LDC1614: Cross talking between PCB coils connected to LDC1614

Jo,

There is an app note that may help:

The title is EMI Considerations for Inductive Sensing, and is available at: 

Please take a look at it and let me know if you have any more questions. 

Jo,

The URL did no embed in my last reply as expected so may not appear in the alert message.
It is included here:
www.ti.com/.../snoa962.pdf

Jo,

Thank you for the feedback.
Please let me know if you have any more questions.

Hi John,

I made a PCB redesign with 4 layers (solid GND and VCC layers) plus common mode chokes as in LDC1614EMV.
There was really no space left for shunt capacitors- I hope the chokes will be equally effective in decreasing EMI effects.

Some other consideration and project clarifications- would appreciate your improvement comments in general:

The application is a snap-dome button interface for a harsh environment joystick. As mentioned, there are 2 PCBs:

- Top circular PCB (45mm diameter) with (7) 2-layer coils (~ 8mm)  symmetrically distributed plus tank capacitors very close to them.

- Bottom circular PCB (45mm diameter) with 2 LDC1614 devices (one channel unused on one device), LDO and MCU. (This one is 4-layer PCB now).

- Both PCBs are stacked together with standard 2.54mm connectors, arranged  in a circular form at PCB borders. The distance between the PCBs
  is about 10mm.

I fear, that mentioned connectors act as antennas- there is no easy way to replace them with short twisted pair cables because the connectors
are part of the mechanical stack structure (base PCB carries top-PCB).

Another idea is to get rid of connectors by designing a single flexible PCB, which could be bent to an U-shape or some other form fitting
perfectly into the joystick button housing. That would allow to route the coil terminals as differential pair.

But I don't have experience with flexible PCB design, there might some pitfalls as well.

Best Regards,
Jo

Hello Justin,

Got the point, thank you.

Also, I would like to submit a comment on Inductive Sensing FAQ:

It is answered there, that cross talking between PCB coils can be an issue if the coils are too close together (less than 2D from center to center).
I my project case it is slightly less.

One of the suggested solutions in the FAQ is to ensure a LC-tank frequency difference of at least 15% from tank to tank. Indeed, that is solving
my issue (one button was about 100% more sensitive than expected, and a neighbor button far less sensitive than it should be). After introducing
different LC-tank frequencies, the problem was entirely gone.

However, I am facing a practical issue with this approach: While this method is fine for prototyping or small series, it is very time consuming on any
production scale. While one could use 1% or 2% COG capacitors for the tank, there is still the problem of induction margin of the coils introduced
by the PCB service provider, and the adjustment process can end up with considerable manual work.

It would just be more convenient, if two LC-tanks could also oscillate with almost the same frequency- I am wondering if common mode chokes
between coils and LDC1614 chip would be of any help in this particular case.

Thank you for your support,

Jo

Hello Jo,

I have a few questions about your application.

1. What is the diameter of your coils?
2. What is the frequency of your design before any shift is applied?
3. How are the coils laid out? Array or grid?
4. What is the distance between coils?

If you don't have the space to increase the distance between the coils, alternating frequencies might be enough so that you may only need 2 or 3 different frequencies. Assuming the capacitance and frequency don't go out of spec in the process, you should be able to pick a capacitor that shifts the frequency by about 20% to accommodate any tolerances.

I don't think a common mode choke between the coils would make too much impact since as the coils oscillate, they generate a magnetic field at their resonate frequency, if a coil that shares the same resonate frequency is too close, the field could couple onto the other coil which is why we recommend the frequency shift or increased distance between coils of the same frequency.

Best Regards,

Hello Justin,

Thank you for your time looking into this.

1. What is the diameter of your coils?
   10mm for all seven coils
   
   
2. What is the frequency of your design before any shift is applied?
   I have experimented with 2 different coils parameters, to find out the best button sensitivity
   Case a.):
   2-layer 10 turns each, 8mil/8mil traces.
   Frequency: ~6.9Mhz @ 270pF tank capacitor
   Q-factor: ~48
   
   Case b.)
   2-layer 14 turns each, 6mil/6mil traces.
   Frequency: ~5 Mhz @ 270pF tank capacitor
   Q-factor: ~40
   
   It turned out that the 8mil/8mil configuration performs about 35% better in terms of sensitivity. I suspect this is
   due to less parasitic capacitance between the coil traces, respective higher Q-Factor, as opposed in case b.)
   
   
   
3. How are the coils laid out? Array or grid?
   The coil PCB is 45mm in diameter.
   There is one center coil, and 6 other coils around it evenly spaced apart.
   
   
4. What is the distance between coils?
   5mm between all coils.
   

I entirely appreciate your explanation on the coil resonance problem. What I have also observed, if the resonance case is present,
the button sensitivity can be very different from the expected values, plus cross-talk between neighbor buttons in question. However,
what I find somehow surprising, the resonance case does not increase the sensor noise floor in any way.

If you don't mind, I would like to ask one more question in this thread:
This particular project was requiring max. possible button sensitivity, because the metallic snap-domes are relatively far away from the
PCB coils (about 4 mm, separated by a plexiglass sheet. That is why I have chosen the LDC1614 part, because of the 28-bit resolution.
After scaling down the rcount result, it gives me a noise-less dynamic range of about 300 (d) with the 8mil/8mil coils. This enables even
z-force detection.

My question, could I achieve same sensitivity with the more button-specific LDC2114 part ? I may have missed it, but on a first glance
I did not see a mentioning of the internal resolution of this device family.

Best Regards,

Jo 

Hello Justin,

The application is a submersible joystick, therefore a relatively large media gap between PCB coils
and metal domes was required. 

Its a different construction than gluing a film onto the snap dome buttons. Submerged, water can
circulate directly underneath the metal domes through drainage channels milled into the plexi sheet.
This has a self-cleaning effect, mud or sand which could accumulate underneath the buttons will
be flushed through the drainage channels when a button is pressed. Plexiglass was chosen  because
behind-button LED indicators are also used. Button gain and base-line algorithm are implemented in
firmware. The prototype tests of the whole construction are promising, even with 4mm media gap.
In production, this might be reduced to 3mm though. The overall design challenge was to bring 7
buttons relatively close together.

Thanks for the keypad link.

Merry Christmas!

Regards,

Jo

Hello Justin,

I thought the application is clearer to communicate with a small, simplified sketch. It is not proportional
and shows only one button, but the principle is the same for all buttons.

Before the project started, I was thinking about a keypad similar to Ti's application note. One of the identified problem
was the cavity below the snap dome. When it is completely sealed, water pressure would compress the dome, depending
on the immersion depth of the joystick. The app note is certainly fine for products resistant to splash water or very shallow
water, but I had some doubts to implement it for our application that way.

Regards,

Jo

Sorry, picture was not there..