[Solved] ESP 32 Timer Interrupts – ‘Unlimited’ Programmatic User Definable Timers

@ronbentley1 - I might have...  exposed a similar kind of capability in my library, but only for ESP8266 devices.  I prefer event style programming to the massively convoluted loop() methods.  Even when they call subroutines, it's still a mess to debug down the road.  Nothing worse than looking at my Arduino programs from five years ago!  😝 😫 

Here's how a user needs to implement a repeating callback of 1 millisecond.  

#include <InqPortal.h>

InqPortal svr;

void setup() 
{
    svr.onInterval(callback, 1);
}

void callback(void*)
{
    Serial.println(micros());
}

void loop(){}

... and the output.  It's not microsecond accurate... but good enough for most anything I can think of.

389573
390573
391567
392570
393565
394567
395562
396564
397559
398554
399556
400551
401553
402546
403549
404544
405547
406541
407544
408539
409541
410536
411538
412533
413535
414530
415533
416528
417530
418525
419520
420522
421517
422520
423515
424517
425512
426514
427508

@inq So is that using one of the hardware timers, or just a millis() loop.

Posted by: @zander

@inq So is that using one of the hardware timers, or just a millis() loop.

No, these are software timers.  I wrote the library directly on top of Espressif's code base.  I did not write it on top of the Arduino Core ESP libraries that ride on top of Espressif code. 

I use os_timer_setfn() and os_timer_arm() methods under the covers.  I did not design this interface to be used for things like bit-banging that requires more accuracy.  As you can see from the results... it is millisecond accurate.  I've even written sketches that send out data to browsers at 1 ms intervals.  I've also seen peak rates of 200 kB/sec total going out to multiple browsers... while reading sensors and setting PWM on multiple servos.  Remembering... that PWM is bit-banging software on the ESP8266!  It's why I like them so much.  

And yes... I'm sure I'd love the ESP32 even more, if I could dive into it as deeply as I'm invested in the ESP8266.

@inq Ok, I have been recently working one level down at the hw_timer level and having a problem, maybe I will use this api and see if the problem goes away. I don't see an esp32 equivalent document, but I imagine the esp32 and esp8266 are similar enough that the api is the same, I guess I will find out. 

@inq I just found another esp32 doc. I think this is one level above the hardware, but I have seen so many different esp32 timer/interrupt api's I am disoriented. I haven't tried this one yet so I will give it a try.

https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/system/esp_timer.html#api-reference

@zander - I think my cursory look over Espressif documents convinced me it might as well be two different companies of engineers creating the ESP8266 versus the ESP32.  Almost everything is different.  It's like they didn't even use the same core concepts.  Almost like Intel versus Motorola chips.  Which gave me serious pause to porting my library to ESP32's.  

@zander 

Hi Ron,

no millis(), instead it uses one of the 4 readily available hardware timers - timer 0 by default, but this can be 1,2 or 3 by changing the default timer number value (a macro).

I could have used a millis() approach but I am on the ESP 32 learning curve so wanted to explore its timer interrupt capabilities a little.

Ron B

@inq 

Hi Inq - The ETA framework design ensures that it does not matter if a timer interrupt is triggered during main code execution - it separates interrupt handling from the processing of any other part of the framework, so no clashes should be possible - it operates asynchronously to the timer interrupt ISR. The 'interface' between the ISR process and the framework is simply an interrupt counter - incremented by each interrupt and decremented as part of the framework's main loop processing.  The rules for reading and updating this counter are followed always which ensure no clash or conflict.

As for switch debounce, then yes, this is a pitfall for anyone starting out and trips up so, so many.  There is much on the net about this and I have suggested that this be a subject for one of Bill's videos - both hardware and software debouncing solutions. There are effective methods and libraries for handling this.

Cheers 

Ron

@inq 

Hi again,

Yours observations are sound and noted.

I would not, however, recommend use of the ETA framework for use requiring sub-millisecond processing, eg communications time out monitoring, etc.  Everything has its limitations and constraints!

Ron B

@inq 

Hi again,

Yes, its all down to preference isn't it.  I have never been a big fan of the call back approach but if the 'shoe fits' and all of that.

The important thing is that we have at our disposal a broad range of techniques.

You 'pays your money and you makes you choice'.

Ron B

Posted by: @ronbentley1

I was recently finalising some work after my hols relating to ESP 32 timer interrupts and thought some members may find an interest in it.

Hey Ron,

    Thanks for sharing...

Tom

@thrandell 

Thanks Tom, most grateful for your note.

Ron B

@ronbentley1 I spent a few weeks trying to get the hardware timers to work. My problem is I would set an alarm on timer 1 and get the interrupt on timer 2. NOT all the time, sometimes it took days. Found several esp examples that were slightly different in RAM macros, semaphore code, critical section code etc. 

I finally gave up which is not my usual way, I normally am like a dog with a bone.

If you have experienced anything like the above and found the solution I would be happy to hear about it.

FYI the entire logic is on button press set a 30 min alarm. When alarm fires set a boolean variable. The main code does all the rest. 

@zander 

Hi Ron,

No, I've not experienced anything like you describe, all very odd! Sounds like  something is getting its wires crossed? Perhaps some corruption somewhere during run time? Impossible to say at a distance.

Just for interest, can you post your code or is it the code you've previously posted on this thread?

Finally, you mention a button switch and I realise I have an outstanding update for you. When we last were discussing switches you had some difficulty installing my switch library. Well now it is available under the Arduino IDE library manager along with it documentation. Do a search for 'ez_switch_lib' if you still have an interest.

In the mean time, how is your button switch wired? I assume you have a pull down 10k resistor connected to the pin and the other side to +3.3v? Ie pinMode(pin, INPUT)?

How are you handling debounce?

Regard

Ron B