Problems with conducted noise and isolated input

[e.sotillet]

Hi all. I am currently reviewing the hardware of a device that counts the pulses from a mechanical sensor, generally located at 5 to 20 meters and close to power lines of three-phase pumps.

I have the conducted noise issue resetting the LCD (uC keeps working) and little / no input hysteresis. I tried with a zener (see attached image) and improved the performance a little bit.

I thought about isolating the GND of that input but it would imply adding an isolated DC / DC converter and I think that the reliability of the circuit will be negatively impacted since it needs to operate safely for at least 5 years. (Also I need to keep power consuption low because the circuit has a battery backup).

It is worth say that the sensor signal is between 0.1Hz and 5Hz and I use CAT6 UTP cable to make the connection.


Is a common mode filter a good option? Or should I necessarily isolate that input and use a circuit with hysteresis?

Best regards

[fourtytwo42]

What sort of switch is it ?
If open mechanical contacts you need a lot more than your 2mA wetting current and as much voltage available at the contact as possible to breakdown oxides.
Much better to use a resistor directly from the supply to the contact wire of maybe 560R or less (>=20mA) and then deal with your noise immunity elsewhere.
Increase the operating current of the led to at least 10mA (so the total contact current will be ~30mA) and keep your 5V zener and consider also shunting the led with a capacitor, whatever you can get away with at your intended operating frequency, ensure the cable is at least frequently twisted pair (not phone stuff but cat5 style).
You can use the same ground BUT NOT the logic ground for the switch side, return it independently to the psu instead.

[e.sotillet]

The signal comes from a reed switch, it's inside a flow meter so I cant see its part number.

I dont increase the current for 3 reasons:

1.- I dont know the technical specs of internal reed switch. I need to keep this setup running at least 5 years.

2.- Extend life of optocoupler's led.  It is possible that the reed switch keep closed for long periods of time (days or months) so during that time the LED will be on and aging. The higher If, less life.

3.- Because the device can be powered by the battery and the goal is to extend as much as possible the backup time without incresing too much the battery size.

Also, I keep having the problem of conducted noise.

Thanks for reply!!

[penfold]

Could you try replacing the 3k3 with a 2k2 and a 1k1, placing one of the two on the ground return side (doesn't matter which one as the opto has some common mode rejection) and the other one stays where the 3k3 currently is

[e.sotillet]

Ok, I'm gonna try it. What is the best way to emulate that kind of noise in the lab (ac motor starting)? The problem occurs only in that moment.

Thanks!

[EEauroro]

The fact that your problem only occurs on motor startup means that it is most probably an EFT phenomena. EFT tests is defined in EN 61000-4-4 and if you have access to such an equipment you can simulate your problem and troubleshoot your circuits.

What you can do if you do not want to have galvanic isolation between the input interfacing circuit to the mechanical sensor to your logic circuits (MCU / LCD)

1) Increase the impedance path to ground by inserting a 1.5kOhm on the bottom side of the reed switch to ground (right now you have direct connection), replace the 3.3kOhm with 1.5kOhm to maintain the appropriate DC current. You can remove the zener diode and use higher valued resistors as well to maintain the appropriate current.

2) Insert common mode chokes on the input signal lines to the reed switch.

3) Make sure in the PCB design that the input circuit of the reed switch is far away from your Logic Ground and other sensitive components.

4) Use a shielded cable with the cable shield grounded to your metal enclosure / earth connection

[AndyC_772]

It would be helpful to see a photo showing your complete project, especially including the display and any wiring connected to it.

Do you have an isolated, differential scope probe? If so, it could be useful to monitor your low voltage supply rails during motor start-up.

There's a clue in that your logic input is connected to the MCU, but it's not the MCU that crashes, it's the LCD.

Try some systematic fault finding. Start by disconnecting this input circuit completely, at the end near your board, and see if it still fails when the motor starts.

Next, attach the cable (only) at your board, but leave it disconnected at the flow meter. This shows whether the cable is acting as an antenna and picking up a harmful RF spike.

Then, connect the ground (only) to the flow meter, and try again. If that's still OK, then connect to the other end of the switch.

[e.sotillet]

Hi, I've been gathering more information about the circuit under review. The manufacturer asked me to cover the device.

First, I found that there is no common mode filter on the device's DC power input. The power cables are close to the AC power input cables.



Second, verify in the field that the earth connection is not made. The installer comments that the connection points are old and deficient. There is no PE , so the potential of the entire metal cabinet is floating. Possibly with residuals of 50Hz by the AC decoupling capacitor that comes in the SMPS.

Third, I don't have access to electrical transient simulators. However, use a 4kV DC source in conjunction with a spark gap to simulate such transients somewhat. I found the "hot spot" to be in the cable connecting the LCD to the main board. My hypothesis is that since the metal gabinet is not at earth potential, some capacitive coupling injects the line noise into the LCD wire.

Fourth, the 4-20mA sensor includes a third wire for shielding but remains open.

So, with this new background, what would be the recommendations?

I'm going to include a common mode filter on the DC input.

[penfold]

Photo helps a lot there, that particular meanwell PSU is nobody's friend when it comes to "noise" or transients, I'd have to search back through my notes for anything exact, but I had some rather strange compliance failures on a system with that in.

Common mode filter on the DC side helps, as does a ferrite on the AC (as many turns as possible on preferably a solid ring, but a split core will suffice with only one turn if needs be). IIRC there isn't a huge amount of filtering on the battery interface in that PSU and that can cause some interesting behaviour, giving a bit more distance between the battery wires and any mains would also help.

Does the "device" have any connection to earth itself? Perhaps regardless of the "potential" of the chassis, it may be forming a return path for noise via capacitive coupling and the noise current causing some ground bounce, seen by the LCD. A symptomatic solution (removes the problem not the cause) may be to use a ferrite or some filtering on the LCD connector (I'm imagining a ribbon cable for which you can buy specialised "long" ferrite).