MEMS microphone modules have become an essential part of modern audio technology. They are used in a wide range of devices, from smartphones to wearables and IoT devices. As the quality and reliability of these microphones is crucial, a reliable and efficient testing process is needed. In this paper, we introduce an automated tester specifically designed for testing MEMS microphone modules, which replaces an already existing system.
We focused on the following requirements:
The heart of the tester is a Raspberry Pi 4B, on which the Python script for controlling the peripherals and evaluating the measurements is performed. It is connected to a multichannel audio interface and via a BroadR-Reach interface to an infotainment system, which is used to configure the microphone boards. The GPIOs of the computer are used to read out position sensors and to control status LEDs in order to clearly assign the measurement result to the corresponding test object in the associated database and unambiguous for the operating personnel.
In order to achieve the highest possible robustness against electromagnetic interference from the production environment, it was essential to design all analogue signal paths differentially and shielded. In order to measure the inherent noise, it was also necessary to ensure that the amplifier could be electronically deactivated and that the outputs of the infotainment unit and the inputs of the audio interface were so low-noise that the inherent noise of the microphones was not masked by the background noise of the measurement system.
A key aspect in the evaluation of microphone quality is the frequency response. The tester can be used to precisely measure the frequency response of the microphone modules to ensure that they meet the specified requirements.
For this purpose, the measurement signal is output from the measurement script via the multichannel audio interface and transmitted via an in-house developed amplifier to the test loudspeaker in the measurement chamber (see Figure 3). There, the acoustic signal is recorded by the test items 1-4 and transferred via A2B bus to the infotainment system, which splits the bus signal again into 4 separate microphone channels. These 4 individual signals are then recorded by the audio interface and evaluated by the Python script on the Raspberry Pi.
For technological reasons, the microphones have a low temperature dependency, which is, however, already critical in connection with the required measurement tolerance. In order to achieve the targeted measurement accuracy of ± 0.1 dB, it would be necessary to stabilise the temperature of the test environment to approximately ± 1.5 °C. In order to achieve a correspondingly high accuracy even in non-air-conditioned rooms, the tester was designed for the simultaneous testing of 4 assemblies. Through the relative evaluation of the measurement results, the requirement for high reproducibility with simultaneously very high accuracy of the measurement results was fulfilled (see Figure 4).
Through this parallelisation of the measurement and an optimisation of the measurement script, the throughput could be increased by a factor of 4 with the same measurement time compared to the first tester. This is particularly advantageous in mass production, where efficiency and throughput are essential.
In addition, the tester is used to measure the inherent noise of the assemblies to ensure an interference-free microphone signal. For a valid test, no external noise may be recorded by the test object, which makes extremely effective sound isolation between the test chamber and the test environment essential.
Since in the production environment, e.g. due to transport carts or activities next to the tester, the mechanical input of structure-borne sound is a problem, a way had to be found to effectively interrupt this coupling path. Usually, damping rubbers or spikes are used here, but they never allow complete decoupling. For optimal mechanical decoupling, we have decided to use a magnetic suspension in which the entire tester rests suspended on a field generated by neodymium magnets. The lower cut-off frequency of this spring-mass system is so low that mechanical interference can be excluded.
For maximum damping of acoustic disturbances, multi-layered damping measures have been implemented to reduce acoustic coupling and the mechanical transmission of these signals to a maximum.
All surfaces of the outer housing were mechanically damped to minimise acoustic coupling into the housing. Furthermore, a special isolation housing was developed that surrounds the actual measuring chamber and was manufactured according to the box-in-box principle with multi-layer absorption layers. This housing was mounted on dampers in the tester and mechanically decoupled as far as possible in order to minimise the mechanical transmission from the outer housing of the tester to the measuring chamber. The inner of the two nested aluminium housings is floating in a layer of rubber granulate. To reduce resonances and as a further insulation layer, all surfaces were fully coated with a further 4mm thick damping material.
The isolation housing consists of two housing parts that are pressed together by the closing mechanism of the tester and sealed by two circumferential seals (see figure 5).
The combination of these various methods lead to the attenuation curve shown in Figure 6 and enables the measurement of the microphone's own noise in a noisy production environment.
Figure 7 shows the effect of the acoustic isolation on the basis of the measured inherent noise of 200 test items. In the noise spectrum, individual residual components of acoustic disturbances can be recognised which, without the isolation measures, would lead to incorrectly evaluated test items in the self-noise measurement. This is now effectively prevented, while microphones that exceed the permissible limit continue to be reliably detected.
The measurement system is integrated into the test environment of our supplier, so that the administration and backup of data in their measurement database is guaranteed. This ensures that only test items that have passed the test with PASS are labelled and delivered after the test. For faulty assemblies, label printing is prevented by the database (see figure 8).
The automatic test sequence is started via a uniform GUI on a test computer from our supplier. In addition to the integration into their test environment, this has the advantage of not having to train the operating personnel additionally, so that the measurement can be carried out by all persons usually responsible for tests there. After completion of the measurement, the test status is visualised via status LEDs next to the test items.
The test script is started via an SSH connection on the Raspberry Pi and the measurement data is downloaded to the control computer after the measurement is completed. The tester is thus self-sufficient from the rest of the test environment and protected against disturbances that are difficult to calculate, e.g. due to system updates of the test computer. Another advantage of this clear separation is that we are able to deliver a complete and previously thoroughly tested image of the Raspberry Pi as an SD card in case of updates of the test procedure. This means that the call of the test script and the transfer of the measurement results are the only and clearly defined interfaces, which enables us to ensure optimal portability and uninterrupted deliverability.
Precise Measurements and Effective Noise Isolation
MEMS microphone modules have become an indispensable part of modern audio technology. They are used in smartphones, wearables or IoT devices. What is the best way to test the quality and reliability of these microphones? Read our technical article on this topic.
Pioneering work for voice control, communication and artificial intelligence
Fascination for speech technology, entrepreneurial spirit and the vision to revolutionise human-machine communication with voice control: This gave birth to voice INTER connect in 2001.
Sounds, background noises, rough surrounding conditions and high demands for service quality, failure protection, and protection against cyber attacks: modern control and information systems must be capable of fulfilling a diverse set of specifications. Learn more how this can be achieved with vicCOM IP.
+49 351 40752650