hw timer

[debugger3056]

Dear colleagues,

I'm not as savvy as everyone else on this forum because of that
Programming MC's is "just" a hobby of mine. That's why I would be
grateful for a useful tip!
I have the following problem:

I initialize a hwTimer with:

Code: [Select all] [Expand/Collapse]

1. hw_timer_t *timer = NULL;
2. portMUX_TYPE timerMux = portMUX_INITIALIZER_UNLOCKED;
3. volatile int interruptCounter;
4. volatile int totalInterruptCounter;
5. void IRAM_ATTR onTimer();

GeSHi © Codebox Plus Extension

the Setup

Code: [Select all] [Expand/Collapse]

1. // Timer
2.   timer = timerBegin(0, 80, true); // Teiler 80MHz / 80 = 1MHz
3.   timerLed = timerBegin(1, 800, true); // Teiler 80MHz / 80 = 100kHz
4.   timerAttachInterrupt(timerLed, &onTimerLed, true);
5.   timerAlarmWrite(timerLed, 20000, true);
6.   timerAlarmEnable(timerLed);

GeSHi © Codebox Plus Extension

the interrupt routine

Code: [Select all] [Expand/Collapse]

1. void IRAM_ATTR onTimer() {
2.   portENTER_CRITICAL_ISR(&timerMux);
3.   interruptCounter++;
4.   totalInterruptCounter++;
5.   portEXIT_CRITICAL_ISR(&timerMux);
6. }

GeSHi © Codebox Plus Extension

and the loop

Code: [Select all] [Expand/Collapse]

1. if (stateCounter == 4) {
2.     stateCounter = 5;
3.     LedState = 0;
4.        
5.     display.clearDisplay();
6.     display.setCursor(0,0);
7.     display.println("measuring... ");
8.     display.display();
9.    
10.     myFile = SD.open(fileString, FILE_WRITE);
11.     if(!myFile){
12.       DEBUG_PRINTLN("file dose not exist");
13.     } else {
14.       myFile.println(dataString);
15.     }
16.     runtime = 0;
17.     timestamp = 0;
18.     interruptCounter = 0;
19.     totalInterruptCounter = 0;
20.     timerAttachInterrupt(timer, &onTimer, true);
21.     timerAlarmWrite(timer, 5000, true);
22.     timerAlarmEnable(timer);
23.     millisTimer = millis();
24.    
25.   } // end of measuring
26.   // Interrupt completed 25 Jan 2021
27.   if (interruptCounter > 0) {
28.     portENTER_CRITICAL(&timerMux);
29.     int sensorWert = 0;
30.     timestamp += 0.005;
31.      
32.     dataString = String(timestamp, 3) + ","; // drei Dezimalstellen
33.     // read sensors and append to the string:
34.     for (byte i = 0; i < AnalogInAnzahl - 1; i++) {
35.       sensorWert = analogRead(AnalogIn[i]) - AnalogInInit[i];
36.       dataString += String(sensorWert, DEC) + ",";
37.     }
38.     sensorWert = analogRead(AnalogIn[AnalogInAnzahl - 1]) - AnalogInInit[AnalogInAnzahl - 1];
39.     dataString += String(sensorWert, DEC); // the last data without komma
40.      
41.     if (myFile) {
42.       myFile.println(dataString);
43.     } else {
44.       SDerror("error opening filestring");
45.     }
46.        
47.     if (totalInterruptCounter >= 1201) {
48.       timerAlarmDisable(timer);
49.       runtime = millis() - millisTimer;
50.       if (myFile) {
51.         myFile.close();
52.       }
53.       DEBUG_PRINT("Messdauer [ms]: ");
54.       DEBUG_PRINTLN(runtime);
55.       LedState = 1;
56.       stateCounter = 6;
57.     }
58.     interruptCounter--;
59.     portEXIT_CRITICAL(&timerMux);
60.   }

GeSHi © Codebox Plus Extension

if I now fade out in the loop,

Code: [Select all] [Expand/Collapse]

1. //myFile.println(dataString);

GeSHi © Codebox Plus Extension

I get a runtime of exactly 6000ms.
But if I transfer the data (approx. 40Byte per string) to the SD card
write results in a runtime (different) of approx. 6600ms!
At first I thought the SD routine couldn't do that in 5ms per
Record and tried 10ms, but inaccurate with the same
Result ?
Please please what can be the reason, what do I not understand?

Thanks in advance !!!

[ESP_Sprite]

Your SD-card is a complicated bit of hardware, and any write to it can potentially take a fairly long while as it may need to do things like re-shuffle sector content for wear levelling.

[debugger3056]

Dear,

so how should I save 50Byte per 5ms (without BLE or WLan) ???

[ESP_Sprite]

The easiest way in traditional ESP-IDF would be to use a separate thread for writing, and a big buffer between the two threads to hold any data when writing is slow. Not sure how to solve it in Arduino, I'm afraid.

[debugger3056]

thank you very much,

could you tell me at last how you would realize a big buffer with ESP-IDF ?

[ESP_Sprite]

The simplest way to do that is to make a large FreeRTOS queue. This can function both to shuttle the data from the sensor reading task to the writing task, as well as as a buffer to smooth out SD-card writing delay times.

[debugger3056]

thank you very much !