CHAPTER 4 Astronomical Detectors I. Detectors

1
CHAPTER 4 Astronomical Detectors I. Detectors

• The light collected by telescopes have to be
  detected, and more importantly, be recorded.
• All detectors transform energy from
  electromagnetic radiation to electron (e.g. CCD),
  or to some other particles (e.g. photographic
  plates) and molecules (e.g. retina).
• Major functions of detectors in astronomy
• To increase sensitivity
• To increase signal by integration over time
• To respond linearly to input photon signal
• To store signal for permanent record
• Will focus on optical light detection in this
  chapter
• A significant portion of the discussion on CCD in
  this Chapter is modified from the lecture notes
  of Dr Steven R. Majewski of University of
  Virginia. (http//www.astro.virginia.edu/class/maj
  ewski/astr313/lectureindex.html)

2
II. Major Characteristics of Detectors

• Quantum Efficiency (QE)
• QE No. of photons detected/No. of incident
  photons
• Detections can be crystals formed (photographic
  emulsion), photoelectrons released (PMT), and
  charge-pairs created (CCD)
• Generally a function of wavelength
• Spectral Bandwidth
• Wavelength range over which photons can be
  detected
• Linearity
• Want response to be linearly proportional to
  incident photons, or exposure, which is the
  product of light flux and exposure time.
• Non-linear detectors e.g. photographic emulsion
• Linear detectors e.g. PMT, CCD

3
II. Major Characteristics of Detectors

• Dynamic Range
• Maximum variation in signal over which detector
  output can represent photons without losing
  signal.
• We want the ratio of the largest measurable value
  to the smallest value to be as large as possible
• Time Response
• Minimum time interval over which changes in
  photon rate are detectable (CCD readout time).

Courtesy Dr Steven R. Majewski
Courtesy Dr Steven R. Majewski
4
II. Major Characteristics of Detectors

• Noise
• Ideally, the output signal should have a definite
  relation with the input photons.
• However, there is always uncertainty in the
  signal that will actually be detected.
• Sources of noise photon statistics, sky noise,
  Johnson noise, readout noise, etc.
• Spatial Resolution
• Determine the extent of detail that can be
  resolved in data
• Should be well matched to the telescope and
  instrument
• Ability to Integrate
• The ability to collect photons for an extended
  period of time is one of the most important
  advantage of any detectors over human eye.

5
III. The Human Eye
rod

• A thin convex lens
• Focal length 14 17mm
• Aperture 2 7mm
• Dynamic range 100,0001
• Photon sensor on the retina cones (photopic
  vision) in daylight, rods (scotopic vision) at
  night
• QE 3 (cone) 10 (rod)
• Cones 6-7 million
• Rods 100 million

cone
Courtesy John P Oliver, http//www.astro.ufl.edu/
oliver/ast3722/ast3722.htm
6
IV. Photographic Plates

• Photography was invented in 1840s, but use in
  astronomy only started to become popular in
  1900s.
• A thin coating of silver halide (e.g. AgBr) on a
  glass plate.
• The micron size crystals of AgBr suspended in a
  galetin emulsion.
• When a photon strikes,
• The silver ion can then combine with the
  electron to produce a silver atom
• The free silver produced in the exposed silver
  halide makes up the latent image
• The latent image is later amplified in the
  developing process. The deposit of silver
  produces a dark area in the film.
• A non-linear detecting device with low QE (4)

7
Nonlinearity of Photographic Plate
Linear region longer exposure time, higher
density of Ag deposited
Note Exposure in logarithm scale, NOT linear
scale
Just like human eye
8
Photoelectric Effect

• Most detectors in astronomy work on the principle
  of the Photoelectric Effect or related phenomena.
• Photons of sufficient energy hitting surface of
  metal releases electrons (photoelectrons)
• Energy of released electrons depends NOT on
  intensity of light (if we think of light as a
  wave), but rather on the frequency of light
  (particle nature of light).
• There is a minimum frequency of light before any
  photo-electrons can be emitted from a particular
  metal
• KEe Ephoton W hf W h(f-fmin)
• where KEe is the KE of photoelectron, is photon
  energy, W is the work function of the metal, h is
  Plancks constant, f is the photon frequency,
  fmin is the minimum photon frequency of the metal

9
V. Photomultiplier Tube (PMT)

• A photo-emissive detector photoelectrons leave
  the metal surface to be measured elsewhere.
• The photocathode can emit photoelectrons in
  response to incident photons. If placed in vacuum
  with a high positive voltage electrode (anode) to
  collect emitted electrons, we can measure photon
  arrival rate measuring current.
• Advantages low noise, high sensitivity
• linear over wide range of signal
• Disadvantages short lifetime (1-2 years),
• limited imaging ability with its bulky size

Courtesy John P Oliver, http//www.astro.ufl.edu/
oliver/ast3722/ast3722.htm
10
Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/ digitalimagin
g/concepts/concepts.html

• Electrons, just like photons, when moving with
  sufficient KE, not only release electrons from
  metals, but there is also amplification, with
  more electrons coming out of the metal than
  entering.
• Photomultiplier Tube combines this effect with
  photoelectric effect to amplify weak incoming
  light to strong electrical signal
• Individual photons detected and measured (photon
  counter)

11
Total electrons per photoelectron
Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/ digitalimagin
g/concepts/concepts.html
12
Courtesy Dr Steven R. Majewski

• Usually operated in low temperature dry ice,
  frozen CO2, 195K liquid N2, 77K (slightly too
  cold) to minimize dark current (thermal electrons
  which can be confused with photoelectrons)
• Maximum QE attained 30
• Used for detection of extremely faint source

13
VI. Charge-Coupled Device (CCD)

• CCDs are silicon-based integrated circuits
  consisting of a matrix of photodiodes which
  convert light energy in the form of photons into
  an electronic charge
• Invented in 1960s, revolutionized modern
  astronomy in 1970s.
• Standard detector for digital imaging from UV to
  infrared
• A non-photo-emissive detector photoelectrons are
  released by semiconductors, but freed
  photoelectrons stay inside
• Advantages high sensitivity, low noise,
  linearity, decent dynamic range (104 to 1), broad
  spectral response, spontaneity, ease of
  computerized data storage and analysis
• Disadvantages Relatively small field-of-view
  (rapidly improving situation with large format
  CCD being developed, the state-of-the-art now is
  40962)

14
Semi-conductors
Courtesy Dr Steven R. Majewski

• Elemental semiconductors column IVa of periodic
  table. Most popular Si, Ge
• Compound semiconductors elements in columns Ib,
  IIb, IIIa, Va, Via, VIIa of periodic table
• Compound semiconductors are made from diatomic
  molecules symmetric spanning column IVa in the
  periodic table, e.g. GaAs, InSb, HgCdTe. They
  have similar behavior as column IVa
  semiconductors.

15
Band Theory of Solid

• In semiconductor crystal lattice, the allowed
  quantum states occupy bands of closely packed
  energy levels (Band Theory of Solid)
• Valence band ground states that are normally
  completely filled
• Conduction band excited states that are normally
  completely unfilled

Courtesy Dr Steven R. Majewski

• The energy levels between the conduction and
  valence bands are forbidden
• Minimum distance between allowed states is
  represented by an energy bandgap (Eg) for
  insulators and semiconductors.
• Electron must absorb (e.g. in form of photons) at
  least Eg for it to be excited to the conduction
  band.

16
Electric Conduction in Semiconductors

• The conduction bands in semiconductors are
  normally unfilled. Electrons in the valence band
  need to absorb photon energy to lift it into
  unpopulated energy levels in the conduction band.
  
• The key for the usefulness of semiconductors for
  visible and infrared photon detectors is that
  their bandgap energies match those of visible/IR
  photons.
• For each semiconductor, long wavelength cutoff
  llong,cut hc/Egap

17
Doping of a Semiconductor

• When electrons are excited to the conduction
  band, they leave behind empty positions, or
  holes.
• Moving of electrons along one direction is
  accompanied with the moving of holes in the
  opposite direction.
• Can drastically change the conductivity of
  semiconductor by preloading it with excess of
  electrons and holes (doping)
• N-type add column Va to column IVa
  semiconductor, surplus electrons, leading to
  reduced energy bandgap, thus changing spectral
  bandwidth
• P-type add column IIIa to column IVa
  semiconductor, shortage of electrons, surpluse of
  holes. Holes moves and conductivity also
  increases

18
Metal Oxide Semiconductor (MOS) Capacitor
Courtesy Dr Steven R. Majewski

• Foundation of a single CCD pixel
• Made of semiconductor covered with thin layer of
  insulator, e.g. SiO2, with electrode (gate) on
  top.
• If semiconductor is P-doped and put the gate at
  voltage V, then holes will move away from the
  gate, but no free electrons exist to move towards
  SiO2 (depletion zone/region).
• The depletion region act as a well, or photon
  bucket, where if there are no thermally created
  electron/hole pairs, only photoelectrons will be
  stored

10mm
depletion region
Courtesy Dr Steven R. Majewski
photoelectron
19
MOS Capacitor as a electron well

• The depletion region can therefore be thought of
  as a potential well, where photoelectrons can be
  stored.
• The size of the bucket is proportional to the
  voltage of the gate electrode (bias voltage)
• Maximum charge a pixel can hold is called well
  capacity/ well depth ( 100k 300k e-/pixel). It
  affect the dynamic range of the CCD.

Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/ digitalimagin
g/concepts/concepts.html
20
Methods of collecting charges in each pixel

• 1. Charge Injection Device Switch the gate
  voltage to negative, thus repelling the electrons
  collected into the Si, where they can be
  collected and measured as a current.

21
Methods of collecting charges in each pixel

• 2. Charge coupling the principle behind CCD
• By building multiple gates on the same piece of
  Si, we can generate a series of depletion zones.
• If the gates are far enough apart (gt 1mm), then
  the wells will be independent of each other.

Courtesy Dr Steven R. Majewski
22
Methods of collecting charges in each pixel

• 2. Charge coupling the principle behind CCD
• By adjusting voltages, we can transfer charges
  from one zone to another because electrons are
  attracted to the larger gate voltage (or, seek
  the deeper well)
• Therefore we are able to move charge on Si along
  the rows of gates.

Courtesy Dr Steven R. Majewski
23
Charge Transfer in CCD

• A group of gates with a common electrical link is
  called a phase
• For each CCD pixel, it will have one gate of each
  phase
• Each phase alters its voltage with a repeat
  pattern of high (opening a well) and low
  (closing a well) states
• Need accurately timed sequence to drive stored
  electrons collected.

Voltage of P(1) V1, etc
V1 V2 V3 V1 V2 V3
- - - -
- - - -
- - - -
- - - -
Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/digitalimaging
/concepts/concepts.html
24
Full Frame Architecture

• The last row of a CCD is called the serial
  register (also called multiplexer)
• Line address readout
• Shift all columns by one pixel into multiplexer
• Readout all the multiplexer pixel electrons to an
  amplifier by shifting charges
• When multiplexer is empty, repeat 1
• Problem Still collecting photons while array
  being emptied! Resulting in smearing of image
• Solution Cover CCD with shutter during readout
  (not big problem in astronomy because sources are
  faint)

Courtesy Dr Steven R. Majewski
25
Putting it all together......
SiO2 10mm
Depletion zone 5mm
Si thickness 0.3 - 0.5mm
Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/digitalimaging
/concepts/concepts.html
26
VII. Properties of CCD

• CCD started to revolutionize astronomy in 1970s,
  full fledge effect in 1980s, spreading to amateur
  astronomy in 1990s.

Hubble Space Telescope Wide Field/ Planetary
Camera Texas Instrument 800x800 CCD
27
A. Plate Scale

• The field of view of the CCD is related to the
  physical size w of the CCD and the focal length f
  of the telescope.
• The angular field of view f 2q, where tanq
  w/2f. In most cases, f gtgtw, thus we get the
  formula fw/f.
• Usually, the plate scale is expressed in terms of
  f/M, where M is the number of pixels on that side
  of CCD. The commonly adopted unit is arcsec/pixel

28
B. Quantum Efficiency
Courtesy Apogee Instruments

• Why is it important to improve QE?
• Every fractional increase in QE means an
  equivalent reduction in either light gathering
  power (no need to spend big to build big
  telescope) or integration time (takes shorter
  time to graduate!) needed to get the same S/N
• Peak CCD QE 40 80

CCD used in ST-8XE (Kodak KAF-1602E)
29
B. Quantum Efficiency

• QE of CCD varies with wavelength, mostly due to
  the different penetrating power through Si of
  different energy photons.
• The long wavelength cutoff llong,cut is caused by
  the energy bandgap in semiconductors.
• The short wavelength cutoff lshort,cut is caused
  by the weak penetration of photons, leading to
  many photons absorbed before reaching the
  depletion zone.

30
Photon Penetration in Si

• Consider incoming photon flux F(0) on the surface
  of a CCD chip. The flux at depth z is given by
• where a is called the coefficient of intrinsic
  absorption, which is a function of temperature T
  and wavelength l.

31
Photon Penetration in Si

• The distance 1/a, is called a scale height (or
  optical depth).
• Photons are stopped by about 4 scale heights
• i.e. f(4/a) 0.02f0
• Therefore blue photons are totally absorbed by
  1mm
• To increase efficiency in blue, decrease
  thickness of Si
• For infrared photons of 1mm, one scale height gt
  200 mm.
• Sensitivity in red requires Si thick enough to
  have enough opportunity to absorb the photons.
• However, Si that is too thick can cause loss of
  resolution (i.e. photoelectrons generated may
  travel to depletion zones of other pixels). Also,
  more Si means higher dark current.

32
To improve QE

• (1) Backlit CCD Photons incident from the back,
  need to be built very thin (15mm versus 300mm
  for regular front-lit CCD)
• Difficult to make (fragile and need even
  thickness of Si to order of 1mm or get large QE
  variation over the chip) and therefore expensive

Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/digitalimaging
/concepts/concepts.html
33
To improve QE

• (2) Florescent coating applications of
  substances to CCD that act as a wavelength
  converter, i.e. release a longer wavelength
  photon upon an incident photon, to improve QE in
  blue and even UV.
• Common UV coating include molecules called
  Polycyclic Aromatic Hydrocarbons (PAHs)
• Problems PAHs are usually carcinogenic

Courtesy Dr Steven R. Majewski
34
Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/digitalimaging
/concepts/concepts.html
35
C. Charge Transfer Efficiency (CTE)

• Some electrons are lost during transfer from one
  pixel to another
• CTE No. of charges transferred/ No. of original
  charges
• E.g. A CTE of 99.9 Is it good enough?
• Consider a CCD with a 3 phase charge transfer,
  recording data from a 1024x1024 CCD require
  3x1024 transfers
• Ratio of charges left (0.999)3072 0.05!
• Need CTE of at least 99.999

Courtesy Dr Steven R. Majewski
36
D. CCD Output

• At the end of the multiplexer, an
  Analog-to-Digital Converter converts the
  electrons to a digital signal.
• The Gain (G) is the number of electrons combined
  to generate one signal count, called a
  Analog-to-Digital Unit, or ADU. So ADU Ne/G,
  where Ne number of electrons arriving at
  amplifier
• Normally, G is set as a positive number larger
  than 1. e.g. for KAF-1602E, G 2.3 e-/ADU.
• The dynamic range of CCD is limited by the number
  of digital bits of the output, e.g. KAF-1602E
  produces a maximum 16-bit output (allowing 216
  65536 values).
• Question What is the depth of electron well of
  KAF-1602E?
• Answer Electron well 65535 x 2.3 151,000 e-
• Error associated with the readout process is
  called read noise (details associated with the
  detector electronics).

37
E. Binning

• An effective method to reduce readout noise is
  pixel binning.
• Combine signals from adjacent pixels before
  arriving at the readout amplifier, improve
  sensitivity at low signal level.
• Reduce angular resolution of final image, i.e.
  increase plate scale
• Reduce total readout time for the whole CCD chip,
  e.g. pixel digitization rate (pixel readout rate)
  for KAF-1602E 30,000 Hz
• Common binning modes 1x1 (no binning) 2x2, 3x3
  (for imaging) 2x1, 3x1 (for spectroscopy)

Courtesy Molecular ExpressionsTM
http//micro.magnet.fsu.edu/primer/ digitalimagin
g/concepts/concepts.html
2x2 binning
38
VIII. Noise Considerations for CCD

• All astronomical measurements come with noise, or
  uncertainties.
• Important quantity Signal-to-Noise ratio (S/N)
• (A) Photon Noise Also known as Poisson noise.
  This is a law of nature that for any naturally
  occurring random events N, the standard deviation
  is equal to the square root of the number of
  events observed, i.e. N1/2
• Therefore, if one repeatedly counts the number of
  photoelectrons Ne collected integrated over time
  t, the standard deviation of these counted
  numbers will be Ne1/2
• If only photon noise exists, then S/N Ne /Ne
  1/2 Ne1/2
• This represents the biggest S/N we can ever
  achieve!

39
VIII. Noise Considerations for CCD

• (B) Read Noise also known as readout noise
• For a given CCD circuitry, the read noise is a
  constant, independent of the signal received,
  expressed in terms of e- root-mean-square (e-
  RMS)
• E.g. KAF-1602E has a read noise of 15 e- RMS
• All sources of noise can be combined in
  quadrature, i.e., total noise s2Total
  s2Poisson s2RN
• Effect of binning Consider 4 pixels and we are
  interested with the combined signal from the
  addition of these pixels
• (a) With no binning and readout of 2x2 pixels
  there are four readouts s2Total Ss2Poisson
  4s2RN
• (b) With 2x2 binning of those same pixels, there
  is only one readout s2Total Ss2Poissons2RN
• S remains the same, but N ?. Therefore S/N ?

40
VIII. Noise Considerations for CCD

• (C) Dark Noise also known as dark current
• Due to thermal emissions of electrons which
  cannot be distinguished from the photoelectrons.
• Expressed in terms of (DN) e-/pixel/second
• Can be reduced by operating CCD at lower
  temperature
• For KAF-1602E, dark current at 0oC 1
  e-/pixel/second (quite high)
• Therefore, for a signal taken over a time t, dark
  current DNt, therefore dark noise sDN
  (DNt)1/2
• Total signal Ntot Ne DNt
• Total noise s2Total s2Poisson s2RN s2DN
• If photon noise dominates the remaining noise,
  then we say that the data is sky limited

41
IX. Reduction of CCD Data

• Goal Remove systematic effects in data
  introduced by the detection process itself
• We want to correct for systematic error present
  in the data without introducing additional random
  error.
• We use separate calibration frames to isolate
  each individual aspect of the systematic effect

Courtesy John P Oliver, http//www.astro.ufl.edu/
oliver/ast3722/ast3722.htm
Flatfield frame to remove systematic fluctuations
Dark frame to remove dark current
Bias frame to remove bias counts
42
1. Bias frame

• A zero second integration of the CCD
• ADC amplifier produces non-zero readings even in
  the absence of photoelectrons. The bias frame
  measures this zero level.
• Subtracted from all observed frames
• Used mainly when dark current is low (e.g. cooled
  CCD)
• No needed to be subtracted if dark frame is taken
  and subtracted though.

43
2. Dark frame

• Exposure of the CCD with shutter closed.
• The dark frame measures the number of thermal
  electrons accumulated in each pixel.
• Bias will be automatically included!
• Subtracted from all observed frames.
• Need to scale this frame with the exposure time
  of the data.

44
3. Flatfield frame

• Exposure of the entire optical system to a source
  which evenly illuminates each pixel of CCD
• The flatfield frame measures the systematic
  fluctuations in the CCD data due to (1)
  pixel-to-pixel variation in the CCD sensitivity
  (2) optical defects (3) shadows of dust speckles
  on optical components.
• Divided from frames after bias or dark frame
  subtraction
• Types of flatfield (a) Domeflat a uniform
  surface mounted inside the dome (b) Skyflat
  images of a blank piece of sky taken around
  dusk or dawn. Better than domeflats because it is
  closer to the observed data. But takes much
  longer time to prepare
• Difficult to prepare for a good one

45
Before flatfielding
Flatfields
After flatfielding
Courtesy Dr Steven R. Majewski
Data MOSAIC 4x2 CCD array camera at Kitt Peak
National Observatory
46
Image Combination

• Want calibration frames with small random errors
• Combine (stack) images to reduce random errors.
• Measure same signal N times and average the
  results, then noise of combined result will be
  reduced by N1/2 (why?)
• For example, taking 25 bias/dark frames can
  reduce the random error by factor of 5
• Method of combining median-stacking
• For each pixel (i,j), rank values of frame1(i,j),
  frame2(i,j), ..., frameN(i,j) in order, take the
  median value to be the value of the value of the
  combined frame (i,j) pixel
• Least affected by huge spike in signal due to
  cosmic rays

47
Bias frame 1
Bias frame 2
Bias frame 3
Cosmic ray hits
Median-combined Bias frame
Courtesy Dr Steven R. Majewski
48
Courtesy Dr Steven R. Majewski
49
Comparisons of Detectors (Summary)
50
Comparisons of Detectors (Summary)