Computational requirements for real-time ptychographic image reconstruction

Kaushik Datta; Andrew Rittenbach; Dong-In Kang; John Paul Walters; Stephen P. Crago; John Damoulakis

doi:10.1364/AO.58.000B19

Applied Optics
Vol. 58,
Issue 7,
pp. B19-B27
(2019)
•https://doi.org/10.1364/AO.58.000B19

Computational requirements for real-time ptychographic image reconstruction

Kaushik Datta, Andrew Rittenbach, Dong-In Kang, John Paul Walters, Stephen P. Crago, and John Damoulakis

Get PDF
Email
Share
Get Citation
Copy Citation Text
Kaushik Datta, Andrew Rittenbach, Dong-In Kang, John Paul Walters, Stephen P. Crago, and John Damoulakis, "Computational requirements for real-time ptychographic image reconstruction," Appl. Opt. 58, B19-B27 (2019)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Check for updates

Related Topics
Optics & Photonics Topics
?

The topics in this list come from the Optics and Photonics Topics applied to this article.

About this Article
History
- Original Manuscript: October 15, 2018
- Revised Manuscript: December 30, 2018
- Manuscript Accepted: January 9, 2019
- Published: February 19, 2019

Accepted Manuscript
Download Accepted Manuscript
Access to this article is made available via and is subject to OSA Publishing Terms of Use.

Abstract

Ptychographic imaging techniques can be coupled with tomographic image reconstruction techniques to obtain cross-sectional 3D images with resolution on the nanometer scale. However, such ptychographic x-ray computed tomography (PXCT) techniques require the collection of a large number of diffraction patterns. This work derives a set of equations that can be used to calculate the rate at which data can be collected given an experimental setup. It also determines the computational system requirements needed to process ptychographic data in real time as soon as it has been collected. This will expedite the ptychography step of PXCT. These theoretical results are then applied to performance data collected from reconstructing simulated diffraction patterns in order to determine the computational resources needed for real-time ptychographic processing for representative experimental setups. All of our results are independent of any specific ptychographic reconstruction algorithm.

Full Article | PDF Article

More Like This

Three-dimensional iterative multislice reconstruction for ptychographic X-ray computed tomography

Kei Shimomura, Makoto Hirose, Takaya Higashino, and Yukio Takahashi
Opt. Express 26(24) 31199-31208 (2018)

Parallel ptychographic reconstruction

Youssef S. G. Nashed, David J. Vine, Tom Peterka, Junjing Deng, Rob Ross, and Chris Jacobsen
Opt. Express 22(26) 32082-32097 (2014)

Coupled ptychography and tomography algorithm improves reconstruction of experimental data

Maik Kahnt, Johannes Becher, Dennis Brückner, Yakub Fam, Thomas Sheppard, Tobias Weissenberger, Felix Wittwer, Jan-Dierk Grunwaldt, Wilhelm Schwieger, and Christian G. Schroer
Optica 6(10) 1282-1289 (2019)

Previous Article Next Article

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (7)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (3)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (10)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

	Experimental Setup 1	Experimental Setup 2	Experimental Setup 3
Scan field dimensions	$100 μm \times 100 μm$
Scan point spacing	$S_{x} = 140 nm$ , $S_{y} = 640 nm$
Sample-to-detector distance	1.97 m
Detector pixel size	75 μm
Detector frame rate	3.5 kHz
Diffraction pattern size	$360 \times 360 pixels$
Scanning probe energy	9 keV
Scanning probe diameter	2 μm
Settling time between fly-scan lines, $t_{s}$	10 ms
Time to collect each scan field, $t_{c}$	34 s
Subscan field dimensions	$5 μm \times 5 μm$	$10 μm \times 10 μm$	$20 μm \times 20 μm$
# of subscan fields per scan field, $n$	400	100	25
Time to process each subscan field, $t_{p}$	13 s	52 s	245 s
# of GPU batches needed, $⌈ \frac{t_{p}}{t_{c}} ⌉$	1	2	8
Total GPUs needed, $g$	400	200	100

Subscan Field Size		Scan Point Count		Total Scan Points in Subscan Field	Number of Subscan Fields per Scan Field
$x$ -dim (μm)	$y$ -dim (μm)	$x$ -dim	$y$ -dim	Total Scan Points in Subscan Field	Number of Subscan Fields per Scan Field
5	5	36	32	1,152	400
5	10	36	63	2,268	200
5	20	36	126	4,536	100
10	5	72	32	2,304	200
10	10	72	63	4,536	100
10	20	72	126	9,072	50
20	5	143	32	4,576	100
20	10	143	63	9,009	50
20	20	143	126	18,018	25

	GeForce GTX 1080 Ti	Tesla P100	Tesla V100
Microarchitecture	Pascal	Pascal	Volta
Single GPU Computation
Max clock rate (GHz)	1.58	1.33	1.38
Peak SP rate (GFLOPs/s)	11.34	9.53	28.26
Single GPU Memory
Size (GB)	11	16	16
Type	GDDRX5	HBM2	HBM2
Peak bidir. BW (GB/s)	484	732	898
Node Description
Vendor/node type	SuperMicro DeepLearning 10	Microway NumberSmasher 4U Server
# GPUs per node	8	4	4

	Experimental Setup 1	Experimental Setup 2	Experimental Setup 3
Scan field dimensions	$100 μm \times 100 μm$
Scan point spacing	$S_{x} = 140 nm$ , $S_{y} = 640 nm$
Sample-to-detector distance	1.97 m
Detector pixel size	75 μm
Detector frame rate	3.5 kHz
Diffraction pattern size	$360 \times 360 pixels$
Scanning probe energy	9 keV
Scanning probe diameter	2 μm
Settling time between fly-scan lines, $t_{s}$	10 ms
Time to collect each scan field, $t_{c}$	34 s
Subscan field dimensions	$5 μm \times 5 μm$	$10 μm \times 10 μm$	$20 μm \times 20 μm$
# of subscan fields per scan field, $n$	400	100	25
Time to process each subscan field, $t_{p}$	13 s	52 s	245 s
# of GPU batches needed, $⌈ \frac{t_{p}}{t_{c}} ⌉$	1	2	8
Total GPUs needed, $g$	400	200	100

Subscan Field Size		Scan Point Count		Total Scan Points in Subscan Field	Number of Subscan Fields per Scan Field
$x$ -dim (μm)	$y$ -dim (μm)	$x$ -dim	$y$ -dim	Total Scan Points in Subscan Field	Number of Subscan Fields per Scan Field
5	5	36	32	1,152	400
5	10	36	63	2,268	200
5	20	36	126	4,536	100
10	5	72	32	2,304	200
10	10	72	63	4,536	100
10	20	72	126	9,072	50
20	5	143	32	4,576	100
20	10	143	63	9,009	50
20	20	143	126	18,018	25

Abstract

Cited By

Figures (7)

Tables (3)

Equations (10)

Applied Optics