The highly sensitive TVIST black-and-white television-computer measuring system has been developed on ... cell by comet assay for DNA repair method, quantitative determination of .... There is also a possibility of remotely varying the gain ...
INSTRUMENTATION Highly sensitive TVIST television-computer measuring system for fluorescence endoscopy and microscopy G. V. Papayan S. I. Vavilov State Optical Institute, All-Russian Scientific Center, St. Petersburg, Russia
A. A. Mantsvetov St. Petersburg Electrical-Engineering University, St. Petersburg, Russia
共Submitted June 10, 1999兲 Opticheski Zhurnal 67, 70–77 共January 2000兲
The highly sensitive TVIST black-and-white television-computer measuring system has been developed on the basis of a personal computer and a digital camera, which employs a 1/2-inch photosensitive area-array CCD from Sony with 582⫻752-pixel resolution. The system can operate in both accumulation and television broadcast modes. In both modes the frames are shown on a computer display. The frame rate lies in the range from 25 to 0.1 Hz. When the slow scan mode is employed, the frame rate is reduced by a factor of 4. In the accumulation mode with both fast and slow scanning there is no smear, and thus electromechanical shutters are not needed. The maximum possible spatial resolution is achieved in the accumulation mode with slow scanning. The system exhibits stability and linearity, permitting its use for precision measurements. The low intrinsic noise of the system (7e) 1兲 and the possibility for operation in a prolonged accumulation mode permit raising its sensitivity in comparison to ordinary television systems by two to three orders of magnitude. The software supports the TWAIN-32 protocol, permitting the use of the system with Windows-95 applications intended for image processing and analysis. The results of estimation of the principal parameters of the system are presented. Areas for possible application of the system are indicated. © 2000 The Optical Society of America. 关S1070-9762共00兲01401-9兴
INTRODUCTION
Television-computer systems are being used increasingly more widely in various areas of science and technology and are replacing ocular observation techniques and traditional recording, measurement, and analysis methods. There is especially strong activity in this regard in microscopy and some areas of medical technology. The following is a list of some of the trends and objectives of such applications: luminescence endoscopy for the early diagnosis of inflammatory diseases and malignant tumors;1–3 video-intensified fluorescence microscopy 共VIM兲 on the cellular and subcellular levels, including methods for determining the calcium concentration and pH within cells, measuring the content and distribution of DNA in chromosomes, detecting antibodies using immunohistochemical reactions, finding definite nucleotide sequences using fluorescent DNA probes and in situ hybridization 共the FISH method兲, DNA sequencing, investigating genome damage in an individual cell by comet assay for DNA repair method, quantitative determination of proteins and DNA, as well as ions, receptors, cellular components, genes, etc.;4–9 Allen and Inou’e video-enhanced contrast 共VEC兲 microscopy, which makes it possible to study subdiffraction structures;5 quantitative morphometric and densitometric analysis of 58
J. Opt. Technol. 67 (1), January 2000
cells and tissues, including television cytophotometry methods;10,11 evaluation of the quality of microscope objectives on the basis of a television-computer analysis of Airy patterns.12 In many of the applications just cited it is necessary to record patterns under the conditions of light deficiency and the impossibility 共in contrast to astronomic applications兲 of using very large signal accumulation times. The time restrictions can also be of a different nature. In the case of an investigation of a living object, they can be associated with the mobility of the object, the dynamics of the processes taking place in it, and the time of exposure of the object to radiation, and in the case of nonliving structures they can be associated with the limited time allocated to finding objects, the need for frequent focusing, the occurrence of a photobleaching effect, etc. In order to ensure the possibility of selecting the most advantageous recording parameters under given conditions, a television-computer system must have a broad range for regulating the accumulation time and, accordingly, a frame rate from fractions of a hertz up to the broadcast standard. Apart from the high sensitivity and speed requirements, in solving measurement problems, increased requirements are also placed on the resolution, signal-to-noise (S/N) ratio, linearity, stability, and other characteristics of televisioncomputer systems. This article provides a brief description of the TVIST
1070-9762/2000/010058-07$18.00
© 2000 The Optical Society of America
58
FIG. 1. Structural diagram of the TVIST system.
system, which we developed for solving the problems enumerated above. Special attention is devoted to estimating the metrological characteristics of the system.
BRIEF DESCRIPTION OF THE SYSTEM
A structural diagram of the TVIST system is presented in Fig. 1. The system consists of two principal units, which are connected to one another by a connecting cable: a camera head, which is essentially a digital television camera, and a high-speed specialized frame grabber, which is inserted into an ISA bus slot of a personal computer. The photoelectric converter in the television camera is a Sony ICX039DLA uncooled black-and-white area-array 59
J. Opt. Technol. 67 (1), January 2000
charge-coupled device 共CCD兲 共with increased sensitivity兲 with a 1/2-inch optical format and 582(V)⫻752(H) elements, each measuring 8.3(V)⫻8.6(H) m. The signals for controlling the CCD are formed by a digital control unit consisting of a driving oscillator 共28.375 MHz兲, a synchronizing generator, a timing generator, and horizontal and vertical drivers. The synchronizing generator provides for the formation of the synchronizing and quenching pulses needed for the operation of the camera head in the standard television mode and is also used to synchronize its operation with the operation of the frame grabber. The timing generator provides for operation in the field/frame accumulation modes, control of the electronic shutter with accumulation times from 1/50 to 1/10 000 s, realization of the G. V. Papayan and A. A. Mantsvetov
59
prolonged accumulation of charges in the CCD, and the slow scan mode. Thus, the system under consideration has provisions for two fundamentally different modes: a standard television broadcast mode with interlaced scanning and an accumulation mode with pseudo-row-by-row readout of the charge packets, which provides a sharp increase in sensitivity, as well as a higher resolving power in the transverse direction. In both modes the frames are shown on a computer display. The maximum number of accumulated frames is 255, which provides for an exposure time up to 10 s. Precision measurements are performed using a slow scan mode with lowering of the frequency of the driving oscillator by a factor of 4, which permits improvement of the longitudinal resolution of the system without altering the transmission band of the video system, as well as an improvement in sensitivity when an additional low-pass filter 共not shown in Fig. 1兲 is included. In the television broadcast mode, the frame rate is 25 Hz, and in the accumulation mode it is determined by the accumulation time and lies in the range from 25 to 0.1 Hz. When the slow scan mode is employed, the frame rate is lowered by a factor of 4, and the readout time of a single frame increases from 40 ms to 160 ms. In the accumulation mode with both fast and slow scanning there is no smear during readout, and thus electromechanical shutters are not needed. The output video signal of the area-array CCD is processed and amplified by a low-noise double correlated decision 共LDCD兲 circuit and an amplifier with an automatic gain control 共AGC兲 circuit and then fed into a video processor, which performs the following functions: gamma correction, restriction of the white level, and regulation of the black level. Fixation, i.e., reconstruction of the constant component of the video signal, is performed in all the cascades of the video processor. The output buffer conveys the video signal to the frame grabber through the connecting cable. The I2C bus controller in the camera permits control of the operating modes of the amplifier and the video processor: switching of the mode from manual amplification to AGC; variation of the value of the gamma characteristic of the video system 共0.45–0.7–1.0兲; control of the gain of the video system in the manual mode; and regulation of the black level bias by supplying a constant voltage of opposite sign with respect to the signal to the input of the video processor. The last property permits an artificial increase in the contrast of the video signal and thereby the real-time observation of weakly contrasted objects on the screen. In addition, the I2C bus controller permits control of the accumulation time, the activation of various delays of the electronic shutter, switching of the field/frame accumulation mode, and activation of the slow scan mode. There is also a possibility of remotely varying the gain over a broad range (K⫽1 – 50) and of varying the gain regulation range. The I2C bus is controlled from the frame grabber and is accessible to programs from the existing driver. The frame grabber contains an analog-to-digital converter 共ADC兲 with an amplifier and a low-pass filter, which restricts the frequency band of the video signal to half of the discretization frequency. The ADC performs 8-bit conversion with the frequency of frame readout from the area-array 60
J. Opt. Technol. 67 (1), January 2000
FIG. 2. Fragment of the window of the controlling program of the TVIST system with the image of a preparation 共chromosomes stained with the fluorescent dye FITC兲.
CCD, which guarantees the absence of frequency beating. The digitized data are stored in a 512 kbyte RAM buffer, from which they can be read into the computer through the ISA bus controller or outputted on the real time scale to the computer monitor through the future connector 共FC兲 of the VGA adapter. In this case the image is reproduced on the computer screen at the television rate in the television mode. In the modes with accumulation the memory is utilized to duplicate the image, i.e., during the accumulation of a frame the preceding frame is always seen on the screen. The frame grabber also contains an I2C bus controller, which provides for transmission of the necessary data to the camera. The processes in the frame grabber are controlled and synchronized by an Altera programmable logic array, which provides for counting the accumulation time, controlling the RAM, servicing the FC, capturing a frame in the computer, and several other service functions. Power-supply voltages needed for operation of the area-array CCD and the framing microcircuits are produced by a secondary powersupply unit, which, in turn, is powered by the computer. The control of the basic modes of the television camera, as well as the visualization of images 共including real-time visualization兲, are performed by a program driver operating in a Windows-95 environment in accordance with the TWAIN-32 protocol. The form of a fragment of the window of the controlling program with an image of an object is shown in Fig. 2. The driver is triggered by Windows-95 applications intended for visualizing and processing images 共Adobe Photoshop, Paint Shop Pro, etc.兲. The images can be stored in any format supported by the application. The camera head measures 55⫻45⫻50 mm.
ESTIMATION OF CHARACTERISTICS
The spectral sensitivity s() of the area-array CCD is given only in relative units in the Sony catalog. It follows from a number of sources that the quantum efficiency ( m ) at the spectral sensitivity maximum in similar devices amounts to about 50% with allowance for losses 共see, for G. V. Papayan and A. A. Mantsvetov
60
TABLE I. Spectral characteristics of the area-array CCD. , nm
400
450
500
550
600
650
700
750
800
850
900
950
1000
s(), % (), %
45 28
90 50
100 50
95 43
85 35
60 23
45 16
35 12
25 7.8
18 5.3
10 2.8
5 1.3
1 0.3
example, Ref. 13兲. If this value is taken as the true value, the quantum efficiency of the detector can be calculated over the spectrum using the following relation:
共 兲 ⫽ 共 m 兲 s 共 兲 s /. The catalog values of the relative spectral characteristic s() and the values of the quantum efficiency calculated on the basis of ( m )⫽50% in the 400–1000 nm range are given in Table I. The dark current was determined at an ambient temperature of 18 °C after warming up the camera at an accumulation time of 10 s for 30 min. Conversion of the values obtained to the time of a single frame gives a mean current per pixel over all the array elements equal to 0.8e. In a very small percentage of the elements 共⬃0.4% of all the elements兲 the current reaches 7e, and in a few elements (⬃0.01%) it reaches 14e. They are the so-called ‘‘hot pixels,’’ which become noticeable at large times, creating the effect of stars in the sky. Software can be used to eliminate them, or a single-cascade Peltier cooler can be installed, which makes it possible to reduce the dark current by more than two orders of magnitude. The spatial resolution of the system was estimated by two different methods: by directly measuring the aperturefrequency characteristic 共AFC兲 using a special testing table according to the method described in Ref. 14 and by twodimensional transformation of Fourier signals from the ‘‘hot pixels,’’ which served as an estimate of the pulsed response of the system. The second method is based on determining the ideal response of the area-array CCD in one element and yields the AFC component formed during the readout, amplification, and processing of the charge. At the same time, this method does not take into account the AFC components appearing during photoelectric conversion within the areaarray CCD. Determination of the AFC by the second method is possible only in the accumulation mode. A comparison of the AFC’s obtained by both methods under identical conditions gave good results, pointing out the possibility of using the second, simpler method for estimating the resolution in practice. Figure 3 shows the results of measurements of the AFC H(m), where m is the number of television lines 共tvl兲 for two directions, viz., the transverse 共the vertical AFC兲 and the longitudinal 共the horizontal AFC兲, for the broadcast mode and the accumulation mode with fast and slow scanning. An interlaced field accumulation principle with summing of the charge packets accumulated in neighboring rows is employed in the broadcast mode. Summation of the charges causes the first zero of the broadcast transverse AFC 共curve 1兲 to be at a frequency which is two times lower than the value for the transverse AFC’s in the accumulation mode 共3, 5兲 using a row-by-row readout principle. The longitudinal 61
J. Opt. Technol. 67 (1), January 2000
AFC’s for the broadcast mode 共2兲 and the accumulation mode with fast scanning 共4兲 coincide with one another. The form of the longitudinal AFC in the accumulation mode with slow scanning and a broad band 共6兲 shows that this mode provides the best television resolution. To calculate the penetrating spatial frequency response together with the optical systems, the AFC must be converted into the conventional system of units in optics, and the following relationship between the television resolution m 共tvl兲 and the optical resolution p (mm⫺1) must be taken into account: p⫽m/2h, where h is the dimension of the instrument along a vertical 共for a 1/2-inch array h⫽4.8 mm兲. The signal-to-noise (S/N) ratio was measured using the following method. The matrix was illuminated using a brightness-regulated quasiuniform light source. Then two successive frames were recorded, and a difference frame was formed after the frame readout operation. Such a procedure permits elimination of the influence of the residual illumination nonuniformity in the calculation of the noise. To obtain purely positive values, a constant shift equal to the mean value of the signals was introduced in calculating the difference frame, and to take into account the additional noise from the second frame, the calculated value of the standard deviation of the noise was divided by 1.41. Figure 4 shows the results of measurements of the S/N ratio for the minimum (K⫽1) and maximum (K⫽50) values of the gain of the video system in the camera, which were obtained in the accumulation mode with slow scanning
FIG. 3. Amplitude-frequency characteristics of the TVIST system in various modes: 1, 2 — broadcast mode; 3, 4 — accumulation mode with fast scanning; 5, 6 — accumulation mode with slow scanning and a broad band. 1, 3, 5 — transverse AFC; 2, 4, 6 — longitudinal AFC. G. V. Papayan and A. A. Mantsvetov
61
FIG. 4. Signal-to-noise ratio for various gain values: K⫽50 共1, 2兲, K⫽1 共3, 4兲: 1, 3 — calculated curves for the case where the intrinsic noise of the camera can be neglected; 2, 4 — experimental curves.
and a narrow band. The figure also shows calculated plots of the S/N ratio for the ideal case, where the intrinsic noise of the camera can be neglected and the output noise is caused only by the shot noise of the photoelectrons. The latter was estimated by geometrically subtracting the intrinsic noise 共the noise of the darkened camera兲 from the noise obtained with illumination. The value of the intrinsic noise for K ⫽50 amounted to 3.5 ADC codes and scarcely depended on the accumulation time, whence it follows that the dark current does not have a significant effect on the intrinsic noise under these conditions and is caused mainly by the noise of the readout and amplifying systems 共the readout noise兲. The curve for the ideal case at K⫽50 shows that the largest value of the S/N ratio under these conditions is about 22, which corresponds to a maximum charge packet in the photodiodes of the CCD array equal to roughly 500e at this gain. When K⫽1, the signal achieves saturation at about 25 000e, which corresponds to ⬃80% of the size of the maximum charge packet on one photodiode. These data allow us to measure the ADC output signal in electrons (2e/code for K⫽50 and 100e/code for K⫽1兲. Hence, the intrinsic noise of the camera can be estimated in electrons: for K⫽50 it amounts to 7e, and for K⫽1 it is 30e. A comparison of the curves in Fig. 4 shows that the actually achieved values of the S/N ratio approach the ideal values for both the maximum gain and for K⫽1. Scrutiny of the character of the dark images reveals that there is pure Gaussian noise in the former case, while periodic interference, which is caused mainly by the use of a long cable 共3 m兲 connecting the camera head to the frame grabber, plays an appreciable role at the low gain. The threshold sensitivity was determined by a computational method from the experimental data obtained above on the intrinsic noise of the system at the maximum gain and the values of the quantum efficiency according to Table I. Conversion of the intrinsic noise into the threshold illuminance in the case of illumination by a monochromatic radiant flux with a wavelength of 550 nm, an accumulation time t⫽1 s, and S/N⫽1 gives E thr⫽8.2⫻10⫺8 W/m2⫽5.6⫻10⫺5 lx. 62
J. Opt. Technol. 67 (1), January 2000
With consideration of the weak dependence of the intrinsic noise on the accumulation time, to determine the threshold for other values of t, the values of E thr must be divided by the corresponding value of t. The linearity of the light-signal characteristic was verified with quasiuniform illumination of the array by a stabilized light source. The possibility of increasing the chargepacket integration time by a strictly fixed number of times in the accumulation mode permits the use of a simple method for estimating the linearity of the entire video system. The maximum output electrical signal for the summation of 25 frames was achieved by initial regulation of the level of the light signal. Then the dependence of the output signal increment on the number of frames accumulated was plotted in the range from 1 to 25 frames. Since this parameter is of interest mainly at high values of the S/N ratio, the measurements were performed at K⫽1. The measurement results show that the deviation of the increments from the mean values in different parts of the scale does not exceed 0.2– 0.4% of the total signal spread and that the smaller of these values can be assigned to the end of the scale. The stability was estimated under experimental conditions similar to the preceding conditions after warming up the camera for 1 h. The maximum deviation of the signal from the mean value over the course of 2.5 h did not exceed ⫾0.5%. The contrast L at the output of the system is related to the contrast of the object L 0 in the following way: L/L 0 ⫽1 共 1⫺C/A 兲 , where A is the signal in the light parts of the object and C is the regulated bias voltage supplied to the camera output. In particular, when A amounts to ⬃80% of the saturation level, the contrast can be increased 10 fold in comparison to the original value. Since such an increase is achieved by a purely electrical technique, it is not difficult to attain an even larger increase in contrast, but the real limits are determined by the possibility of stabilizing the light source and the quality of the preparation. CONCLUSION
In creating the TVIST system, we strove primarily to satisfy the requirements introduced by fluorescence endoscopy and microscopy for recording black-and-white television images, although these areas far from exhaust the possibilities of its use in scientific instrumentation, medical technology, and industrial control. The problem of obtaining a sufficiently high-quality system that is, at the same time, economic was posed. This, in particular, is the reason for constructing it on the basis of a personal computer using the future connector 共FC兲. To optimize the structure of the system, it was also taken into account that in many cases a 580⫻750⫻8 digital representation is fully sufficient and that there are practical restrictions on increasing the exposure time. The system developed has the high stability and linearity needed for measuring tools, which permit its use for solving precision photometric problems. In addition, it provides for high-quality morphometric measurements owing to the abG. V. Papayan and A. A. Mantsvetov
62
sence of geometric distortions in the area-array CCD, the rigid matching of pixel coordinates to the positions of the raster elements, the high S/N ratio, and the possibility of obtaining the maximum values of the spatial resolution in the accumulation mode with slow scanning. The introduction of a submode with fourfold narrowing of the transmission band in this mode permits lowering of the intrinsic noise of the system roughly by a factor of 2. The possibility of a rapid passage from the accumulation mode to the television broadcast mode and back provides for the on-line transition from visual inspection of an object and focusing to recording of the image. The presence of an AGC circuit and an electronic shutter makes it possible to easily compensate variations of the illuminance over a broad range, and on-line variation of the contrast permits the real-time observation of weakly contrasted objects, such as unstained cytological preparations. The program driver provides for complete remote control of the camera operation from the computer, and the driver menu is combined with the window in which the current image is displayed. Placement of the camera head directly on an endoscope is possible owing to its small dimensions and the presence of a long flexible cable connecting the camera to the computer. The functioning of the driver in accordance with the TWAIN-32 protocol permits the use of the system together with numerous Windows-95 applications intended for processing and quantitatively analyzing images, and the stored image can have practically any format. The absence of smear during readout permits dispensing with electromechanical shutters, which is especially important in endoscopic applications. The estimates of the metrological characteristics of the system presented in this article are based on methods that do not require special equipment in most cases, permitting their employment for checking by users. Wherever it was possible cross checking was performed, and independent data confirming the correctness of the results presented were used. The following improvements to the system are proposed for the future: 1. Introduction of controlled frame summation for increasing the sensitivity, S/N ratio, and frame rate in cases where there is a margin for resolution. 2. The possibility of transmitting a frame of reduced dimensions will also be utilized to increase the frame rate. 3. The introduction of external digital signal averaging to increase the maximum attainable S/N ratio and to further reduce the intrinsic noise. Reduction of the intrinsic noise, however, is not always accompanied by a significant increase in the possibilities of recording weak signals. Appreciable improvement can occur only for signals recorded at a very low value of the S/N ratio. In cases where the signals must be measured with S/NⰇ1 and not simply detected or an image of sufficiently high quality for investigating fine-structure features of the object must be obtained, fluxes significantly more intense than the threshold counterparts are required. In such cases the shot noise in the photocurrent begins to play a dominant role in the total noise of the camera. To illustrate this, we present 63
J. Opt. Technol. 67 (1), January 2000
the following numerical example. Let the permissible value be S/N⫽5. It follows from Fig. 4 for real and ideal systems operating at K⫽50 that complete elimination of the intrinsic noise permits lowering of the minimum detectable signal by roughly 2 fold in this case. At the same time, if S/N⫽15 is required, the signal can be reduced by only 25%. Since in many applications the requirements for the minimum permissible S/N ratio are even more stringent, the advantage in reducing the intrinsic noise of the camera can turn out to be essentially imperceptible. Until recent times, television systems with brightness amplifiers were widely used to solve problems in recording weakly luminous images. Preliminary amplification of the light flux permits significant reduction of the intrinsic noise of the camera. However, the quantum efficiency of the sensitive elements with an external photoeffect employed in these cameras is several times smaller than the values for solid-state sensors. The loss of quantum efficiency cannot be compensated by decreasing the intrinsic noise of the system when S/NⰇ1. If the preceding example is extended to take into account that the quantum efficiency of the photocathode of a brightness amplifier is 3 times poorer on the average than that of a similar CCD sensor, we find that the systems being compared will have approximately equal possibilities for recording weak signals only when S/N⬇3. In all cases where larger values of the S/N ratio are required, systems with a brightness amplifier will be inferior to the TVIST system, and the disadvantage will be equal in the limit to the loss in quantum efficiency.
1兲
Here e is an electron.
1
V. A. Lisovski, V. V. Shchedrunov, I. Ya. Barski, G. V. Papayan et al., Luminescence Analysis in Gastroenterology 关in Russian兴, Nauka, Leningrad 共1984兲, 234 pp. 2 M. Kriegmair, R. Baumgartner, A. Ehsan et al., ‘‘Detection of early bladder cancer and dysplasia by fluorescence cystoscopy,’’ J. Urol. 共Baltimore兲 153, 457 共1995兲. 3 ‘‘Xillix LIFE-Lung Fluorescence Endoscopy System,’’ in Focus on Life, Olimpus Optical 共1998兲. 4 Light Microscopy in Biology: A Practical Approach, A. J. Lacey 共Ed.兲, IRL Press, Oxford–New York 共1989兲 关Russ. transl., Mir, Moscow 共1992兲, 464 pp.兴. 5 S. Inou’e, Video Microscopy, Plenum Press, New York 共1986兲, 340 pp. 6 J. S. Ploem and H. J. Tanke, Introduction to Fluorescence Microscopy, Oxford University Press, Oxford 共1987兲, 310 pp. 7 W. T. Mason, Fluorescent and Luminescent Probes: A Practical Guide to Technology for Quantitative Real-Time Analysis, Academic Press, New York 共1999兲, 350 pp. 8 N. Tomilin, Yu. Rosanov, V. Zenin, V. Bozhkov, and B. Vig, ‘‘A new and rapid method for visualizing DNA replication in spread DNA by immunofluorescence detection of incorporated 5-iododeoxyuridine,’’ Biochem. Biophys. Res. Commun. 190共1兲, 257 共1993兲. 9 V. A. Tronov and I. I. Pelevina, ‘‘Comet assay of DNA repair for individual cells,’’ Tsitologiya 38, 427 共1996兲. 10 G. V. Papayan, L. S. Agroskin, and I. Ya. Barski, ‘‘Instruments for analytical microscopy,’’ Opt. Zh. 60共12兲, 16 共1993兲 关J. Opt. Technol. 60, 836 共1993兲兴. 11 L. S. Agroski and G. V. Papayan, ‘‘Experience in digital television cytophotometry,’’ Tsitologiya 30, 503 共1988兲. G. V. Papayan and A. A. Mantsvetov
63
12
G. V. Papayan, L. S. Agroskin, and R. M. Larina, ‘‘Evaluating microscope-objective quality on the basis of a television-computer analysis of Airy patterns,’’ Opt. Zh. 62共6兲, 21 共1995兲 关J. Opt. Technol. 62, 357 共1995兲兴. 13 E. G. Stevens, B. C. Burkey, D. N. Nichols et al., ‘‘A 1-megapixel, progressive-scan image sensor with antiblooming control and lag-free op-
64
J. Opt. Technol. 67 (1), January 2000
14
eration,’’ IEEE Trans. Electron Devices ED-38, 981 共1991兲. A. A. Mantsvetov and Sh. Sh. Tsinadze, ‘‘Measurement of the aperturefrequency characteristic of solid-state image converters,’’ in Radioelectronics in the St. Petersburg Electrical-Engineering University, No. 1 关in Russian兴, St. Petersburg Electrical-Engineering University, St. Petersburg 共1995兲, pp. 37–44.
G. V. Papayan and A. A. Mantsvetov
64
