Audio Test Performance and Accuracy Benchmarks

When performing test and validation of audio signals from devices such as loudspeakers, cell phones, or hearing aids, it is important to understand the quality of the measurement being made. Traditional instrumentation systems sometimes specify measurement accuracy and uncertainty for common tests. National Instruments audio measurement instrumentation such as the NI 4461 and 4462 Dynamic Signal Acquisition (DSA) devices provide specifications for hardware performance (i.e. analog input and analog output quality) rather than in terms of measurement performance (accuracy of THD and gain measurements). This document provides measurement quality benchmarks for NI DSA products as expectations for overall system (hardware and software measurement) performance.

Background Information

The analog input (AI) and analog output (AO) specifications of the NI 4461 and 4462 Dynamic Signal Acquisition (DSA) devices alone do not indicate the measurement uncertainty and accuracy that you can expect when using the devices.  The AI and AO performance coupled with signal processing and analysis software results in the measurement performance of the complete system.   

This document provides a series of benchmarks on the measurements of frequency, amplitude, total harmonic distortion (THD), THD+noise (THD+N), AC level, and DC level that would be performed to quantify audio electronics.  These benchmarks provided are typical for the entire system inclusive of an NI 4461 DSA and the NI Sound and Vibration Measurement Suite.

Products Used and Hardware System Setup

A LabVIEW VI was designed to evaluate the total measurement system performance for single-tone measurements. The AO and AI subsystems of the NI 4461, NI LabVIEW, and the NI LabVIEW Sound and Vibration Measurement Suite comprised the measurement system.  Figure 1 shows the basic wiring scheme for the test.  NI-DAQmx VIs were used to control the NI 4461 and NI Sound and Vibration Measurement Suite VIs were used to perform the measurement analysis.

Figure 1. NI 4461 External Connections

 

Single-tone measurements are measurements where the test signal and DUT output can be considered a sum of distinct components: DC bias, fundamental tone, harmonics of the fundamental, and noise. For this investigation no DC bias was generated in the test signal so this signal was centered around 0V. The amplitude of the fundamental tone, the amplitude of the harmonics, and the noise level were each swept through the dynamic range of the NI 4461 for a single input range. With each of these sweeps, the test measurements were observed.  The hardware configuration is summarized in Table 1.

Sampling Rate	100.0 kS/s
AO Output Range	± 10 V
AO Terminal Configuration	Default (Differential)
AI Input Range	± 10 V
AI Terminal Configuration	Default (Pseudodifferential)
AI Coupling	DC
IEPE	Off

Table 1.  Hardware Configuration

The test signal was generated on AO0 and acquired on AI0, then Sound and Vibration Measurement Suite VIs were used to measure the parameters of the acquired test signal. These measured values were then compared with both the values written to the output and analytical processing of these digital values. The digital values were measured by applying the Sound and Vibration Measurement Suite measurement VIs to the computed test signal. The analytical values were computed from the specified amplitudes for the fundamental amplitude and harmonics as well as the specified level for the noise.

Detailed Process

Two block sizes were used to perform measurements to look at the effect that a larger block size may have on improving accuracy and decreasing uncertainty. The short data block included 4096 samples (~0.04 s of data) and the long data block included 96000 samples (~1 s of data). In both cases, the block size provided sufficient frequency resolution to perform all measurements for the selected fundamental frequency of 5500 Hz.

Figure 2 shows the typical report format of the testing. These results are for the measurement of the fundamental frequency when sweeping the level of the noise floor.  The top graph shows the identified frequency from the analysis and the bottom graph shows various indicators for the accuracy and precision of the measurement.  Measurements such as these were used directly and indirectly to derive the results reported in this document.

Figure 2.  Fundamental Frequency vs Noise Floor

The legend shown in Figure 2 is used for all of the figures in this article.  Please refer to this legend for the remainder of this discussion unless otherwise specified.

For accuracy and uncertainty measurements, two states were found:

1. 1. Accuracy for typical test signal parameters:
      The typical accuracy and uncertainty will be measured when the swept parameter is roughly in the middle of the dynamic range for the measurement. For example, in the graph above (Figure 2), a noise level of 10 mV rms was used as the reference for the typical measurement.
   2. Measurement range when the measurement accuracy fails to meet the required value:In the graph above, the measurement range was determined by drawing horizontal lines in the bottom graph at values of ± 0.1 % error) and looking for the max or min error lines to cross this accuracy threshold. The state of the test signal at a threshold crossing is one limit of the measurement range.

Benchmark Results

The final results for typical accuracy found during the benchmark testing are listed in Table 2.

Measurement	Typical Accuracy	Typical Standard Deviation
Frequency*	±50 μHz, ±9e-7%	30µHz, 0.003% of df
Amplitude	±1 mV, ±0.05%	60µV, 0.003%
Total Harmonic Distortion (THD)	±0.05 dB	0.01 dB
THD+Noise	±0.12 dB	0.06 dB
AC Level	±0.5 mV, ±0.03%	0.013 mV, 0.0009%
DC Level	±1 mV	20 µV

Table 2.  Typical Accuracy for Benchmark Testing

* AO and AI subsystems share common clock, so clock accuracy of hardware is not evaluated by synchronized AO-AI measurement. Clock accuracy of hardware is specified at 20 ppm.

Additional Measurement Factors and Discussion

It was found that the larger block size produced more accurate, more repeatable results. Only the results for the larger block size 96000 samples (~1 s) will be reported here.  This is because block size was not shown to be a factor in measurements dominated by hardware limitations. 

There are two common behaviors of the tests.  The first is where hardware limitations could be seen as a deviation of the hardware measurement from the digital measurement as the hardware measurement approached an asymptotic limit. Figure 3 shows data exemplary of this.

Figure 3.  Total Harmonic Distortion (THD) vs Harmonic Amplitude

For decreasing harmonic amplitudes, it can be seen that the hardware traces (green and purple) begin to deviate from the analytical and digital traces as the measurement at these THD levels becomes dominated by the distortion in the NI 4461. The digital measurements also begin to deviate from the analytical solution as the harmonic components drop to a level where they become masked by the noise added to the test signal.  

The second behavior that can describe many of the tests is based on the hardware limitation as well.  The behavior is when the hardware and the software asymptotes are at different levels.  This can be seen in the Fundamental Amplitude, DC Level, and THD.  This was only seen when sweeping noise level as well.  Figure 4 shows how this looks.  The reason for this occurring is due to the limitations of the hardware and is expected.  The values of these offsets are within the listed specifications of the NI 4461.  They can also be minimized with regular calibrations and using the devices within the recommended environment. 

Figure 4. DC Offset vs Noise

The results of this benchmark show that the hardware is by far the largest cause of discrepancy between calculated values of a theoretical signal and those of an actual signal.

Conclusion

This benchmarking data is designed to show an overview of system components often used in audio test instrumentation.  For more complex systems with additional measurement types such as video or RF, you can use this benchmarking data to as a baseline determine the complete type of system you want to implement based on the current and future requirements of your application.

Related Links

• Sound and Vibration Measurement Devices and Software
• Audio Solutions and Case Studies
• Audio Test Technical Resource Library

 

Reader Comments | Submit a comment »