False triggers of limit endstop - very sensitive input

[sn4k3]

First i had 5 NPN sensors in parallel on same input (Each one have a diode on signal line), that worked but from time a time i had a false endstop trigger, so i decide to split that into 3 inputs instead, ending in 2 sensors in parallel per input. But still same happens. Now it start to happen more often, so i decide to debug the problem, i disconnected all L- from sensors and left open, disable all limit end stops on software but keep only 1 active, and for my surprise it also happen if i left input (L-) open. Trigger is so fast that UCCNC doesn't show any LED nor on software DIAG, it only shows the Limit end stop trigger error and trigger the RESET.
When L- is open it meassure 23.5V from L- to GND and +/-0.5V from L- to L+, so it's normal, but when i touch the wire ferrule with my hand it trigger the input in a matter of seconds and sometimes milliseconds! I tried to meassure voltage while i touch it, but the voltage keep constant at 23.5V on my Multimeter, it is so fast that i guess my multimeter cant update that fast, maybe this require an oscilloscope?
My 24V and 5V power supplies are isolated. All L+ shares the 24V from psu.
If i connect L- to GND it does it job and trigger the endstop as expected and don't setback while grounded. So i'm a bit lost here, what can cause such problem and what i should check?

Image below with all my 3 endstops disconnected with only L3- configured on software, L6 is reset, and L5 is Z probe (those never fail nor random trigger)
https://i.imgur.com/0V8PCXY.jpeg

[cncdrive]

It depends on which motion controller and BOB you using.

For example on the AXBB-E the inputs are super-fast, so they are ofcourse sensitive, because short spikes will also trigger them.
If it was not like this then the inputs could not read high frequency signals like incremental encoders for example.
What you could do if your system has a noise level which false triggers the inputs is to setup a Debounce filter in the configuration, it is a software filtering, like if you placing an RC filter on the input, but the filtering is done in software, you can define the time constant to make the input less sensitive.

[sn4k3]

It depends on which motion controller and BOB you using.

For example on the AXBB-E the inputs are super-fast, so they are ofcourse sensitive, because short spikes will also trigger them.
If it was not like this then the inputs could not read high frequency signals like incremental encoders for example.
What you could do if your system has a noise level which false triggers the inputs is to setup a Debounce filter in the configuration, it is a software filtering, like if you placing an RC filter on the input, but the filtering is done in software, you can define the time constant to make the input less sensitive.

I'm with AXBB-E and UCCNC only. Does UCCNC offer any filtering option on software? Otherwise how can i do a Debounce filter on hardware? i guess a simple 100nf alone wont do it?
It can be coincidence but it start worse on winter than on summer, maybe because humidity, and when CNC is working is happen less often than idle.

[cncdrive]

The AXBB-E has superfast input optocouplers with 5MHz bandwidth, this is why it is sensitive, because they can throughput high frequency, so short signals like short spikes can also get through. This is an advantage compared to cheap BOBs with slow optos, because the AXBB-E can handle encoders and can do probing with higher speeds with higher accuracy etc. but is also a disadvantage when it comes to EMI noise rejection.
This is the reason why we implemented the debouncing and yes, debouncing is a software filter in which the motion controller is using the set time constant to filter the inputs. You set for example 1 milliseconds then all input signals which are shorter than 1 milliseconds get filtered out and ignored.
So, it acts the same as a hardware RC low pass filter, but it is implemented in the software (in the firmware of the motion controller)

And yes, you could do hardware filtering with low pass RC filter (resistor + capacitor network), but as I wrote the debouncing doing the same implemented in software.

[MikeM]

You could also try adding pull-up resistors between 24V and the inputs to the controller.

NPN sensor are open collector, meaning that in the "off" state the output is floating and prone to picking up noise.
In the "on" state the output is connected to 0V.

By adding a pull-up resistor between 24V and the output of the NPN sensor you pulling the "off" state floating output to 24V, this making it less prone to pick up noise and trigger the sensitive input of the controller.
You only need one resister per controller input for your limit switches.

[sn4k3]

The AXBB-E has superfast input optocouplers with 5MHz bandwidth, this is why it is sensitive, because they can throughput high frequency, so short signals like short spikes can also get through. This is an advantage compared to cheap BOBs with slow optos, because the AXBB-E can handle encoders and can do probing with higher speeds with higher accuracy etc. but is also a disadvantage when it comes to EMI noise rejection.
This is the reason why we implemented the debouncing and yes, debouncing is a software filter in which the motion controller is using the set time constant to filter the inputs. You set for example 1 milliseconds then all input signals which are shorter than 1 milliseconds get filtered out and ignored.
So, it acts the same as a hardware RC low pass filter, but it is implemented in the software (in the firmware of the motion controller)

And yes, you could do hardware filtering with low pass RC filter (resistor + capacitor network), but as I wrote the debouncing doing the same implemented in software.

Ok i found the option on software, i will try that and report back.

You could also try adding pull-up resistors between 24V and the inputs to the controller.

NPN sensor are open collector, meaning that in the "off" state the output is floating and prone to picking up noise.
In the "on" state the output is connected to 0V.

By adding a pull-up resistor between 24V and the output of the NPN sensor you pulling the "off" state floating output to 24V, this making it less prone to pick up noise and trigger the sensitive input of the controller.
You only need one resister per controller input for your limit switches.

Dont the inputs already have one? Because if i left it float it measure 23.5V from Input- to GND

[MikeM]

Dont the inputs already have one? Because if i left it float it measure 23.5V from Input- to GND

Not if you can trigger the input by touching it. By touching it you are ground the input enough to trigger it.

A multimeter has a very high impedance, if a multimeter did not have a very high impedance it would draw some small amount of current and that would influence the voltage reading you get.

[MikeM]

I have noticed from the photo in your post that the controller is fixed to what looks like a VSD.

If that is the case. It's not a good location for the controller.
VSD's generate a lot of EMI noise that can influence your controller inputs.

[sn4k3]

Dont the inputs already have one? Because if i left it float it measure 23.5V from Input- to GND

Not if you can trigger the input by touching it. By touching it you are ground the input enough to trigger it.

A multimeter has a very high impedance, if a multimeter did not have a very high impedance it would draw some small amount of current and that would influence the voltage reading you get.

Ok i will do a test with a 10K from + to LX.
I will also try with a better multimeter to confirm this

I have noticed from the photo in your post that the controller is fixed to what looks like a VSD.

If that is the case. It's not a good location for the controller.
VSD's generate a lot of EMI noise that can influence your controller inputs.

Yes is a VDF, it's the only place i had, case is very small, but VDF is OFF and the false triggers are worse when machine is not working. When working it can trigger only about 1h, but is randomly, still is much better than idle

[sn4k3]

From L+ to L- i measure 1.8Mohm.
I started with 47k pullup, down to 10k with same result - trigger with my hand. I go down to 1k it fix the hand trigger but when i touch it with only a probe wire it trigger so i didn't go lower than ohm value as it start to seen to much current to sink, then i removed the pullup and set the debounce to 1ms as adviced, with that problem is gone on all my tests, i will leave it on for a while to see if is enough, but seens like it will do. Thanks