nRf24L01+ connection quality meter


  • Hero Member

    A little frustrated by radio's showing inconsistent behaviour I built a simple connection quality meter with a leftover of a previous project.

    0_1464549715094_upload-8cefcb84-2528-42e3-98e1-792b63fc0b50

    Not too complicated 😉 a nano, I2C lcd display, radio adapter board (guarantees a stable power supply for the radio) and the radio to be tested.

    0_1464549601260_upload-0bca4a04-80b0-4f70-9263-cef5d75e7702

    a little hot glue and it fits nicely in the box. The radio can be swapped on the outside. Power comes in from a power bank connected to the nano USB so that I can walk around the house.

    0_1464549797943_upload-328cbdd6-bd88-417c-9bbb-0160b1a2d55a

    The sketch sends values from 0..99 to a gateway with "ack" enabled. The LCD shows the "failed" sends and the number of not acknowledged messages. i.e. the messages which did not arrive at the gateway or did not make it on their return flight.

    It now takes less than a minute to sort the radio's :bowtie:


  • Mod

    @AWI is that hardware ack or software ack? (The MySensors ack confusion is driving me insane)



  • Nice and very useful, also to test if there is signal in a special area in your house.

    Can you put the code here?



  • @AWI
    Thanks for the idea! This can become very useful in deciding at which locations a repeater can be used to solve my coverage problem. In my setup I'm facing difficulties getting the signal through a glass pane at the rear of my house.
    Same question as @flopp: putting the code in here would be very useful

    Thanks,

    Boozz


  • Hero Member

    @mfalkvidd my interpretation and based on experience:

    • fail (hardware ack..) is the node to node acknowledgement. (i.e. from node to repeater or gateway). Can not be switched on/off from the API. This is the return value from the send() function (or gw.send()). If a fail occurs the node tries multiple times to resend (this can take quitte a lot of time..) This acknowledgement is only from the next node in line.
    • missed (software ack) is the node to destination acknowledgement (i.e. node - repeater - gateway) . Can (and should) be enabled in the send() function. The acknowledgement is send automatically by the destination node in the form of an "echo" of the message sent. No automatic resend happens. Eventual action should be taken by the sender (controller / sketch in node).

    The "connection quality meter" measures the number of fails and averages this for 100 samples (percentate) and similar for not acknowledged "software ack's".

    Ik will post the sketch later today..


  • Hero Member

    The sketch for the connection quality meter.

    /*
     PROJECT: MySensors / Quality of radio transmission 
     PROGRAMMER: AWI (MySensors libraries)
     DATE: 20160529/ last update: 20160530
     FILE: AWI_Send.ino
     LICENSE: Public domain
    
     Hardware: ATMega328p board w/ NRF24l01
    	and MySensors 2.0 (Development)
    	
    Special:
    	
    	
    Summary:
    	Sends a radio message with counter each  x time to determine fault ratio with receiver
    Remarks:
    	Fixed node-id & communication channel to other fixed node
    	
    Change log:
    20160530 - added moving average on fail/ miss count 
    */
    
    
    //****  MySensors *****
    // Enable debug prints to serial monitor
    #define MY_DEBUG 
    #define MY_RADIO_NRF24									// Enable and select radio type attached
    #define MY_RF24_CHANNEL 80								// radio channel, default = 76
    
    #define MY_NODE_ID 250
    #define NODE_TXT "Q 250"								// Text to add to sensor name
    
    // #define MY_RF24_CE_PIN 7								// Ceech board, 3.3v (7,8)  (pin default 9,10)
    // #define MY_RF24_CS_PIN 8
    #define DESTINATION_NODE 0								// receiving fixed node id (default 0 = gateway)
    
    #include <SPI.h>
    #include <MySensor.h>  
    
    // display
    #include <Wire.h>											// I2C
    #include <LiquidCrystal_I2C.h>								// LCD display with I2C interface
    
    
    // helpers
    #define LOCAL_DEBUG
    
    #ifdef LOCAL_DEBUG
    #define Sprint(a) (Serial.print(a))						// macro as substitute for print, enable if no print wanted
    #define Sprintln(a) (Serial.println(a))					// macro as substitute for println
    #else
    #define Sprint(a)										// enable if no print wanted -or- 
    #define Sprintln(a)										// enable if no print wanted
    #endif
    
    
    // MySensors sensor
    #define counterChild 0
    
    // send constants and variables
    int messageCounter = 0 ; 
    const int messageCounterMax = 100 ; 					// maximum message counter value 
    const unsigned counterUpdateDelay = 100 ;				// send every x ms and sleep in between
    
    // receive constants and variables
    boolean failStore[messageCounterMax] ;					// moving average stores & pointers
    int failStorePointer = 0 ;
    boolean missedStore[messageCounterMax] ;
    int missedStorePointer = 0 ;
    int newMessage = 0 ;
    int lastMessage = -1 ;
    int missedMessageCounter = 0 ; 							// total number of messages in range (messageCounterMax)
    int failMessageCounter = 0 ; 							// total number of messages in range (messageCounterMax)
    
    // Loop delays
    const unsigned long displayInterval = 1000UL ;			// display update in ms
    unsigned long lastDisplayUpdate = 0 ;					// last update for loop timers
    
    // standard messages
    MyMessage counterMsg(counterChild, V_PERCENTAGE);		// Send value
    
    // ***** LCD
    LiquidCrystal_I2C lcd(0x27, 2, 1, 0, 4, 5, 6, 7, 3, POSITIVE);  // Set the LCD I2C address
    
    void setup() {
    	Wire.begin();  // I2C
        // ** LCD display **
        // LCD 2 lines * 16 char.
        lcd.begin(16, 2);
        lcd.setBacklight(HIGH);
        lcd.setCursor(0, 0);
    	lcd.print("AWI Quality nRF24");
    
    	for(int i= 0 ; i <  messageCounterMax ; i++){		// init stores for moving averages
    		failStore[i] = true ;
    		missedStore[i] = true ;
     	}
    	missedStorePointer = failStorePointer = 0 ;
        delay(1000);
    }
    
    void presentation(){
    // MySensors
    	present(counterChild, S_DIMMER, "Quality counter " NODE_TXT) ;  // counter uses percentage from dimmer value
    }
    
    
    void loop() {
    	// Sprint("count:") ; Sprintln(messageCounter) ;
    	LCD_local_display();
    	missedStore[failStorePointer] = false  ; 			// set slot to false (ack message needs to set) ; 
    	boolean succes = failStore[failStorePointer] = send(counterMsg.setDestination(DESTINATION_NODE).set(failStorePointer), true);  // send to destination with ack
    	if (!succes){
    		failMessageCounter++ ; 
    		Sprint("Fail on message: ") ; Sprint(failStorePointer) ;
    		Sprint(" # ") ; Sprintln(failMessageCounter);
    	}
    	failStorePointer++ ;
    	if(failStorePointer >= messageCounterMax){
    		failStorePointer =  0	;						// wrap counter
    	}
    	wait(counterUpdateDelay) ;							// wait for things to settle and ack's to arrive
    }
    
    void receive(const MyMessage &message) {  				// Expect few types of messages from controller
    	newMessage = message.getInt();						// get received value
    	switch (message.type){
    		case V_PERCENTAGE:
    			missedStore[newMessage] = true ;			// set corresponding flag to received.
    			if (newMessage > lastMessage){				// number of messages missed from lastMessage (kind of, faulty at wrap)
    				Sprint("Missed messages: ") ; Sprintln( newMessage - lastMessage - 1) ;
    				missedMessageCounter += newMessage - lastMessage - 1 ;
    			}
    			lastMessage = newMessage ;
    			break ;
    		default: break ;
    	}
    }
    
    
    // calculate number of false values in array 
    // takes a lot of time, but who cares...
    int getCount(boolean countArray[], int size){
    	int falseCount = 0 ;
    	for (int i = 0 ; i < size ; i++){
    		falseCount += countArray[i]?0:1 ;
    	}
    	return falseCount ;
    }
    
    void LCD_local_display(void){
    /* prints all available variables on LCD display with units
    */
    	
        char buf[17]; 											// buffer for max 16 char display
        lcd.setCursor(0, 0);
        snprintf(buf, sizeof buf, "Failed:%4d %3d%%", failMessageCounter, getCount(failStore, messageCounterMax));
        lcd.print(buf);
        lcd.setCursor(0, 1);
        snprintf(buf, sizeof buf, "Missed:%4d %3d%%", missedMessageCounter, getCount(missedStore, messageCounterMax));
    	lcd.print(buf);
    }
    
    


  • Thnx!

    👍


  • Hero Member

    I have been walking through and around the house with the meter to find the best radio solutions. So far I have the best reach with the nRF24L01+PA+LNA with the plastic/ aluminum foil (the ugly fix ;-), credits @Oitzu).
    0_1464961145010_upload-8716e90d-606f-4d9e-8dcb-697cb95fbbd6

    0_1464959045642_upload-85c16c8d-ac9f-40a9-a1cb-8464cc0bba0b

    b.t.w. I added "parent" and "destination" node to the display to show where the meter is connected to. Destination is fixed in the sketch but the "parent" will change if the meter connects to another node (i.e. repeater)

    To give you a real life impression . The transmitter (also nRF24L01+PA+LNA) is where the red arrow is behind two thick brick walls and a few trees. So, unless you live in a big mansion this should be sufficient range. Open field performance would be much better.
    0_1464961025398_upload-b1ddebfd-d124-433f-a00e-6296301a71f1

    For the rest of the tests I had a large variance in results. Some general observations:

    • nRF24L01+PA+LNA without the "shielding" is pretty similar (and sometimes even worse) than the standard radio with the pcb antenna.

    • all nRF24L01+ in SMD version have a comparable performance to the standard ones

    • The standard radio's I have tested with a silk screen (white lines) and pin designation print on the back seem to be fine.
      0_1464960195473_upload-9ff7ca2d-75e5-45b7-bea3-8824b6c19dd2

    • The (around 18) radio's is have tested without silk screen (3 different batches) are either "bad" performers (few meters) or don't perform at all. ( I noticed earlier and already put them apart. I was able to sort out 6 that can be used for short distance.).
      0_1464960320567_upload-b0a4caa9-7764-485b-a12e-8b9577dbe3e5

    • I don;t have any "blob" samples..

    I know this is far from scientific proof, but hope it gives you some guidance..



  • @AWI there also 2 versions of the PA/LNA Modules out there. Which one do you tested?


  • Mod

    @AWI said:

    The standard radio's I have tested with a silk screen (white lines) and pin designation print on the back seem to be fine

    In the picture, these radios have the notorious 1242AF datecode on the chip (if you didn't take the pictures yourself, do you have the same date code?). I also have a batch of these and they perform very well!



  • Thanks @awi for this, might have to build one myself to test a potential bad batch of radios i have.

    Do you need the LCD for this or is the data also print in serial?

    Thanks,


  • Hero Member

    @drock1985 It will print the data on serial but not very well formatted. You don't need the LCD but it is easy to carry around the house. (a laptop too 😃 but be aware of disturbances coming from the laptop i.e. Wifi ) I thought off blinking led's also. You can easily change the sketch, there is not much to it..


  • Hero Member

    @Oitzu which versions do you relate to? I also have some with ceramic antenna (smd connection)



  • @AWI

    Kijken je buren niet gek op wanneer je met je meter rondloopt? 😉

    gr.

    Boozz


  • Hero Member

    @boozz my neighbors ask me every two weeks what I am doing and what all those strange looking things in my garden and on my roof are... 😳



  • @AWI

    And when you start explaining they look somewhat odd don't they?



  • I'm having a problem in this part off the code:
    "present(counterChild, S_DIMMER, "Quality counter " NODE_TXT) ; // counter uses percentage from dimmer value"

    I'm getting this error:
    "exit status 1
    'present' was not declared in this scope"

    How can i fix this ?


  • Mod

    @mrc-core sounds like you are using the stable (1.5.x) version of MySensors. This sketch requires the development version which is available on github



  • Thanks for the quick reply.
    i'm going to change the versions 🙂



  • Can this be extended to also measure NRF24l01 standby current? Maybe some "fake" modules have a high current consumption in stadnby mode, this could drain batteries quickly when used in battery operated devices


Log in to reply
 

515
Online

6.7k
Users

7.6k
Topics

80.5k
Posts

Looks like your connection to MySensors Forum was lost, please wait while we try to reconnect.