Willem M. Beltman

A 2.3 nJ/Frame Voice Activity Detector-Based Audio Front-End for Context-Aware System-On-Chip Applications in 32-nm CMOS

Advanced human-machine interfaces require improved embedded sensors that can seamlessly interact with the user. Voice-based communication has emerged as a promising interface for next generation mobile, automotive and hands-free devices. Presented here is such an audio front-end with Voice Activity Detection (VAD) hardware targeted for low-power embedded SoCs, featuring a 512 pt FFT, programmable filters, noise floor estimator and a decision engine which has been fabricated in 32 nm CMOS. The dual-VCC, dual-frequency design allows the core datapath to scale to near-threshold voltage (NTV), where power consumption is less than 50 uW. At peak energy efficiency, the core can process audio data at 2.3 nJ/frame - a 9.4X improvement over nominal voltage conditions.

Modal analysis for BGA shock test board and fixture design

This work serves to evaluate solder joint shock reliability for Intels desktop motherboard and OEM desktop systems. A component level dynamic test system was developed to represent motherboard system behavior and modal analysis was introduced to evaluate and calibrate the test board fundamental frequency. By mounting a mass to the test board, the test system was adjusted to simulate the BGA solder joint dynamic behavior at the motherboard and desktop system level. Modal testing was performed with various mounting masses on the test board, which represents designed heatsink masses on the motherboard. The resonance frequencies and the corresponding mode shapes of the test boards were measured. Finite element modal analysis was performed for the test board/fixture. Numerical results showed excellent agreement with the modal testing data. The testing/modeling data provides the ground work for the component level shock test and the solder joint dynamic failure strength evaluation.

A 2.3nJ/frame Voice Activity Detector based audio front-end for context-aware System-on-Chip applications in 32nm CMOS

An audio front-end with Voice Activity Detection (VAD) hardware targeted for low-power embedded SoCs, featuring a 512pt FFT, programmable filters, noise floor estimator and a decision engine has been fabricated in 32nm CMOS. The dual-VCC, dual-frequency design allows the core datapath to scale to near-threshold voltage, where power consumption is less than 50μW. At peak energy efficiency, the core can process audio data at 2.3nJ/frame - a 9.4X improvement over nominal voltage conditions.

Differential Prefrontal Response during Natural and Synthetic Speech Perception: An fNIR Based Neuroergonomics Study

Synthetic speech has a growing role in human computer interaction and automated systems with the emergence of ubiquitous computing such as smart phones, car multimedia control and navigation systems. Cognitive processing costs associated with comprehension of synthetic speech relative to comprehension of natural speech have been demonstrated with behavioral (reaction time, accuracy, etc.) and self-reported (ratings, etc.) measures. In this neuroergonomics study, we have used optical brain imaging (fNIR: functional near infrared spectroscopy) to capture the brain activation of participants while they were listening to speech with varied quality, as well as natural speech. Results indicated a differential hemodynamic response with speech quality. As fNIR systems are safe, portable and record brain activation in real world settings, fNIR is a practical and minimally intrusive assessment tool for user experience researchers and can provide an objective metric for the design and development of next generation synthetic speech systems.

Applications of psychoacoustics to information technology products

Emerging usage models for computing devices require low acoustic noise, for example in home entertainment systems. Studies have shown that not only the overall level, but also the psychoacoustic aspects matter. This paper provides an overview of testing techniques that are used in the information technology industry and outlines two specific case studies. First, an extensive subjective psychoacoustic study was designed and conducted in multiple geographies to determine the aspects of sound that best describe the annoyance to sound from information technology products in a home type environment. Over 200 participants in four countries participated in this carefully controlled experiment and rated typical steady state sounds on a 5 point annoyance scale. The relevant sound quality metrics were extracted and geographical variations quantified. Second, in a paired comparison study the influence of modulation on annoyance was investigated by superimposing different frequency and amplitude modulated sounds onto a baseline sound. The results indicate that modulation can have a significant effect on subjective perception.

Radiated noise from structural vibrations of a PC

Computer systems typically contain several noise sources like fans, hard drives, and optical drives. The noise emission of these devices can be characterized on the component level, for example with sound power measurements. For fans, the backpressure effect due to chassis airflow impedance can be characterized to capture an installation effect. However, the induced vibration levels and resulting radiated structural noise when sources are placed in a system are less well understood. This paper focuses on a quantification of the radiated noise from enclosure vibrations due to vibration of a source. The surface acceleration distribution on the enclosure panels was measured. This measured vibration pattern was then used to calculate the radiated noise with several models: the Rayleigh integral model, the radiation mode model, and finally a boundary element model. The panel radiation noise was compared to the measured total noise level. This allowed a quantification of the enclosure vibration noise contributi...

Low power all digital acoustic noise suppression technique for switching voltage regulators

An all-digital method of pseudo randomly varying switching frequency of a buck VR in non-linear low power modes (hysteretic and pulse frequency modulation) is proposed in order to reduce acoustic noise. The switching frequency is maintained within the audio range, instead of being increased, and hence switching losses do not increase. The random variation in switching frequency in this scheme converts tonal noise to white noise which has lower peak power and is less noticeable to humans.

Aero‐acoustic simulations of small scale electronic cooling blowers.

The noise emission of small radial blowers is important for the cooling of notebook systems. It has been demonstrated that airflow simulations for these radial blowers can be performed to predict the blower performance, even in the presence of inlet restrictions. The current paper presents an extension of this work to include predictions of the radiated acoustic noise. The calculated blower flow field is used as an input for a second stage aero‐acoustic prediction. This paper outlines two methods. The first method uses a free field approach based on the Ffowcs‐Williams and Hawkings analogy. The second approach constructs a rotating fan source, the magnitude of which is determined by integrating the pressure across a number of blade segments. The radiated field is then determined in a boundary element calculation, both for a free field and in the presence of the fan enclosure. Results for both methods are presented and compared to experimental data.

Constant acoustic fan curves for 92 mm axial fans.

The noise emission of fans changes as a function of the operating point. Airflow and acoustic fan plenum experiments can be used to characterize these effects to estimate noise under installed conditions. Recently, the concept of iso‐acoustic fan curves was introduced by various authors, and results were presented for small radial blowers and some axial fans. The current paper presents an analysis of the iso‐acoustic fan curves for an extensive set of 92 mm axial fans. An analysis is presented of the airflow and acoustic performance of these fans, and expressions for the iso‐acoustic fan curves are derived and presented. This allows a direct comparison between the various fans based on operating point conditions.

An overview of models for viscothermal wave propagation, including fluid structure interaction

In acoustics, the standard wave propagation models neglect the effects of viscosity and thermal conductivity. When waves propagate in narrow tubes or thin layers these simplifications might not be accurate. This paper presents an overview of models that take the effects of inertia, viscosity, thermal conductivity and compressibility into account. Based on the use of dimensionless parameters, three classes of models are outlined. The most important dimensionless parameter is the shear wave number, an unsteady Reynolds number that indicates the ratio between inertial and viscous effects. These viscothermal wave propagation models can be coupled to structural models to capture the fluid structure interaction. Analytical solutions can be found for these coupled analysis cases only for simple geometries and boundary conditions. For more complex geometries, numerical models were developed. Examples of applications of these models are also presented.