Group Actions
Engagement
  • Author Author: sunnyiut
  • Date Created: Date Created
  • Views 1385 views
  • Likes 1 like
  • Comments 11 comments
Related
Recommended

Part3.1 - Head Impact Detection [ADXL377]

sunnyiut

Title: Total Player Monitoring

By: Md. Kamrul Hussain

Project Category: Design Challenge

Project Name: Sudden Impact Wearables Design Challenge

Blog post: 08


Intro:


This is a short update on detecting head impact injury using ADXL377 3 axis accelerometer. The goal is to detect potential concussion type collisions during a game.


Components:



image


The sensor provides analog outputs for individual 3-axis which can be processed by PIC12F1840 microcontroller to detect the impact due to collisions. PIC12F1840 is a small [8 pins only], cost effective and low power Microcontroller with all the functions needed for this design. It has 4 channel 10 bit ADC for capturing analog data from ADXL377. The UART module is used to communicate with the BLE [RN4020]. It has internal oscillator up to 32MHz and XLP version ensures low power consumption [34uA @ 1MHz]. So, the full impact detection module can be constructed in a small circuit board and can be powered for long time without changing the Battery. This is necessary because I am going to mount it on a HEAD BAND, so it should be small and have long battery support.


In this short blog post I am just uploading the images of experimenting with the ADXL377 module. For demonstration purpose I am using Mikromedia dsPIC board from Mikroelektronika, to display the output of the ADXL377.


Setup:

imageimage

This setup is pretty straight forward. If the ADC reading exceeds a certain value then send impact data through BLE. Here, I'll just display the data on the TFT screen for demonstration.


Calibration:

After powering up the ADXL377 sensor, it has to be calibrated for 0g in each of the axis. In general, it produces around Vs/2 at 0g and varies up Vs and down to 0v respectively for positive and negative variation of acceleration value. It means, for 3.3v supply the 0g value would be around 1.65v or 512 ADC value. So, before taking actual measurements the controller takes 20 readings for each axis and averages the readings to get the ADC value for 0g state.

image


Output at stationary condition:


The calibrated values for each axis slightly vary from expected value [512]. It depends on the position of the sensor and the surface flatness where the sensor is placed.

For calculating the g value, I am simply counting the half value of the difference between ADC and calibrated value.

Vs = 3.3v

ADC resolution = 10bit -> 1024

So, each step = 3.2226 mV

Therefore, 2 steps = 6.445 mV ~= typical output voltage change of ADXL377 for per g variation.

image


Output at low impact:

image

By shaking the sensor with hand or hitting with a rubber pad it gives a low impact and the output stays below 10g. The helmet icon stays in green [lime green indeed]. But if the sensor experiences a free fall or hit by the rubber pad directly on to the sensor surface, the z-axis shows huge variation !!!!


Output at high impact:

image

For higher impact the helmet has changed its color into RED, which indicates an impact detection. I have placed a threshold value of 12g just for demonstration purpose.


But I have got unexpected output / behavior couple of times, where I hit the sensor not so hard but it showed higher values [no way I can produce -31g with that weak hit] image and still the helmet didn't change the color [stays green]!!!

It's weird.... I think it's a bug in my coding, but it's a simple code and if the sensor somehow shows higher values, the helmet has to change its color???? So, I'll try to figure it out later on. Hopefully it's a silly mistake image

image


Circuit Diagram:

image


next step - I'll make the board and mount it into a Head Band to check out the performance.

  • in reply to Former Member

    hello Nayak,

    As Hendrik said, the Z value cannot be 2v for stationary condition. May be you need to check out the voltage again on Z axis to be sure.

    Hendrik mentioned previously that you may need a high resolution ADC because you are using 5v as reference which will give low ADC span for 0 - 3.3v sensor output. I've checked the Arduino mega [i think it's AtMega2560, right?] pinout and schematic. It seems that the board has "AREF" pinout at the 'top right' corner which is left open. So, if you want, you can connect the on board 3.3v supply directly into the "AREF" pin, which will give you an ADC reference of 3.3v. It will eventually increase the resolution a little bit. You can take help from the tutorial for setting up the external ADC reference.

    there is another advantage of using 3.3v as ADC reference, it makes the 'g' calculation easier. the sensor produces an output of 6.5mv per 'g'. For 3.3v ADC reference,

    one ADC step = 3.3v / 1024 ~= 3.223mV. so two ADC steps = 3.223mV * 2 = 6.446mV which is close to 6.5mV = 1 g

     

    So, the easy way to check 'g' values is to take every two ADC steps as 1 g, but that won't give you the precise output. The right way should be the one explained by Hendrik. However, I am just giving you an example on the basis of 3.3v ADC reference with 10bit [0-1023] ADC output.

     

    1. setup your controller for external ADC reference

    2. connect 3.3v into AREF pin

    3. take ADC sample for stationary condition. lets say you got ADC_X0 = 510, ADC_Y0 = 513, ADC_Z0 = 515

         for example we are taking only X values. Y and Z values will follow same calculations.

    4. for X axis, ADC_X0 = 510 will be our reference

    5. take ADC sample while moving the sensor. lets say you got ADC_Xnew = 586

         [to get rid of noise or frequent change in values you can take around 1000 consecutive values [Hendrik suggested 1024], add them and divide by 1000 to take average value. lets say you got the average value ADC_Xnew = 586]

    6. measure the deviation of X axis, say Xdeviation = ADC_Xnew - ADC_X0 = 586 - 510 = 76, means it has +ve acceleration in X axis.

    7. calculate the deviation of X axis in voltage, Xvolt = Xdeviation * (3300mV / 1024) or Xdeviation * 3.223mV = 244.9 mV

    8. calculate the acceleration of X axis, Xg = Xvolt / 6.5mV = 244.9mV / 6.5mV = +37.67 g

    ** if the acceleration is in opposite direction, you will get -ve 'g'

    ** remember that you may need to take 'signed integer' data types for 'Xdeviation' and 'float' for 'Xvolt' and 'Xg'

    let me know if you have any confusion.

  • in reply to Former Member

    Something in you numbers is wrong. The Z voltage is way off with 2V 8this would correspond to about 60g. And then you say the ADC result is the same as the others.

    Regarding the calibration: the sum of 1024 values of 352 is not 352, its 360448, so the result of your calibration reading should be around that mark. Take the sum of 1024 values! Then divide by 1024 to get the average.

    (But even here your numbers are off: when the values are {352, 352, 353} then a division by 1024 cannot result in {0.350, 0.349, 0.351} - the first two numbers must be equal. Something in your code is wrong.) Whether you want to calibrate so that you readings end up with 1g for the z axis depends on what you are doing - in my case I was interested only in the resulting acceleration, so I removed the Z axis force of 1g out of my data (this also meant that my system is not depending on its orientation).

    You calculate the resulting G force like this: (assuming that you are using calibrated values, so they should be in the range of about -350 to +350): ADC_result*(4.7/1024)/0.0065. The part 4.7/1024 calculates how many mV one LSB represents, multiplying this with the ADC result gives you the reading in mV. Then you divide this by 6.5 to convert mV into g, as per the ADXL377 data sheet.

  • If I Use your method:
    have raw values = (352,352,353)
    1)Now say, CalX = 0 , CalY = 0, CalZ = 0 and

    2)1024 Values are (352,352,353)

    3) Dividing by 1024 we get (0.350,0.349,0.351)
    4)Now , CalX = 0.350 , CalY = 0.349, CalZ = 0.351
    5)Now Next reading Substract Calx,CalY, CalZ from new averaged raw datas(i mean after dividing by 1024)
    6)for each change of one LSB , we add or substract 0.75g , so for each 0.1 change in ADC datas in step 5, we add or substract 0.075.


    Is it correct? I have a doubt , If i get nest raw datas averaged as (0.350,0.349,0.351) and after substracting CalX, calY, CalZ, we get (0,0,0)G, instead of (0,0,1) G.

    Should i manupulate the reading so as to get (0,0,1)G here?

  • in reply to sunnyiut

    Hallo Md. Kamrul,

     

    Thank you very much. I have now a clear Idea of this operating voltage and would try to play with changing the code and give a try as per your suggestion. Yes, I am using Mega which is operating at 5 Volt although it is getting 4.7 (you are right) volt as i just checked. The arduino has two pins for 5V and 3.3 V and i have connected sensor it to 3.3 Volt.

     

    The ADXL377 breakout board (Sparkfun) I am using is operating at 3.3 Volts. Yes, I am using the Arduino from USB Port of my Laptop (USB2). I also checked with the Multimeter at the following point and the datas i got are as follows; (Stationary case)

     

    1) VCC = 3.332 Volt
    2) X      = 1.659 Volt
    3) Y      = 1.631 Volt
    4) Z      = 2.012 Volt

     

    so i calculates for 4,7 volt it should be approximately 718 instead of 675. After changing it i get the following raw datas

    X: 352

    Y: 352

    Z: 354

     

    X: 352

    Y: 351

    Z: 353

     

    X: 351

    Y: 351

    Z: 353

     

    X: 352

    Y: 352

    Z: 353

    If I take Offset_X = 352 , Offset_Y = 352 , Offset_Z = 353  then i get 0, 0, 0. I am not sure if you mean I should substract the 352 from each and then it is 0, 0, 1 and afterwards for each 6.5 mv i should change 1g ?

     

    Can you please tell me Which offset would be substracted at Z axis? How we convert it to g values?

    I am relaxed to know this sensor is working correctly image.  Thank you very much for the support. I am really glad that this post has been very very helpful to me. image

  • Regarding the calibration: I wrote it for the PSoC4 BLE, and used a ADXL375. So maybe its not really useful to paste my code. The logic I used was:

    1. set the calibration values to 0
    2. read 1024 values for each axis and add them
    3. divide each result by 1024 (shift-left 10 bits) to get the average
    4. store the result in the calibration values
    5. afterwards, when reading values, substract the calibration values from the result