Friday, January 23, 2015

Moteino to Raspberry Pi over Serial v3

I was able to get my Moteino host talking to my Raspberry Pi over the Pi's serial port.  Last time, it was working but the information coming over did not make sense.  The problem was in my sprintf statement in the Arduino sketch.  I had a %d parameter in the command.  Text was getting converted to numbers.  I replaced the %d with a %s and now the sketch sends the correct data to the Pi.

The other problem I was having is that the host was not getting all of the data from the node.  The node was sending three measuring points to the host in with about 100ms between each send.  The host was getting the first and third MPs but not the second.  The easy fix was to put a 500ms pause in between each send.  That gave the host enough time to process each piece of data.  However, that might not be a good solution.  What if two nodes send data to the same host within a half-second of the other?  I need to find a better solution.

After some experimentation, it looks like the problem was the time it took for the Moteino to blink the LED.  After each receipt, I was blinking the LED three times with a 100ms pause between each on-off cycle.  That was apparently too much time.  The middle transmission was getting lost.  I turn the blinking down to one per transmission and the host now receives all transmissions.

Also, the node was sending all three transmissions with three LED blinks with a 100ms pause.  I set the blinks to one per transmission and the host could once again not keep up.

I changed the node to blink once per transmission and pause for 150ms.  This limited transmissions to one every 1/4 of a second.  The host was able to read those.

After all of this, I set the node to only transmit one piece of data every five seconds.  That way the HESA will have time to process each reading.

More time experimenting
The HESA takes 10 seconds to look for water.  Plus, it needs a few seconds to get the temperature and humidity readings from the DHT22.  Finally, it has to update the web page every hour.  That takes a few seconds.

I set my test program on the Pi to sleep for ten seconds if there is nothing in the serial's input buffer.  I also set the node to send some data every few seconds.  When I do that, the host misses 2/3 transmissions each cycle.  That is not good.  Here is the code I am using:
import serial
import time
port = serial.Serial("/dev/ttyAMA0", baudrate=115200, timeout=3.0)
while True:
    rcv = ''
while port.inWaiting() > 0 :
    rcv = port.readline(255).strip();
if (rcv != "") :
    print ("Incoming transmission: " + rcv);
else :
    print ("Sleeping");

 I tried something different.  I set the timeout to 0 and read the port every ten seconds.  Here is the code in this case:
import serialimport time
port = serial.Serial("/dev/ttyAMA0", baudrate=115200, timeout=0.0)
while True:    rcv = ''    rcv =
if len(rcv) > 0 :    print ("Incoming transmission: " = rcv);else :    print ("Sleeping");    time.sleep(10);
Now, the program reads all of the data in the buffer, but it reads it as one big string.  Since I am sending the data in three transmissions with EOL, the data looks like this:

Incoming transmission: 002|101~3369.00

I will change the node program to do one transmission with all data in one long string.  That is for another day.

Friday, January 16, 2015

Connect to Raspberry Pi Over Serial Connection

Tonight at my weekly me time at the Milwaukee Makerspace, I am going to attempt to get my Moteino connected to my Raspberry Pi using the serial connection on the Raspberry Pi.  Also, to be a bit different, I am going to try to take notes on the blog as I go.  This will either help me take better notes or make this blog entry completely unintelligible to any  reader who does not think like me.

There are instructions to work with the Raspberry Pi's serial port at this web page:  I scrolled down to the section titled "Connection to a microcontroller or other peripheral".  There are several things to do to stop the Linux kernel from using the serial port.  This is a prerequisite.  It looks like there is a script someone made to do all of these things automatically.  I am going to try to do the steps myself.

I edited the /boot/cmdline.txt file.  Also, commented out a line in /etc/inittab that opens a login prompt ("getty") on the serial port.

Rebooted.  Logged in with no problems.  There are two checks on the web page that you can run to verify that you did it right.  Both checks passed.  

The next step is to try interacting with the serial port using Python code.  I installed the PySerial package by typing 
sudo apt-get install python-serial 
at the prompt.  While the install was running, I noticed a message saying that there were some packages that were no longer used that could be uninstalled.  I ran 
sudo apt-get autoremove
to get rid of those packages.

This page has a very small python program to test the serial connection.  I used nano to create it. 

The first test I did was to connect the Raspberry Pi's serial pins to the serial pins on the Moteino --
 Tx to Rx, Rx to Tx, and Gnd to Gnd.  Then, I connected the Moteino to my Linux laptop using the FTDI adapter.  I loaded the Arduino IDE on the laptop, uploaded the HESA_Host sketch, plugged in the USB to the Moteino, and ran the python program.  Nothing happened.  I don't know if the problem was that the Moteino does not know what to do when there are two serial connections or if my sketch is bad.

The next test I did was to bypass the Moteino and connect the serial wires from the Raspberry Pi directly to the FTDI adapter.  That test worked.  The terminal screen on the Arduino displayed the output from the Pi and the Pi received the text I typed in on the terminal screen.  Cool!

Tried again with the FTDI connected to the Moteino and the Moteino connected to the Pi.  Made some small modifications to the sketch and tried uploading it to the Moteino.  The sketch would not upload.  

Unplugged the TxRx wires from the Pi.  Now, the sketch will upload.

For the next test, I changed the python code so it will send the letter "t" to the serial port every second, read the results on the serial port, and print the results, if any to the screen.  I also connected a wire from one of the 5v outputs on the Pi to the VIn pin on the Moteino.  I ran the python code and it worked perfectly.  The Moteino faithfully sent its temperature back to the Pi everytime it received a T.

Time to try the whole solution.  I setup my node with the DHT11 temperature sensor.  Then, I started the python program and waited a few seconds.  Next, I powered up the Moteino host.  It transmitted its messages from the setup part of the sketch.  Finally, I powered up the Moteino node.  It started transmitting its data every few seconds.  Everything worked.  The host received the transmissions and sent them to the Raspberry Pi.  The python program displayed the transmissions on the screen.

One problem is that it looks like the python program started displaying the ascii text as hex characters, ex. 0\xe0\x00.  That seems to have started after it received a "]" character.

Tried changing the baudrate on both devices to 9600.  That did not help.
Tried adding a \n to the end of each transmission to the serial port.  That did not help.
Tried putting a 1ms delay after every transmission.  That did not help.

The weird thing is that if I send a T to the Moteino, the sketch returns the temperature text with no problems.  Anything else looks like hex.

Figured it out.  I had some code in the sketch turn off the Moteino in between transmissions to save power.  This was the cause of the garbled code.  Turning off the power save feature fixed the problem.  I don't really need the power save feature since the host Moteino will get power from the Pi.

The new problem is that the data being sent to the PI does not appear to be the data the node is sending.  The data should be something like "002|102~87", but it is some other combination of number.  I'll have to look into that later.  

Also, need to figure out how to know when a batch of data is from one node.  Maybe I could put some text around each end of the package.  For example, ">>>002|102~67<<<".

I'll have to figure that out some other time.  That's enough for one night.

Saturday, January 3, 2015

Moteino Node Talking to Host

After about two weeks off for the holidays, I spent time at the Milwaukee Makerspace working on the Moteinos for my Home Environmental Sensor Array.

3D Print of a Battery Holder
As part of the node solution for the HESA, I want to have a complete stand-alone package that includes a holder for the battery.  Also, I want to have 6v battery for the power supply since the Moteino needs at least a 3.2v input.

I have a 6v Lithium battery from Energizer.  I made a case for it but it is not ideal and not reproducible.  I found a Thing on Thingiverse to 3D print a holder for a Panasonic Lithium battery.

I printed the holder on the Cube Pro Trio 3D printer at the Makerspace.  The printer is very easy to use.  I was hoping it would be more reliable than the other printers we have but my first print failed.  I printed the holder in draft mode and it finished but the corners pealed up a bit.  Also, I found out that the Panasonic battery is longer than the Energizer and the terminals are in different spots.  So, I won't be able to use the holder and will have to find a different solution.

On the plus side, I learned how to use the CubePro Trio printer.  This is a professional-grade 3D printer that should make high quality prints.

ACK Testing
The last time I played around with the Moteino, the node was sending data to the host but the node was never getting an ACK back from the host.  I tested combinations of all three Moteinos with the same results.  I posted something on the Moteino forum and the response was that it should work. I finally got back to testing the ACK feature and was able to get it working -- sort-of.

I think the original problem is that I did not have the antennas attached to the Moteinos.  I soldered antennas onto all three Moteinos, and they now send and receive ACKs properly.

However, this gets more and more unreliable as the distance between the node and host increases.  When the distance is less than 20 feet, the ACKs are almost always received.  At 50 feet, ACKs are received about 50% of the time.  If I go over 100 yards away, almost no ACKs are received.

I posted a message on the Moteino forum to see what the problem is.  However, I also decided that I don't care about ACKs for the HESA.  For now, I can continue without this feature.

Finally, I created a sketch for a node on the HESA network.  The sketch will read and transmit the following information to the host:
  • Notification that it is on-line on bootup
  • Battery voltage
  • Temperature from the RF69 chip or from a DHT sensor if installed.
  • Humidity from DHT sensor if installed
The code is flexible enough to request ACKs or not when sending data.  The device will also sleep between sends.  This should make the battery last longer.

The next step is to connect the host to a Raspberry Pi and have the Pi read the serial inputs.  Then, I need design a circuit board and complete solution for the node.  Finally, I need to make the HESA python program on the RPi send node data to the SQL database.