Post Go back to editing

How ADXL354 Calculates Bandwidth and Sample Rate Correctly to Obtain Low Noise

I have an EVAL-ADXL354 and am using the ADS1256 for signal sampling.
Figure (1).

I don't understand how the so called noise density is calculated.

I found this formula " Noise = Noise Density * sqrt(C * 1.6)"

What I want to ask is, does the BW in the formula refer to the frequency width between DCHz and (Low Pass) Hz? Or does it depend on the Sample Rate of the ADC?

If it refers to frequency width between DCHz and (Low Pass) Hz, then EVAL-ADXL354 itself has a 32K internal resistance and an external 0.1uf capacitor, so the bandwidth (BW) is DCHz-49.8Hz.So "Noise = 20ug * sqrt(49.8Hz * 1.6)" = 178.5ug=0.000178g= 1.75×10^-3 m/s²=0.175cm/s2

So why does the noise floor of the data I observe fall about 0.6cm/s2s(ADS125 Sample Rate 100) 0.9 cm/s2s (ADS125  Sample Rate 500)

What am I thinking wrong? Or do you think the reference voltage is unstable or the ADC itself has too much noise?

sorry i really don't understand~

  • The bandwidth is equal to 1500Hz, you can check that in datasheet.

    This is due to the physical characteristic of the chip. ADXL354 has inner low pass filter of 1500Hz.

    I don't know what is your ADS1256, I don't think we have such an ADC part number.

     You can share your circuit or schematic here.

  • In your datasheet it says: ADXL354 has user adjustable analog output bandwidth.

    Sorry, ADS1256 (www.ti.com/.../ads1256)is an ADC from Texas Instruments, because I have no experience with your company's ADC, so I don't know how to choose a suitable ADC for ADXL354.

    I am using the EVAL-ADXL354 module(www.analog.com/.../EVAL-ADXL354-355-UG-1030.pdf) and I would like to be able to measure acceleration <= 0.6cm/s^2 with a sampling rate of 100Hz or 200Hz, and Low pass filter 20Hz. is this possible?

  • I suggest you using AD7682 and V1P8ANA providing the voltage reference, which can cancel noise. Your  calculation is correct. Knowing how much noise you accept, you can calculate the bandwidth and choose cap value.

    I don't know why your noise is big, maybe it is from external ADC.

    Another thing to notice is the sampling of ADC actually form another low-pass filter with frequency of ODR/2.

    I strongly recommend you using the ADC I suggested. You can now read ADC and use a data acquisition board to acquire real value. Compare them to see if it is an ADC issue.

    By the way, you can  also averaging the output of ADC to lower down the noise. For white noise, 4 samples average can make the noise half.

  • Thank you very much for your reply.

    ----------------------------------------------------
    (1.)"By the way, you can also average the output of ADC to lower down the noise. For white noise, 4 samples average can make the noise half. "

    1.Question:Do you mean to output 4 ADXL354 sensors in parallel to the ADC?

    ----------------------------------------------------

    (2.)Another thing to notice is the sampling of ADC actually form another low-pass filter with frequency of ODR/2.

    2.Question:I'm not very knowledgeable in this area can you give a general overview? Thank you very much

  • No, I mean average four output samples to one sample, it is from one ADC. This method is called over-sample.

    That is to say, with a ODR at 100Hz, you can only get data within 50Hz, higher frequency data will be filtered. So, it serves like a low pass filter.

  • thank you~I understand.
    By the way, do you have a recommendation for a 24BIT ADC suitable for ADXL354?
    Requirements: Measure the seismic frequency of 20Hz (I will directly configure a 0.25UF capacitor in the EVAL-ADXL354), and finally obtain seismic data of 100SPS or 200SPS.

    Thank you for your help

  • Why 24BIT,20BIT means 1ppm, For 24bits, the dynamic range is 135dB+, I don't think you can get that level of accuracy.

    If you are an enterprise account with volume, you can send email to China.Support <China.Support@analog.com> for answers.