In-pixel AI for lossy data compression at source for X-ray detectors

Manuel B. Valentin; Giuseppe Di Guglielmo; Danny Noonan; Priyanka Dilip; Panpan Huang; Adam Quinn; Thomas Zimmerman; Davide Braga; Seda Ogrenci; Chris Jacobsen; Nhan Tran; Farah Fahim

doi:10.1016/j.nima.2023.168665

In-pixel AI for lossy data compression at source for X-ray detectors

Manuel B. Valentin, Giuseppe Di Guglielmo^*, Danny Noonan, Priyanka Dilip, Panpan Huang, Adam Quinn, Thomas Zimmerman, Davide Braga, Seda Ogrenci, Chris Jacobsen, Nhan Tran, Farah Fahim

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

5 Scopus citations

Abstract

Integrating neural networks for data compression directly in the Read-Out Integrated Circuits (ROICs), i.e. the pixelated front-end, would result in a significant reduction in off-chip data transfer, overcoming the I/O bottleneck. Our ROIC test chip (AI-In-Pixel-65) is designed in a 65 nm Low Power CMOS process for the readout of pixelated X-ray detectors. Each pixel consists of an analog front-end for signal processing and a 10b analog-to-digital converter operating at 100KSPS. We compare two non-reconfigurable techniques, Principal Component Analysis (PCA) and an AutoEncoder (AE) as lossy data compression engines implemented within the pixelated area. The PCA algorithm achieves 50× compression, adds one clock cycle latency, and results in a 21% increase in the pixel area. The AE achieves 70× compression, adds 30 clock cycle latency, and results in a similar area increase.

Original language	English (US)
Article number	168665
Journal	Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment
Volume	1057
DOIs	https://doi.org/10.1016/j.nima.2023.168665
State	Published - Dec 2023

Funding

We acknowledge the Fast Machine Learning collective as an open community of multi-domain experts and collaborators. This community was important for the development of this project. GDG, DN, AQ, DB, NT and FF are supported by Fermi Research Alliance, LLC under Contract No. DE-AC02-07CH11359 with the Department of Energy (DOE), Office of Science, Office of High Energy Physics. NT and FF are also supported the DOE Early Career Research Program. NT is also supported by the DOE Office of Science, Office of Advanced Scientific Computing Research under the “Real-time Data Reduction Codesign at the Extreme Edge for Science” Project (DE-FOA-0002501). DB and AQ are also supported by the DOE Office of Science under the Microelectronics Co-Design Research Project “Hybrid Cryogenic Detector Architectures for Sensing and Edge Computing enabled by new Fabrication Processes” (LAB 21-2491).

Keywords

Data-compression at source
Machine-learning in pixel sensors
Microscopy
Ptychography
Read-out integrated circuits
X-ray detectors

ASJC Scopus subject areas

Nuclear and High Energy Physics
Instrumentation

Access to Document

10.1016/j.nima.2023.168665

Cite this

B. Valentin, M., Di Guglielmo, G., Noonan, D., Dilip, P., Huang, P., Quinn, A., Zimmerman, T., Braga, D., Ogrenci, S., Jacobsen, C., Tran, N., & Fahim, F. (2023). In-pixel AI for lossy data compression at source for X-ray detectors. Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 1057, Article 168665. https://doi.org/10.1016/j.nima.2023.168665

@article{ef9ce456a2744f57b562890ff1438d22,

title = "In-pixel AI for lossy data compression at source for X-ray detectors",

abstract = "Integrating neural networks for data compression directly in the Read-Out Integrated Circuits (ROICs), i.e. the pixelated front-end, would result in a significant reduction in off-chip data transfer, overcoming the I/O bottleneck. Our ROIC test chip (AI-In-Pixel-65) is designed in a 65 nm Low Power CMOS process for the readout of pixelated X-ray detectors. Each pixel consists of an analog front-end for signal processing and a 10b analog-to-digital converter operating at 100KSPS. We compare two non-reconfigurable techniques, Principal Component Analysis (PCA) and an AutoEncoder (AE) as lossy data compression engines implemented within the pixelated area. The PCA algorithm achieves 50× compression, adds one clock cycle latency, and results in a 21% increase in the pixel area. The AE achieves 70× compression, adds 30 clock cycle latency, and results in a similar area increase.",

keywords = "Data-compression at source, Machine-learning in pixel sensors, Microscopy, Ptychography, Read-out integrated circuits, X-ray detectors",

author = "{B. Valentin}, Manuel and {Di Guglielmo}, Giuseppe and Danny Noonan and Priyanka Dilip and Panpan Huang and Adam Quinn and Thomas Zimmerman and Davide Braga and Seda Ogrenci and Chris Jacobsen and Nhan Tran and Farah Fahim",

note = "Publisher Copyright: {\textcopyright} 2023 Elsevier B.V.",

year = "2023",

month = dec,

doi = "10.1016/j.nima.2023.168665",

language = "English (US)",

volume = "1057",

journal = "Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment",

issn = "0168-9002",

publisher = "Elsevier B.V.",

}

B. Valentin, M, Di Guglielmo, G, Noonan, D, Dilip, P, Huang, P, Quinn, A, Zimmerman, T, Braga, D, Ogrenci, S , Jacobsen, C, Tran, N & Fahim, F 2023, 'In-pixel AI for lossy data compression at source for X-ray detectors', Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 1057, 168665. https://doi.org/10.1016/j.nima.2023.168665

TY - JOUR

T1 - In-pixel AI for lossy data compression at source for X-ray detectors

AU - B. Valentin, Manuel

AU - Di Guglielmo, Giuseppe

AU - Noonan, Danny

AU - Dilip, Priyanka

AU - Huang, Panpan

AU - Quinn, Adam

AU - Zimmerman, Thomas

AU - Braga, Davide

AU - Ogrenci, Seda

AU - Jacobsen, Chris

AU - Tran, Nhan

AU - Fahim, Farah

PY - 2023/12

Y1 - 2023/12

N2 - Integrating neural networks for data compression directly in the Read-Out Integrated Circuits (ROICs), i.e. the pixelated front-end, would result in a significant reduction in off-chip data transfer, overcoming the I/O bottleneck. Our ROIC test chip (AI-In-Pixel-65) is designed in a 65 nm Low Power CMOS process for the readout of pixelated X-ray detectors. Each pixel consists of an analog front-end for signal processing and a 10b analog-to-digital converter operating at 100KSPS. We compare two non-reconfigurable techniques, Principal Component Analysis (PCA) and an AutoEncoder (AE) as lossy data compression engines implemented within the pixelated area. The PCA algorithm achieves 50× compression, adds one clock cycle latency, and results in a 21% increase in the pixel area. The AE achieves 70× compression, adds 30 clock cycle latency, and results in a similar area increase.

AB - Integrating neural networks for data compression directly in the Read-Out Integrated Circuits (ROICs), i.e. the pixelated front-end, would result in a significant reduction in off-chip data transfer, overcoming the I/O bottleneck. Our ROIC test chip (AI-In-Pixel-65) is designed in a 65 nm Low Power CMOS process for the readout of pixelated X-ray detectors. Each pixel consists of an analog front-end for signal processing and a 10b analog-to-digital converter operating at 100KSPS. We compare two non-reconfigurable techniques, Principal Component Analysis (PCA) and an AutoEncoder (AE) as lossy data compression engines implemented within the pixelated area. The PCA algorithm achieves 50× compression, adds one clock cycle latency, and results in a 21% increase in the pixel area. The AE achieves 70× compression, adds 30 clock cycle latency, and results in a similar area increase.

KW - Data-compression at source

KW - Machine-learning in pixel sensors

KW - Microscopy

KW - Ptychography

KW - Read-out integrated circuits

KW - X-ray detectors

UR - http://www.scopus.com/inward/record.url?scp=85172181828&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85172181828&partnerID=8YFLogxK

U2 - 10.1016/j.nima.2023.168665

DO - 10.1016/j.nima.2023.168665

M3 - Article

AN - SCOPUS:85172181828

SN - 0168-9002

VL - 1057

JO - Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment

JF - Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment

M1 - 168665

ER -

In-pixel AI for lossy data compression at source for X-ray detectors

Abstract

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this