Processing-in-memory startup gains microphone design win
Cette publication existe aussi en Français
Startup company PIMIC Inc. (Cupertino, Calif.) has gained a design win for its neural networking noise cancellation chip inside a digital microphone from ZillTek Technology Corp. (Hsinchu, Taiwan).
The Clarity NC100 is designed for use in single-microphone applications in edge and wearable devices where it can suppress environmental noise such as wind noise or industrial background noise.
The Clarity NC100 makes use of a processing-in-memory architecture to perform AI inference, separate out noise at a current draw of 150 microamperes.
“Developed with the latest advancements in deep neural networks and PIMIC’s innovative PIM (processing-in-memory) silicon architecture, the NC100 chip is a major breakthrough in noise cancellation technology,” said Subi Krishnamurthy, founder and CEO of PIMIC, in a statement.
Krishnamurthy claims PIMIC’s neural network approach is superior to more traditional DSP-based audio noise cancellation.
PIMIC calls its PIM architecture ‘Jetstreme’ and it is currently implemented in TSMC’s 22nm ultra-low leakage (ULL) manufacturing process using SRAM as the in-memory array. The company claims several advantges for Jetstreme
- 10x PPA advantage over existing near-memory processors
- Dynamic AI model updates
- Low power model reconfiguration
PIMIC has already implemented an always-on ultra-low-power keyword-spotting (KWS) chip based on Jetstreme called ‘Listen’.
The Clarity chip integrates into a microphone saving PCB space and implementation and is software-free, the company said.
ZillTek’s digital AI MEMS microphone will showcase the Clarity NC100 at CES 2025 in Las Vegas starting January 7, 2025. Samples are due to be available in 1Q25.
PIMIC was founded in 2022.
Related links and articles:
News articles:
Automotive-qualified MEMS microphone with analog interface for active noise cancellation
Samsung, SK Hynix partner on processing-in-memory
Samsung shows MRAM can support in-memory computing
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
