Help -
Help for Webmasters
« back to results for ""
Below is a cache of http://sales.hamamatsu.com/assets/applications/SD/imagem-tech.pdf. It's a snapshot of the page taken as our search engine crawled the Web.
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content. EM-CCD technical note EM-CCD technical note 1. CCD structures and characteristics 1.1. Interline transfer CCD (IL-CCD) 1.2. Full Frame (FFT-CCD) and Frame Transfer CCD (FT-CCD) 1.3. Back-thinned CCD 2. Noise components of CCD 2.1. Dark current 2.2. Readout noise 3. Electron Multiplier CCD (EM-CCD) 3.1. Principle of Electron Multiplication gain in EM-CCD 3.2. EM gain dependence on temperature 3.3. Gain aging characteristic 4. CCD/EM-CCD noise calculation 4.1. CCD noise components calculation 4.2. EM-CCD noise components calculation 4.3. EM-CCD noise dependence on EM-gain and Input photons 4.3.1 EM gain vs. S/N 4.3.2 Input Photons vs. S/N 4.3.3 S/N Crossover point between normal CCD readout and EM-CCD readout 5. EM-CCD ImagEM Camera Technical note 5.1. Outline of ImagEM Features 5.1.1. Cooling capability improvements 5.1.2. Multi-modes readout 5.1.3. Image reversal function 5.1.4. Multiple clock speed selection 5.1.5. Photon Imaging mode (Patent pending) 5.1.6. Real-time Image Processing features 5.1.7. External trigger / Synchronous readout trigger (Patent Pending) 5.1.8. Programmable Sync output 5.1.9. Multiple camera heads control feature 5.2. Specification 2 P3 P4 P5 P7 P10 1. CCD structures and characteristics There are three types of CCDs which are well known in scientific im- aging. One is the interline transfer CCD (IL-CCD), a second is the Full Frame Transfer CCD (FFT-CCD) and a third is the Frame transfer CCD (FT-CCD). 1.1. Interline transfer CCD (IL-CCD) The structure of IL-CCD is shown in Figure 1(a). The interline CCD has vertically paired columns consisting of imaging Photodiodes (PD) and a readout register (Vertical charge transfer register (V-CCD). Electric charges generated in all the PD by incoming photons are shifted simultaneously to the adjacent V-CCD register. The V-CCD register is covered by a mask of aluminum or other opaque material to prevent photons from creating additional charges in this area dur- ing readout. Readout is accomplished by transferring each horizontal row of information in the V-CCD, line by line, up the CCD to the Hori- zontal serial register (H-CCD). There, charges are transferred hori- zontally and converted into charge voltage by AMPFDA. The design of an IL-CCD has the advantage that the signal accumu- lation (exposure) and readout can be done simultaneously because PD can accumulate charges for the next frame right after the previ- ously generated electric charges in the PD are shifted to the V-CCD. There is no possibility of image smearing in this device. Traditionally, the design of the IL-CCD has had the disadvantage that the open ratio of the light sensitive area (fill factor) is reduced be- cause of the presence of the masked V-CCD area. Recently this dis- advantage has been dramatically improved by on-chip lenses (as shown in Figure 2) and improvement of sensor structures that allow detection of photons deeper in the PD than previous models. Overall Quantum efficiency has increased to over 70%. New IL-CCDs like the Hamamatsu ER-150 CCD (Figure 3) used in the ORCA series of cameras, offer characteristics ideally suited to many scientific applica- tions. 1.2. Full Frame Transfer CCD (FFT-CCD) and Frame
Transfer CCD (FT-CCD) As shown in Figure 1, Frame transfer CCDs are divided into two types, Full frame transfer (FFT-CCD)(b) and Frame transfer (FT-CCD)(c). In the case of a FFT-CCD Figure 1(b), charges generated in pixels are transferred vertically, row by row, to the horizontal serial register for readout. Unless a shutter is used during this transfer and readout, im- age smearing will occur. While this design offers 100% open area ratio with full collection of incoming photons, the shutter limits the frame rate and photons falling on the shutter are lost when it is closed for readout. It is not possible to acquire signal and readout at the same time. FT-CCDs Figure 1(c) offer both 100% fill factor and simultaneous signal acquisition and readout. Like the FFT-CCD, the chip has no charge transfer regions in the signal acquisition area. Rather, the FT-CCD has two separate but equal regions, one with pixels exposed to the incom- ing photons and another region with an equal number of pixels but en- tirely masked to eliminate photons from being detected. As shown in Figure 1(c) one area works as the detection area and the other works as the storage area. Accumulated charges detected in the detection area are rapidly transferred to the masked storage area, and the accu- mulated charges are transferred vertically, line by line, to the horizontal serial register for readout. The detection area and storage area are driv- en individually so the next exposure can start right after the completion of the rapid vertical charge transfer from the detection area to the stor- age area. This design eliminates the need for a mechanical shutter, al- lowing signal accumulation (exposure) and readout simultaneously like an IL-CCD. 1.3. Back-thinned CCD As mentioned before, in FFT-CCDs and FT-CCDs the light sensitive pixels have a charge transfer function as well. This function requires the front surface of light sensitive pixels to be covered by a semi- transparent Poly-Si electrode for the charge transfer to function, as shown in Figure 4. Even with 100% fill factor, the effective quantum efficiency (QE) drops into the 40% range because the Poly-Si electrode absorbs some per- centage of incoming photons depending on their wavelength. To over- come this disadvantage, Back-Thinned CCDs (BT-CCD) are becom- ing popular. In a BT-CCD the CCD is turned upside down and this back side of the CCD is thinned to 10um to 15um in thickness as shown in Figure 4(b). Incident photons now enter the CCD from this back thinned side, without the Poly-Si electrode in the light path. QE values of greater than 90% can be achieved. Figure 5 compares typi- cal QE curves of the same CCD in front- illuminated and back-illumin- ated versions. 3 FDA H-CCD V-CCD PD Pixel (a) (b) (c) Interline CCD Full Frame Transfer CCD Frame Transfer CCD Fig. 1 Old-model Interline CCD Photons Photons ER-150 Interline CCD (a) (b) AL shield sensor sensor Gate Electrode Gate Electrode V-CCD V-CCD Onchip lens Fig. 2 Fig. 3 2. Noise components of CCD As seen chapter 1, innovations in silicon based CCD technology have created many kinds of CCDs. With the Quantum efficiency of the IL- CCD reaching about 70% and the BT design achieving more than 90%, detection limits are nearing their theoretical limits. Signal detec- tion in modern CCDs is often limited by how much camera noise (due to dark current and readout noise) must be overcome before the sig- nal is apparent on the CCD. These values determine the camera per- formance of CCD, especially in low light applications. 4 Fig. 4 Quantum efficiency (%) 0
200 400 600 800 1000 1200 20 10 40 30 60 50 80 70 90 100 Wavelength (nm) Back-thinned illuminated type Front-illuminated type Fig. 5 2.1. Dark current A CCD is made from Silicon, and the dark current caused by thermal mi- gration of electrons in silicon is a main noise factor for a CCD sensor. The Dark current of a CCD depends on the temperature, and it decreases by half when the temperature drops by approximately 7 to 8 degree C. It is apparent that cooling a CCD is a very good way to reduce the dark cur- rent noise. Figure 6 shows the CCD dark current vs. Temperature. Dark current also depends on the type of CCD. In most applications an IL-CCD can normally achieve good performance with -30 to -50 degree C cooling. In the case of an FFT-CCD or FT-CCD, most require -50 to - 90 degree C cooling for low light applications. Cooling is most effective when the CCD is placed in a vacuum chamber. Temperature (;) Dar k current (electron/pix els) Dark Current vs Temperature (S5466) 10 5 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 - 50 - 40 - 30 - 20 - 10 0 10 20 30 Fig. 6 Hamamatsu Photonics K.K. have developed an unique hermetic va- cuum-sealed chamber with high performance cooling capability based on many years of experience with high vacuum technologies used in Photo Multiplier Tube (PMT) and Image Intensifier (I.I.) tech- nologies (shown in Figure 7). The CCD chip and a multi-stage peltier element (ThermoElectric cooler) are built into a welded metal cham- ber with a special window on the front to create the camera head. The chamber of the vacuum head is vacuumed and hermetically sealed to retain a high degree of vacuum (10 -8 torr or less), to ensure great cooling performance over many years. In comparison with simpler va- cuum sealed models using ordinary gaskets or o-rings, there is no need for periodic re-evacuation or maintenance. With several hundred such vacuum heads in daily use over many years, the unique Hama- matsu hermetic vacuum-sealed head reliably provides significantly better and more stable cooling performance than other designs. VIS UV Gate Electrodes (a) Front-Illuminated FT-CCD Absorption of photons in Gate
Electrodes reduces Quantum
Efficiency in Front- illuminated type. Back-thinned FT-CCD No Absorption of photons in Gate
Electrodes increases Quantum
Efficiency in Back-thinned type. (b) 2.2. Readout noise The largest factor influencing the detection limit of a CCD is the readout noise caused by the on-chip Floating Diffusion Amplifier (FDA) that con- verts accumulated charges into voltage. Accumulated charges transferred into Horizontal serial register are serially transferred into the FDA pixel by pixel. Readout noise is primarily caused by the resetting of the amplifier after the accumulated charge in each pixel is converted to a voltage and the amplifier is reset for the next incoming pixel. This reset noise can be dramatically reduced by an external correlated dual sampling (CDS) cir- cuit. Additionally readout noise depends on pixel clocking frequency and is generally lower with slower CCD clocking speeds. However, slower pix- el clock speeds may limit the camera use for dynamic real-time imaging. In summary, camera readout noise performance depends greatly on the external circuit design of the camera manufacturer and the readout speed. As an example, there are only 3 electrons r.m.s. readout noise in cam- eras such as the Hamamatsu ORCA2-ER cooled CCD. The detection limit of such cooled CCD cameras is about 10 electrons, making it an ul- tra-sensitive camera. More details of noise are explained in chapter 4 : CCD/EM-CCD noise calculations. 3. Electron Multiplier CCD (EM-CCD) As mentioned above, CCD technological innovations are making dramat- ic progress. As a result of various approaches, very high sensitivity and low noise CCDs are readily available. Despite all the advances, readout noise is still the dominant factor limit- ing weak signal detection. Detecting signals below the readout noise lev- el of a camera is possible with various special methods or technologies. Detecting a lower signal than the readout noise is possible by signal inte- gration on CCD chip. Over time the signal will accumulate and become greater than the readout noise. Other techniques that involve signal multiplication are done with Micro- channel Plates (MCP) in an Image Intensifier (I.I.) or direct electron bom- bardment of a CCD (EB-CCD). In these cases signal electrons are cre- ated at a photocathode and then multiplied by a high voltage in a vacuum tube before signals are readout A mechanism for direct multiplication of electrons on the CCD itself has been known for many years. A host of technological problems associated with this on chip multiplication process prevented the technique from be- ing useful until just recently. In the last few years solutions to the prob- lems have been developed and it is now becoming an effective means of ultra low light detection in biological and scientific imaging. This exciting technology has become known as an Electron Multiplier CCD (EM- CCD). 5 CCD Light input window Peltier cooler Fig. 7 3.1. Principle of Electron Multiplication gain in EM-
CCDs Figure 8(a) shows the structure of an EM-CCD. The basic structure is the same as a normal FT-CCD and it is shown as a back-thinned ver- sion. Accumulated charges detected in the detection area are rapidly transferred to the storage area, and then the accumulated charge is transferred line by line to the horizontal serial register for readout, just as in a normal FT-CCD. At this point in an EM-CCD a multiplication (Charge Multiplier) register is built into the horizontal serial register. With this charge Multiplication register, signal multiplication is done by supplying a higher voltage than normal to each horizontal transfer electrode. Figure 8(b) shows the principle of signal multiplication in the charge Mul- tiplication register. When a signal electron charge is transferred from stage to stage, the signal charge is accelerated by high electric field generated under the multiplication gate by applying a high voltage (30 to 40V) to each multiplication electrode (multiplication gate). This high vol- tage is much greater than the normal horizontal transfer electrode vol- tage, and it generates an occasional extra electron-hole pair. This is called an impact ionization event. The probability of such an event is very small, typically about 1.0% to 1.6% at each stage. This is the value (g) in the formula shown below. Electrons are multiplied from stage to stage repeatedly in the gain register and high multiplication gain is ach- ieved. Normally, there are 400 - 600 stages (N) in the gain register. Total multiplier gain (M) can be expressed by the following formula : M =(1+g) N g : probability to generate an electron-hole pair at each stage N : total number of charge Multiplication stages 3.2. EM gain dependence on temperature In the case of EM-CCD, as mentioned above, the Multiplication gain factor in the Multiplication register greatly depends on the tempera- ture. It is obvious that the stabilization of temperature of a sensor be- comes a very important issue. Figure 10 shows an example of an E2V CCD97 CCD and the EM gain vs. temperature. A change of 70 degrees at the CCD changes the EM gain by about 10 times in Figure 10. In addition, as the temperature is decreased, the slope of the change increases. Temperature stability becomes in- creasingly important at lower cooling temperatures to maintain con- stant gain in an EM-CCD. Here, probability of (g) depends not only on the supply voltage of the charge multiplier register but the temperature of a sensor also has a great influence. Control and stability of both the supply voltage and the CCD temperature are very important factors when EM-CCD tech- nology is used for quantitative measurement. The key technical point and advantage of an EM-CCD is signal char- ges in the CCD are multiplied in the multiplication register before it is converted to voltage by the FDA. As mentioned before, the readout noise caused by the FDA is the limiting factor to low signal detection. Multiplication of the signal before the FDA makes the readout noise of the FDA relatively smaller as the multiplication of the original signal increases. Even at moderate gain settings, the relative readout noise becomes less than 1 electron, enabling detection of even single pho- ton events in the signal. As readout speed increases, the readout noise increases by a square function of the frequency. To observe 6 Fig. 8 x 2 x 10 x 50 x 500 Output images obtained when gain
was varied with light level kept constant Fig. 9 Fig. 10 1000 100 10 1 GAIN 39 40 41 42 43 44 45 46 R 2HV T = 20: T = -0 : T = -20: T = -40: T = -50: Decrease of 70
degrees in
temperature yields 10
times gain increase low intensity objects at high speeds, the EM-CCD can overcome the increase in readout noise by additional multiplication gain; again re- ducing the relative readout noise to less than 1 electron. This ability to use the Multiplication gain register to overcome the readout noise even at high readout speeds is the chief advantage of the EM-CCD cameras for fast, scientific, low light imaging. Figure 9 shows sequential images taken while increasing the Multipli- cation gain factor from no Multiplication gain to higher Multiplication gain. With higher Multiplication gain, the multiplied signal becomes larger than the readout noise and an image is visible. (a) (b) Horizontal serial Resister Horizontal serial Register Multiplier Register Supplying higher than normal
voltage to each multiplication
electrode results in an extra
electron hole-pair generated
in an Impact Ionization event. Generated electron Image area Storage area Image area Storage area Normal FT-CCD Output Output EM-CCD 7 4. CCD and EM-CCD noise calculations 4.1. CCD noise components calculation There are four kinds of important noise factors that determine the S/N of a CCD. (1) Readout noise r The reset noise in the electric charge - voltage conversion AMP (FDA) on the CCD is the main source. It is expressed as the disper- sion r. (2) Noise caused by dark current d= (DT) This noise is the fluctuation in the dark current generated in a CCD, and it depends on cooling temperature and exposure time. It is expressed as the square root of the product of D: Dark current (electrons/pixel/sec) and T: Exposure time (sec). (3) Photon shot noise s= (QP) This noise is the fluctuation in the number of incident photons (pho- ton shot noise). It is expressed as the square root of the product of P: Input photon and Q: Quantum efficiency. (4) Spurious Noise cic Spurious noise is a charge generated by signal charge transfer pro- cess in the CCD called Clock Induced Charge. (CIC) This is a fixed value when readout clock and clock duty cycle is fixed, but it is small enough and usually it is possible to ignore it from the following calculation. From our measurement result, it is 0.001 e-/pix- el/frame-readout. Total noise N is calculated by the following expression: Total noiseN = {(Readout noise) 2 +(Dark current noise) 2 +(Photon shot noise) 2 } = ( r 2 + d 2 + s 2 ) = ( r 2 + DT + QP) On the other hand, Signal S is expressed by P: incident Photon and Q: Quantum efficiency as follows. Signal S = QP Thus, signal to noise ratio (S/N) is expressed by the following expression. Signal / noise = QP / ( r 2 + DT + QP) Figure 11 shows the change in noise (N) based on the number of in- cident photons under the following conditions: Readout noise: 8 electrons Exposure time: 100 milliseconds Quantum efficiency: 90% Dark current: 0.01e/p/s 3.3. Gain aging characteristic EM-CCDs have been widely used for high sensitivity applications due to the advantages of the charge multiplication register. But it was not widely known until recently that the multiplication gain tends to suffer gain aging, a slow decrease in gain over time based on the total elec- tric charge passed through the multiplication register. This is changing and has now been officially announced by the CCD manufacturers. To use an EM-CCD for stable quantitative measurement over long periods, it is necessary to consider this EM gain aging. The multipli- cation principle of EM-CCD is achieved by impact ionization effects of a high voltage (30 to 40V) applied to each multiplication electrode (multiplication gate). While the exact cause of the gain degradation is not known, it is thought that the higher than normal voltages used in the process traps accelerated electrons in the bottom of the transfer electrode. These trapped electrons may change the electric field at this point and thus create the gain aging phenomenon. This gain aging occurs exponentially over time and is most prominent in the early use of an EM-CCD. In order to reduce this to a minimum in ac- tual use, every C9100 camera is factory aged for more than 100 hours and readjusted before being shipped. As a result, serious gain deterioration should not occur in C9100 series cameras. However, since this phenomenon depends on total electric charge through the multiplication register, there may be some applications where addi- tional gain degradation can occur. Reducing gain and light intensity when the camera is not being used can help prevent this. It is wise to check the gain with a standard sample occasionally if long term stan- dardization is required. Tips for operating EM-CCD cameras (1) Keep the gain adjusted to a level that offers just enough gain to overcome the readout noise. There is no increase in the Signal to Noise ratio once the readout noise becomes less than one and add- ing gain past this point only increases the rate of gain degradation. (2) Reduce illumination to the detector as much as possible. Even if the gain is not increased, an increase in the signal intensity will create a larger charge in the multiplication register and increase the rate of gain degradation. (3) Reducing the gain to the minimum setting and blocking illumina- tion to the CCD when the camera is not being used for measurement can help maintain stable EM-CCD gain over long periods. As shown in Figure 11, the noise characteristics can be divided into two domains. A change in the slope of the noise value occurs at 70 photons. Below this value the graph indicates that when the incident photon number is less than 70 photons the camera readout noise is the dominant factor in the noise calculation. When the incident photon number is greater than 70 photons the photon shot noise is the domi- nant noise factor. (In this example with the comparatively short ex- posure time, dark noise is too small to influence the result.) In practi- cal terms, 70 photons define the point at which the camera noise no longer has an influence on the S/N and the number of incident pho- tons defines the S/N. 8 Fig. 11 4.2. EM-CCD noise components calculation In the case of EM-CCD, multiplication of the signal in the multiplica- tion register has noise associated with that process and this will influ- ence the S/N. Noise caused by signal multiplication is called Excess noise (F) and is added to the signal. Excess noise: F is calculated by the multiplication factor M and a ratio of the dispersion of a multiplica- tion register input signal: in and a dispersion of the multiplication register output signal: out. F 2 = out 2 / M 2 in 2 It is important to note that not only are the detected signal electron charges multiplied but also any electron charge in the CCD from other sources such as dark current are multiplied with the same multi- plication and noise factors. Calculations for signal level and total noise in an EM-CCD must in- clude multiplication noise, and are shown below. Signal : S is expressed as the product of detected signal : QP and multiplication gain : M. Signal S = QPM Total noise: N is expressed as the Square root of the product of the sum of signal charge :S, dark current :D, time :T, multiplication factor : M and Excess Noise :F as follows. Total noise N = { r 2 + F 2 M 2 ( d 2 + cic 2 + s 2 )} In this expression, cic is small enough to ignore from S/N calcula- tion, total noise becomes following. Total Noise N = { r 2 +F 2 M 2 (DT + QP) } S/N is calculated as shown below. S/N = QPM/ { r 2 +F 2 M 2 (DT + QP) } = QP/ { r 2 /M 2 +F 2 (DT + QP) } 4.3. EM-CCD noise dependence on EM-gain vs. In-
put photons Excess Noise factor is estimated at 1.41 for calculations of multiplica- tion register noise. This value is included in the examples shown below. 4.3.1. EM gain vs. S/N Figure 12 shows how S/N is influenced by the number of the incident photons (10, 100, and 1,000 (photon / pixel / frame), and increasing multiplication gain. An Exposure time of 33 msec, Quantum efficiency 90% , and readout noise of 80 electrons are used for this example.. The graph shows S/N clearly improves with signal multiplication but there is a limit to the S/N improvement in each case. Even if higher EM gain is applied, there is a certain level at which no further im- provement can be expected. When the input photon level exceeds 1000 photons/pixel in 33 msec the EM gain feature offers almost no benefit. The S/N at that point is limited only by the number of input photons (QP) and the multiplication noise (F) This can be seen in expression (A). Supposing the dark current is small enough, it is possible to ignore r 2 / M 2 as multiplication gain increases. In this case S/N is simplified as in Expression (A) below. S/N = QP/ {F 2 (QP) } = ( QP/F 2 ) From this calculation S/N is constant since the incident photon num- ber P is constant. This example clearly shows there is no advantage to increasing gain more than required according to the numbers of incident photons. This is very important point to be considered because it effects the gain stability in the long term. S/N vs. EM gain EM Gain S/N 0.1 1 10 100 1000 10000 100 10 1 Input photon=10
Input photon=100
Input photon=1000 1 10 100 1000 1 10 100 1000 Noise (e - ) 10000 100000 1000000 Input photon number Input photon / Noise Fig. 12 9 4.3.2. Input Photons vs. S/N Figure 13 shows how S/N is affected by the number of incoming pho- tons at different gain settings. The conditions are otherwise the same as in Figure 12. In this graph, it is shown that the S/N does not im- prove at gain settings over 200 but does improve with the number of incoming photons. The point is that using more gain than necessary is not a benefit and will only increase the rate of gain degradation in the multiplication register. Different conditions will change the gain setting at which this happens but the lesson is the same. Input photon number S/N 0.01 0.1 1 10 100 1000 1 10 100 1000 10000 100000 1000000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000 S/N vs. Input photon Fig. 13 Figure 14 shows how readout noise becomes relatively smaller as Gain is increased. 0 Gain 0.001 0.1 10 100 1000 200 400 600 800 1000 1200 1 1400 Relativ e noise Noise / Gain 10000 0.01 Fig. 14 4.3.3. S/N Crossover point between normal CCD
readout and EM-CCD readout Figure 15 shows how S/N changes with the number of incident pho- tons in normal CCD readout and EM-CCD readout respectively. As shown in Figure 15, using 100 photons/pixel/frame as a reference (exposure time: 100msec), when the incident photons are 100 pho- tons/pixel/frame or more, normal CCD readout provides better S/N. When incident photons are less than 100 photons/pixel/frame, EM- CCD readout provides better S/N. Notably, even with less than 100 photons/pixel/frame, normal CCD readout offers higher S/N if the EM gain is less than 50 times. Another way to look at this is that in order to improve S/N with less than 100 photon/pixel/frame conditions, it is necessary to use at least 50 times EM gain. This value is confirmed in Figure 14 by observing that the point at which the noise becomes less than one photon - at a gain of 50 times. Since the ImagEM camera has both normal CCD and EM-CCD read- out modes it is recommended to switch readout modes to maximize the S/N based on the number of incident photons. EM-CCD out / Normal CCD out Input photon number S/N 0.01 0.1 1 10 100 1000 1 10 100 1000 10000 100000 1000000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000
LN-AMP Input photon number S/N 100 10 1 0.1 10 100 1000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000
LN-AMP Enlargement Fig. 15 5. EM-CCD ImagEM Camera Technical note 5.1. Outline of ImagEM Features Hamamatsu Photonics K.K. uses the very latest cooling and CCD camera design technologies for the ImagEM EM-CCD camera. The ImagEM camera is the result of many years of experiments and ex- perience with deep cooling, high vacuum and low noise CCD technol- ogy. It provides optimum signal and noise characteristics for a wide range of applications. With one camera, it is now possible to capture both low and high light level images, wide dynamic range images and single photon binary images, long integration and high frame rate im- ages and it offers a very high Quantum efficiency over a broad range of wavelengths. The features below confirm that virtually any applica- tion can be addressed with this new camera. ' -90 degree C cooling with hermetic sealed head A newly designed hermetic sealed head includes a specially devel- oped 4 stage peltier cooler for the CCD and the entire camera head has been designed to optimize heat radiation. With both air and water cooling capabilities built-in, temperature at the CCD can be main- tained as low as -90 degrees C with the water cooling operational and a water temperature of 10 degrees C or less. ' Gain stability/Temperature stability Temperature stability is the key to gain stability in an EM-CCD cam- era so the ImagEM provides temperature stability to within +/-0.05 de- grees at -80 degrees C with water cooling. ' Dual readout mode ImagEM has both an EM-CCD readout and a normal CCD readout. With recent innovations, the normal CCD readout offers very low noise readout and very low dark current for long integration expos- ures. This creates tremendous flexibility in applications. From ultra low luminescence samples to routine fluorescence microscopy, the con- ventional readout provides high S/N images. With the EM-CCD read- out, high frame rates at low intensities are possible for live cell imag- ing and spinning disk confocal applications. ' Multiple pixel clock selection ImagEM offers a selection of pixel clock speeds in both EM-CCD readout mode and normal CCD readout mode. According to the appli- cation, it is possible to select a clock speed that offers the best S/N. Faster clock speed offer faster frame rates and lower clock speed pro- vides lower noise. ' Photon imaging mode (Patent Pending) Due to the Excess noise factor in the multiplication register, EM-CCD cameras have traditionally had limited use in photon counting applica- tions. Intensified CCDs have dominated this field for many years. Us- ing over 20 years of experience in the design and production of image intensifiers and electronic circuit technology, Hamamatsu Photonics K.K. has incorporated a unique photon counting imaging mode into the ImagEM. 10 ' Real time Image processing Traditionally, background correction and sensor non-uniformity correc- tions in images required separate software operations and processing time in a computer. ImagEM is equipped with on-board Digital Signal Processing functions that offer real time image processing that repla- ces the software and computer processing. When implemented, high- ly corrected images emerge directly from the camera at full frame rates. A special recursive filter function is also included to dramatically reduce the effects of Excess noise when the EM-CCD operation is se- lected. This real time filter provides full frame rate images of averaged images; creating exceptional quality EM-CCD images. ' External trigger/Synchronous readout trigger
(Patent pending) When an EM-CCD camera is combined with a real-time confocal mi- croscope, it is very important to synchronize the camera readout with the rotation of the disk in the case of spinning disk type or with the galvanometer in case of a mirror scanning type. Vertical smear or a non-uniform intensity (banding) appears in images unless the same number of scan points are created in every point in the image. To overcome this, the ImagEM is designed with a specially developed synchronous readout trigger that assures even intensity in every im- age. ' Programmable trigger signal output To simplify the control and synchronization of peripheral devices, the ImagEM is equipped with a versatile programmable trigger signal out- put. It is possible to freely control delay time, pulse width, and polarity of the trigger signal output with external commands. ' Multiple heads capability Scientific imaging often requires the simultaneous acquisition of data from multiple cameras. Until now this has been difficult but a new fea- ture of the ImagEM is the ability to synchronously drive two or more cameras. Synchronization is accurate to within one pixel clock opera- tion. To take full advantage of this feature over a wide range of appli- cations, each camera is able to operate synchronously even with indi- vidual exposure settings and individual EM gain settings. With this option, multiple wavelength imaging and multiple polarization angle imaging is simply and reliably done with high precision. 11 5.1.1 Increased cooling capacity ImagEM provides dramatically reduced dark current with new -90 de- gree C cooling capability. A newly designed hermetic sealed vacuum head with a 4 stage peltier element and a highly efficient heat radia- tion structure means great cooling using only air or even lower cooling using water; all in the same camera head and with no modifications. Considering the importance of cooling to the gain and especially the gain stability in EM cameras, ImagEM provides temperature stability of +/-0.05 degrees at -80 degrees C with water cooling. To compensate for the difference in gain values when switching be- tween the water cooled operating temperature of -80 degrees C and the air cooled operation at -65 degrees C, each camera has a built-in gain correction table. This table ensures a constant gain factor regard- less of the cooling mode used. Due to the superb cooling of the ImagEM, the dark current of the CCD has been reduced by 100 times. When used with a circulating water chiller, (F25-ED from Julabo is recommended) the CCD temper- ature is maintained and regulated at -80 degrees C at water tempera- tures under 20 degrees C. When using the fan assisted air cooling feature, the temperature is maintained and regulated at -65 degrees C in air temperatures up to 30 degrees C. Combining this very low dark current with the slow scan readout from the normal CCD mode, the ImagEM is able to capture images over a wide range of applica- tions including those that require very long integration times. 5.1.2. Multiple readout modes ImagEM offers two readout modes. The traditional electron multiplication mode has usually been used for rapid imaging of live cells in low light sit- uations. The addition of a normal CCD amplifier creates the normal CCD readout mode and extends the capabilities of the camera into additional areas of imaging where long term integration with low noise readout is re- quired. Figure 16 is a diagram of the CCD with both readout modes shown. Accumulated charges are transferred from the image storage area of the CCD into the Register Elements as part of the readout process. When the camera is operated in the EM-CCD mode, the accumulated charges would normally be transferred to the right in the diagram and then through the multiplication register to the EM-AMP. In the case of normal CCD readout, the accumulated charges are transferred to the left in the dia- gram, directly into the CCD-AMP from the register elements. This reduces readout noise and creates high S/N values in the image. Fig. 16 point of these two modes (EM-CCD readout mode at 11MHz clock speed and Normal CCD readout mode at 2.75MHz clock speed) is where the incident photon number is around 100 photons/pixel/frame. (See Figure 15) Comparison of readout modes (images of labeled intracellular proteins) EM-CCD readout mode High sensitivity live imaging Exposure time : 30 msec NORMAL-CCD readout mode High precision imaging Exposure time : 3 sec Images of a fluorescently labeled HeLa cell. Objective lens x100 Yokogawa CSU22 Excitation 488nm laser (3mW) ND 10%T Multiplication Elements Register Elements NORMAL-CCD
readout AMP EM-CCD readout AMP 5.1.3. Image Reversal function Since ImagEM includes both normal CCD readout and EM-CCD readout the images from these two methods would normally be mirror images of each other due to the directional readout of each mode. To simplify the data handling by the end user, an automatic image rever- sal function is applied to the EM-CCD mode and the images from the two modes will be created with the same orientation. It is easy to select the readout mode in ImagEM according to the ex- perimental conditions. EM-CCD readout works well for dynamic and real time image acquisition in low light and normal CCD readout works well for imaging normally done with a cooled CCD camera, even for ultra-low light luminescence imaging. The S/N cross-over 12 5.1.4. Multiple clock speed selection ImagEM is equipped with multiple pixel clock speeds in both the EM- CCD readout mode and the Normal CCD readout mode. Since the pixel clock speed changes the readout noise characteristics of the camera, it is possible to select a pixel clock speed that best suits your S/N requirements in either mode. Typical readout noise figures at various pixel clock speeds are listed in Figure 17. Readout noise shown below is the figure at minimum EM gain (4 times). LN-readout Pixel clock 2.75 MHz 0.69 MHz readout noise 17 electrons 8 electrons Fig. 17 Using the 0.69MHz pixel clock speed, it is possible to readout through the EM-AMP register with approximately 10 electrons readout noise. In this case, the necessary EM gain to bring the relative readout noise to less than 1 elec- tron becomes very small, meaning that even a small EM gain setting provides high S/N. This can be a very sensitive imaging method. It is important to note that due to the structure of the FT-CCD, the minimum exposure time is limited by the time required to readout the accumulated charges in the storage area. Figure 18 shows the EM gain setting at which the noise becomes less than 1 electron. Figure 19 shows the minimum exposure time related to different pixel clock speeds. Fig. 18 Figure 20 and Fig 21 show the relationship of different pixel clock speeds to the readout noise, minimum exposure time and frame rate in the Normal CCD readout mode. LN-readout min. exposure time 122 ms (10 frame/s) 488 ms (2 frame/s) Pixel clock 2.75 MHz 0.69 MHz EM-readout (
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content. EM-CCD technical note EM-CCD technical note 1. CCD structures and characteristics 1.1. Interline transfer CCD (IL-CCD) 1.2. Full Frame (FFT-CCD) and Frame Transfer CCD (FT-CCD) 1.3. Back-thinned CCD 2. Noise components of CCD 2.1. Dark current 2.2. Readout noise 3. Electron Multiplier CCD (EM-CCD) 3.1. Principle of Electron Multiplication gain in EM-CCD 3.2. EM gain dependence on temperature 3.3. Gain aging characteristic 4. CCD/EM-CCD noise calculation 4.1. CCD noise components calculation 4.2. EM-CCD noise components calculation 4.3. EM-CCD noise dependence on EM-gain and Input photons 4.3.1 EM gain vs. S/N 4.3.2 Input Photons vs. S/N 4.3.3 S/N Crossover point between normal CCD readout and EM-CCD readout 5. EM-CCD ImagEM Camera Technical note 5.1. Outline of ImagEM Features 5.1.1. Cooling capability improvements 5.1.2. Multi-modes readout 5.1.3. Image reversal function 5.1.4. Multiple clock speed selection 5.1.5. Photon Imaging mode (Patent pending) 5.1.6. Real-time Image Processing features 5.1.7. External trigger / Synchronous readout trigger (Patent Pending) 5.1.8. Programmable Sync output 5.1.9. Multiple camera heads control feature 5.2. Specification 2 P3 P4 P5 P7 P10 1. CCD structures and characteristics There are three types of CCDs which are well known in scientific im- aging. One is the interline transfer CCD (IL-CCD), a second is the Full Frame Transfer CCD (FFT-CCD) and a third is the Frame transfer CCD (FT-CCD). 1.1. Interline transfer CCD (IL-CCD) The structure of IL-CCD is shown in Figure 1(a). The interline CCD has vertically paired columns consisting of imaging Photodiodes (PD) and a readout register (Vertical charge transfer register (V-CCD). Electric charges generated in all the PD by incoming photons are shifted simultaneously to the adjacent V-CCD register. The V-CCD register is covered by a mask of aluminum or other opaque material to prevent photons from creating additional charges in this area dur- ing readout. Readout is accomplished by transferring each horizontal row of information in the V-CCD, line by line, up the CCD to the Hori- zontal serial register (H-CCD). There, charges are transferred hori- zontally and converted into charge voltage by AMPFDA. The design of an IL-CCD has the advantage that the signal accumu- lation (exposure) and readout can be done simultaneously because PD can accumulate charges for the next frame right after the previ- ously generated electric charges in the PD are shifted to the V-CCD. There is no possibility of image smearing in this device. Traditionally, the design of the IL-CCD has had the disadvantage that the open ratio of the light sensitive area (fill factor) is reduced be- cause of the presence of the masked V-CCD area. Recently this dis- advantage has been dramatically improved by on-chip lenses (as shown in Figure 2) and improvement of sensor structures that allow detection of photons deeper in the PD than previous models. Overall Quantum efficiency has increased to over 70%. New IL-CCDs like the Hamamatsu ER-150 CCD (Figure 3) used in the ORCA series of cameras, offer characteristics ideally suited to many scientific applica- tions. 1.2. Full Frame Transfer CCD (FFT-CCD) and Frame
Transfer CCD (FT-CCD) As shown in Figure 1, Frame transfer CCDs are divided into two types, Full frame transfer (FFT-CCD)(b) and Frame transfer (FT-CCD)(c). In the case of a FFT-CCD Figure 1(b), charges generated in pixels are transferred vertically, row by row, to the horizontal serial register for readout. Unless a shutter is used during this transfer and readout, im- age smearing will occur. While this design offers 100% open area ratio with full collection of incoming photons, the shutter limits the frame rate and photons falling on the shutter are lost when it is closed for readout. It is not possible to acquire signal and readout at the same time. FT-CCDs Figure 1(c) offer both 100% fill factor and simultaneous signal acquisition and readout. Like the FFT-CCD, the chip has no charge transfer regions in the signal acquisition area. Rather, the FT-CCD has two separate but equal regions, one with pixels exposed to the incom- ing photons and another region with an equal number of pixels but en- tirely masked to eliminate photons from being detected. As shown in Figure 1(c) one area works as the detection area and the other works as the storage area. Accumulated charges detected in the detection area are rapidly transferred to the masked storage area, and the accu- mulated charges are transferred vertically, line by line, to the horizontal serial register for readout. The detection area and storage area are driv- en individually so the next exposure can start right after the completion of the rapid vertical charge transfer from the detection area to the stor- age area. This design eliminates the need for a mechanical shutter, al- lowing signal accumulation (exposure) and readout simultaneously like an IL-CCD. 1.3. Back-thinned CCD As mentioned before, in FFT-CCDs and FT-CCDs the light sensitive pixels have a charge transfer function as well. This function requires the front surface of light sensitive pixels to be covered by a semi- transparent Poly-Si electrode for the charge transfer to function, as shown in Figure 4. Even with 100% fill factor, the effective quantum efficiency (QE) drops into the 40% range because the Poly-Si electrode absorbs some per- centage of incoming photons depending on their wavelength. To over- come this disadvantage, Back-Thinned CCDs (BT-CCD) are becom- ing popular. In a BT-CCD the CCD is turned upside down and this back side of the CCD is thinned to 10um to 15um in thickness as shown in Figure 4(b). Incident photons now enter the CCD from this back thinned side, without the Poly-Si electrode in the light path. QE values of greater than 90% can be achieved. Figure 5 compares typi- cal QE curves of the same CCD in front- illuminated and back-illumin- ated versions. 3 FDA H-CCD V-CCD PD Pixel (a) (b) (c) Interline CCD Full Frame Transfer CCD Frame Transfer CCD Fig. 1 Old-model Interline CCD Photons Photons ER-150 Interline CCD (a) (b) AL shield sensor sensor Gate Electrode Gate Electrode V-CCD V-CCD Onchip lens Fig. 2 Fig. 3 2. Noise components of CCD As seen chapter 1, innovations in silicon based CCD technology have created many kinds of CCDs. With the Quantum efficiency of the IL- CCD reaching about 70% and the BT design achieving more than 90%, detection limits are nearing their theoretical limits. Signal detec- tion in modern CCDs is often limited by how much camera noise (due to dark current and readout noise) must be overcome before the sig- nal is apparent on the CCD. These values determine the camera per- formance of CCD, especially in low light applications. 4 Fig. 4 Quantum efficiency (%) 0
200 400 600 800 1000 1200 20 10 40 30 60 50 80 70 90 100 Wavelength (nm) Back-thinned illuminated type Front-illuminated type Fig. 5 2.1. Dark current A CCD is made from Silicon, and the dark current caused by thermal mi- gration of electrons in silicon is a main noise factor for a CCD sensor. The Dark current of a CCD depends on the temperature, and it decreases by half when the temperature drops by approximately 7 to 8 degree C. It is apparent that cooling a CCD is a very good way to reduce the dark cur- rent noise. Figure 6 shows the CCD dark current vs. Temperature. Dark current also depends on the type of CCD. In most applications an IL-CCD can normally achieve good performance with -30 to -50 degree C cooling. In the case of an FFT-CCD or FT-CCD, most require -50 to - 90 degree C cooling for low light applications. Cooling is most effective when the CCD is placed in a vacuum chamber. Temperature (;) Dar k current (electron/pix els) Dark Current vs Temperature (S5466) 10 5 10 4 10 3 10 2 10 1 10 0 10 -1 10 -2 - 50 - 40 - 30 - 20 - 10 0 10 20 30 Fig. 6 Hamamatsu Photonics K.K. have developed an unique hermetic va- cuum-sealed chamber with high performance cooling capability based on many years of experience with high vacuum technologies used in Photo Multiplier Tube (PMT) and Image Intensifier (I.I.) tech- nologies (shown in Figure 7). The CCD chip and a multi-stage peltier element (ThermoElectric cooler) are built into a welded metal cham- ber with a special window on the front to create the camera head. The chamber of the vacuum head is vacuumed and hermetically sealed to retain a high degree of vacuum (10 -8 torr or less), to ensure great cooling performance over many years. In comparison with simpler va- cuum sealed models using ordinary gaskets or o-rings, there is no need for periodic re-evacuation or maintenance. With several hundred such vacuum heads in daily use over many years, the unique Hama- matsu hermetic vacuum-sealed head reliably provides significantly better and more stable cooling performance than other designs. VIS UV Gate Electrodes (a) Front-Illuminated FT-CCD Absorption of photons in Gate
Electrodes reduces Quantum
Efficiency in Front- illuminated type. Back-thinned FT-CCD No Absorption of photons in Gate
Electrodes increases Quantum
Efficiency in Back-thinned type. (b) 2.2. Readout noise The largest factor influencing the detection limit of a CCD is the readout noise caused by the on-chip Floating Diffusion Amplifier (FDA) that con- verts accumulated charges into voltage. Accumulated charges transferred into Horizontal serial register are serially transferred into the FDA pixel by pixel. Readout noise is primarily caused by the resetting of the amplifier after the accumulated charge in each pixel is converted to a voltage and the amplifier is reset for the next incoming pixel. This reset noise can be dramatically reduced by an external correlated dual sampling (CDS) cir- cuit. Additionally readout noise depends on pixel clocking frequency and is generally lower with slower CCD clocking speeds. However, slower pix- el clock speeds may limit the camera use for dynamic real-time imaging. In summary, camera readout noise performance depends greatly on the external circuit design of the camera manufacturer and the readout speed. As an example, there are only 3 electrons r.m.s. readout noise in cam- eras such as the Hamamatsu ORCA2-ER cooled CCD. The detection limit of such cooled CCD cameras is about 10 electrons, making it an ul- tra-sensitive camera. More details of noise are explained in chapter 4 : CCD/EM-CCD noise calculations. 3. Electron Multiplier CCD (EM-CCD) As mentioned above, CCD technological innovations are making dramat- ic progress. As a result of various approaches, very high sensitivity and low noise CCDs are readily available. Despite all the advances, readout noise is still the dominant factor limit- ing weak signal detection. Detecting signals below the readout noise lev- el of a camera is possible with various special methods or technologies. Detecting a lower signal than the readout noise is possible by signal inte- gration on CCD chip. Over time the signal will accumulate and become greater than the readout noise. Other techniques that involve signal multiplication are done with Micro- channel Plates (MCP) in an Image Intensifier (I.I.) or direct electron bom- bardment of a CCD (EB-CCD). In these cases signal electrons are cre- ated at a photocathode and then multiplied by a high voltage in a vacuum tube before signals are readout A mechanism for direct multiplication of electrons on the CCD itself has been known for many years. A host of technological problems associated with this on chip multiplication process prevented the technique from be- ing useful until just recently. In the last few years solutions to the prob- lems have been developed and it is now becoming an effective means of ultra low light detection in biological and scientific imaging. This exciting technology has become known as an Electron Multiplier CCD (EM- CCD). 5 CCD Light input window Peltier cooler Fig. 7 3.1. Principle of Electron Multiplication gain in EM-
CCDs Figure 8(a) shows the structure of an EM-CCD. The basic structure is the same as a normal FT-CCD and it is shown as a back-thinned ver- sion. Accumulated charges detected in the detection area are rapidly transferred to the storage area, and then the accumulated charge is transferred line by line to the horizontal serial register for readout, just as in a normal FT-CCD. At this point in an EM-CCD a multiplication (Charge Multiplier) register is built into the horizontal serial register. With this charge Multiplication register, signal multiplication is done by supplying a higher voltage than normal to each horizontal transfer electrode. Figure 8(b) shows the principle of signal multiplication in the charge Mul- tiplication register. When a signal electron charge is transferred from stage to stage, the signal charge is accelerated by high electric field generated under the multiplication gate by applying a high voltage (30 to 40V) to each multiplication electrode (multiplication gate). This high vol- tage is much greater than the normal horizontal transfer electrode vol- tage, and it generates an occasional extra electron-hole pair. This is called an impact ionization event. The probability of such an event is very small, typically about 1.0% to 1.6% at each stage. This is the value (g) in the formula shown below. Electrons are multiplied from stage to stage repeatedly in the gain register and high multiplication gain is ach- ieved. Normally, there are 400 - 600 stages (N) in the gain register. Total multiplier gain (M) can be expressed by the following formula : M =(1+g) N g : probability to generate an electron-hole pair at each stage N : total number of charge Multiplication stages 3.2. EM gain dependence on temperature In the case of EM-CCD, as mentioned above, the Multiplication gain factor in the Multiplication register greatly depends on the tempera- ture. It is obvious that the stabilization of temperature of a sensor be- comes a very important issue. Figure 10 shows an example of an E2V CCD97 CCD and the EM gain vs. temperature. A change of 70 degrees at the CCD changes the EM gain by about 10 times in Figure 10. In addition, as the temperature is decreased, the slope of the change increases. Temperature stability becomes in- creasingly important at lower cooling temperatures to maintain con- stant gain in an EM-CCD. Here, probability of (g) depends not only on the supply voltage of the charge multiplier register but the temperature of a sensor also has a great influence. Control and stability of both the supply voltage and the CCD temperature are very important factors when EM-CCD tech- nology is used for quantitative measurement. The key technical point and advantage of an EM-CCD is signal char- ges in the CCD are multiplied in the multiplication register before it is converted to voltage by the FDA. As mentioned before, the readout noise caused by the FDA is the limiting factor to low signal detection. Multiplication of the signal before the FDA makes the readout noise of the FDA relatively smaller as the multiplication of the original signal increases. Even at moderate gain settings, the relative readout noise becomes less than 1 electron, enabling detection of even single pho- ton events in the signal. As readout speed increases, the readout noise increases by a square function of the frequency. To observe 6 Fig. 8 x 2 x 10 x 50 x 500 Output images obtained when gain
was varied with light level kept constant Fig. 9 Fig. 10 1000 100 10 1 GAIN 39 40 41 42 43 44 45 46 R 2HV T = 20: T = -0 : T = -20: T = -40: T = -50: Decrease of 70
degrees in
temperature yields 10
times gain increase low intensity objects at high speeds, the EM-CCD can overcome the increase in readout noise by additional multiplication gain; again re- ducing the relative readout noise to less than 1 electron. This ability to use the Multiplication gain register to overcome the readout noise even at high readout speeds is the chief advantage of the EM-CCD cameras for fast, scientific, low light imaging. Figure 9 shows sequential images taken while increasing the Multipli- cation gain factor from no Multiplication gain to higher Multiplication gain. With higher Multiplication gain, the multiplied signal becomes larger than the readout noise and an image is visible. (a) (b) Horizontal serial Resister Horizontal serial Register Multiplier Register Supplying higher than normal
voltage to each multiplication
electrode results in an extra
electron hole-pair generated
in an Impact Ionization event. Generated electron Image area Storage area Image area Storage area Normal FT-CCD Output Output EM-CCD 7 4. CCD and EM-CCD noise calculations 4.1. CCD noise components calculation There are four kinds of important noise factors that determine the S/N of a CCD. (1) Readout noise r The reset noise in the electric charge - voltage conversion AMP (FDA) on the CCD is the main source. It is expressed as the disper- sion r. (2) Noise caused by dark current d= (DT) This noise is the fluctuation in the dark current generated in a CCD, and it depends on cooling temperature and exposure time. It is expressed as the square root of the product of D: Dark current (electrons/pixel/sec) and T: Exposure time (sec). (3) Photon shot noise s= (QP) This noise is the fluctuation in the number of incident photons (pho- ton shot noise). It is expressed as the square root of the product of P: Input photon and Q: Quantum efficiency. (4) Spurious Noise cic Spurious noise is a charge generated by signal charge transfer pro- cess in the CCD called Clock Induced Charge. (CIC) This is a fixed value when readout clock and clock duty cycle is fixed, but it is small enough and usually it is possible to ignore it from the following calculation. From our measurement result, it is 0.001 e-/pix- el/frame-readout. Total noise N is calculated by the following expression: Total noiseN = {(Readout noise) 2 +(Dark current noise) 2 +(Photon shot noise) 2 } = ( r 2 + d 2 + s 2 ) = ( r 2 + DT + QP) On the other hand, Signal S is expressed by P: incident Photon and Q: Quantum efficiency as follows. Signal S = QP Thus, signal to noise ratio (S/N) is expressed by the following expression. Signal / noise = QP / ( r 2 + DT + QP) Figure 11 shows the change in noise (N) based on the number of in- cident photons under the following conditions: Readout noise: 8 electrons Exposure time: 100 milliseconds Quantum efficiency: 90% Dark current: 0.01e/p/s 3.3. Gain aging characteristic EM-CCDs have been widely used for high sensitivity applications due to the advantages of the charge multiplication register. But it was not widely known until recently that the multiplication gain tends to suffer gain aging, a slow decrease in gain over time based on the total elec- tric charge passed through the multiplication register. This is changing and has now been officially announced by the CCD manufacturers. To use an EM-CCD for stable quantitative measurement over long periods, it is necessary to consider this EM gain aging. The multipli- cation principle of EM-CCD is achieved by impact ionization effects of a high voltage (30 to 40V) applied to each multiplication electrode (multiplication gate). While the exact cause of the gain degradation is not known, it is thought that the higher than normal voltages used in the process traps accelerated electrons in the bottom of the transfer electrode. These trapped electrons may change the electric field at this point and thus create the gain aging phenomenon. This gain aging occurs exponentially over time and is most prominent in the early use of an EM-CCD. In order to reduce this to a minimum in ac- tual use, every C9100 camera is factory aged for more than 100 hours and readjusted before being shipped. As a result, serious gain deterioration should not occur in C9100 series cameras. However, since this phenomenon depends on total electric charge through the multiplication register, there may be some applications where addi- tional gain degradation can occur. Reducing gain and light intensity when the camera is not being used can help prevent this. It is wise to check the gain with a standard sample occasionally if long term stan- dardization is required. Tips for operating EM-CCD cameras (1) Keep the gain adjusted to a level that offers just enough gain to overcome the readout noise. There is no increase in the Signal to Noise ratio once the readout noise becomes less than one and add- ing gain past this point only increases the rate of gain degradation. (2) Reduce illumination to the detector as much as possible. Even if the gain is not increased, an increase in the signal intensity will create a larger charge in the multiplication register and increase the rate of gain degradation. (3) Reducing the gain to the minimum setting and blocking illumina- tion to the CCD when the camera is not being used for measurement can help maintain stable EM-CCD gain over long periods. As shown in Figure 11, the noise characteristics can be divided into two domains. A change in the slope of the noise value occurs at 70 photons. Below this value the graph indicates that when the incident photon number is less than 70 photons the camera readout noise is the dominant factor in the noise calculation. When the incident photon number is greater than 70 photons the photon shot noise is the domi- nant noise factor. (In this example with the comparatively short ex- posure time, dark noise is too small to influence the result.) In practi- cal terms, 70 photons define the point at which the camera noise no longer has an influence on the S/N and the number of incident pho- tons defines the S/N. 8 Fig. 11 4.2. EM-CCD noise components calculation In the case of EM-CCD, multiplication of the signal in the multiplica- tion register has noise associated with that process and this will influ- ence the S/N. Noise caused by signal multiplication is called Excess noise (F) and is added to the signal. Excess noise: F is calculated by the multiplication factor M and a ratio of the dispersion of a multiplica- tion register input signal: in and a dispersion of the multiplication register output signal: out. F 2 = out 2 / M 2 in 2 It is important to note that not only are the detected signal electron charges multiplied but also any electron charge in the CCD from other sources such as dark current are multiplied with the same multi- plication and noise factors. Calculations for signal level and total noise in an EM-CCD must in- clude multiplication noise, and are shown below. Signal : S is expressed as the product of detected signal : QP and multiplication gain : M. Signal S = QPM Total noise: N is expressed as the Square root of the product of the sum of signal charge :S, dark current :D, time :T, multiplication factor : M and Excess Noise :F as follows. Total noise N = { r 2 + F 2 M 2 ( d 2 + cic 2 + s 2 )} In this expression, cic is small enough to ignore from S/N calcula- tion, total noise becomes following. Total Noise N = { r 2 +F 2 M 2 (DT + QP) } S/N is calculated as shown below. S/N = QPM/ { r 2 +F 2 M 2 (DT + QP) } = QP/ { r 2 /M 2 +F 2 (DT + QP) } 4.3. EM-CCD noise dependence on EM-gain vs. In-
put photons Excess Noise factor is estimated at 1.41 for calculations of multiplica- tion register noise. This value is included in the examples shown below. 4.3.1. EM gain vs. S/N Figure 12 shows how S/N is influenced by the number of the incident photons (10, 100, and 1,000 (photon / pixel / frame), and increasing multiplication gain. An Exposure time of 33 msec, Quantum efficiency 90% , and readout noise of 80 electrons are used for this example.. The graph shows S/N clearly improves with signal multiplication but there is a limit to the S/N improvement in each case. Even if higher EM gain is applied, there is a certain level at which no further im- provement can be expected. When the input photon level exceeds 1000 photons/pixel in 33 msec the EM gain feature offers almost no benefit. The S/N at that point is limited only by the number of input photons (QP) and the multiplication noise (F) This can be seen in expression (A). Supposing the dark current is small enough, it is possible to ignore r 2 / M 2 as multiplication gain increases. In this case S/N is simplified as in Expression (A) below. S/N = QP/ {F 2 (QP) } = ( QP/F 2 ) From this calculation S/N is constant since the incident photon num- ber P is constant. This example clearly shows there is no advantage to increasing gain more than required according to the numbers of incident photons. This is very important point to be considered because it effects the gain stability in the long term. S/N vs. EM gain EM Gain S/N 0.1 1 10 100 1000 10000 100 10 1 Input photon=10
Input photon=100
Input photon=1000 1 10 100 1000 1 10 100 1000 Noise (e - ) 10000 100000 1000000 Input photon number Input photon / Noise Fig. 12 9 4.3.2. Input Photons vs. S/N Figure 13 shows how S/N is affected by the number of incoming pho- tons at different gain settings. The conditions are otherwise the same as in Figure 12. In this graph, it is shown that the S/N does not im- prove at gain settings over 200 but does improve with the number of incoming photons. The point is that using more gain than necessary is not a benefit and will only increase the rate of gain degradation in the multiplication register. Different conditions will change the gain setting at which this happens but the lesson is the same. Input photon number S/N 0.01 0.1 1 10 100 1000 1 10 100 1000 10000 100000 1000000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000 S/N vs. Input photon Fig. 13 Figure 14 shows how readout noise becomes relatively smaller as Gain is increased. 0 Gain 0.001 0.1 10 100 1000 200 400 600 800 1000 1200 1 1400 Relativ e noise Noise / Gain 10000 0.01 Fig. 14 4.3.3. S/N Crossover point between normal CCD
readout and EM-CCD readout Figure 15 shows how S/N changes with the number of incident pho- tons in normal CCD readout and EM-CCD readout respectively. As shown in Figure 15, using 100 photons/pixel/frame as a reference (exposure time: 100msec), when the incident photons are 100 pho- tons/pixel/frame or more, normal CCD readout provides better S/N. When incident photons are less than 100 photons/pixel/frame, EM- CCD readout provides better S/N. Notably, even with less than 100 photons/pixel/frame, normal CCD readout offers higher S/N if the EM gain is less than 50 times. Another way to look at this is that in order to improve S/N with less than 100 photon/pixel/frame conditions, it is necessary to use at least 50 times EM gain. This value is confirmed in Figure 14 by observing that the point at which the noise becomes less than one photon - at a gain of 50 times. Since the ImagEM camera has both normal CCD and EM-CCD read- out modes it is recommended to switch readout modes to maximize the S/N based on the number of incident photons. EM-CCD out / Normal CCD out Input photon number S/N 0.01 0.1 1 10 100 1000 1 10 100 1000 10000 100000 1000000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000
LN-AMP Input photon number S/N 100 10 1 0.1 10 100 1000 Gain=4
Gain=50
Gain=100
Gain=200
Gain=500
Gain=1000
LN-AMP Enlargement Fig. 15 5. EM-CCD ImagEM Camera Technical note 5.1. Outline of ImagEM Features Hamamatsu Photonics K.K. uses the very latest cooling and CCD camera design technologies for the ImagEM EM-CCD camera. The ImagEM camera is the result of many years of experiments and ex- perience with deep cooling, high vacuum and low noise CCD technol- ogy. It provides optimum signal and noise characteristics for a wide range of applications. With one camera, it is now possible to capture both low and high light level images, wide dynamic range images and single photon binary images, long integration and high frame rate im- ages and it offers a very high Quantum efficiency over a broad range of wavelengths. The features below confirm that virtually any applica- tion can be addressed with this new camera. ' -90 degree C cooling with hermetic sealed head A newly designed hermetic sealed head includes a specially devel- oped 4 stage peltier cooler for the CCD and the entire camera head has been designed to optimize heat radiation. With both air and water cooling capabilities built-in, temperature at the CCD can be main- tained as low as -90 degrees C with the water cooling operational and a water temperature of 10 degrees C or less. ' Gain stability/Temperature stability Temperature stability is the key to gain stability in an EM-CCD cam- era so the ImagEM provides temperature stability to within +/-0.05 de- grees at -80 degrees C with water cooling. ' Dual readout mode ImagEM has both an EM-CCD readout and a normal CCD readout. With recent innovations, the normal CCD readout offers very low noise readout and very low dark current for long integration expos- ures. This creates tremendous flexibility in applications. From ultra low luminescence samples to routine fluorescence microscopy, the con- ventional readout provides high S/N images. With the EM-CCD read- out, high frame rates at low intensities are possible for live cell imag- ing and spinning disk confocal applications. ' Multiple pixel clock selection ImagEM offers a selection of pixel clock speeds in both EM-CCD readout mode and normal CCD readout mode. According to the appli- cation, it is possible to select a clock speed that offers the best S/N. Faster clock speed offer faster frame rates and lower clock speed pro- vides lower noise. ' Photon imaging mode (Patent Pending) Due to the Excess noise factor in the multiplication register, EM-CCD cameras have traditionally had limited use in photon counting applica- tions. Intensified CCDs have dominated this field for many years. Us- ing over 20 years of experience in the design and production of image intensifiers and electronic circuit technology, Hamamatsu Photonics K.K. has incorporated a unique photon counting imaging mode into the ImagEM. 10 ' Real time Image processing Traditionally, background correction and sensor non-uniformity correc- tions in images required separate software operations and processing time in a computer. ImagEM is equipped with on-board Digital Signal Processing functions that offer real time image processing that repla- ces the software and computer processing. When implemented, high- ly corrected images emerge directly from the camera at full frame rates. A special recursive filter function is also included to dramatically reduce the effects of Excess noise when the EM-CCD operation is se- lected. This real time filter provides full frame rate images of averaged images; creating exceptional quality EM-CCD images. ' External trigger/Synchronous readout trigger
(Patent pending) When an EM-CCD camera is combined with a real-time confocal mi- croscope, it is very important to synchronize the camera readout with the rotation of the disk in the case of spinning disk type or with the galvanometer in case of a mirror scanning type. Vertical smear or a non-uniform intensity (banding) appears in images unless the same number of scan points are created in every point in the image. To overcome this, the ImagEM is designed with a specially developed synchronous readout trigger that assures even intensity in every im- age. ' Programmable trigger signal output To simplify the control and synchronization of peripheral devices, the ImagEM is equipped with a versatile programmable trigger signal out- put. It is possible to freely control delay time, pulse width, and polarity of the trigger signal output with external commands. ' Multiple heads capability Scientific imaging often requires the simultaneous acquisition of data from multiple cameras. Until now this has been difficult but a new fea- ture of the ImagEM is the ability to synchronously drive two or more cameras. Synchronization is accurate to within one pixel clock opera- tion. To take full advantage of this feature over a wide range of appli- cations, each camera is able to operate synchronously even with indi- vidual exposure settings and individual EM gain settings. With this option, multiple wavelength imaging and multiple polarization angle imaging is simply and reliably done with high precision. 11 5.1.1 Increased cooling capacity ImagEM provides dramatically reduced dark current with new -90 de- gree C cooling capability. A newly designed hermetic sealed vacuum head with a 4 stage peltier element and a highly efficient heat radia- tion structure means great cooling using only air or even lower cooling using water; all in the same camera head and with no modifications. Considering the importance of cooling to the gain and especially the gain stability in EM cameras, ImagEM provides temperature stability of +/-0.05 degrees at -80 degrees C with water cooling. To compensate for the difference in gain values when switching be- tween the water cooled operating temperature of -80 degrees C and the air cooled operation at -65 degrees C, each camera has a built-in gain correction table. This table ensures a constant gain factor regard- less of the cooling mode used. Due to the superb cooling of the ImagEM, the dark current of the CCD has been reduced by 100 times. When used with a circulating water chiller, (F25-ED from Julabo is recommended) the CCD temper- ature is maintained and regulated at -80 degrees C at water tempera- tures under 20 degrees C. When using the fan assisted air cooling feature, the temperature is maintained and regulated at -65 degrees C in air temperatures up to 30 degrees C. Combining this very low dark current with the slow scan readout from the normal CCD mode, the ImagEM is able to capture images over a wide range of applica- tions including those that require very long integration times. 5.1.2. Multiple readout modes ImagEM offers two readout modes. The traditional electron multiplication mode has usually been used for rapid imaging of live cells in low light sit- uations. The addition of a normal CCD amplifier creates the normal CCD readout mode and extends the capabilities of the camera into additional areas of imaging where long term integration with low noise readout is re- quired. Figure 16 is a diagram of the CCD with both readout modes shown. Accumulated charges are transferred from the image storage area of the CCD into the Register Elements as part of the readout process. When the camera is operated in the EM-CCD mode, the accumulated charges would normally be transferred to the right in the diagram and then through the multiplication register to the EM-AMP. In the case of normal CCD readout, the accumulated charges are transferred to the left in the dia- gram, directly into the CCD-AMP from the register elements. This reduces readout noise and creates high S/N values in the image. Fig. 16 point of these two modes (EM-CCD readout mode at 11MHz clock speed and Normal CCD readout mode at 2.75MHz clock speed) is where the incident photon number is around 100 photons/pixel/frame. (See Figure 15) Comparison of readout modes (images of labeled intracellular proteins) EM-CCD readout mode High sensitivity live imaging Exposure time : 30 msec NORMAL-CCD readout mode High precision imaging Exposure time : 3 sec Images of a fluorescently labeled HeLa cell. Objective lens x100 Yokogawa CSU22 Excitation 488nm laser (3mW) ND 10%T Multiplication Elements Register Elements NORMAL-CCD
readout AMP EM-CCD readout AMP 5.1.3. Image Reversal function Since ImagEM includes both normal CCD readout and EM-CCD readout the images from these two methods would normally be mirror images of each other due to the directional readout of each mode. To simplify the data handling by the end user, an automatic image rever- sal function is applied to the EM-CCD mode and the images from the two modes will be created with the same orientation. It is easy to select the readout mode in ImagEM according to the ex- perimental conditions. EM-CCD readout works well for dynamic and real time image acquisition in low light and normal CCD readout works well for imaging normally done with a cooled CCD camera, even for ultra-low light luminescence imaging. The S/N cross-over 12 5.1.4. Multiple clock speed selection ImagEM is equipped with multiple pixel clock speeds in both the EM- CCD readout mode and the Normal CCD readout mode. Since the pixel clock speed changes the readout noise characteristics of the camera, it is possible to select a pixel clock speed that best suits your S/N requirements in either mode. Typical readout noise figures at various pixel clock speeds are listed in Figure 17. Readout noise shown below is the figure at minimum EM gain (4 times). LN-readout Pixel clock 2.75 MHz 0.69 MHz readout noise 17 electrons 8 electrons Fig. 17 Using the 0.69MHz pixel clock speed, it is possible to readout through the EM-AMP register with approximately 10 electrons readout noise. In this case, the necessary EM gain to bring the relative readout noise to less than 1 elec- tron becomes very small, meaning that even a small EM gain setting provides high S/N. This can be a very sensitive imaging method. It is important to note that due to the structure of the FT-CCD, the minimum exposure time is limited by the time required to readout the accumulated charges in the storage area. Figure 18 shows the EM gain setting at which the noise becomes less than 1 electron. Figure 19 shows the minimum exposure time related to different pixel clock speeds. Fig. 18 Figure 20 and Fig 21 show the relationship of different pixel clock speeds to the readout noise, minimum exposure time and frame rate in the Normal CCD readout mode. LN-readout min. exposure time 122 ms (10 frame/s) 488 ms (2 frame/s) Pixel clock 2.75 MHz 0.69 MHz EM-readout (
