Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Machine Learning Applications to Static Timing Analysis
Modeling complex cell behavior is critical for accurate static timing analysis. Effective current source model, ECSM, and composite current source, CCS, waveform data compression became a necessity to reduce the size of technology files and increase the accuracy of the cell characterization data. We...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
AUC Knowledge Fountain
2022
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613421731577856 |
|---|---|
| access_status_str | Open Access |
| author | Raslan, Waseem Mohamed |
| author_browse | Raslan, Waseem Mohamed |
| author_facet | Raslan, Waseem Mohamed |
| author_sort | Raslan, Waseem Mohamed |
| collection | Thesis |
| description | Modeling complex cell behavior is critical for accurate static timing analysis. Effective current source model, ECSM, and composite current source, CCS, waveform data compression became a necessity to reduce the size of technology files and increase the accuracy of the cell characterization data. We used deep learning nonlinear Autoencoders to compress voltage and current waveforms and compared them with singular value decomposition, SVD, approach. Autoencoders gave ~1.67x compression ratio for voltage waveforms better than SVD approach and gave 45x to 55x better compression ratio compared to other lossless techniques like bz2 and gzip. Autoencoders achieved ~1.7x compression ratio for complex rising-edge current waveforms. However, SVD remains more computationally efficient than Autoencoders. Deep learning non-linear delay model, DL-NLDM, is proposed to replace the standard 7x7 non-linear delay modeling lookup tables, NLDM-LUT. The proposed DL-NLDM performed better than the standard 7x7 NLDM-LUT tables in percentage errors compared to SPICE simulation. In addition, deep learning waveform delay model, DL-WFDM, is proposed to radically change transition/delay propagation to a full waveform propagation that can be used to measure the delay or perform ECSM delay calculations.
Obtaining accurate and less demanding computational reduced models is a continuous challenge for complex systems. We propose structured recurrent neural network, S-RNN, that can model LTI single-input-single-output, SISO, and multiple-input-multiple-output, MIMO systems of any order. We showed how to obtain the continuous time transfer function of the reduced system from the trained S-RNN weights. These S-RNN models outperformed other model order reduction techniques reported in selected literature. |
| format | Thesis |
| id | oai:fount.aucegypt.edu:etds-2950 |
| institution | American University in Cairo (Egypt) |
| last_indexed | 2026-06-10T12:35:53.165Z |
| license_str | Not specified — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from AUC Knowledge Fountain — bepress |
| publishDate | 2022 |
| publishDateRange | 2022 |
| publishDateSort | 2022 |
| publisher | AUC Knowledge Fountain |
| publisherStr | AUC Knowledge Fountain |
| record_format | dspace |
| source_str | AUC Knowledge Fountain — bepress |
| spelling | oai:fount.aucegypt.edu:etds-2950 Machine Learning Applications to Static Timing Analysis Raslan, Waseem Mohamed Modeling complex cell behavior is critical for accurate static timing analysis. Effective current source model, ECSM, and composite current source, CCS, waveform data compression became a necessity to reduce the size of technology files and increase the accuracy of the cell characterization data. We used deep learning nonlinear Autoencoders to compress voltage and current waveforms and compared them with singular value decomposition, SVD, approach. Autoencoders gave ~1.67x compression ratio for voltage waveforms better than SVD approach and gave 45x to 55x better compression ratio compared to other lossless techniques like bz2 and gzip. Autoencoders achieved ~1.7x compression ratio for complex rising-edge current waveforms. However, SVD remains more computationally efficient than Autoencoders. Deep learning non-linear delay model, DL-NLDM, is proposed to replace the standard 7x7 non-linear delay modeling lookup tables, NLDM-LUT. The proposed DL-NLDM performed better than the standard 7x7 NLDM-LUT tables in percentage errors compared to SPICE simulation. In addition, deep learning waveform delay model, DL-WFDM, is proposed to radically change transition/delay propagation to a full waveform propagation that can be used to measure the delay or perform ECSM delay calculations. Obtaining accurate and less demanding computational reduced models is a continuous challenge for complex systems. We propose structured recurrent neural network, S-RNN, that can model LTI single-input-single-output, SISO, and multiple-input-multiple-output, MIMO systems of any order. We showed how to obtain the continuous time transfer function of the reduced system from the trained S-RNN weights. These S-RNN models outperformed other model order reduction techniques reported in selected literature. 2022-06-21T07:00:00Z dissertation application/pdf https://fount.aucegypt.edu/etds/1920 https://fount.aucegypt.edu/context/etds/article/2950/viewcontent/waseem_raslan_thesis.pdf Theses and Dissertations AUC Knowledge Fountain Electronic Design Automation Composite Current Source Extended Current Source Model Deep Learning Autoencoders Waveform compression Recurrent Neural Networks Model Order Reduction Electrical and Electronics Electronic Devices and Semiconductor Manufacturing VLSI and Circuits, Embedded and Hardware Systems |
| spellingShingle | Electronic Design Automation Composite Current Source Extended Current Source Model Deep Learning Autoencoders Waveform compression Recurrent Neural Networks Model Order Reduction Electrical and Electronics Electronic Devices and Semiconductor Manufacturing VLSI and Circuits, Embedded and Hardware Systems Raslan, Waseem Mohamed Machine Learning Applications to Static Timing Analysis |
| title | Machine Learning Applications to Static Timing Analysis |
| title_full | Machine Learning Applications to Static Timing Analysis |
| title_fullStr | Machine Learning Applications to Static Timing Analysis |
| title_full_unstemmed | Machine Learning Applications to Static Timing Analysis |
| title_short | Machine Learning Applications to Static Timing Analysis |
| title_sort | machine learning applications to static timing analysis |
| topic | Electronic Design Automation Composite Current Source Extended Current Source Model Deep Learning Autoencoders Waveform compression Recurrent Neural Networks Model Order Reduction Electrical and Electronics Electronic Devices and Semiconductor Manufacturing VLSI and Circuits, Embedded and Hardware Systems |
| url | https://fount.aucegypt.edu/etds/1920 https://fount.aucegypt.edu/context/etds/article/2950/viewcontent/waseem_raslan_thesis.pdf |
| work_keys_str_mv | AT raslanwaseemmohamed machinelearningapplicationstostatictiminganalysis |