FEDEVEL
Platform forum

User , 02-25-2025, 07:01 PM

Hello; I have an immunity problem at 20V/m during testing at spot frequencies, I have very sensitive analog lines (uV measurements) running from a sensor to an unshielded cable (15cm long) to a PCB; adding capacitors right after the connector of the PCB helped but not sufficiently; can someone give advices on what to do in such case? Would you try adding a shield to the cable? It seems that at higher frequencies, we pick noise on the PCB, so tackling the cable only should not be a good enough solution at higher frequencies

QDrives , 02-25-2025, 07:51 PM

Did you add a resistor in the sense line too before you get to the capacitor?
If not, the inductance from the cable together with the capacitor will cause a resonance.

User , 02-25-2025, 07:59 PM

not yet, we have indeed resonances frequencies at which we fail immunity, what value would you add?

QDrives , 02-25-2025, 08:16 PM

It depends.
You get an RC filter, so how much delay can you have?
The resistor will add some losses too, but how much can you have?

User , 02-25-2025, 08:46 PM

Delay should not be a problem; very slow system, for voltage drop I'll check it out, so I guess you advise to use the maximum resistor possible

QDrives , 02-25-2025, 08:56 PM

If delay is no problem you could use a 1k resistor.
For the voltage drop you may need to adjust the scaling a bit too.

User , 02-25-2025, 08:58 PM

Thanks, would you try shielding the cable?

QDrives , 02-25-2025, 08:59 PM

https://www.youtube.com/watch?v=Hj-HVve1QC4

User , 02-25-2025, 09:00 PM

I watched that video, conclusion is it's tough!!!

QDrives , 02-25-2025, 09:35 PM

You can make a simulation with the parasitic elements and see how that effects things.
Do not forget the power supply to your sensor too.

User , 02-26-2025, 11:37 AM

The noise I pick is sensed on differential analog lines, we have cm caps and diff caps on these lines, any idea on what is really happening? I assume the noise is picked up as Common Mode through the cable (cable is ywisted,there are 4 others lines in this cable btw), how can I have diff noise

QDrives , 02-26-2025, 04:33 PM

I assume the signal is on a twisted **pair**?
If so, and the signal is differential, but you still get measurement errors, then I would say you need to look at the sensor, power or measurement device.
You could add a common mode choke to help with the CMRR.

User , 02-26-2025, 05:08 PM

The cable from the sensor contains 6 wires; among which 2 are for power; 2 others are for the analog differential pair (I don't recall what are the others), the wires are inside a shielded cable, but the insulation and shielding stop mid cable to expose the wires over 10cm, all the wires are twisted together, signal lines / power lines are not specifically twisted together, so it may help to twist them together separately, makes sense

QDrives , 02-26-2025, 05:16 PM

Yes, you need something like the 4x2x0.25mm² cable -- also known as CAT6.
Power is a pair, your analog signal is a pair.

You mentioned 15cm cable and 10cm bare wires? And shielded but also unshielded.

User , 02-26-2025, 05:47 PM

Yes, first half of the cable contains a shield, second half has its shield and insulation removed to expose the wires, shield is connected to the structure through a 10cm long wire, it seems (not sure) that unconnecting the shield wire from the structure removes a resonance at a particular frequency (I don't know about the other frequencies)

User , 02-26-2025, 05:48 PM

Why is power pain? We only need 20mA or so

QDrives , 02-26-2025, 06:17 PM

Sorry, power is a **pair**.

QDrives , 02-26-2025, 06:19 PM

You have a 10cm pig tail? Then I think you are better off without 'shielding'.

User , 02-26-2025, 07:09 PM

Exactly, 10cm of pigtail, so you confirm that removing the shield would be better in your opinion? Letting the shield floating would make things worse

User , 02-26-2025, 07:10 PM

In reality, the shield is connected on one the other side to the chassis with screws, so maybe removing the pigtail (on the other side) only would be better

QDrives , 02-26-2025, 07:52 PM

If the shield is connected on one side to chassis and on the other end to Gnd, both with a good connection, then use that.
A pigtail is useless and you may be better of without shield if you do.

User , 02-26-2025, 08:57 PM

One side of the shield is connected to the chassis, the chassis is itself grounded somewhere else. I guess removing the pigtail would not hurt

User , 02-27-2025, 06:56 PM

Update: Cutting off the shield wire on one side, leaving the shield connected to the structure on the other side made things worse at our frequency of interest, any explanation for this phenomenon?

QDrives , 02-27-2025, 07:09 PM

"*any explanation for this phenomenon*" -- not without much more details -- schematic, layout, photos, etc.

Although one simple response is to improve the Gnd connection a lot.
Use some wide conductor (eg. copper tape) to connect both Gnds together.

User , 02-27-2025, 09:23 PM

I understand your need for the schematic, but unfortunately I cannot share it online...

User , 02-27-2025, 09:25 PM

Good idea, I use a Cell Load sensor, its metallic structure is connected to a braid shield which contain the analog lines, it seems that when we radiating over the Cell Load is what produces the most noise picked up. Could it be that the metallic structure of the load directly couples the noise inside the shield to the analog lines? The shield is cut halfway it would be costly to remove the braid completely
The cell load is connected to the chassis which is grounded

QDrives , 02-27-2025, 09:47 PM

Radiated immunity is one of the hardest things to solve, especially without details.
Now I understand that you do not want/allowed to share, but then do not expect a useful answer.
Shielding can make things better, **but** it can also make things worse.

One of the reasons immunity is so difficult is that it is not always clear which structure is (partly) responsible for the problem. You cannot place a near field probe somewhere.

User , 02-27-2025, 10:14 PM

Yes I am not allowed to share the details and I understand it is difficult for you to help in such case; thank you very much for your help

User , 02-27-2025, 10:14 PM

Definitely harder than radiated emission!