Post Go back to editing

ADXL103 / ADXL203 accelerometer with AD7887 ADC for vibration sensing

I want to measure vibration with an ADXL103/ADXL203 accelerometer and sample the analog output voltage with an AD7887 ADC at a rate of around 2.5kHz. The samples are then used to calculate an RMS value (based on 5000 samples). The bandwidth of the accelerometer will be limited to 500Hz with a 10nF capacitor as described in the datasheet.

The ADXL103/ADXL203 has an output impedance of 32k and this leads me to my question: Can I connect the accelerometer output directly to the AD7887 input or do I need a buffer amplifier between the two? The datasheet of the ADC only says that the source impedance should be kept low. Would I still be OK with the 32k impedance? I only need around 1% accuracy/resolution of the calculated RMS value and I really would like to avoid the buffer amplifier if at all possible (for space reasons).

Could someone give me some guidance please. Many thanks!

  • Hi MaVo, 

    I would recommend using a buffer between the XL and the ADC. I think when the AD7887 datasheet says low impedance means in the tens or low hundreds of Ohms. Higher impedances will result in too much harmonic distortion on the AD7887. 

    I hope this helps, 


  • Hello Pablo,

    Many thanks for the quick reply. I agree, in an ideal world there should be a buffer, but I am struggling with real estate on the PCB. In the AD7887 datasheet there is a graph showing THD versus input frequency for different source impedances and it ends at 1k. So 32k seems way above.

    However, my input signal is bandwidth limited to 500Hz and I am sampling only at 2.5kHz, so I don't know how relevant THD is at these low frequencies. Unfortunately I have no experience what the THD value actually means in practice. I understand the definition, but not the practical implications. As I said I only need a 1% FS accuracy. Any further thoughts?


  • Hi Matt, 

    One important thing to notice is that in the AD7887 datasheet says: " If the on-chip reference is to be used externally in a system, it must be buffered before it is applied elsewhere. [..] When the internal reference is disabled, SW1, shown in Figure 14, opens and the input impedance seen at the AIN1/VREF pin is the input impedance of the reference buffer, which is in the region of gigaohms. When the internal reference is enabled, the input impedance seen at the pin is typically 10 kΩ." 

    If you disabled the internal reference, then you are good. If not, then the XL signal would be severely attenuated. 

    P/S: THD gives you an idea of linearity (or non-linearity) of the system. Say you input a pure sine wave of a 1kHz frequency to the ADC, in an ideally linear system you expect to have only the 1kHz sine but digitalized at the output of the ADC. In the real world, there are non-linearities and harmonic will be seen at the output of the ADC. The THD is a way to quantify the additional signal content with respect to the input signal.


  • Hello Pablo,

    Many thanks for your further reply, much appreciated. The datasheet could be a bit clearer. The way I read it is that this high impedance only applies when the AIN1/VREF pin is used as a reference input. I don't think it applies when the pin is used as a signal input??

    Regarding THD, all I would need to know is how many effective bits (ENOB) are left, if I connect without a buffer, i.e. through the 32k impedance and a 10nF capacitor to ground, limiting the bandwidth to 500Hz. As I understand it THD gets worse the higher the input frequency and I am at the low end. If I end up with 8 bits, that's all I need.

    I picked this ADC because I need a 125°C rating, a 5V supply voltage and a minimum sampling rate of 5kHz. Do you have any suggestions for an alternative ADC? What ADC do people typically use with these accelerometers?

    Many thanks again for your help!