VBUS - Low Power - Interrupt

[battosai30]

Hi,
I searched in all files and git issues but I could not find this information : is it possible to generate an interruption from VBUS ?
I think you understand why : I want to immediatly etablish a link with CMWX (in order to flash it or communicate with it) as he is in stop mode.
Thanks

[hallard]

Well, just depending on which I/O pin is connected VBUS, so depending on your board.

check variants files of your board for STM32L0_CONFIG_PIN_VBUS defined symbol GPIO.

For GNAT board, you can see variant.h it's STM32L0_GPIO_PIN_PA8

then in variant.cpp you search for STM32L0_GPIO_PIN_PA8 and see it's D24 (comments are false last is 24 not 25) and it has IRQ capabilities PIN_ATTR_EXTI

So attachInterrupt(24, usb_handler, RISING); may do the trick. Never tested, so let us now

[GrumpyOldPizza]

I am still pondering that one.

Is it good enough to add 2 APIs:

USBDevice.enableHotPlug()
USBDevice.disableHotPlug()

Idea being that at reset the code detects whether USB is connected or not. After the initial determination it only checks if hotplug got enabled.

[hallard]

Sot sure what's we are able to do on this point, you know better than I.

The idea is to be able to wake from sleep on hot plug and then do some things (like wait 5s to be able to flash the device or whatsoever )
Bu we need to know that we've been waked by hotplug, do able?

[GrumpyOldPizza]

Still toying with that. Key questions:

• do you just need a way to say "please keep USB_VBUS detection in STOP mode"

• do you need a way to detect via callback whether USB_VBUS has something attached to ?

• do you need a way to detect via callback whether a USB host has connected to your USB device ?

N.b. it is possible that a battery charger attaches to your USB port. So depending upon whether you need to know that the answers to the last 2 questions might be interesting to me.

[battosai30]

do you just need a way to say "please keep USB_VBUS detection in STOP mode"

It could be a solution and I think :

do you need a way to detect via callback whether USB_VBUS has something attached to ?

would be the same thing no ?

do you need a way to detect via callback whether a USB host has connected to your USB device ?

In my precise case, VBUS detection would do the trick, but It would be even better to detect a "real" USB connection ! But I guess it's not the same quantity of work ^^'

I precise that it's a custom board and I have PA8 well connected to VBUS from my USB connector.

[GrumpyOldPizza]

Mind checking out the code I dropped this evening on github ?

USBDevice:

bool begin();
void end();
bool attach();
bool detach();
void wakeup();

bool attached();
bool connected();
bool configured();
bool suspended();

void onConnect(void(*callback)(void));
void onDisconnect(void(*callback)(void));
void onSuspend(void(*callback)(void));
void onResume(void(*callback)(void));

void enableWakeup();
void disableWakeup();

begin() is called automatically as well as the first attach().

So in a nutshell, begin() sets up the whole internals and get USB_VBUS detection going. end() nukes everything. attach() attaches the protocol stack, so that when USB_VBUS gets detected, USB gets going. detach() removes the stack, but keeps USB_VBUS detection alive. wakeup() is the remote wakeup should you be in suspended USB bus mode.

attach() means USB_VBUS present, connected() means attached() and the USB stack active, configured() means that the USB host has set the USB configuration (speed, power ...), suspended() means the bus got suspended.

enableWakeup() means your USB_VBUS is tracked during STOP, and disableWakeup() disables that again.

[battosai30]

Waow ! I will test that tomorrow ;)

Big thanks !

[hallard]

Thomas,

Just to be sure to do the correct way

A)

• On startup call enableWakeup() set onConnect() and onDisconnect() callbacks
  • if usb is connected it works as usual USBSerial
  • if not all USB stuff is disabled and set to low power

B)

• if USB was connected at startup and we disconnect later
  • onDisconnect is fired, do we need to call end() or it's done by your end?

C)

• if USB is reconnected when cpu is sleeping or not
  • onConnect is fired, we need to call attach() and wait for connected() and configured() to get all USB stuff correct?

[GrumpyOldPizza]

A) Yes. If USB_VBUS is present at startup everything works like before. The USB stack is running. Depending upon the timing you might call the USBDevice.onConnect before the USB_VBUS logic says that the signal is good. So that is race condition I perhaps need to avoid.

B) Yes, then onDisconnect() fires. The USB stack is stopped automatically. If you call USBDevice.end() it also nukes the USB_VBUS detection (which you may or may not want to do).

C) Yes, onConnect() fires, but the stack automaitcally gets going. No need to wait for anything.

Looks like I need to rename attach() / detach() to something else. The idea is that "attach()" is really "if USB_VBUS is present, start the stack, otherwise do so when USB_VBUS is detected in the future". "detach()" is really "if connected, stop the STACK, if not connected, disable the automatic start of the USB stack on detecting USB_VBUS".

The APIs are meant for 2 purposes:

With onConnect() and onDisconnect() you could control for example as to whether your sketch owns DOSFS, or USB/MSC.

With attach() / detach() you can control STOP mode. Remember, as long as USB is connected L0 cannot enter STOP mode (it lingers then in SLEEP). So if it matters, you could call USBDevice.detach() before you enter STOP (for a longer time) and USBDevice.attach() after you leave STOP mode.

[GrumpyOldPizza]

An alternative scheme would be that "setup()" gets called via "USBDevice.begin()" not not with "USBDevice.attach()". Only when SerialUSB.begin() is called USBDevice.attach() is called from there.

This would allow you then to install the callbacks before you call SerialUSB.begin().

However in that variant I needed to add a "USBMSC" class, so that a USBMSC.begin() would also call USBDevice.attach() as side-effect.

[hallard]

Thomas,

With attach() / detach() you can control STOP mode. Remember, as long as USB is connected L0 cannot enter STOP mode (it lingers then in SLEEP). So if it matters, you could call USBDevice.detach() before you enter STOP (for a longer time) and USBDevice.attach() after you leave STOP mode.

Interesting, was not aware of. I saw when measuring power consumption that calling detach() was the trick to reduce consumption to minimal. I saw also that when I measure consumption when I upload and sketch start (so with usb connected) and few seconds later I remove USB the consumption was not going to "minimal'. Something related to USB stuff was still consuming.
So the only way to measure was then to reset the board with no usb connexion so the consumption was the one expected. Kris confirmed that measuring consumption should be done starting with no USB connected.
May be with these new features, it's solved, need to try that.

I'll try to write and post a sketch demonstrating this new USB Stuff, it will talk more than writing text.

[hallard]

Thomas,

I tried with mitigated success this following sketch on a gnat board (but should work on any other as soon as VBUS is wired on the correct pin and defined in the variant).

It output debug one Serial and also on SerialUSB if it's connected of course

See my comments below

#include "STM32L0.h"

// Choose FTDI Serial debug Port
// On Gnat D9 (Serial3) connected to RX of FTDI to we can
// see debug even if no USB
#define SERIAL_DEBUG  Serial3

#define debug(VAL)    { SERIAL_DEBUG.print(VAL);    if (debug_USB) {Serial.print(VAL);} }
#define debugln(VAL)  { SERIAL_DEBUG.println(VAL);  if (debug_USB) {Serial.println(VAL);} }
#define debugflush()  { SERIAL_DEBUG.flush();       if (debug_USB) {Serial.flush();} }

void ledOn() { digitalWrite(LED_BUILTIN, LOW); }
void ledOff() { digitalWrite(LED_BUILTIN, HIGH); }

bool debug_USB ;
bool VUSB;
bool configuredUSB;
bool connectedUSB;
bool suspendedUSB;
volatile bool changedUSB;

void usbSuspend()   
{ 
  debugln("** USB Suspend **");     
}

void usbResume() 
{ 
  debugln("** USB Resume **"); 
}

void usbConnect()   
{ 
  debugln("** USB Connected **"); 
  changedUSB = true;     
}

void usbDisconnect()
{ 
  debugln("** USB Disconnected **"); 
  changedUSB = true;
}

bool manageUSBState() 
{ 
  bool _usb = USBDevice.attached();
  connectedUSB = USBDevice.connected();
  configuredUSB = USBDevice.configured();
  suspendedUSB = USBDevice.suspended();

  // Enable debug output thru USB port if all is fine
  if ( configuredUSB ) {
    debug_USB = true;
  } else {
    debug_USB = false;
  }

  debug("VUSB attached="); debug(_usb);
  debug(" connected=");  debug(connectedUSB);
  debug(" configured="); debug(configuredUSB);
  debug(" suspended=");  debug(suspendedUSB);
  debug(" serialUSB="); debugln(debug_USB);

  return _usb;
}

void setup() 
{
  debug_USB = false;

  pinMode( LED_BUILTIN, OUTPUT);
  ledOff();

  SERIAL_DEBUG.begin(115200);
  while(!SERIAL_DEBUG);
  debugln("\r\n*** Setup Serial Debug OK ***");

  USBDevice.onConnect(usbConnect);
  USBDevice.onDisconnect(usbDisconnect);
  USBDevice.onSuspend(usbSuspend);
  USBDevice.onResume(usbResume);

  // Don't try to get USBDevice.attached() it's false
  // all related stuff may not be ready so classic wait 
  // USB serial port ready (or not) need to be done
  // Wait connected in 8s max
  Serial.begin(115200);
  while (!Serial && millis()<=8000 ) {
      ledOn();
      delay(10);
      ledOff();
      delay(750);
  }

  if (USBDevice.configured() ) {
    debug_USB = true;
    debugln("*** Serial USB OK ***");
  } else {
    debugln("*** Serial USB not configured ***");
  }

  // Enable usb plug to wake up
  USBDevice.enableWakeup();
}

void loop() 
{
  // Get and show USB State
  if (changedUSB) { 
    changedUSB = false;
    debugln("USB Change");       
    VUSB = manageUSBState();

    if (VUSB) {
      ledOn();
    } else {
      ledOff();
    }
  } else {
    debugln ("timer");
  }

  debugln("Going to sleep...");
  debugflush();
  STM32L0.deepsleep(10000);
  debug("Waked by ");
} 

So with this sample everything is working fine I can plug and unplug the USB, it's detected.
BUT the callback usbConnect does not wakeup the CPU, thus when I plug the USB I'm waked after timer expiration only (in the trace below it's happing 10s after plug).

So except I need to wait to be able to wake and do stuff, all is working fine. And once waked debug goes thru Serial and SerialUSB fine because as you see in the logs USB is configured()

So I decided to change callback usbConnect() to add a wakeup() as follow

void usbConnect()   
{ 
  STM32L0.wakeup();
  debugln("** USB Connected **"); 
  changedUSB = true;     
}

And I'm waked immediately but then, I don't have anymore debug on USBSerial, looking at the logs looks like the resume is happening after and thus my internal flag debug_USB is not positioned that's why I don't have any data on SerialUSB, check log below

So I decided to fire the wakup() on the callback usbResume()

void usbResume() 
{ 
  STM32L0.wakeup();
  debugln("** USB Resume **"); 
  changedUSB = true;     
}

I'm waked after resume, but still no luck, usb never get configured(), looks like wakeup() may have a side effect?

[hallard]

Ok, looks like I got it fixed added a delay in loop() after wakeup because of changedUSB and leaving the wakeup() into usbResume callback

  if (changedUSB) { 
    changedUSB = false;
    debugln("USB Change");       
    delay(2000);
    VUSB = manageUSBState();

    if (VUSB) {
      ledOn();
    } else {
      ledOff();
    }
  } else {
    debugln ("timer");
  }

I tried to replace this delay with a loop until configured() but sometimes it works sometime not, so the delay is the only thing working each time on my side;

here is the final sketch working on my side

#include "STM32L0.h"
#include "stm32l0_gpio.h"

// Choose FTDI Serial debug Port
// On Gnat D9 (Serial3) connected to RX of FTDI to we can
// see debug even if no USB
#define SERIAL_DEBUG  Serial3

#define debug(VAL)    { SERIAL_DEBUG.print(VAL);    if (debug_USB) {Serial.print(VAL);} }
#define debugln(VAL)  { SERIAL_DEBUG.println(VAL);  if (debug_USB) {Serial.println(VAL);} }
#define debugflush()  { SERIAL_DEBUG.flush();       if (debug_USB) {Serial.flush();} }

void ledOn() { digitalWrite(LED_BUILTIN, LOW); }
void ledOff() { digitalWrite(LED_BUILTIN, HIGH); }

bool debug_USB ;
bool VUSB;
bool configuredUSB;
bool connectedUSB;
bool suspendedUSB;
volatile bool changedUSB;

void usbSuspend()   
{ 
  debugln("** USB Suspend **");     
}

void usbResume() 
{ 
  STM32L0.wakeup();
  debugln("** USB Resume **"); 
  changedUSB = true;     
}

void usbConnect()   
{ 
  debugln("** USB Connected **"); 
}

void usbDisconnect()
{ 
  debugln("** USB Disconnected **"); 
  debug_USB = false;
  changedUSB = true;
}

bool manageUSBState() 
{ 
  bool _usb = USBDevice.attached();
  connectedUSB = USBDevice.connected();
  configuredUSB = USBDevice.configured();
  suspendedUSB = USBDevice.suspended();

  // Enable debug output thru USB port if all is fine
  debug_USB = configuredUSB;

  debug("USB attached="); debug(_usb);
  debug(" connected=");  debug(connectedUSB);
  debug(" configured="); debug(configuredUSB);
  debug(" suspended=");  debug(suspendedUSB);
  debug(" serialUSB="); debugln(debug_USB);

  return _usb;
}

void setup() 
{
  debug_USB = false;

  pinMode( LED_BUILTIN, OUTPUT);
  ledOff();

  SERIAL_DEBUG.begin(115200);
  while(!SERIAL_DEBUG);
  debugln("\r\n*** Setup Serial Debug OK ***");

  USBDevice.onConnect(usbConnect);
  USBDevice.onDisconnect(usbDisconnect);
  USBDevice.onSuspend(usbSuspend);
  USBDevice.onResume(usbResume);

  // Don't try to get USBDevice.attached() it's false
  // all related stuff may not be ready so classic wait 
  // USB serial port ready (or not) need to be done
  // Wait connected in 8s max
  Serial.begin(115200);
  while (!Serial && millis()<=8000 ) {
    ledOn(); delay(10); ledOff(); delay(750);
  }

  if (USBDevice.configured() ) {
    debug_USB = true;
    debugln("*** Serial USB OK ***");
  } else {
    debugln("*** Serial USB not configured ***");
  }

  // Enable usb plug to wake up
  USBDevice.enableWakeup();
}

void loop() 
{
  // Get and show USB State
  if (changedUSB) { 
    changedUSB = false;
    delay(2000);
  }

  VUSB = manageUSBState();
  if (VUSB) {
    ledOn();
  } else {
    ledOff();
  }

  debugln("Going to sleep...");
  debugflush();
  STM32L0.deepsleep(10000);
  debug("Waked by ");
  if (changedUSB) { 
    debug("Change");       
  } else {
    debug ("Timer");
  }
} 

[GrumpyOldPizza]

I need to recheck the code here. I ran into something similar yesterday where the code went throu a suspend/resume sequence, but did not properly restart up SerialUSB (and/or USB/MSC).

The USBDevice.configured() tracks directly the state of the USB Stack. So I am puzzeled why that would not show things properly.

With STM32L0.wakeup(). The rule is simple. If there is a callback and you have attached it, you need to call STM32L0.wakeup(). If there is a callback and you have nothing attached to it, it will wake you up. The idea is to allow calllback drive code that does not bail out of STM32L0.sleep() / STM32L0.deepsleep(). I am constantly forth and back on this one. On one hand that is a powerful tool. On the other hand, the set of events that will wake you up is so big to begin with that unless you attach callbacks to every single one, you'll be woken up most of the time anyway.

[hallard]

Thomas,

I tested the consumption on gnat with the above sketch, the graph below is when It finished to start and setup of course. Then I removed USB. As you can see you have peak every 10s (it's the wakeup) and between then some small ones may be related to USB looking to happen every 200ms

now reseting the same board (with no usb connected), wait the setup to finish and here the new graph, as you ca see all "glitch" in between the 10s sleep are not present anymore