MAGIS 100

04/08/2021

Murtaza, Maxime

Lightfield imaging

Stanford Prototype

Window 5.6cm from chamber center

13.7 cm diameter window aperture

Just short of 10mm width

2

Number of Views & Light Collection

3

Number of Views & Light Collection

4

Number of Views & Light Collection

5

Number of Views & Light Collection

6

Simulation results - Stanford Setup

7

• Simulations include quantum & readout noise at sensor.

Simulation results - Stanford Setup

8

• Simulations include quantum & readout noise at sensor.

Simulation results - Stanford Setup

9

• Simulations include quantum & readout noise at sensor.

Both now ~ F16.3 object side

Simulation results - In Air Setup

10

m=.109, 1.1” (12.3x12.3) mm

2.74 um pixels

F1.6; 100um features

​

m=.1504, FF (36x24) mm

3.76 um pixels

F1.6; 100um features

[38% more light in max]

60 x 60 pixels

60 x 60 pixels

Simulation results - Conclusion

• Smaller sensor is better adapted for the Stanford setup.
• Smaller light cone leads to less blurring
• 200um features seem reasonable for a proof of concept.
• Larger sensor gives us more angular coverage
• Can mitigate blurring by stopping lens down
• Results in air are pretty much identical (modulo light collected)
• May require quantitative assessment?
• Blur & 3d reconstruction.
• Do we want to add Lens model to this?

11

12

Number of Views & Light Collection

13

Stanford Prototype

Assuming window 5.6 cm from cloud either side

Window aperture diameter is 13.7 cm, radius 6.9 cm

We should be able to capture views within +/- 51^o

​

55 mm lens could be too tight,

Zeiss makes 85 mm (m=.129) & 100 mm (m=.116) lenses

https://www.zeiss.com/consumer-products/us/photography/videography/otus-lenses.html

​

14

Stanford Prototype - Camera 1 FF Sensor

• Keep m = .1504 since these are 3.76 um pixels
• We ask for densest mirror packing, and want to know number of views with mirrors past x = 5.6 cm
• Most extreme view that’s accommodated is +/- 48^o

​

15

Stanford Prototype - Camera 4 - Global Shutters!

Based on the Sony IMX530/540/541 [4/3 or 1.1” sensors]

• https://www.flir.com/products/blackfly-s-usb3/?model=BFS-U3-244S8M-C
• https://www.flir.com/products/oryx-10gige/?model=ORX-10G-245S8M
• https://www.baslerweb.com/en/products/cameras/area-scan-cameras/ace2/a2a4504-18umpro/
• https://www.baslerweb.com/en/products/cameras/area-scan-cameras/ace2/a2a4504-5gmpro/
• https://thinklucid.com/product/triton-24-mp-imx540/
• https://thinklucid.com/product/triton-20-4-mp-imx541/

GigE/USB3.0 variants. 2.74 um pixels, m=.109, can use all 3 lenses, O(250) views

Angular acceptance = +/- 40^o

All seem to quote “Temporal Dark Noise” ~ 2-3 e, some in addition say ~2e/s ?

QE is lower at ~ 70%

Don’t look cooled, but I might be wrong. USB3.0/GigE interface should allow for triggering, global shutters are a good bonus here.

Look like they’re all in the $1.9k to $3k range

16

Do we want to develop this?

https://www.sony.net/SonyInfo/News/Press/202103/21-021E/

• 3.45 um pixels, 13,400 x 9,528
• Diagonal 56.73 mm (3.6-type)
• Main functions -

Global shutter,

trigger synchronization, ROI,

gradation compression,

multi-exposure, short exposure,

pixel binning readout

• Chris: It uses the Sony SLVS-ED interface. Xilinx offers IP for this https://www.xilinx.com/products/intellectual-property/1-yx5i51.html
• Ryan Herbst will get back to us about how much effort it will take to read the sensor

17

A little bit of 1D fourier space maths...

What does our density function look like?

​

18

In Fourier space this is a convolution of the FTs from N and sin²

A little bit of 1D fourier space maths...

19

Real

Imaginary

setting b=0

In Fourier space this is a convolution of the FTs from N and sin²

Density used

20

3D fourier picture of our cloud

21

Real

Imaginary

Absolute

How to best represent clouds?

22

Colors - Interpolation

Glyphs - Points

How to best represent gradients?

23

Vector flow?

https://docs.enthought.com/mayavi/mayavi/mlab_case_studies.html#mlab-case-studies

A little bit of 1D fourier space maths...

Ray tracing approach lets us estimate the density of the atom cloud directly

→ Estimate the density as a Gaussian mixture

24

A little bit of 1D fourier space maths...

→ Can we estimate the FT of the density as a Gaussian mixture?

We’ll need to take the complex part into account somehow...(Add mixing angle parameters?)

25

Here x is the fourier space variable

Inverse FT to recover the density of the cloud

Lens Numbers

• 55mm F1.4 Zeiss Otus BB sent by Zeiss
• OTFs evaluated by Engineer-Michael
• Consider now +- 6mm on the image sensor
• MTF cyc/mm corresponds to one full period of the sine wave modulation
• Worst case MTF with 100um wavelength sine waves ~ 30%
• Worst case MTF with 200um wavelength sine waves ~ 75%
• With 200um cycles, overall contrast ratio ~ 67%

→ brightest pixel = 5 x dimmest pixel value

• With 100um cycles, overall contrast ratio ~ 20%

→ brightest pixel = 1.5 x dimmest pixel value

26

Approximate Light Collection Numbers

10⁶ atoms & 10⁸ 𝛾/s emitted per atom in 4𝜋

Imaging time 10us → 10³ 𝛾/10us/atom emitted in 4𝜋

• For F10 object side system for each individual view, we can find the ratio of the solid angle in 4𝜋
• Angle ratio = ( 𝜋 r² / d² ) / 4*4𝜋 ~ 6 10^-4 [NB 2r/d =1/10]
• → 10³ * angle_ratio = 0.6 𝛾/10us/atom/view

​

• 40x40x40 voxel space, individual views are 40x40 bins
• Assume 1mm cloud, and 10 equal fringes per cloud, 4x40 pixels per cycle (sin²), head-on view:
• 10⁶ atoms → 10⁵ atoms/cycle → 4 10⁴ /40 atoms in one bright pixel & 9/40 10³ atoms in one low pixel

→ 500 - 100 𝛾/10us in one pixel of the given view

• Contrast ratio of 2 pixels from sin² → 64% contrast ratio
• For 5 fringes → 300 - 15 𝛾/10us in one pixel of the given view & 90% contrast ratio
• NB actual number of atoms in each fringe changes the number of photons, not contrast ratio

​

Do these numbers look reasonable?

Can we get away with 5 fringes per cloud instead of 10?

(ie, Period of one full oscillation (0 to 𝜋 for sin²) = 200um)

​

27

Stanford Prototype - Camera 1

• Zero amp glow, low noise CMOS camera
• O(10^-8) e/px from dark current for 10us imaging
• 1-4 e/px from Readout Noise
• 5 sigma above noise is ~ 14 e/px

⟶ 18𝜸/px should be good with QE ~ 80%

• Electric rolling shutter

⟶ Potentially need to image for O(ms) to insure we’re simulating global shutter - Low FPS imaging

​

https://www.qhyccd.com/index.php?m=content&c=index&a=show&catid=94&id=55

QHY600M PRO/PH/L ($8k/$4.6k)

​

28

EM: 80000e

∼82% @ 461 mm

1.0e-3.7e

Dual Stage TEC

2GB PRO, 1GB(8b) PH/L

/Fibre Port PRO 2*10Gb/s

FPS 2.5 PH/ 4 PRO

Exposure Range: 40μs-3600s

Zero Amp Glow Circuitry

Camera Numbers

QHY600M PRO/PH/L

​

29

“QHY600 Performance Curves in Readout Mode #0 (Photographic Mode). In this mode there is a drop in the noise between Gain 25 and Gain 26. We recommend setting the Gain to 26 to begin. At this setting the full well is 27ke- and readout noise is 2.7e-. For every long exposures you can lower the gain from this point to increase the full well capacity.

QHY600 Performance Curves in Readout Mode #1 (High Gain Mode). Please note there is a HGC/LGC switch point at gain55 to gain56. Gain0-55 uses LGC and Gain55-100 uses HGC.

​

QHY600 Performance Curves in Readout Mode #2 (Super Fullwell Mode).

​

Now QHY600 adds #3 mode Extend Fullwell 2CMSIT (yellow curve). The advantage of this mode is that it has the same full well value and system gain as the #2 mode Extend Fullwell, but the read noise is reduced by about 1.3 times”

​

Camera Decision

QHY600M PRO

• Larger DDR3 buffer @ 16 bit
• “It has more flexible triggering so you should have an easier time integrating it with EPICS control and synchronizing it with other aspects of your experimental setup, and the frame rate and data transfer rates are higher than the other model”
• PRO supports 2 x 10 Gigabit Optical Fiber ports. Faster, high stability, allows for long transfer distance (up to 300 meters).

​

30