Hawkeye Systems Design and - PowerPoint Presentation

Slide1

HawkeyeSystems Design and Optical Snapshot

Compiled February 26

th

, 2016

Slide2

The SeaHawk ProgramGoal of program is to fly two high resolution ocean color instruments carried by two CubeSat PlatformsClyde Space provides the CubeSat bus

Cloudland Instruments provides the Hawkeye Sensor

Program is funded by the Gordon and Betty Moore Foundation

The program is administered by Dr. John Morrison

Affiliated with the University of North Carolina – Wilmington

Details are available at

http://www.uncw.edu/socon

Slide3

Systems and Optical Design Snapshot OrganizationHigh level summary of approachOptical detailsRadiometry

Slide4

Optical Parameters120 meter pixel footprint on ground Minimum image size is 1800 pixels wide x 4000 rows tall x 8 bandsMaximum number of rows will be 6000 – nominally 96 seconds of data

Angular swath width is 22.6 degrees

Nominal 540 Kilometer orbit

8 spectral bands – derived from

SeaWiFS

bands

Sensitivity comparable to

SeaWiFS

Unit is designed for smaller targets, such as:

Large lakes

Coastal zones

Bays and estuaries

rivers

Slide5

Field of View from 540 Km Orbit-Santa Barbara ChannelMODISimage

1800 x 5000 pixels illustrated

Slide6

Practical ConstraintsSpacecraft is defined to be a CubeSatSevere size limitationsShorter lifetimes than larger satellitesMoney was limited for initial effort

“Off-The-Shelf” CCD arrays had to be used

Custom arrays would have been too expensive and time consuming to obtain

Guiding principle was to do the best we can in a small package with readily available parts

Excellent performance can be obtained!

Slide7

Nominal CubeSat Configuration and Coordinate System

Slide8

Orbital Operation ParametersNominal orbital height = 540 kmGround speed is 7600 meters per secondData nominally collected between 10AM and 2PM local timeLatitudes from +?? to –?? degrees will be surveyed

Orbital lifetime of 1 year is baseline

Lunar calibration is possible by rolling the spacecraft attitude

No solar calibrator is planned

Slide9

Hawkeye Spectral Bands

Slide10

Pushbroom Linear Array Concept chosen as BaselineLinear array chosen is On-Semi (Kodak) KLI-4104

Includes t

hree rows of 4080 pixels, 10 microns square

Includes one row of 8160 pixels, 5 microns square

Array is available with no RGB filter matrix over CCD sites

Full well performance is good

120K electrons for Chroma arrays

110K electrons for

Luma

array

Electronic shutter can be implemented

We do not use

Luma

array in any way

Slide11

Overview of OnSemi KLI-4104 CCD Quad-Linear Array (G, R, B, L

)

High Resolution:

Luma

(

Monochrome) Array

with

5 um

Pixels with 8,160

Count (

Active

Pixels)

Luma

Channel has 4

Outputs

High Resolution: Color (RGB) Array

with 10 um

Pixels with 4,080

Count (

Active

Pixels). For

our application

we use CCDs with no RGB filter

Each Color (Chroma) Channel has 1 OutputTwo-Phase Register ClockingElectronic Exposure Control is possible

Slide12

Design Concept8 Lens assemblies and 8 filters feeding light to 4 linear arrays, each with dedicated readout circuitry“Finderscope

” 648x486 pixel array for attitude determination

Shutter for dark frame collection

One computer collects data

Separate preamp used for every single CCD output

Four sets of clock drivers are used

Twelve 16 bit A/D converters, each with 3 channels, are used

Slide13

Optical Design Concept – Single Optical Path Shown

CCD Array

Filter

Lens assembly

Scrambler

Slide14

Optical Design Concept – 8 Arrays

Slide15

“Finderscope” Assembly

RG 830 Filter

Lens Elements

Micron 752x480 Pixel CMOS

Sensor

Field

Of

View

Slide16

Mechanical Design Concept

Solenoids for shutter

Circuit Cards

Science Apertures

Finderscope

Aperture

Interface to

CubeSat

Debug Port

Shutter Vane

Slide17

Electrical Configuration Overview

CCD Board

Analog

Board

Motherboard

Picozed

Board

Interface

Board

Slide18

CubeSat Envelope Illustration

Slide19

CubeSat Frame with Prop added for Scale

Slide20

Xilinx Zynq System on a Chip is Current Choice for Control Computer

Hardware ARM CPU and Peripherals (PS) Section

FPGA (PL) Section

Interconnect system between PS and PL Sections

Slide21

Overview of KLI-4104 GeometryKLI-4104 Linear Array Arrangement

Slide22

Multiple arrays within KLI-4104 allow for Improving Sensitivity through AggregationKLI-4104 Channel Alignment

Slide23

A Considerable amount of Raw Data is Collected for each Line of DataThe HawkEye Camera has 4 linear array CCDs in individual packages, one for each two bandsEach CCD package has 4 linear CCD arrays, 3 with 4080 pixels (Chroma arrays) and 1 with 8160 pixels (

Luma

array)

The camera will only use the Chroma arrays

Therefore, a total of 4080 pixels from 3 channels will be read from each CCD for one line of data

Three channels are read in parallel at a 2 MHz rate for each

CCDs are oversampled to improve the signal to noise ratio, so a line of data is read out 4 times during a pass over 1 ground pixel

The total number of pixels read per channel in one FOV (16

mS

long) is 16320 pixels, for an overall sampling rate near 1.02

MPixels

per second per channel

Slide24

On-Board Averaging is used to Improve the SNREach CCD channel is sampled 4 times per ground pixel, and the values summedThe 3 oversampled channels of data per CCD are further aggregated

to improve the

SNR.

So, a total of 12 raw pixel values are averaged per pixel

downlinked

On-board

parsing and aggregation is

used

to

only save 1800

pixels per line of data per band (two bands per CCD)

For all 8 bands, a total of

14,400

pixels

data

is

saved

approximately every 16

milliseconds, for a data rate saved of 900

Kpixels

per second

Slide25

A Considerable Amount of Data is Captured in one 4000 Row FrameOversampling and averaging of the raw data will be performed in processor’s FPGA Aggregation of the three channels of data will take place in real time in the ARM CPU processor

Total number of pixels for all channels in an aggregated image = 900Kpixels * 64 seconds = 58M pixels, 13 bits per pixel

Aggregated image would require 13 minutes to downlink at 1 Mbit per second

Slide26

Advantages of using only Four CCD Arrays are MultipleSmaller package – more space for GPS board and X-Band RadioReduced analog power (by 2X)Better fit to capability of single computer, saving more power

120 meter resolution (216 km swath) can be achieved

Data generated more in line with

Cubesat

download capability

1800x4000x8 pixels nominal

R

equires 12 minutes to downlink at 1 Mbit/sec

With 2x2 binning could be downlinked in single overpass using S-band

900x2000 pixel image – still good resolution

X-Band baselined for CubeSat downlink

Slide27

A Shutter is used to Provide a Dark Frame Reference Immediately before Data CaptureStrip to block light from CCD

Solenoids

Shutter

Mask

Slide28

Lightweight Metal Mask blocks Direct Path with Redundant DesignDeflection of 3 mm is adequate to block light

Mechanism is redundant

s

o either solenoid can

block the light

Normally two solenoids are actuated at same time to minimize the magnetic dipole

Slide29

Philosophy behind Shutter DesignTwo solenoids are usedShutter is normally openEither one is strong enough to close shutter individually

When used together their magnetic fields mostly cancel

Weight of solenoid plungers balances weight of shutter vane

Balanced to within a few percent

Random vibration is nearly all translational, not rotational

(mostly acoustical in nature)

Vane should not chatter

Held by spring exerting force equal to about 3X weight of vane or plungers

Cannot chatter in one direction – up against a stop

Slide30

Optical Design DetailsLens design and Performance

Radiometry and Signal to Noise Ratios

Passbands and Filter Specification

“

Finderscope

” Design

Slide31

How Optical Parameters were ChosenLens focal length is set by pixel size and desired ground footprintIntegration time is set by how long it takes spacecraft to fly over one pixel footprint on groundLens aperture is chosen so that Chroma array (5 micron) pixels) is 80% of CCD blooming charge after integration time for

Lcloud

, on a band by band basis

For

Lcloud

, Band 3 through 6 have the greatest signal

These bands employ a reduced aperture to balance the signal with the other bands

Band 2 needs a slight reduction also

Slide32

Lens Design

Triplet lens with 45 mm focal length, F/5 (9.0 mm aperture)

Max lens diameter is constrained to 20 mm by packaging limitations

Field of view is +/- 11.3 degrees in extent

Filter

CCD Plane

Lens elements

(Scrambler Front to CCD is 66 mm length)

Scrambler

Slide33

Current Plan is to use a Polarization Scrambler for each BandPlan is to use a single wedge polarization scrambler for each band to reduce CCD’s sensitivity to polarization

Crystal quartz

Wedge angle of 1.5 degrees

Thickness of 4 mm

Some deviation will result

(very small)

(Illustration from Karl

Lambrecht

Web Site)

Slide34

Lens Performance Goal Illustrated67% of the energy is a 30 micron circle (10 micron pixels)100% of the energy in a 50 micron circleThese goal are based on my experience with lots of CCD ImagesTypical MTF spec ignores sampling issues, which are significant

(Note that

pixelization

is still visible)

Slide35

Blurring due to 4X oversampling and Polarization State Splitting is Tolerable

Simulated image with

n

o added

blur

Simulated image with

h

orizontal

oversampling and polarization splitting

Binned

2x2 version

Binned

2x2

version

Slide36

Spot Sizes from 412 to 443 nm

Slide37

Spot Sizes from 490-555 nm

Slide38

Spot Sizes from 670-865 nm

Slide39

KLI-4104 Quantum Efficiency

(Derived from

OnSemi

Responsivity

D

ata)

Slide40

Pixel Values from 3 Chroma arrays and Luma array are added after shifting

Path of Ground Point

8 Pixels

8 Pixels

Note: Pixels here refers to Chroma pixels

Note: 3.2 degrees of Yaw error will result in one pixel of blurring

Slide41

Expected Radiances From Orbit

Ltypical

,

Lmax

(Clear air), and

Lcloud

are derived from Gerhard Meister’s data

Lmax

is max radiance for brighter ocean surface waters

Slide42

Expected Signal

To Noise Ratios at

Meister’s

Ltypical

,

120

Meter

Pixels

*Note:

SeaWiFS

Ltypical

is 2X higher than Meister’s in the near infrared. The CubeSat SNR is for Meister’s values.

Slide43

Bilinear Gain will be used to Fit data Within 12 bit Range (0 to 4096 ADU)

Slide44

Bilinear Curve will be defined by location of Knee and Gain below Knee

Gain above knee is nominally 42 electrons per count for all bands (since all are set to nearly saturate on a white cloud)

Llow

signal is

Ltypical

at 60 degrees north latitude

Bilinear curve is determined by values in columns in yellow

Slide45

Note that Noise in Signal is nowhere Limited by Digitization

Electrons/ADU is less that

Llow

photon noise below break

Digitization noise is less than photon noise at break

Slide46

Important points regarding Bilinear GainKnee is set to occur above Lmax, so it should never complicate ocean dataData above knee will mostly be from land area or clouds

Lunar calibration values are on both sides of knee

At knee signal resolution is everywhere better than one part in 150

Bilinear gain does not result in any loss of SNR!

Slide47

Response to Full Moon during Lunar Calibration is Mid-Range

Important Values

Slide48

Gerhard’s Filter Specification

Gerhard Meister Specification

Band Center

Nominal

Lower 50%

Upper 50%

Lower 1%

Upper 1%

Out-of-

Required

Bandwidth

Edge

Edge

Point

Point

Band Max

OOB

(nm)

(nm)

(%)

Rejection

412

15

404.5

419.5

394.5

429.5

1

0.000325

443

15

435.5

450.5

425.5

460.5

1

0.0004

490

15

482.5

497.5

472.5

507.5

1

0.000463

510

15

502.5

517.5

492.5

527.5

1

0.000476

555

15

547.5

562.5

537.5

572.5

1

0.000453

665

10

660.0

670.0

650.0

680.0

1

0.000228

765

40

745.0

785.0

735.0

795.0

1

0.000536

865

40

845.0

885.0

835.0

895.0

1

0.000257

 

Means tolerance is +/- 2 nm

Slide49

Proposed Hawkeye Specification

We have to fit in a 4 inch (10cm) cube, so our specs may need to be tailored!

Slide50

Example Filter Performance – Passband Shape (10 nm filter)(Provided by Gary Carver at Omega Optical)

Slide51

Example Filter Out-of-Band Performance(Provided by Gary Carver at Omega Optical)

Slide52

What I am Trying to Avoid – SeaWiFS Band 5

Not good!

Slide53

Filters will have Passband Shifts with AngleShift is approximately 1.4 % at 15 degrees off-axis8 nm at 555nm for a 20 nm passbandShift toward shorter wavelengths changes as the square of the incidence angle

At 15 degrees the variation with polarization is less than 0.1%

About 0.5 nm at 555 nm

Shift will require correction to data before use with ocean color algorithms

Temperature shifts of the passband will be less than 0.018 nm per degree C at 555 nm

Slide54

Optical Barrels support the Lens Elements and are Threaded for FocusingStainless steel is preferred material for barrel due to lower thermal e

xpansion coefficient

Prevents lenses from cracking due to temperature extremes

Venting is done by

drilling small

holes

in

the barrel

to allow outgassing between

optical elements

Slide55

Optical Barrels need to be BlackenedBarrels will be painted using Aeroglaze Z306 space grade flat black paint

Rims of selected optics

may need to be painted to reduce reflection

Optics

will be

mounted using

a low

outgassing adhesive that will adhere to the paint such

as

:

Hard Epoxy

3M Scotch Weld 2216, 2-part epoxy

Space Grade RTV 93500

Slide56

Array AlignmentLenses will have provisions for focusing, and X/Y adjustmentLenses will be adjusted during assembly to make the center pixel of each array coincidentHowever, arrays may not be parallel, so ends of the array may be +/- 10 pixels out of registration

Also, lens focal length may vary slightly with wavelength, so scale factor may vary slightly from band to band

Bottom line – any error greater than ½ pixel requires resampling, and it doesn’t much matter if it is resampled ½ pixel or 10 pixels

Slide57

“Finderscope” Sub-systemConcern is spacecraft may be flying with some amount of yawParameter can be measured using a small area array surveying the area of regard – the “

Finderscope

”

An exposure is captured about every 1000 lines

Near IR response is baselined (RG830 filter)

Enhances visibility of clouds and land features

Not deterministic if spacecraft is rolling

Area array chosen is Micron CMOS part MT9V-034

We have

breadboarded

and tested array and performance is good

Slide58

Finderscope Optical System CharacteristicsMicron CMOS Sensor is used: 752 x 480 pixels, 6 microns square

Focal length is 11.5

mm

F/number is F/11.5

Field of view of CCD is 22.2 x 14.3 degrees

22.2 degrees in

crosstrack

direction

Exposure captured every 1000 scans

Working

exposure is

10

milliseconds

Gives signal of

90%

of full well from a

cloud

Lens has

0.4

% barrel

distortion

Slide59

Optical System is Very Simple: Two Edmund Lenses and RG 830 filterRG 830Filter

Micron

Sensor

Slide60

Image Quality is Good mostly because it is F/11.5 and only works at Near IR Wavelengths