Using more accurate temperature sensor with my Raspberry Pi

When I stumbled upon the MCP9808 precision temperature sensor, I was sold based upon its promise of up to 0.25 degrees Celsius accuracy. Just like my Freescale Xtrinsic sensor board, there’s a Python library that that allows me to use the MC9808 temperature sensor with my Raspbbery Pi.

My one and only challenge was doing a proper soldering work because the MCP9808 temperature sensor board comes as a breakout board. I referred to this tutorial to prepare the header strip (with the pins) that comes with the sensor board, inserted it into my breadboard, placed the breakout board over the pins, and solder! It was tough. Being a software guy, there is a big mental hurdle because there is not much undo feature or “let’s try to step into this code to see if it works or fails”. It’s a big challenge to hold a soldering iron and try not to apply too much heat when trying to apply solder onto the pins, else the sensor board would be toast.

The end result: I did a better job with the first breakout board (I bought 3 sensor boards as contingency), and the other two boards were not so lucky under the soldering iron.

Here’s what the MCP9808 sensor looks like on my breadboard with the 4 connections to my Raspberry Pi through the very usefulT-Cobbler Plus connector.

mcp9808

Here’s a photo of the GPIO cable nicely coming out of my Raspberry Pi clear case and connecting to the T-Cobbler Plus.

mcp9808-whole

Here’s a “creative way” of making sure the clear case cover still protects most of my Raspberry Pi even with the protruded Xtrinsic sensor board. Priceless :)

xtrinsicsensorboard-rubberband

In my follow up post, I will describe how I send the temperature data stream to an Azure Event Hub, and then do simple analytics of the data on transit using Azure Stream Analytics.

How to send sensor data from Raspberry Pi to Azure Event Hub using a Python script

This post follows what I intended to do which is to pump sensor data consisting of temperature and altitude readings from the Xtrinsic sensor board to my Azure Event Hub named getfityall. The sensor board comes with some Python scripts already as you have seen in my earlier posts. Coupled with Microsoft Azure Python packages from the SDK, I could easily reuse a very nifty Python script I found from Kloud’s blog, a very competent Microsoft cloud partner, to send sensor data to my Azure Event Hub. The repurposed Python script looks like the following:

import sys
import azure
import socket

from azure.servicebus.servicebusservice import (
  ServiceBusSASAuthentication
  )

from azure.http import (
  HTTPRequest,
  HTTPError
  )

from azure.http.httpclient import _HTTPClient

class EventHubClient(object):

  def sendMessage(self,body,partition):
    eventHubHost = "youreventhubname.servicebus.windows.net"

    httpclient = _HTTPClient(service_instance=self)

    sasKeyName = "yourprocessorname"
    sasKeyValue = "yourprocessoraccesskey"

    authentication = ServiceBusSASAuthentication(sasKeyName,sasKeyValue)

    request = HTTPRequest()
    request.method = "POST"
    request.host = eventHubHost
    request.protocol_override = "https"
    request.path = "/youreventhubname/publishers/" + partition \
    + "/messages?api-version=2014-05"
    request.body = body
    request.headers.append(('Content-Type', \
    'application/atom+xml;type=entry;charset=utf-8'))

    authentication.sign_request(request, httpclient)

    request.headers.append(('Content-Length', str(len(request.body))))

    status = 0

    try:
        resp = httpclient.perform_request(request)
        status = resp.status
    except HTTPError as ex:
        status = ex.status

    return status

class EventDataParser(object):

  def getMessage(self,payload,sensorId):
    host = socket.gethostname()
    body = "{ \"DeviceId\" : \"" + host + "\""
    msgs = payload.split(",")

    for msg in msgs:
      sensorType = msg.split(":")[0]
      sensorValue = msg.split(":")[1]
      body += ", "
      body += "\"SensorId\" : \"" + sensorId \
           + "\", \"SensorType\" : \"";
      body += sensorType + "\", \"SensorValue\" : " \
           + sensorValue + " }"
    return body

I saved and named this Python script as sendtelemtry.py.

Then in the mpl3115a2.py script, add the following import statements:

import socket
import sendtelemetry

At the end of the script, replace the Python code with the following:

mpl = mpl3115a2()
mpl.initAlt()
#mpl.initBar()
mpl.active()
time.sleep(1)
while 1:

        #print "MPL3115:", "\tAlt.", mpl.getAlt(), "\tTemp:", mpl.getTemp()
        hubClient = sendtelemetry.EventHubClient()
        parser = sendtelemetry.EventDataParser()
        hostname = socket.gethostname()
        message = "temperature:"+repr(mpl.getTemp())+",altitude:"+repr(mpl.getAlt())
        body = parser.getMessage(message,"mpl3115")
        hubStatus = hubClient.sendMessage(body,hostname)
        print "[RPi->AzureEventHub]\t[Data]"+message+"\t[Status]"+str(hubStatus)
        time.sleep(0.1)

Then execute this script by doing

sudo python mpl3115a2.py

The Azure Event Hub REST API returns HTTP status code to indicate the result of the send event REST call. A HTTP status code of 201 means success. Read more about the send event REST call here. You can monitor your Azure Event Hub dashboard to see the incoming messages.

In the next post I will share more about what I’m doing with the preview feature of Azure Stream Analytics to do a toll-gate analysis of event data in transit. Thus far I had only been doing descriptive analytics of data at rest. This ought to be interesting because it requires a different understanding of what I would like to analyze.

Fresh raspberry pi at my service

I am super glad that the items I ordered from Element14 arrived overnight. I got not one but 3 MEMS Sensor board, and it is working well. Proof that I somewhat “overused” the previous board so much so that the temperature and altitude reading stopped working.

pi@raspberryclay ~/rpi_sensor_board $ sudo python mpl3115a2.py
MPL3115: Alt. 100.24 Temp: 28.144
MPL3115: Alt. 100.24 Temp: 28.144
MPL3115: Alt. 100.24 Temp: 28.144
MPL3115: Alt. 100.24 Temp: 28.144
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128
MPL3115: Alt. 101.32 Temp: 28.128

I got a new Raspberry Pi too, it’s for my co-worker, Clayton. It’s pretty simple to clone the micro-SD card from raspberryfai and write it on to the new one. Just use the Win32DiskImager. It works like a charm.

Just to make sure my ISS agent which sends the right telemetry data which consists of MEMS sensor temperature/altitude reading, here’s a good OData feed of the data all captured in the cloud.

raspclayiss

I’m thinking of the next steps in my experimentation such as:

  • Revert to plain old way of sending telemetry data from my Raspberry Pis to Azure Event Hubs using a friendly AMQP client/library.
  • Do real-time analytics of multiple data streams from Fitbit, Strava, Raspberry Pi (temperature/altitude). I’ve been mucking around with the new preview of Azure Stream Analytics to some success.

I have 2 more packages of nice toys to be delivered. I’d ordered an Arduino Starter Kit, and 2 Intel Galileo 2, among other nifty breakout boards, sensors, and kits. Stay tuned.

 

Sending sensor board data from Raspberry Pi to Intelligent Systems Service (ISS)

I have been mucking around with the Xtrinsic-sense board which is an add-on to my Raspberry Pi. In my previous post, I had executed the Python scripts that came with the sensor library that I cloned out of this github. There is a good article that explains about how to enable Xtrinsic-sense board in Raspbian.

However my aim is to get the Intelligent Systems Service (ISS) sample C application to invoke the sensor board shared library to retrieve the altimeter and temperature values, assign that to a string property within the data model, then send it to ISS. The sensor board library comes in the form of a shared library called sensor.so.

Hence I started on a journey to try to invoke the functions within the sensor.so out of mirroring the Python code, but in the sample C application. Now I must admit that Python coding skills is non-existent, and my C coding skills were really rusty. The last time I wrote a C/C++ application was in 2000/2001! What’s surprising is that I haven’t completed forgotten C, it’s like riding a bike, suddenly remnants of makefile, linking with dynamic/static libraries, include files, and function pointers in C slowly came back to memory.

But I’m still hopeless in Python, I gave up because I didn’t understand what the following code does except that it does some byte shifting, but for what reason, it beats me.

def getTemp(self):
    t = self.readTemp()
    t_m = (t >> 8) & 0xff;
    t_l = t & 0xff;

    if (t_l > 99):
        t_l = t_l / 1000.0
    else:
        t_l = t_l / 100.0
    return (t_m + t_l)

I knew there has to be an easier way to invoke the sensor board library in C, but where’s the sample code? After much searching, I found a reverse-engineered sensor.so with C source code! How amazing, thanks to this bloke named Lars Christensen. It was just hot off the oven, to be specific just last Saturday 13 Sept. To get started, all I had to do was to clone his source from github. In my raspberrywifai SSH session, I issued the following command (provided you have already installed git too).

git clone https://github.com/larsch/rpi-mems-sensor.git

There was a slight problem. The mpl3115a2.h header file was empty. No worries, just go to the github repository at https://github.com/larsch/rpi-mems-sensor, click on mpl3115a2.h, copy the content, and paste it in the file in your working directory on RaspPi.

You also need to download and extract the bcm2835 library. Then just follow the steps on how to make it. This creates a static library called libbcm2835.a. Next, make the reverse-engineered sensor board library so that you create a new sensor.so, and libmemssensor.a. I was only interested in the static libraries of libbcm2835.a and libmemssensor.a so that I can compile test.c. There wasn’t a proper make statement for test so I just issued my own gcc command.

gcc -Wall -I /home/pi/rpi-mems-sensor  -L. -lmemssensor -lbcm2835 -static test.c -o test libmemssensor.a libbcm2835.a

Do not just run ./test because initializing bcm2835 requires elevated permission. If you do that you will get the following error:

pi@raspberryfai ~/rpi-mems-sensor $ ./test
bcm2835_init: Unable to open /dev/mem: Permission denied

So do this instead:

sudo ./test

And voila! I get the altimeter and temperature reading.

raw alt = 11968
alt: 46.750000, temp: 23.812500
raw alt = 11968
alt: 46.750000, temp: 23.812500
raw alt = 11968
alt: 46.750000, temp: 23.812500
raw alt = 11968

The Python code which I just couldn’t comprehend is the equivalent of the following C code:

 

double getTemp() {
    int t = MPL3115A2_Read_Temp();
    int t_m = (t >> 8) & 0xFF;
    int t_l = t & 0xFF;
    if (t_m > 0x7f) t_m = t_m - 256;
    return t_m + t_l / 256.0;
}

It still beats me what it does, but that’s alright, I just wanted this code to work within the ISS sample C application. Next I embedded this code within the DATAPROVIDER_RESULT DATA_PROVIDER_INIT function. I didn’t want to create a new data model so I reused the sample data model, and chose to assign the altimeter and temperature reading to the stringProperty, as shown in the following code:

// setting StringProperty which contains the alt and temperature
bcm2835_init();
MPL3115A2_Init_Alt();
MPL3115A2_Active();
sleep(1);
char tempinfo[1024];
snprintf(tempinfo, sizeof(tempinfo), "alt: %f, temp: %f\n", getAlt($
stringProperty = tempinfo;

Compile (also includes modifying the makefile to ensure the sensor board header files and static libraries are correctly linked), and then run the ISSAgent app, and wait for it……..

Info: Microsoft Intelligent Systems Agent Console
Info: Using device endpoint: getfityall.device.intelligentsystems.azure.net
Info: Registering device: Name=raspberrywifai, FriendlyName=raspberrywifai, ModelName=Contoso.Device, Description=raspberrywifai
Info: x-ms-activity-id: 557374f8-2342-4bfb-bb7b-ce19199fa77e
Info: Getting endpoints for device raspberrywifai
Info: x-ms-activity-id: 229227bf-483b-4914-a613-e447bb7caca5
Info: Got ingress queue namespace=uswepm01c02x0sb20-0.servicebus.windows.net, name=getfityall/raspberrywifai
Info: Got commands queue namespace=uswepm01c02x0sb20-0.servicebus.windows.net, topic path=DeviceBank1, subscription name=raspberrywifai
Info: Got per device token=[secret]
raw alt = 11152
Info: Uploading message 2e5f2634-bee2-4e88-b1c3-3ed991f25f41 for device raspberrywifai (payload={"@iot.devicename":"raspberrywifai", "value":[ {"@odata.context":"Contoso\/$metadata#Device\/$entity", "@odata.id":"Device('raspberrywifai')", "@Microsoft.IntelligentSystems.Vocabulary.V1.Timestamp":"2014-09-17T00:34:41Z", "structProperty":{"simpleField":17, "structField":{"int32Field":10, "int64Field":34, "doubleField":40.000000000000000, "stringField":"Wed Sep 17 10:34:40 2014\u000A", "guidField":"0F1E2D3C-4B5A-6978-8796-A5B4C3D2E1F0", "binaryField":"AAAA", "dateTimeOffsetField":"2014-09-17T10:34:40Z", "booleanField":true}}, "stringProperty":"alt: 43.562500, temp: 22.812500\u000A", "dateTimeOffsetProperty":"2014-09-17T10:34:40Z"} ]}) ...Info: Success

Or more specifically:

"stringProperty":"alt: 43.562500, temp: 22.812500\u000A"

Woo hoo! I got it. Now moving on to the next steps. This opens up other possibilities such as sending temperature as an event on a pre-determined interval, and also set up alarm when temperature falls below or goes above a certain value. The fun begins! :)