Post Go back to editing

Cannot avoid noise on AD7147 when touching sensor traces

Dear community,


I have a doubt related to the PCB layout for a capacitance sensor stretchable PCB, that needs to be connected to a separate PCB provided with AD7147 for capacitance acquisition. 

The sensor board is a TPU 2 layers stretchable PCB. A simple graphical representation of sensors, traces and AC Shield is given in the image below.

The structure followed for the AC Shield is the one suggested in the AD7147 PCB Guidelines technical document (apart from the fact that the AC Shield is not a plane but it replicates the trace shape (only larger)):

"If there is no space on the sensor PCB to allow routing around the sensors, use an ACSHIELD plane layer directly under the sensor layer."

My problem is the presence of noise when I touch the sensor traces. In the graph below you can note 2 peacks in the capacitance values:

- the first one (higher) appears when I touch the sensors (expected and correct)

- the second one (lower) appears when the sensor traces are touched (undesired)

My question is:

how is it possible to avoid the noise related to traces touch? Is it only a matter of improving the shielding? In that case, which is the correct way? Otherwise, what am I missing?

I was thinking of adding a dielectric film on top, with GND circle pads again on top (since my application does not contemplate the use of finger),  could this be a possible structure? IDoes this have a relationship with the noise issue?


Thank you very much,

best regards,

Gianluca Milani

  • Hi Gianluca,

    The sensor traces will act as a sensor but with less sensitivity, as the area is small. To prevent the traces responding to touches, either put a ground or AC shield layer between the traces and the fingers. Adding a ground layer will increase parasitic capacitance, AC shield layer will not. An alternative if you have an available channel is to configure the AD7147 inputs as differential inputs and connect a differential channel which runs a trace alongside the sensor trace, this will cancel out the effect of the fingers on the sensor traces, the differential trace should run up to the sensor electrode, but not be connected to a sensor electrode. Below is an example in a proximity detection application, the differential trace is referred to as a dummy trace.



  • Thank you very much Maurice, 

    very interesting!

    I will keep you updated about the news.

    Only two minor doubts:

    - do you think there could be problem using TPU as material? It is not listed in the AD7147 technical document but it has been produced by industrial grade machinery.

    - do you think there  any restrinctions about copper thickness? our is 180um

  • Hi Gianluca, I am not familiar with TPU, is it used as a PCB substrate material ? what is its dielectric constant, Similar to FR4 ? Copper thickness is not a concern, 180um is relatively thick but I don't think there will be any issue.



  • - Yes it is used for stretchable PCB, here a technical  document

    - The dielectric constant is 4,4 (10 MHz); 3,2 (1 GHz)

    Another doubt. If I have more than one AD7147 connected to the sensors (for isntance 12 sensors connected to a AD7147 and other 12 to another) the AC Shield foil between traces and finger that you suggested me has to be more than one? (each trace has to be shielded by a plane connected to the related AD7147, or it can be an AC Shield plane from the same AD7147 for all?)

  • Thanks Maurice,

    yes it is a PCB substrate material, the dielectric constant is 4,4 (10 MHz); 3,2 (1 GHz)

    One doubt regarding your previous answer:

    I have more than one AD7147 since the sensor number is higher than 12. In this case, how should I behave with the AC Shield between traces and fingers you suggested? Each trace should be placed behind the AC Shield plane connected to its related AD7147? Or one single AC Shield plane connected to a rondomly chosen AD7147 is ok for all traces?

    Thanks again,



  • Hi Gianluca, Using multiple devices simultaneously can be a challenge, creating interference between the devices as the excitation signals will not be synchronised.

    Depending on your application needs, maybe you can tolerate the noise due to the interference, if not, then the AC shields and sensors for each device must be separated from each other. If separation is not possible then you could think about enabling one AD7147 at a time by sequencing their power states and having just one device active at any one time, again all this depends on the needs of your application and whether there is interference taking place that affects the functionality.

    If AC shield of device 1 is shielding a sensor trace of device 2 then I would almost certainly say you will see noise issues.



  • Thank you very much Maurice, 

    based on what you said in all your detailed answers, do you think that these tentative layout would make sense?

    • Which of the two options would you prefer?

    • Could the vias of the second image represent a problem for signal integrity?

    • Could the fact of having the sensors in the bottom layer as in the first image represent an issue? The PCB is soft and the stackup is this:

    • In the PCB guidelines technical document they suggest an AC Shield plane on top of the traces to protect from noise as also you suggested me. Otherwise, in one sentence there is also written:

    TOP LAYER: Put Layout sensor electrodes and CIN connection traces; surround by 2 mm of an ACSHIELD plane. Place a ground plane around board edges to protect from ESD.

     how do you think they manage to avoid noise issues due to traces in this case?

    Thank you very much,

    best regards,


  • Hi Gianluca,

    Vias are not an issue, I have not known them to cause problems, so either of the layouts look good, which one you would use depends on which best serves your application, i.e. sensor pads on top layer or on bottom layer.

    My concern as mentioned earlier is interference between the two AD7147 devices, the close proximity of the two AC shields and the two sets of sensor electrodes could cause high noise in the capacitance measurements. AC shield functions only when its signal is synchronised with the sensor excitation signals, so an AC shield signal or sensor excitation signal in the system that is not synchrionised to other AC shields or sensor signals can cause high noise. I would recommend distance isolation or ground isolation between the signals associated with each AD7147 device, again the needs of your application will dictate how this can be implemented, if you could place a ring of ground around each AC shield plane, making the ground ring as wide as the application will allow.