• State Resolved
  • Locked Locked
  • Replies 10 replies
  • Answers 2 answers
  • Subscribers 31 subscribers
  • Views 1380 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

LDC1614: Noise impulses in the output values

Pradeepa Senanayake
Intellectual 715 points
Part Number: LDC1614

Hello,

I'm evaluating LDC1614 for a metal object sensing application in automotive industry. For our application, we are not interested in precise inductance measurements. It is more or less a proximity detection application. 

I'm currently reading the sensor output from the LDC1614 EVM and developing the required algorithms. However, during the tests I noticed that the output data is always contaminated with impulses. I cannot determine the reason for these random high peaks which are not sustained but repeated.

I plotted the output in excel to make things clear. Please find following two graphs.

Graph1:

Graph 2:

Graph 2 has the required waveform. As you can see, the graph 1 is contaminated with high amplitude noise signals.

I have implemented a software filter to remove the impulses. However, it would be comforting to know the reason for those impulses and any input from the experts would be appreciated.

Thanks!

  • 0
    TI__Genius 13335 points
    Hello,
    Based on your graphs it looks like you are operating at 166kHz which is a very low sensor frequency. If you are using the default LDC1614EVM sensor and settings then there is something damaged with your board. If you have connected a different sensor can you share with us the sensor characteristics (inductance, capacitance, frequency, amplitude, etc)? Also, can you share with us the register settings you are using?
    Typically if we see jumps like these it implies there is something physically wrong with the sensor like the amplitude is too low, Q factor is too low, or there is interference from an outside source directly at the sensor resonant frequency.
    Regards,
    Luke
  • 0 in reply to Luke LaPointe
    Intellectual 715 points

    Hello Luke,

    Thanks for the response.

    I'm not using the EVM sensor, I've broken it from the cutout and connected the sensor that we are working on.

    We are doing something which is not recommended by TI but is demanded by the dynamics of our application. That is, the distance between the sensor and the LDC1614 IC is around 8~10 m. We connected the sensor using a shielded cable and we are still have not finalized the production unit cable. We are completely aware of the fact that there will be high interference due to the long cable, however, we are not focusing on detecting the inductance values accurately, we just want proximity detection. As long as we can filter out the interference, it will keep us happy.

    The designed coil parameters are as follows.

    L = 695nH (This is a typo, it should be 695uH)
    F = 498 kHz
    C = 150 pF
    Q = 19.6

    However, I'm using Ferrite shielding to improve range and reduce direct interference to the sensor. Moreover, the long cable will affect all the parameters for sure.

    Following are the register values. I've set them to get 2000 samples/second/channel.

    DATA_MSB_CH0 0x0011
    DATA_LSB_CH0 0x4658
    DATA_MSB_CH1 0x0011
    DATA_LSB_CH1 0x35c8
    DATA_MSB_CH2 0x0
    DATA_LSB_CH2 0x0
    DATA_MSB_CH3 0x0
    DATA_LSB_CH3 0x0
    RCOUNT_CH0 0x0172
    RCOUNT_CH1 0x0172
    RCOUNT_CH2 0xffff
    RCOUNT_CH3 0xffff
    OFFSET_CH0 0x0000
    OFFSET_CH1 0x0000
    OFFSET_CH2 0x0000
    OFFSET_CH3 0x0000
    SETTLECOUNT_CH0 0x0100
    SETTLECOUNT_CH1 0x0100
    SETTLECOUNT_CH2 0x0400
    SETTLECOUNT_CH3 0x0400
    CLOCK_DIVIDERS_CH0 0x1001
    CLOCK_DIVIDERS_CH1 0x1001
    CLOCK_DIVIDERS_CH2 0x1001
    CLOCK_DIVIDERS_CH3 0x1001
    STATUS 0x404c
    ERROR_CONFIG 0x0001
    CONFIG 0x1e01
    MUX_CONFIG 0x820c
    RESET_DEV 0x0000
    DRIVE_CURRENT_CH0 0x8c40
    DRIVE_CURRENT_CH1 0x8c40
    DRIVE_CURRENT_CH2 0x8800
    DRIVE_CURRENT_CH3 0x8800
    MANUFACTURER_ID 0x5449
    DEVICE_ID 0x3055

    Please let me know your hypothesis.

    Thanks!

  • 0 in reply to Pradeepa Senanayake
    TI__Genius 13335 points
    Hello,
    As you mention, the 8-10m distance is not a recommended length. Having such a distance means the reflections on the line may interfere with the switching behavior of the LDC1614. Also, based on your inductance value (695nH) and capacitance values (150pF), the expected oscillation frequency should be 15.59MHz, however you are measuring 0.498MHz. This means that your cable is significantly loading the sensor. You may consider placing the sensor capacitor near the device so that the inductance of the cable is included as part of the LC tank (normally not recommended for LDC1614). You could also try adding some small capacitors (~50pF) to ground near the device in case there's common mode noise coming from the cable. Using a shielded cable is highly recommended and to potentially use a twisted pair (IN0A/IN0B) for the return routing. Also, since the sensor inductance is only 695nH there will be a large amount of noise pickup from the wiring. Typically we recommend to use an inductance value >2µH to minimize this effect, so if it's possible to increase this value that may help.
    Lastly, if you're able to measure with an oscilloscope the IN0A and IN0B pins to ground near the device and at the sensor, that will help us determine if there is something else to explore.
    Thanks!
    Luke
  • 0 in reply to Luke LaPointe
    TI__Genius 13335 points
    One additional point is your deglitch filter is set to 3.3MHz. If the sensor oscillation frequency really is 0.498MHz then you should consider dropping this to 1MHz which will provide additional filtering. However, if the sensor frequency is truly 15.59MHz, you should first reduce it to be within the specified range of the datasheet (<10MHz) then use a higher deglitch filter setting like 33MHz or 10MHz, otherwise you will be clipping the signal.
    Regards,
    Luke
  • 0 in reply to Luke LaPointe
    TI__Genius 13335 points
    Also, I just noticed that you are setting your current drive to 195µA (DRIVE_CURRENT_CH0 0x8c40). It's very likely that you are not driving the sensor with enough amplitude to oscillate properly and you are picking up extra noise as a result. You can verify this by measuring the amplitude on the IN0A/IN0B pins and increase the IDRIVE setting until you are between 1.2Vpk and 1.8Vpk for best performance.
    Regards,
    Luke
  • 0 in reply to Pradeepa Senanayake
    Intellectual 715 points

    Hello,

    After further testing I could confirm that the noise impulses are directly related to the RCOUNT value. In this application we want the sensors to work very fast. However, if I increase the RCOUNT to some arbitrary value, say 5000, the impulses get disappeared. The original RCOUNT value is 270, combined with SETTLECOUNT of 370 to get 0.5ms sampling time.

    Please let me know whether there is anything that I can do to improve performance.

    Thanks!

  • 0 in reply to Pradeepa Senanayake
    TI__Genius 13335 points
    Hello,
    Thanks for the confirmation! Having an RCOUNT setting too low can certainly introduce extra noise. The only other way to increase the sample rate is to use a faster reference frequency. If you're using the internal oscillator you're already at the maximum frequency, however you could connect an external reference frequency of 40MHz if you want more precise control.
    Additionally, please see the above replies for more improvement suggestions.
    Regards,
    Luke
  • 0 in reply to Luke LaPointe
    Intellectual 715 points
    I'm really sorry for my typo. Inductance is 695uH not nH.
  • 0
    Intellectual 715 points

    Hello,

    Since increasing the RCOUNT is not an option for our application, I used one other suggestion you have made. That is moving the sensor capacitor to the device end. In this method, I think now the cable also becomes part of the LC tank. This drastically reduced the impulse noise.

    However, I can't understand how this can reduce the noise. This method solved the problem, but, if someone can explain the theory behind this it is much appreciated.

    Thanks!

  • 0 in reply to Pradeepa Senanayake
    TI__Genius 13335 points

    Hello,

    I'm glad to hear that your system is working better. 

    Typically if you have a long wiring between LC tank and the device this represents a complex load which could introduce a second resonant frequency to the circuit. When you place the capacitor close to the device you incorporate the wiring inductance into the LC tank. Now this reduces the effectiveness of the variable inductor because it's now in series with a fixed inductor. This is one of the reasons we typically recommend not to do it. However it is a valid approach to add a series inductor especially when the LC tank inductance is too low to oscillate. Take a look at the following blog for more info: e2e.ti.com/.../inductive-sensing-how-to-use-a-tiny-2mm-pcb-inductor-as-a-sensor

    Regards,

    Luke