1. CCDs in Astronomy History of CCDs How do CCDs work ?
Charge Coupled Device
History of CCDs
How do CCDs work ?
Advantages of CCDs
Calibrations
Observations with a CCD

2. Photographic Imaging
Photographic Plates: 100mm thick emulsion spread over a glass base
Solidified gelatin w/silver halide salt grains
“Pixel size” (grain size): 6 mm
How it works: multiple photons strike silver halide grains – give off electrons and form silver atoms within grains
Developing converts exposed grains into opaque silver

3. Photomultiplier Tubes
“Single pixel” detector made of alkali metals
Photon strikes detector and knocks off one electron
That electron creates a cascade effect that ends in many electrons at the end of the detector
The incoming photon must have enough energy to knock loose the first electron or it is not “seen”

4. CCD History CCD image of Uranus from 1975 (JPL & UofA) 8900 Å
First developed in the 1960s as memory storage devices
Sensitivity to light suggested imaging possibilities
In the 1970s, CCDs were used primarily as experimental devices
In the 1980s, their use became more widespread
By the 1990s they’d essentially completely taken over almost all imaging applications
video and still cameras, scanners etc.
Astronomy is a highly demanding application
low light
noise
cosmetics
CCD image of Uranus from 1975 (JPL & UofA)
8900 Å
61”, Mt. Bigelow
Methane at the south pole

5. How Does a CCD Work? Rain = Photons
Water = Charge (photon strikes silicon semiconductor surface and knocks an electron loose by the photoelectric effect)
Buckets = pixels (electrons accumulate in “potential wells;” depth represents how much charge each pixel can hold)
The charge in each line of pixels is shifted to the readout register
The charge in each pixel is counted

6. How CCDs Work
Electrodes
To Output amplification
Electrons
(a)
Output
register
Pixel
(b)
9 pixel CCD, an output register, and an amplifier
Pixels divided into 3 regions to create potential wells
during an exposure, the higher potential of the central region (yellow) collects electrons
during readout, the potentials are changed to transfer charges to the next region.
1 The CCD detector
Not only it becomes possible to know the overall quantity of rain which fell on the farm but also to know it‘s spatial distribution. A CCD detector works with the same principles, where rain drops are replaced by photons, the buckets by pixels, etc. (see above and next figure).
1.3 How does a CCD work ?
In order to produce an image, a CCD must accomplish four functions:
1) generate photoelectrons (cf. rain drops),
2) collect electrons (cf. the buckets),
3) transfer the collected charges (cf. the conveyor belts),
4) read the charges (cf. weighting device).
The first function is based on the photoelectric effect. The light absorption in the silicate network of the CCD generates these photoelectrons, in proportion to the number of incident photons. The latter are immedialtly collected at some specific locations, the closest to where the photons fell on the chip. Those sites (cf. the buckets) are called pixels (cf. “picture elements”). Those pixels are defined by means of an electrode network which covers the CCDs surface. The electrodes form some potential wells, to prevent the collected charges of escaping. When collecting the charges is finished, their transfer (cf. displacement of buckets on the conveyor belt) is realized by changing in a synchronized manner the potentiels at the limit of each electrode in such a way that electrons can moove horizontally from one pixel to the other. At the end of each horizontal line of pixels sits a couting device (output register).

7. Electrons transfered from pixel to pixel
(b)
(a)
Impurity (doping)
Electrons transfered from pixel to pixel
Charges guided to the output register
Horizontal charge transfer is paused while charge packages at the output register are transfered vertically to an output amplifier and then read one by one
The cycle repeats until all the charges are read
Readout times can be as long as a few minutes
1 The CCD detector :
Which is a serie of electrodes which lies outside the photosensible zone of the CCD and disposed in a perpendicular way to it. This output register sends one by one the charge packages to an output amplifier where, at the end of this chain, charges are digitalized and stocked into a computer harddisk. The registered signal can be afterwards calibrated, analysed ..ect. We can restitute in the form of a numerical image the brillance distribution of the observed astronomical object.
1.4 Advantages of CCDs
In order to understand why CCDs are so useful and powerful, it is worth enumerating the essential characteristic of this detector. These characteristics are:
1) A good spatial resolution which allows astronomers to see the details at the surface of the studied object.
2) A very good quantum efficiency, which enables the detection of very faint objects. We require that the fraction of detected photons to be very high.
3) A efficient reponse in a very large spectral window. The detector should be sensible to the radiation in a large wavelenght domain.
4) A very low noise. The noise should remain very small compared to the weak signal emitted by faint objects.
5) A large domain for the strenght of detected signals (dynamical domain). The ratio of the fluxes of the faintest detected objects to the fluxes of the brightest one, should be as large as possible.

8. What moves the charge? Charge transfer in a CCD
Changing the voltages φ1 and φ2 moves the location of the potential well to the right, and the electrons follow along

9. CCD Advantages
Wide dynamic range (can measure both very faint targets and very bright ones)
Low noise
High QE
Accuracy (both linearity and stability)
Sensitivity over a wide spectral region (to 1 mm)
Dimensionally stable
Regular grid of pixels
Digital
CCD mosaic from Kitt Peak
Four 2048x2048 pixels

10. Quantum Efficiency
QE is a measure of how efficient a device is in turning input energy (in this case light) into a measurable signal.
Best film
Amateur CCD
Professional CCD
Greater efficiency means that more data can be gathered in a shorter time, or that in the same time you can measure a fainter signal.

11. Comparing Detector QEs
1 The CCD detector
However, CCDs possess advantages which clearly distinguish them from photographic plates, and from all other detectors in general.
2) The above figure compares the quantum efficiency of a CCD with the one of other types of detectors. The eye- the first astronomical detector- possesses at visible wavelenght a quantum efficienty of around one percent. In other words, we can only detect one single photon among one hundred who hit our eye. In contrast, more than 50 ( 80 at certain wavlenghts) of photons falling on the surface of a CCD are detected.
3) Additionnaly, the domain of spectral response of our eye is much more narrower compared to the one of the CCD. This limiation affects also other detectors as photocatods, photographic plates, ...
It is noticable that when a CCD is illuminated from above, it is rather insensitive to ultraviolet light and X rays, the electrodes surrounding each individal pixel beeing opaque to these types of photons. We can improve by about 20 the quantum efficiency of CCDs in the ultraviolet by covering the upper surface by a thin layer of phosphore which main task is to convert ultrviolet photons to photons who have a higher wavelenght. A much more performant method consists in making the CCD very thin and illuminating it from the bottom. The incident photons can then enter in contact with the photosensible area of the CCD, without beeing absorbed by electrodes. However, at this face of the CCD (bottom) appears a potential well who tends to trap the photoelectrons at the bottom surface.
Ingeneers have found two solutions to deal with this problem. The first method ( called “back side charging”) consists in flashing the CCD by means of an ultraviolet light before using it. In this way, an excess of photoelectrons is produced which destroys the potential well at the bottom surface of the CCD.

12. Linearity
CCDs have a linear response to light, i.e. the measured signal is directly proportional to the amount of light which was received. This is not true for film.
A linear response means that if the exposure is doubled, then the measurable signal will double. Also, twice the signal means the source is twice as bright.
CCD linear response
Film non-linear response

13. Designing a CCD Most commercial CCDs are “front-side illuminated”
A 3-d circuit on a base of silicon (the light sensitive layer)
Light has to go through the circuitry, which causes losses
Astronomical CCDs are “back-side illuminated” with QEs of 90% or greater
silicon is be thinned to a few tens of microns
need to support the silicon.
some charge diffusion in the silicon
Anti-reflection coating applied to the CCD surface reduces loses
Front-side illuminated CCDs have low blue QE
devices can be coated with “Lumigen” – an organic substance similar to the “glow” in highlighter markers (Lumigen converts blue/UV photons to 520 nm, where the CCD has higher QE)

14. Readout Noise - A CCD has an analog output
How accurately can we measure the number of photons detected by a CCD pixel?
A CCD has an analog output
Photons are converted to a charge and then to a voltage for measurement.
An amplifier on the CCD boosts the signal to a useful level
Is it possible to measure exactly how many photons fell on each pixel?
No. The noise inherent in the conversion and amplification process introduces some noise
The lowest noise CCDs now used in astronomy have a readout noise (RMS error ~2e–)

15. Bias low-level structure in the bias
The CCD amplifier also introduces a “bias level” to the output voltage
typically a few hundred electrons
The bias level is measured from the “overscan” region and subtracted off
“Bias structure” may also be present in a 2D image
The electronics as well as the physical make-up of a CCD can also imprint a faint background structure on the images.

16. Charge Transfer Efficiency (CTE)
How efficiently can charge be moved across the pixels and the readout register? Will every electron be moved or will some be lost?
The earliest CCDs had a CTE of only ~98%
Today CTE is typically better than 99·995% in commercial devices (“4 nines”)
Much higher in scientific devices - 99·9999% (5-6 “nines”).
Poor CTE means that not all of the photons which arrived on the CCD will be counted, and the further from the readout register the worse the effect.

17. Bleeding from bright stars
Saturation
What if a pixel’s potential well “fills up” with electrons? The physical size of the pixel determines how much charge it can hold. Larger pixels can hold more charge.
Pixels are saturated when their potential wells are full. Electrons “bleed” along columns into nearby pixels
As a CCD pixel nears saturation, the response of the pixel becomes non-linear
The data number read out from a saturated pixel cannot exceed the largest number allowed by the “analog to digital converter” that converts the voltage to a digital signal (typically 16 bit, or 65,536)

18. Thermal Noise or Dark Current
The finite temperature of the CCD leads to the production of thermally induced electrons in the silicon
Dark current increases linearly with time
A function of the temperature of the CCD
CCDs cooled to around 170°K (-100°C) to reduce thermal noise
Dark current can be removed with careful calibration

19. A few more problems with CCDs:
The amplifier in some CCDs glows!
Defects in the silicon wafer can cause cosmetic problems
CCDs are sensitive to cosmic rays
Glowing amplifier
Cosmic rays
Bad columns

20. CCD Characterization
CCDs are characterized before they are put on a telescope.
The parameters which are needed are:
The amplifier gain – how many electrons per count.
The linearity of the amplifier and electronics – there will always be some slight variation from perfection.
QE and CTE – how good is the CCD.
Any cosmetic or electronic blemishes (“trapping sites”, etc.) – every CCD is unique!

21. Gain and Readout Noise Noise (ADU) as a function of the signal (ADU)
1 The CCD detector
1.6 Observations with a CCD
1.6.6 Determination of the gain and the read out noise of a CCD
A typical diagram representing the variance of the pixels as a function of the signal is displayed in the above figure. The slope of the lign which goes throught the majority of the points is 1 / g, and for SADU = 0, we find that BADU = BDL. For large value of SADU, we see that the variance does not grow anymore like SADU.The CCD stops beeing linear.

22. Observing with a CCD 1, 10, 100 and 1000 sec exposures of M100
S/N ratio improves with exposure time
Readout noise dominates in the shortest exposure
Photon noise in the sky dominates for the longest exposure
1 The CCD detector
8). Distorsions du to magnetic fields or other physical distorsions that affect other detectors (Vidicon detectors in televisions). These problems are unexitant for CCDs. The position of each pixel is fixed in a rigid way when the CCD chip is produced.
1.5 The CCD as a 3 dimensional detector
Photons in the ultraviolet domain (  Å), far ultraviolet (  Å) and soft X rays (  1,2-120 Å) are much more energetic than visible photons. When they get absorbed in the silicate network of the CCD, these high energy photons generate much more electrons; the exact number depends on the wavelenght of the photon. For example, an X ray photon which has a wavlenght of 2,1 Å is capable to generate, in average, 1620 electrons. For wavelenghts typically shorter than 100 Å, a single photon is detected each time and its energy is determined by measuring directly the quantity of generated charges.

23. CCD Calibrations Basic calibrations include BIAS, DARK and FLAT FIELD
NGC 2736, part of the Vela SN remnant.

24. CCD Calibrations - Bias
The image is scaled with only a few ADU from black to white
Little structure is evident
Statistical variation is only 0·4 ADU so this is a clean bias frame
CCD Calibrations - Bias
A BIAS frame is a zero-length exposure to show any underlying structure in the image from the CCD or electronics
The bias consists of two components
a non-varying electronic zero-point level
plus any structure present
CCD systems usually produce an overscan region to allow the zero-point for each exposure to be measured
The bias structure is a constant and may simply be subtracted from each image
Because of readout noise, average several (say, 10–20) bias frames to create a master bias frame

25. CCD Calibrations - Dark
To remove dark current, take a series of DARK frames
A dark frame is the same length as a normal exposure but with the shutter closed so no light falls on the CCD
Since CCDs also detect cosmic rays, take several darks and combine them with a median filter to remove cosmic rays from the combined dark frame. Combining several dark frames also minimizes statistical variations.
Subtract the combined dark frame from a normal image, provided they are of the same duration. (After the bias has been removed, of course.)
All images, including darks, contain the bias. A shortcut often used is to not separate out the bias but subtract the dark+bias.
Most research CCDs have very low dark current, so dark frames may not be necessary.

26. CCD Calibrations – flat field
Center to edge variations and donuts are both are about 1%
CCD Calibrations – flat field
Pixel-to-pixel variations are removed with a “flat field” image
A flat field is an image of a featureless, uniform source (such as the twilight sky or a dome projector screen)
A flat field shows the minor pixel variations, as well as all the defects in the optical train (e.g. vignetting and dust spots)
After bias and dark subtraction, divide the image by the “normalized” (image mean reduced to 1.0) flat field
Dividing by the flat field image corrects for variations in sensitivity on the detector and throughput of the telescope and instrument

27. Flat Field Calibration - Divide (a) by (b) to get (c)
1 The CCD detector
CCDs are linear to an accuracy of 0,1, within the range of all their dynamical range. It is therefore relatively simple to correct for the differences in sensitivities of the different pixels of a CCD. In fact, due to this linearity, this relative sensitivity is the same among all pixels, independant of the value of the incident flux.
To correct for these variations in the sensitivity among different pixels of a CCD, the “flat field” technique is used. The three above figures enable the understanding of this method. Figure (a) corresponds to a raw (simulated) CCD image of a galaxy, as we can observe it with a telecope. In this example, the CCD is made of 9 pixels. This raw image looks everything but smooth, essentialy due to the differences in the relative sensitivities between the 9 pixels. Such an image is useless for astronomical usage if not corrected. It should be corrected for this effect of pixels sensitivities irregularity. In order to do this, we observe during the same night a uniform source of light such as a uniform light projected on the dome of the telecope. The image obtained then enable us to see directly the differences in sensitivities between the 9 pixels (cf. Figure (b)). We than need to divide the signal in (cf. Figure (a)) by the “flat field” signal (cf. image CCD (b)). The final result is shown in Figure c. It is identical to an image of the galaxy obtained by a *perfect* CCD for which the sensitivity of all pixels would be the same.

28. Cosmic Rays CCDs are good cosmic ray detectors
Cosmic rays are always found on long exposures
To correct for cosmic rays, take at least three object exposures, and combine them with a median filter
1 The CCD detector
1.6 Observations with a CCD
1.6.4 Cosmic rays
Long exposures increase the probability of seeing another noise intervening in the read out noise : The impact point of cosmic rays which are random and which can cause the saturation of one or many pixels. Cosmic rays, which are highly energetic particles, create in the earth atmosphere secondary muons which are capable of depositing in CCDs charges which can exceed 1000 electrons in a restrained number of pixels. These events occur on the ground as well as in space, particularly at the time of solar eruptions.

29. Other artifacts
poor guiding
donuts

30. Data Histograms
A histogram is a plot of the number of times a particular data number occurs vs. data number

31. Sources
Kodak KAF-1302E(LE) CCD - Image courtesy of Eastman Kodak Company. KODAK is a trademark.
Other images © Steven Lee
Bob O'Connell's Fall 2003 Lecture Notes
U. Florida notes on electronic camerasThe_Electronic_Camera_in_Astronomy.ppt
Notes from the Max Planck Institute: