Page 26 - EngineerIT November 2021

AUTOMATION AND CONTROL

Figure 7: Example DAQ signal chain for a battery-powered edge node sensor solution supporting single-axis vibration sensing using an ADXL1002 accelerometer.

The proposed solution is shown in Figure 7. It consists of single-
channel precision sigma-delta ADC AD7768-1 and ADC driver amplifier
ADA4805-1.

ADC selection
The AD7768-1 is a versatile precision ADC that has many modes of
operation, which allows the trade-off between power, bandwidth and
noise. The programmable digital filter is essential for aliasing rejection,
and different filter types can be used to support frequency and time
domain analyses.

In this design, we’ve chosen to operate the device with the following
configuration:
• Enable integrated reference buffer on REF+ input
• Low power mode
• Low ripple wideband filter with 32 kSPS ODR (filter option A)
• Sinc5 filter with 128 kSPS ODR (filter option B).

The integrated reference buffer allows a very compact design and Figure 8: How different digital filter responses can be used for different
removes the need for an additional buffer amplifier. This design takes measurement requirements.
the advantage of the ratiometric relationship between the ADXL1002’s
output to its supply voltage, and the fact that the AD7768-1’s reference Compared to sending the unfiltered, oversampled data to an external
buffer supports rail-to-rail operation, by sharing the 3.3 V battery digital host for postprocessing, the integrated digital filter on the AD7768-1
supply between the sensor and the ADC and using the same voltage greatly improves the power efficiency of the digital processing. The power
as the ADC’s reference voltage. Not only does this remove the need to consumption of the AD7768-1 in low power mode, with a 3.3 V supply on
generate a dedicated reference voltage for the DAQ signal chain, it also AVDD1, AVDD2, and IOVDD, and reference buffer on REF+ pin enabled, is
removes any measurement signal magnitude variation due to a change estimated to be 10.2 mW for a sinc5 filter with 128 kSPS ODR and 12.6 mW
in supply voltage, such as battery discharge over time. for a wideband low ripple filter with 32 kSPS ODR.
The low power mode operation minimises the power consumption The noise of the AD7768-1 in this configuration is 11.5 μV rms for
of the ADC to maximise the battery life. In low power mode, the filter option A and 49.5 μV rms for filter option B. The input signal to
AD7768-1 can support a brick wall style low ripple wideband filter the ADC in this design is a pseudo differential signal of ±1.32 V. The
with 13 kHz of flat (–0.1 dB) bandwidth at ODR of 32 kSPS (filter effective dynamic range of the ADC with this input range and filter
–6
option A), which is perfect for covering the 11 kHz flat bandwidth of option A is 20 × log(1.32/√(2)/11.5e ) = 98 dB, and 85.5 dB with filter
the ADXL1002 to perform frequency analysis. The brick wall filter has option B. Both are more than sufficient to let the sensor dominate the
a nearly ideal filter profile and is great for frequency analysis, but the overall noise performance.
high order filter makes it less desirable for performing time domain
analysis. For this reason, the sinc5 filter, which has great step response, AFE design
can be used to meet the need for time domain analysis. The AD7768-1 Although the ADXL1002 has a buffered output, its output impedance
in low power mode with a sinc5 filter supports output data rates up is not low enough at the ADC’s sampling frequency (2.048 MHz) to
to 128 kSPS and –3 dB frequency at 26 kHz (filter option B), enough fully settle the ADC’s input during sampling. A wide bandwidth driver
to oversample the sensor’s 21 kHz resonant frequency by more than amplifier is recommended to bridge the sensor with the ADC. The
5×. The digital filter type and the output data rate are all register ADA4805-1 is chosen for this task based on its characteristics of wide
programmable through the SPI interface, which allows the signal bandwidth, low output impedance, low noise, small size and low
bandwidth to be adjusted dynamically based on application needs. power consumption.

EngineerIT | November 2021 | 24