# Reducing Reference Voltage Error in Accelerometer and ADC Circuit

I am utilizing a specific analog Accelerometer ADXL354 and an ADC AD7682. Currently, I drive the REFIN-Pin of the ADC with the reference Voltage supplied by the Accelerometer. According to the Accelerometer's datasheet, the reference Voltage can have a deviation of $$\\pm 10\,\%\$$.

Given this, my understanding is that I need to measure the actual reference Voltage output by the accelerometer to accurately calculate the real accelerometer values using the following formula: $$G_{Acc}[g] = \frac{\left(INT_{16}[bit] \cdot \frac{V_{ref}[V]}{2^{16}[bit]}\right) - \frac{V_{ref}[V]}{2}}{G_{SENS}[\frac{mV}{g}]}$$

Given this, my understanding is that I need to measure the actual reference Voltage output by the accelerometer to accurately calculate the real accelerometer values using the following formula: $$\\pm10\,\%\$$ in my total error, which seems quite high. Is it possible to mitigate this by using an external precision Voltage reference like the MCP1501, making the error negligibly small? Alternatively, can I avoid the use of an external reference by employing the internal reference from the Accelerometer as COM-Input $$\\left(\frac{V_{ref}}{2}\right)\$$ for the ADC and disregard the need to measure the real reference Voltage output by the accelerometer?

The outputs are ratiometric with the internal 1.8 volt supply shown below on the pin (10): -

According to the Accelerometer's datasheet, the reference Voltage can have a deviation of ±10%.

That is correct but, the outputs are ratiometric with this value ($$\V_{1P8ANA}\$$) so, you should use an ADC that can use the 1.8 volts directly as its input reference then, these measurement errors cancel out.

Having said that, you still need to calibrate your final design because of these tolerances listed below: -

Then you are assured a repeatability for the x and y axis of 0.16% and, for the x axis 0.3%. The repeatability figures are valid for ten years as per note 2: -

• Given the ADC I mentioned, it can handle 1.8 Volts from the Accelerometer, as you've described. Thank you for that clarification! Could you assist me with additional information regarding the COM of the ADC? If I supply V(1P8ANA)/2 through a simple OP-AMP buffered voltage divider, is it possible to read out the values from the ADC in second complementary form, like the actual measured values (aside from the scaling sensitivity factor of the respective axis)? When COM is connected to GND, would I need to calculate my own radiometric point with software and subtract it from my measured values? Commented Sep 26, 2023 at 11:34
• @Daniel I'm unsure what this means: is it possible to read out the values from the ADC in second complementary form <-- if you mean you want the xyz values when referenced to a perfect and trustworthy reference voltage then I don't think that's possible without changing the reference voltage on the ADC. Please clarify what you mean. Of course you can use an accurate reference and read xyz plus the 1.8 volts and use code to produce a ratiometric value for xy and z. Commented Sep 26, 2023 at 12:03
• I plan to use the ADC in bipolar mode, utilizing the 1.8 volts reference from the accelerometer at the ADC. I will buffer it internally, as suggested by the datasheet, and use vref/2 = 0.9 volts at the COM port of the ADC. By setting the ADC CFG to [12:10] = 010 and [5:3] = 111, I should be able to read the ADC's output as signed int 16 values, eliminating the additional calculation required to convert from unsigned int 16 values to signed values. Commented Sep 26, 2023 at 12:48
• @Daniel I'm getting nowhere trying to understand what you are suggesting so, maybe ask a brand new question for the site and get answers from many more folk. You can link back the the question/answer above if needs be. Leave a link to the new question please. Commented Sep 26, 2023 at 12:54