesn't check GPIO2 for drop to "0" until AFTER it detects a button press. Because, if it DID check all the time, it would immediately cut power before the script gets executed that sets it to "1"!

cyperghost

@caver01 Yes this behaviour sounds logical.
Man... it could be so easy but I think the SSR is also a good catch. I use one to activate and deactivate the monitor from LCD by switching on/off the video input via GPIO. So in fact it is just the same approach as to toggle a LED on/off.

caver01

It's hard to blame the design--if all you ever do is use the power button, it's fine. But if you do a soft shutdown without using the button, it seems as though the mausberry has no way to detect that, and it sits in a powered state.