This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

FDC2212: Human body proximity sensor electrode design

Part Number: FDC2212
Other Parts Discussed in Thread: TPS610981

Hello,

  I am developing a human body proximity sensor with FDC2212 and have made several tests using a cooper tape of 40cm long and 2.5cm large (100cm2 area) as an electrode. I have reached about 30cm to 35cm distance of sensing wich is good enought for my application. 

  An important information is that is a battery powered system (CR2032).

I am now developing a new electrode using a double face, 0.5mm thickness, FR4, 1oz/ft^2 PCB (see picture below). As it is a double face PCB I can leave copper in top and bottom, but I am not sure what is the best pratice in order to increase distance of sensing :

1) Populate as much vias as possible  in order  to make Top and Bottom fully conected?

2) What happens if I place vias only in the extremities of the board so it closes a big loop with the twice area?

3) Place vias only in the middle so Top and Botom becames as two electrodes in paralel ? 

   

Best regards.

  • Hello,

    Can you clarify about the 2-sided PCB if both sides are connected to the same net or if one side is connected to ground or another net? If you have a ground plane beneath the sensor then you will have a much smaller signal change. If you need the ground plane for shielding purposes then you should increase the thickness of the PCB or use a hatched pattern.

    With regards to your specific questions, typically you want the entire electrode to be at the same potential which means keeping the electrical resistance low so more vias is better. However, there are a number of other factors that will have more impact on the system performance such as the geometry of the sensor electrodes, sensor frequency / transmission line effects, proximity to ground planes, placement of components / routing too close to the electrode resulting in fringing fields, housing material around the sensor, reference clock quality, device settings, etc. Usually the best answer is to build a couple prototypes and to put in the final packaging/housing so you can see the full system-level interaction and then iterate if needed.

    Regards,

    Luke

  • Hello Luke, Thanks for your fast answer.

    Clarifing your doubt, we don't need any shielding because I want to detect proximity from any direction, so my idea is to connect both sides of the PCB to the same net and connect it in the FDC2212 Input pin.

     OK, I understand your suggestion to reduce electrode resistance so vias will help this. Thanks.

     Regarding to the other factors you have mentioned :

    1)  Geometry : Please check geometry in the picture I have sent previusly, any suggestion?

    2)  Transmission line effects: I have placed FDC2212 as close as possible to the electrode, so there is a very small connection line as you can see in the following picture. Is that ok?

    3) Reference Clock Quality: I am using internal clock reference, do you think a external clock will help to improve sensibility?

    4) Other point I need your help. We have made 3374 seconds of aquisition (about 56 minuts) each second we made 12 aquisitions during 5ms and put  FDC in shutdown mode for 955ms. Each point of the following graphic (in blue) is a mean of 12 aquisitions. The parts marked in red are when there is human body close to the sensor, so we can clearly identify the event, but we have noted a strong drift in the signal. Do you have any clue of what is causing this drift ?

    Best regards,

    Renato

  • Hello,

    Thanks for the additional info. The placement and geometry look ok. For the additional questions (if I am reading the chart correctly) it looks like the drift you are seeing is about 600,000 codes / hr or about 10,000 codes / min. Note that each LSB represents about 0.162Hz (assuming 43.4MHz internal oscillator and dividers set to 1). This means that for a drift characteristic of 10,000 codes / min roughly translates to about a 1,620Hz shift / min. Now the output of the FDC is dependent on the frequency values of both the reference clock and the sensor. If either one are drifting it will show up on the output as a code change. There are a number of factors that would influence the drift, namely:

    • Temperature:
      • Reference Clock: The internal oscillator has a -13ppm/degree C shift. This translates to a 564.2Hz / degree C. It may be worth while to monitor the temperature during this test.
      • Sensor: The temperature coefficients of the LC tank capacitor. Typically a C0G type has 30ppm / C, but other caps like X7R have much higher values. We recommend C0G caps for this reason.
    • Frequency stability
      • Reference Clock: The internal oscillator has a nominal frequency of 43.4MHz, but this nominal value may drift over time. Currently there isn't a spec in the datasheet, but you can find very good external oscillators which specify this and perform much better than the internal oscillator of the part. For example, we use a 40MHz external oscillator from CTS Electrocomponents (625L3C040M00000) on the EVM. The frequency stability is specified as ±50ppm which covers initial tolerance at time of shipment, changes in supply voltage, load, temperature, and aging. 
      • Sensor: Primarily the Q factor of the LC tank will determine how stable the oscillation is. For capacitive sensing we recommend discrete inductors with low series resistance or alternatively a PCB coil can be used like what was done in the Capacitive Frost / Ice detection Design: http://www.ti.com/tool/TIDA-01465

    Also if you'd like to learn more about the resolution characteristics of the device, you can review the following application note. 

    It was written for the LDC161x, but the concepts / results are the exact same for the FDC221x.

    Hopefully this helps point you in the right direction.

    Cheers,

    Luke

  • Hello Luke,

    Thanks for your help so far. 

    1) I have designed a new PCB with a 10ppm external oscilator from ECS (ECS-2520S33-400-FN-TR) and the drift remains the same.

    2)  At LC tank we are using a 33pF/C0G/0402 and 18uH/ 0603/1.6 ohms (MLF1608C180KTA00). Do you think those are ok?

    3) Even though the serial resistance is higher (7.6 ohms), do you think a wirewonded inductor like 0603LS-183XGLB can help ?

    4) My electrode is conected to IN0, I am not using IN1 and I was woundering if I could use IN1 to cancel the drift by subtrating (In0 - In1 = In0 without drift and less noise) does it make sense for you? Is there any application note about that? If it make sense, what is the best configuration for In1 ( left floting without any electrode? should I place is close to shield or Ground?)

    5) If I decide to shield one side, where should I connect the shielding plate? On ground or in the IN0B?

    Following is the schematic for your reference:

    Best Regards,

    Renato

    We are 

  • Hi Renato,

    33pF/C0G/0402 and 18uH/ 0603/1.6 ohms (MLF1608C180KTA00) sound reasonable to me. We do not recommend unshielded wirewounded inductors since a change in L also results in a change in sensor frequency. With an unshielded inductor, L could vary with the presence of metal nearby. 

    The unused channel can be used for temperature correction. You could check out our Liquid Level Sensing Reference Design which uses reference and ambient sensors in the design. 

    If you need to shield the sensor, you should consider our FDC1004 device, which comes with shielding drivers.

  • Jiashow and Luke,

    Thanks again for your help so far.

     We are experiencing an strange behavior in our electrode.

     See picture and some information below:

      

    1) That is a 30cm x 6.2cm, 2 layers FR4 PCB with 0.5mm thickness

    2) The electrode is full of vias (as you can see in the picture) in order to keep top and bottom conected with the whole electrode in the same potencial

    3) Electrode area is about 160cm2 

    4) Ground area is about 9cm2  

    5) We are using only one chanel of FDC2212

    6) That is a battery (CR2032) system with a step up converter to 3.3V (TPS610981)

    7) Our target is to identified a humam body proximity with 20cm distance.

     The point is, when I touch the "A area" (border) the readen raw value increases, when I touch the "B Area" (middle to center) the readen raw value decreases and when I touch the ground area, the value decreases even more. We can change electrode and increse groud area if necessary, could you give me a direction to go ?  

    Best regards,

  • Hi Renato,.

    This may be due to transmission line effect and you may need to lower your sensor frequency. Could you provide your measured sensor frequency, sensor amplitude, and idrive setting so we have a better understanding of your system?

    In addition, because your sensor is quite large, we recommend using the balanced configuration by splitting the electrode into two equally sized pieces and connecting one to INA and the other to INB.

    Best,

    Jiashow

  • Jiashow,

      Thanks for your useful answer.

       The Balanced configuration works great and we got a much better SNR comparing to single plate configuration.On the other hand, we have  also tried a parallel finger configuration (with half eletrode conected to InA and other half conected to ground and the behavior looks quite similar to balanced configuration.

       Is there any app notes about balanced configuration? I was not able to find it anywhere. If there is no documentation about it, what is your opinion about the two topologies above (balanced x parallel finger) considering:

       - Humidity and temperature influence

       - General Drift behavior

       - Distance from the target

       - Difereciation between human body, metalic objects and other objects

    Best Regards,

    Renato

  • Hello Renato,

    At this time we don't have an app-note on balanced sensor configuration, but we find it to be superior for a wide range of applications.

    For the sensor design, you should ensure that the two sections of the sensor are equal in area; this provides optimum sensor balancing across temperature and other effects.  The balancing will ensure good matching on the A & B driven phases- with an unbalanced configuration the amplitudes can vary significantly, which can increase the noise level.

    For best performance, a few ferrites are recommended:

    Regards,

    ChrisO