Fried two Sonoff modules - help me understand why

[Andrej]

Hi guys.

Firstly I would like to thank to all of Tasmota contributors - it is a great project. 

Here is my story. I want to use Sonoff Basic units to automate some of the ceiling lights in the house as the module just fits into the wall boxes behind wall switches. I have successfully flashed first few Sonoff modules with Tasmota firmware and managed to configure MQTT with Domoticz and external switch via GPIO14 and GND pin - everything worked as expected. Afterwards I wanted to install first module in the wall box and have prepared the cablings according to the diagram below.

After connecting all the cables to Sonoff module everything seemed to work flawlessly. I could toggle the light using both switches and the button so I fitted the module in the wall box and mounted the wall switch. After turning back power supply the module didn't respond anymore and appears to be dead. It doesn't connect to wifi, led does not respond, even when I try to connect it via serial interface. Is there anything else I should try to bring it back? I thought that the first module was just faulty so have tried to connect another one and the same thing happened again in exactly the same way! 

Obviously I am doing something wrong. What bothers me is the neutral lead connection. Is it ok to have it connected as shown on the diagram or should the neutral lead come to light from Sonoff neutral output? What happens if the sonoff button remains pressed for longer time (theoretically could be the case after I have mounted the wall switch in front of Sonoff module)? 

Any comment would be highly appreciated.      

[Mike Roberts]

Your neutral connection looks fine.  The neutral is in any case connected internally from input to output.

I would look more closely at the GPIO and ground connections - bearing in mind that the internal circuitry will float somewhere between L and N - hopefully with a high enough impedance not to be a problem.  However connecting either GPIO or ground to ground/earth might cause a problem.

It looks like the action of installing the gear was the cause of the failure and not your setup.  So check for potential shorts - including to ground.

Presume you have opened these up and looked for signs of problems?

Mike

[Andrej]

Mike, thank you for your response.

I have visually checked all connection points on both circuits and they look fine. Multimeter confirms there is no continuity between wall switches circuit and high voltage circuit. Don't know if it means anything to you...resistance of complete shorted wall switches circuit seems to be between 3 and 15  Ohms. Being a chemical engineer I don't know how to test if I created any short to earth wire somewhere at the bends in the walls while I pulled extra cables? 

I should have left connected Sonoff module hanging out of wall box for longer time to see if it would fail. I am not sure whether the short occured somewhere in the wall box where all the connections (although should be isolated) were touching.

Could it be that both Sonoff modules (probably same batch) were just bad?

   

Dne nedelja, 07. januar 2018 00.27.51 UTC+1 je oseba Mike Roberts napisala:

[Andrej]

I don't really understand what do you mean when you say that internal circuitry will float somewhere between L and N. What is then the voltage between GPIO14 and GND pin (isn't it 3.3V)? I thought there is only a very small current flowing in internal circuit so I use Arduino jumper cables to connect wall switches to GPIO and GND pin. 

Dne nedelja, 07. januar 2018 00.27.51 UTC+1 je oseba Mike Roberts napisala:

Your neutral connection looks fine.  The neutral is in any case connected internally from input to output.

[Mike Roberts]

Andrej

Yes the relative voltage beween the GPIO and GND will be 0 to 3.3v.  However if you put a meter between GND and neutral (with meter on AC) you will get a surprise!  It will read around half the live voltage.  This comes from how switched mode supplies work.  Plus the basic does not have the PCB clearances to qualify for 'Double isolation' so you should treat the GPIO and GND wires with the same respect as the live.

I was going to suggest that you check the voltage between the GPIO and GND as it should be nearly 3.3v if the switch is open.  If I were doing this I would pull fuses first, then attach meter probes securely and then reapply power, and then wait a few minutes after power off (fuses back out) for the internal high voltage capacitors to discharge before touching any part.  If you have any doubts consult an electrician (I am not one).

Mike

[Cysix]

Andrej-

That circuit looks fine.  It's essentially the same as I describe in my video here: https://youtu.be/l-CM8JQmges

Not sure what wires you're using between the switches, but you shouldn't be using the GND (bare copper) for GPIO14.  The bare copper ground wire should be connected as ground on both switches and to the main earth ground in at least one of the switch boxes.  

Assuming standard US 3-way wire between the boxes, for the GPIO14 switching, you should be using some combination of Black/Red, Black/White, Red/White.  Also make darn sure that the wires you choose to use for GPIO14 ONLY travel between the switches directly and aren't connected to anything else in the wall.  Where this might trip you up (and fry your Sonoff) is if you're using the white wire and it happens to be connected somewhere to other white wires.  There are many ways electricians wire up 3-ways and if you're not absolutely positive how it's working get an electrician to assist (not trying to insult your skills here).

Re: one of the reply comments regarding "floating" voltage, it's possible you need to connect an extra pull-up resistor between GPIO14 and GND.  Something like a 5K ohm will work.  This would eliminate any possible "noise"/"interference."  It's not always required as the ESP8266 has an internal pull up, but it's not very strong, i.e. can't overcome a very electrically noisy situation.