Analog inputs (TMP36) and cable length questions

Question

Hi all, I'm trying to do some data logging on my greenhouse, but I'm getting some inconsistent values from my analog pins. Specifically, I have 6 sensors, using all of the UNO's analog pins A0-A5. I have 2x TMP36 temperature sensors (inside and outside temperature), 2 moisture sensors (nails on wires), a light sensor, and eventually will have a water level indicator for the rain barrel. All the sensors seemed to work well individually when I tested things separately (and with short leads), but now the numbers seem to vary a lot from one reading to the next, at least for my temperature sensors (I think the others need some fine tuning of expected ranges). I take a reading every 15 minutes, and store it in an array. I have a bluetooth adapter hooked up, so I can take my laptop to the room nearest the greenhouse and connect wirelessly to the Arduino that way. I can then type in a command and see the data that was stored, and I can get the current values. To save it from added power drain I thought I would use D12 as my power pin. So D12 is usually low, but just before reading the values I set it high to provide 5V to the two temperature sensors (the other sensors just connect to ground). Come to think of it, I think I'll just keep that set to high to remove that extra unknown So I'm wondering - do the analog pins need to "settle" at all before I take readings? Right now I set D12 high, wait 1 second, and then read each of the analog pins: digitalWrite(power_pin, HIGH);
 delay(1000);
 int insideTemp = readTemp(inside_temperature_sensor);
 int outsideTemp = readTemp(outside_temperature_sensor);
 int moistureLevel1 = readMoistureLevel(moisture_sensor1);
 int moistureLevel2 = readMoistureLevel(moisture_sensor2);
 int waterLevel = readWaterLevel(waterlevel_sensor);
 int sunshineLevel = readSunshine(sunshine_sensor);
 // prints stuff to BTSerial port here
 digitalWrite(power_pin, LOW); The TMP36 sensors are hooked up to 5v, AnalogPin, Gnd (as per the oomlout example). All of the other sensors are connected just via 2 wires each: ground and analog pin. I'm using the internal pull-up resistors as follows: pinMode(moisture_sensor1, INPUT);
 digitalWrite(moisture_sensor1, HIGH); // use pullup resistor The Arduino is set up in my garage, and uses about 15 feet (5 meters) of CAT-5 cable to connect to the sensors in the greenhouse. CAT-5 has 4 pairs, or 8 wires. I use 1 wire for 5v, 1 for Gnd, and the other 6 are for each of the Analog pins. I soldered a header on the greenhouse end to bring out 6x Gnd connections and 2x 5v connection, so that all sensors can be plugged in that way. So my questions are: 1. Do I need to do anything special for the analog reads, or is my procedure ok? For example, would it make sense to wait between readings, or take 15 readings per pin and use the average? 2. Does the length of the sensor wires cause a problem? 3. Does sharing the GND over that distance cause an issue? 4. I realize that the waterlevel sensor is not connected, but I have a pullup resistor in place which puts it to a constant 5v, or 1023. This wouldn't affect any other pin readings, would it? Thanks! -Nico

mcb1 · Accepted Answer

Nico This may give you give the answer. http://forums.adafruit.com/viewtopic.php?f=25&t=11597 I've learned something today so thanks Mark

neilk · Answer

Nico I see Mark has solved this for you, however there are a few other issues which I had when reading from 2 x LM35 (3 wire device, similar to TMP36, I think) temperature sensors: I observed noise with significant amplitude on the Vcc line to the LM35, so I slugged it (at the Arduino end) with 47uF tantalum cap to ground. More than a few inches of lead resulted in noise on the LM35 signal, so I used twin screened - 1 core for Vcc, 1 core for LM35 output and braid to ground My maximum signal from the LM35 was less than 3,3V, so I used the 3.3V reference. I also experienced ADC settling time issues, so, just like you I did dummy reads and with delays, but I also took 10 readings at 100mS intervals and calculated the average. Eventually, my reading were not only solid, but agreed with an array of digital and mercury in glass thermometers sitting alongside the LM35s Good luck Neil

mcb1 · Answer

I decided the TMP way wasn't, and use the DS18B20 chips for temp sensing. One thing you can do to create a less noisy supply line is :- Feed the supply to these through a diode, and then have a reasonably large 100-1000uF as a filter on the sensor side of the diode. You'll be receiving 0.6v less than Vcc, but you can adjust for that when you do the calibration. I'm also wondering if you slow down the readings ie increase the 10mS time to 100mS, whether you will gain anything. Having a capacitor across the Analogue input is essential, especially since you the input is susceptible to any noise, and the whole circuit is high impedance. The 47uF across each input (as close as possible to the input) is good, but I would include a 0.1uF to bypass any high frequency noise as well. You also need to be careful about where the current flows are, so having a common point where everything originates from/to is the best way to stop current flowing in the ground wires. The best way to picture this is two water drains, one small and one large. If you make them both finish at the point they enter the ground, then no problems. If you make the large one join the small one at any point, some water will flow and affect the small drain. Mark

Analog inputs (TMP36) and cable length questions

Top Replies