MEMS Microphone Design

The market for MEMS microphones has been growing rapidly over the past few years. As well as mobile devices, microphones are now being increasingly adopted in consumer electronics. Smart devices now use two or more MEMS microphones to improve directional sensitivity and employ Active Noise Cancellation for better sound quality. Microphone arrays are also being used in other consumer-based products (like Amazon Echo and Google Home), with multi-directional functions to improve performance.

Most modern MEMS microphones are based on air-gap capacitors with a single fixed backplate and movable membrane. More sophisticated configurations can support dual backplates or dual membranes. Recently, there has also been increasing interest in using the piezoelectric effect as a transduction mechanism to convert acoustic waves into an electrical signal.

The Challenges of MEMS Microphone Design

From conceptual design, to optimization and performance enhancement, CoventorMP®can simulate a wide range of key issues related to successful MEMS Microphone design. These include:

  • Capacitive (or PZE sensing) with single backplate, dual backplate or dual membrane configurations
  • Mode frequencies with fluid loading
  • Robust pull-in, contact zipping and lift-off
  • Sensitivity, pre and post-contact
  • Transient pressure overloads
  • Damping and associated thermoacoustic noise analysis in the sensing element to predict Signal to Noise Ratio (SNR)
  • Simulation in an IC design platform using Cadence Virtuoso® or a VerilogA compatible simulator
  • Total Harmonic Distortion (THD) using Cadence Virtuoso

In capacitive microphones, the pull-in voltage is a key parameter that needs to be optimized in the design loop. Often, the high degree of gap non-linearity, combined with contact and stress, make it difficult to simulate pull-in and post contact behavior (zipping) using standard dc sweep-based algorithms. To address this issue, CoventorMP® offers a robust continuation algorithm, that finds both stable and unstable conditions, with voltage driven pull-in and lift-off simulation capabilities.

Quarter condenser MEMS Microphone model actuated into contact. For clarity the model is scaled in Z by a factor of 20 and contact bumpers hidden. The left hand plot shows capacitance as a function of voltage, simulated using a continuation algorithm to determine both stable and unstable states. The right hand plot show the variation in sensitivity at 1 KHz with bias voltage.

Quarter condenser MEMS Microphone model actuated into contact. For clarity the model is scaled in Z by a factor of 20 and contact bumpers hidden. The left hand plot shows capacitance as a function of voltage, simulated using a continuation algorithm to determine both stable and unstable states. The right hand plot show the variation in sensitivity at 1 KHz with bias voltage.

Another key design consideration for microphones is the SNR. One common approach in modeling the SNR is to build an electrical equivalent model from lumped components in a network (Spice-like) simulator. This approach generally requires specialist design expertise and substantial time to build a complete equivalent model. As an alternative, CoventorMP offers the capability to simulate sensitivity and noise directly from the multi-physics element model, from which the SNR can be computed.

Quarter Microphone model viewing squeeze film damping elements, back chamber (scaled for clarity) and vent resistance. Sensitivity is shown in the left hand plot and output capacitance noise generated in the fluidic system in the right hand plot

Quarter Microphone model viewing squeeze film damping elements, back chamber (scaled for clarity) and vent resistance. Sensitivity is shown in the left hand plot and output capacitance noise generated in the fluidic system in the right hand plot

Co-simulation with the Circuit

In fact, the same SNR simulations can be performed by loading the multi-physics model directly in to Cadence Virtuoso. Here, the advantage is that the effect of parasitic loads, and indeed any circuit element or biasing scheme on the response, can be simulated directly using Cadence Spectre®. All the non-linear physics, including damping models for noise analysis, are included by default. Furthermore, as the MEMS+ models are inherently non-linear, the THD can be simulated as well time domain performance, for example to model startup effects.

Condenser microphone model in Cadence Virtuoso biased under constant charge. Parasitic capacitance and leakage resistance loads the output. An A-weighting filter is included in order to compute the A-weighted SNR. The inset image shows Spectre simulated results, post processed to computed Sensitivity, SNR and THD

Condenser microphone model in Cadence Virtuoso biased under constant charge. Parasitic capacitance and leakage resistance loads the output. An A-weighting filter is included in order to compute the A-weighted SNR. The inset image shows Spectre simulated results, post processed to computed Sensitivity, SNR and THD

In conclusion, CoventorMP enables a designer to simulate key issues that are specific to MEMS microphones, from pull-in to sensitivity and noise calculations, including circuit co-simulation using direct model export to Cadence Virtuoso or VerilogA.