enter image description here

I am having trouble with a dht22 sensor. It works well for most of the time but randomly the library I am using can no longer receive a reading until I unplug then plug the sensor back in. (just disconnecting and reconnecting the GND pin is often enough) Then it will work for a while longer until the problem repeats.

I will try to describe my setup pictured in the drawing above in detail. The picture above is not a proper diagram drawing so please ask any questions if my text description falls short as well...

The DHT22 hookup is pretty straight forward 5V to VCC pin, GND pin to ground, DATA pin to a GPIO pin (GPIO 24 in my case). I am pulling up the DATA to 3V3 with a 4.7K Ohm resistor. I am using the pgpio DHT22 library. Here is the code which involve a loop in order to ignore the first couple of "bad" readings from the DHT22. (the DHT22 always gives the first two or three reading for temp and humidity as -999) also An important note is I am using about 10 feet of cat5e cable to connect the sensor directly to the pi.

import pigpio
import DHT22
from time import sleep

# initiate GPIO for pigpio
pi = pigpio.pi()
dht22 = DHT22.sensor(pi, 24) 

def readDHT22():
    #get reading
    humidity = '%.2f' % (dht22.humidity())
    temp = '%.2f' % (dht22.temperature())
    return (humidity, temp)

r = 0

while r < 10:
    humidity, temp = readDHT22()
    print ("trying to read temp...")
    r = r + 1
    if '999' in humidity or '999' in temp:
    print ("H:"+humidity+"")
    print ("T:"+temp+"")

The other part of my diagram shows the connection for a 5V relay. There is actually 5 relays connected. they each share a 5V rail and a Ground rail. Each relay has its own GPIO connection used to send a HIGH signal to a transistor that acts as a switch for the relay.

The reason I mention the relays in my problem is I suspect that my issue could be that I am drawing to much current from the pi.

What basic steps should I take to begin trouble shooting an issue like this?

How can I accurately calculate the amperage of my circuit to determine if I am drawing more than avaible? ( As I understand the mx from the 3v3 rail is 50mA, and the max from 5V is the max of the power supply to the raspberry pi which is 2000mA in my case) I dont know the resistance of the relays or how to figure that out accurately either.

As a secondary question id like to ask about a recommendation for wiring the dht22 that was detailed in the DHT22.py library file:

For 5V operation connect pin 1 to 5V and pin 4 to ground.
The following pin 2 connection works for me.  Use at YOUR OWN RISK.

DHT22 pin 2 -----+
gpio ------------+

Here is a drawing of how I understand to wire this up... Is it correct?

enter image description here

  • 1
    Warning, I wrote the DHT22.py module so will be biassed. Some DHT22 just seem to hang after a random number of readings. They need a power cycle to recover. I have done a million readings without a hang (every 3 seconds I think over a month and a bit). The DHT11 is read in pretty much the same way. My DHT11 never hangs. My DHT22 sometime hangs.
    – joan
    Mar 14, 2016 at 8:41
  • Do you know of a way to trigger a "power cycle" for the pin without having to physically unplug the ground connection? Also do you have a diagram for wiring the DHT22 with a pull-up to 5V with the 5K and 10K resistors?
    – BryanK
    Mar 14, 2016 at 13:19
  • 1
    @BryanK You could break the GND or 5v line using a GPIO-controlled NPN or PNP transistor like you're doing with the relay already. I'd probably use a PNP just to be sure the sensor doesn't just find another path to GND but NPN will most likely work too.
    – jDo
    Mar 14, 2016 at 22:55
  • @joan do you have diagrams for wiring the dht22 documented anywhere. I am very curious about the setup with the 5V pull-up using 5K and 10K ohm resistors. I dont quite understand the part about running the pull-up line to ground since the DHT22 already has a ground pin connected.
    – BryanK
    Mar 15, 2016 at 9:17
  • @BryanK Sorry for tardy response. It acts as a resistor divider also referred to as a voltage divider to safely drop 5V to 3V3. By the way I understand (BUT HAVE NOT TESTED AND I AM NOT AN ELECTRONICS PERSON) you can power the DHT22 from 5V but just use a 3V3 pull up on the data line. I.e. 5V to power pin. Ground to ground pin. Data pin to Pi GPIO with a 4k7 (or thereabouts) resistor between the GPIO and 3V3 at the PI end. None of this will help with the DHT22 crashing as some seem to do at random.
    – joan
    Mar 15, 2016 at 12:41

1 Answer 1


"I suspect that my issue could be that I am drawing too much current from the pi."

Whether or not that's the case, I'm not a fan of the "modern" practice of sourcing relatively high currents from the Pi. In all my setups, I only source power from a PSU - if not to avoid overloading then simply to avoid introducing ripple and noise in the Pi's power rail that might disturb digital signalling. Instead of sourcing from the Pi, use a 12 volt supply or a laptop PSU and get some cheap switching regulators for producing the voltages you need. Rule of thumb in case you're going to do it anyway: never go above 80% capacity in any circuit. Ever.

About the relays... Whenever you're powering coils and other inductive loads (motors, solenoids and relays being the most common examples), you're effectively creating a magnetic field that resists sudden current changes as long as the coil is powered (in practical terms, think of the powered coil as an energy storage area). When you suddenly cut the power, i.e. turn off your relay, the magnetic field collapses and this energy is converted to a huge voltage spike. This phenomenon is typically referred to as "back electromotive force" or "back-EMF" for short. Since the energy can't just disappear, it'll have to travel down the supply line where it hits your controlling circuit. This means that your NPN transistor might be taking a beating whenever you turn your relay off. Transistors and digital logic don't like that one bit!

Oscilloscope output showing the voltage spike in a solenoid being turned off: https://upload.wikimedia.org/wikipedia/commons/8/81/BackEMFWaveform.png


Basically your current relay setup (No back-EMF protection)

To mitigate the problem, it's common practice to add a flyback diode across the terminals of the inductive load (in parallel from GND to the positive terminal). You have to insert it so the little ring/stripe on one end of the diode is closer to the positive terminal than GND. Be careful with the orientation of the diode; put it in the wrong way and you've created a very good connection (low resistance path) from the positive to the negative terminal, also referred to as a short circuit.


How it could/should look (with back-EMF protection in the form of a flyback diode. The term "flyback" only refers to the application, not the diode itself. For your low current setup, use any regular diode capable of handling hundreds of volts. Eg. a 1N4001)

Check the schematic above... Here you see a circuit with voltage spike/back-EMF protection added in the form of a diode. There's a DC power source (Vs) on the left, a coil (L1) at the top that could be your relay, a resistance (R1), a switch (SW1) and a diode (D) placed across the coil. See how the diode is "pointing towards" the positive terminal on the power supply? This means that no current can flow through it during normal DC operation. However, as soon as you turn off the relay the diode will ensure that the power stored in the coil will loop around in the coil itself until the resistance in the copper windings has "eaten up" all the energy; thus sparing your controlling circuit (SW1) from a nasty shock.

The same solenoid setup as before but this time WITH a flyback diode across the load. See that nice, horisontal line? That's safety. https://en.wikipedia.org/wiki/File:FlybackWaveform.gif

But wait, there's more!

Instead of mucking about with individual transistors and diodes, I suggest you buy 10 trusty old ULN2803 ICs for ~£1 on ebay including free shipping. (I'm not affiliated with any vendors. I simply picked the cheapest product).

The ULN2803 is a so-called Darlington transistor array. A single chip effectively gives you 8 individual NPN transistors in one small package (it actually contains 16 transistors but that's hidden away in the chip). They are perfect for your application since they have built-in flyback-diodes and each output can handle up to 500mA (~300mA when controlled from 3v3). They even have built-in input resistors of 2.7 K so you can and should connect them directly to the Pi's 3v3 GPIO.

If you get these ICs, you can safely connect up to 8 relays/motors/whatever and control them via 8 GPIO pins. Providing that all the relays are powered from an external supply - not straight from the Pi - this is a safe, cheap and stable solution. It could be done like this.

"Do you know of a way to trigger a "power cycle" for the pin without having to physically unplug the ground connection?"

Yes! Let a semiconductor do the dirty work! Simply connect GND from your sensor to one of the outputs of your newest investment - the ULN2803 :) To reset your sensor, you simply write a low/0 to the GPIO pin, wait a few seconds and then write a high/1.

  • Using the transistor as a switch for the DHT22s power supply seems to work well. Since I am getting a reading only once a minute I have been switching the DHT22s power off then on before each reading. Thanks for the wealth of information. Cant wait to try a transistor array. Do you know at which points in the circuit I can measure the voltage spike, if it is there, with a multi meter?
    – BryanK
    Mar 15, 2016 at 9:11
  • @BryanK You're welcome. Glad to hear it works. Sometimes, if there are multiple paths to GND and the sensor's current draw is sufficiently small, you might experience that cutting GND using an NPN won't turn it off immediately or at all. That's when you'd have to use a PNP and cut power instead. Anyway, the voltage spike happens too fast for the standard multimeter to see it. In this video you can see that the voltage spike reaches 400 volts in 20 microseconds. That's too fast for most multimeters (they essentially sample slowly to save power).
    – jDo
    Mar 15, 2016 at 11:58
  • @BryanK Perhaps if you connect one multimeter probe to a common GND in your circuit (e.g. at the power supply's GND) and connect the other probe to the relay's GND terminal, you'd be able to see a tiny amount of jitter when you turn off the relay. Maybe a minus sign will appear briefly on the display or a rise from 0.00 Volts to 0.01 volts. I don't know your meter but I wouldn't expect to see bigger changes than that without an oscilloscope.
    – jDo
    Mar 15, 2016 at 12:08
  • 1
    I have been using the darlington array you suggested for a while now. I am hoping you might know of a similar IC that would act as a "latching" switch instead. e.g. send pulse from GPIO -> 5V circuit closes and stays closed (stays closed even if no more current from GPIO) -> send another pulse to open 5V circuit again
    – BryanK
    Jun 2, 2016 at 17:58
  • 1
    @BryanK Hi again. There are probably many options but I'd recommend an MCP23017 i2c GPIO expander. Here's a photo of a dusty old development board that I've been using for years now. The MCP23017 connects to the Pi via just 2 GPIOs (and GND) and gives you 16 inputs or latching outputs. I hard-wired the right side to an ULN2803 via a DIP switch that allows me to disconnect it manually and use all 16 pins as inputs without interfering with the Darlington array. Here's a guide to using the MCP23017 that seems good.
    – jDo
    Jun 2, 2016 at 22:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.