Page 25 - EngineerIT November 2021

AUTOMATION AND CONTROL

Figure 4: Oversampling is required for time domain peak value detection of a
transient signal.

from 1–cos (π/OS), where OS is the oversampling ratio that equals
the effective sampling rate over the frequency of the input signal.
An oversampling by ten times the transient signal oscillation Figure 5: The ADXL1002 accelerometers’ frequency response divide
frequency can limit the peak value detection accuracy to less than
± 5%. The sensor has a noise density of 25 μg/√Hz up to 10 kHz. If the total
• Noise. As the noise contained in each sample can directly impact rms noise over 10 kHz bandwidth is 25 × √(10e ) = 2.5 mg rms with a
3
the amplitude detection accuracy of a time domain waveform, it is ±50 g input range, the sensor’s dynamic range can be calculated by
the total rms noise value that matters in the time domain analysis.
The flatness of the noise spectral density isn’t important, as long
as the total integrated noise over the effective noise bandwidth
meets the required measurement accuracy. Noise improvement DSP
techniques, such as FFT process gains, are no longer available in the The output of the ADXL1002 is a buffered voltage signal, with the
time domain analysis. amplitude proportional to both the sensed acceleration and the
• Step response. The measurement signal chain needs to have a sensor’s supply voltage. The output signal is biased at a DC voltage
good step response in order to properly replicate the profile of the that is equal to half of the sensor’s supply voltage. With a 5 V supply,
transient signal input. This impacts the filter design and selection in the ADXL1002 has a sensitivity of 40 mV/g. With a 3.3 V supply, the
the DAQ signal chain. maximum sensor output signal swing over the ±50 g input range is ±50
× 40e /5 × 3.3 = ±1.32 V, cantered at
–3
DAQ signal chain design examples
In this section, we will use two CM system DAQ signal chain examples
to show how to translate system requirements into signal chain design.

EXAMPLE 1
System requirements
• 3 V to 3.6 V battery-powered system in edge node architecture
• Single-axis vibration sensing with ±50 g range
• Support frequency analysis of up to 10 kHz of (flat) bandwidth
• Dynamic range >80 dB over 10 kHz bandwidth
• Support time domain analysis, including the shock pulse method,
with sample rate of 128 kSPS
• Equal to or less than 0.1% of dynamic nonlinearity over full-scale
range. Figure 6: The full-scale output signal of the ADXL1002
• Able to operate in a noisy environment and able to reject electro-
magnetic interference (EMI) DAQ requirements
The DAQ signal chain to interface with the ADXL1002 sensor needs to
Sensor selection meet the following requirements:
An ADXL1002 MEMS accelerometer is chosen for the task of vibration • Support the full output voltage range of the sensor
sensing. It meets the key performance criteria and has the low power • Have flat frequency response over 11 kHz
and small form factor that is well suited for edge node systems. • Able to oversample the resonant frequency by at least five times
The ADXL1002 has a flat response bandwidth of 11 kHz, which is • Let the sensor dominate the overall AC and DC performance
ideal for frequency analysis over the 10 kHz bandwidth of interest. The • Provide adequate aliasing rejection to signals outside the band of
resonant frequency of the sensor is at 21 kHz. Signals at this frequency interest
can be oversampled to support time domain analysis methods such as • Low power
the shock pulse method. • Small solution size.

EngineerIT | November 2021 | 23