TY - JOUR
AU - Kim, Dong, Hwan
AB - Abstract Noise is a primary issue in obtaining an image in a scanning microscope. This noise needs to be minimized in order to have a clear image of the sample in case of a nanosize level measurement. In this work, we propose a method to improve the image quality by applying dither signal injection to the scanning signal. This method involves minimizing the noise that occurs in scan control circuits, which results in a blurry or distorted image. The collected secondary electrons are first multiplied through a photomultiplier tube and are then converted into digital form using an analog/digital (A/D) converter. We propose a solution for the noise from the scan control circuit that appears on the image by adopting the spread spectrum method. scanning electron microscope, dithering, noise reduction, image improvement, spread spectrum Introduction For a nanosize measurement, scanning electron microscope (SEM) is a well-known primary equipment. The sample image, by the microscope, is obtained by scanning the electron beam emitted from the tungsten filament in horizontal and vertical directions. The image is also obtained, by focusing lenses, using a magnetic field with an electron beam emitted from the filament. To implement this acquiring image, sophisticated electric circuits are designed. With these circuits, many kinds of noise occur even if the correct noise source is clearly identified. This noise is very crucial in obtaining an image, particularly in a high-resolution mode. This work proposes a method to improve the quality of the sample image obtained by an SEM with the help of schemes developed for enhancing the image quality. When the focused electron beam is injected into the sample, the image is obtained by secondary electrons from the sample, and a secondary electron detector (SED) is used as a detector. While obtaining the image, a large amount of noise is involved. This noise can largely stem from its inherent sources, such as coils, scanning process, analog-to-digital (A/D) conversion process, vacuum pressure and vibration. The noise can be minimized using different methods, including digital filters [1], image processing [2] and a high-speed A/D converter. However, all of these methods have issues such as the high price of the equipment and the increasing execution time required for obtaining the sample image. We propose a way of minimizing the noise by avoiding the problems of the above issues by using a dither signal [3] imposition rather than electric circuit modification. We have studied a method to improve the image quality by using dither to minimize the noise. The dither signal is a kind of external signal with extremely high frequency and very small amplitude. Here, we propose a way of minimizing the noise by allowing a dither signal and adopting the spread spectrum [4,5] method. Obtaining an image from an SEM An SEM uses electrons that are emitted from a tungsten filament. These electrons are focused by electromagnetic lenses, being changed to electron beam to intensify injection energy onto the sample by a series of electromagnetic lenses. To obtain an image from the SEM, the electron beam needs to be scanned in the horizontal and vertical directions with high speed. The electron beam is injected onto the sample which in turn generates secondary electrons. These secondary electrons are collected at the photomultiplier tube (PMT) and finally realized as an image using the intensity level. Figure 1 shows the structure of an SEM equipment. An electron beam is emitted from the gun, and magnetic fields are created from the electromagnetic lens located in a serial manner, focusing the electron beam to reduce the beam size. The focused electron beam interacts with the sample and scans using scanning deflection coils. When the electron beam interacts, secondary electrons emit from the sample. The secondary electrons are then collected in the SED, multiplied through the PMT, digitized through the A/D converter and finally expressed as an image intensity for the electron beam-injected location on the sample. During the acquisition of intensity data through an A/D converter, noise occurs, thus distorting the image. Fig. 1. Open in new tabDownload slide An SEM system. Fig. 1. Open in new tabDownload slide An SEM system. To remove white noise, dither signal is intentionally imposed onto the analog input circuit between the PMT and the A/D converter [6]. By reducing the distortion and quantization error, spurious free dynamic range is improved, and it is possible to minimize the white noise in the A/D process. Spread spectrum using dither Even if the white noise can be reduced by adopting the dither signal to the A/D converter, there still other noise involved. To check the sample image, the electron beam should be moved onto the X- and Y-axes using the magnetic field. While performing this scanning process, noise usually comes in from the control circuit, an intrinsic noise source. An intrinsic noise is caused by electrical parts and can be largely classified into four types: thermal noise, short noise, contact noise and popcorn noise [7]. Of these four types, thermal noise, short noise and popcorn noise can be minimized by choosing an appropriate electrical device, as they are specific phenomena of transistors, diodes, semi-conductor devices and integrated circuits. However, contact noise is created as a result of an incomplete contact between two conductors, and the relevant fluctuation in their conductivity. This noise is also referred to as 1/f noise because its noise power is inversely proportional to its frequency [7]: (1) where If is the contact noise current, IDC the average contact noise current, K the integer designated by the material and the shape, B the bandwidth centered at the frequency of contact noise and f the scan frequency. The noise current per unit bandwidth can approximately be expressed as in Eq. (1). According to this formula, the contact noise and the frequency are inversely proportional to each other. While it is not possible to calculate the level of contact noise when the integer K is unknown, the level of noise increases as the frequency decreases and becomes an important source of noise at low frequencies. This noise can be resolved using spread spectrum [6–8]. Spread spectrum is a technology used in communications, which provides security by increasing the bandwidth by inserting a high-frequency signal into the base-band signal bandwidth. By adopting this method, scan signal bandwidth is increased and contact noise can be minimized. Also, the spectrum of a certain frequency band is expanded over the broadband [9] at the same time, and therefore the noise occurring in the scan control circuit is minimized. Noise in scan system To estimate the noise that occurs during scanning, the image is checked without emitting the electron beam, and frequency analysis is carried out. Figure 2 shows several images obtained experimentally by changing only vertical scan frequency without emitting an electron beam. As evident from the results, the noise in the image varies based on the scan frequency. It is understood that the noise has a relatively high value of ∼30 Hz and the frequency changes as the scan frequency increases, which shows the characteristics of white noise. Also, the magnitude of the noise level has a tendency to decrease as the scanning frequency increases, which indicates that the noise in the image is a contact noise. This prediction of the noise type is proved by the fact that the image in (a) is worse than the image in (d), which results from the increased scanning frequencies. With this experiment, it is possible to understand that the noise in the image depends on the frequency, which exemplifies contact noise which is inversely proportional to the frequency as in Eq. (1). Fig. 2. Open in new tabDownload slide An SEM image obtained by changing vertical-axis scanning frequency without the electron emission: (a) 1 kHz, (b) 1.3 kHz, (c) 1.6 kHz and (d) 2 kHz. Fig. 2. Open in new tabDownload slide An SEM image obtained by changing vertical-axis scanning frequency without the electron emission: (a) 1 kHz, (b) 1.3 kHz, (c) 1.6 kHz and (d) 2 kHz. Frequency analysis by the fast Fourier transform (FFT) is shown in Fig. 3, which uses the Y-axis pixel data obtained in Fig. 2 by selecting 512 data points. As shown in the graph, the magnitude of the noise signal decreases as the scan frequency increases. Therefore, the noise in the image is affected by the frequency signal, which ensures that the noise in the SEM image is a contact noise. However, the noise around 30 Hz is not reduced much when compared with other signals corresponding to other frequencies, but still affects the image. On the other hand, 30 Hz could be understood as a power noise. However, the power used in the present system has a switching frequency in the range of ten to hundreds of kilo Hertz, thus the 30 Hz is estimated as a contact noise rather than a power noise. Fig. 3. Open in new tabDownload slide Frequency analysis of the acquired SEM image without emission: (a) 1 kHz, (b) 1.3 kHz, (c) 1.6 kHz and (d) 2 kHz. Fig. 3. Open in new tabDownload slide Frequency analysis of the acquired SEM image without emission: (a) 1 kHz, (b) 1.3 kHz, (c) 1.6 kHz and (d) 2 kHz. Minimizing scan noise By using a data acquisition board, reference signal for the scan is created, and the current is allowed to coil through a multiplier circuit. This creates a magnetic field, making the electron beam be deflected according to the scan and finally injected onto the sample. Figure 4 shows the scan circuit [10] where the dither signal is applied. The dither signal is a kind of signal used to adopt a spread spectrum scheme, which adds a high-frequency signal to the true signal bandwidth. With this signal spread, the noise level can be reduced and can become robust to disturbance. In this work, we adopt the spread spectrum scheme, yielding to noise reduction and moreover contributing to eliminating contact noise the power of which is inversely proportional to the input signal frequency as shown in Eq. (1). The amplitude of this signal must be of a size that does not directly influence the image. The rule of determining a dither frequency [11,12] is related to the sampling frequency. As a result, we chose 1 MHz as the dither frequency which is equal to the sampling frequency. Therefore, the test was carried out by allowing a dither signal with an amplitude of 1.3 mV and a frequency of 1 MHz into the scanning signal of the X-axis. On the other hand, the magnitude of the scan is in the level of 100 mV and its frequency ranges from 1 to 7 kHz. The reason for selecting 1 MHz as a dither frequency is that the maximum sampling frequency of the digital-to-analog converter (DAC) in the present SEM is 1 MHz. To see the dither effect more clearly, the maximum sampling frequency of the DAC was chosen as a dither frequency. Of course, a smaller frequency than the maximum DAC frequency can be selected for the dither frequency for the dither test: (2) where X(t) is the total signal of x-axis scan and a dither signal, Xsawtooth(t) the sawtooth scan wave of the x-axis, Ad the scale factor of a dither signal (1.3 mV), Ax the scale factor of a sawtooth wave (0–100 mV), fd the frequency of a dither signal (1 MHz), fx the frequency of a x-scan signal (1–7 kHz) and k the harmonic factor of the x-axis. Fig. 4. Open in new tabDownload slide A scanning circuit with dither addition. Fig. 4. Open in new tabDownload slide A scanning circuit with dither addition. The final input signal on the scan coil including the dither signal is as shown in Eq. (2). Spread spectrum test and result Since the prime noise for the SEM is a contact noise and its noise level is inversely proportional to the exciting frequency, as shown in Eq. (1), the noise can be reduced by imposing a high-frequency signal to the original scan signal. This inspires the dither signal imposition which is of high frequency but a relatively small amplitude signal, to the original scan signal. The image acquisitions with and without the dither signal were tested. The dither signal was allowed in the scan X-direction. Figure 5 shows an image when no dither was applied. The accelerating power is 15 kV with a magnification of 20 000. It can be seen that there is noise in the sample. The noise is more visible in the boundary than at the center of the sample. Fig. 5. Open in new tabDownload slide An acquired SEM image without dither. Fig. 5. Open in new tabDownload slide An acquired SEM image without dither. Figure 6 shows the result of the FFT to analyze the noise. As a result, a signal of 30 Hz is still included with a high magnitude. Fig. 6. Open in new tabDownload slide Frequency analysis of the acquired SEM image without dither. Fig. 6. Open in new tabDownload slide Frequency analysis of the acquired SEM image without dither. Figure 7 is an image when the dither signal is applied. The noise of 30 Hz which appeared to be a ripple on the boundary in the sample image has faded. Fig. 7. Open in new tabDownload slide An acquired SEM image with dither. Fig. 7. Open in new tabDownload slide An acquired SEM image with dither. Figure 8 shows the FFT graph of the image data where the dither signal is allowed. The graph shows that the signal of 30 Hz is spread into a wideband. Therefore, an almost uniform spectrum is distributed over the whole frequency band, and the noise signal of 30 Hz has faded in the sample image. In order to prove the performance upon dither signal addition, a different sample specimen was utilized. The specimen consists of circular-shaped small balls. Fig. 8. Open in new tabDownload slide Frequency analysis of the acquired SEM image with dither. Fig. 8. Open in new tabDownload slide Frequency analysis of the acquired SEM image with dither. Similar to the case of mesh-type specimens, the boundary of the ball has a fluctuating noise in the horizontal direction when no dither is applied, as shown in Fig. 9a. On the other hand, the addition of the dither signal contributes to eliminating this noise, which is shown in Fig 9b. Fig. 9. Open in new tabDownload slide A resultant image of the ball-shaped sample: (a) image with contact noise without the dither signal and (b) image with dither signal imposition. Fig. 9. Open in new tabDownload slide A resultant image of the ball-shaped sample: (a) image with contact noise without the dither signal and (b) image with dither signal imposition. Conclusion A method to improve the sample image in an SEM is proposed. To reduce the contact noise in the sample image, a dither signal is used. To eliminate this noise, a dither signal is allowed into the input terminal of the scan deflector coil. The amplitude of this dither signal should be of a level that would not influence the image. Contact noise is created and appears on the image due to devices used in an SEM controller. Images due to frequency have different shapes, and noise intensity is inversely proportional to the frequency, which is a typical characteristic of the contact noise. Thus, it is concluded that the primary noise in an SEM is a contact noise. The other noise type is not shown in Fig. 3 because these noise frequencies are far from the frequency we consider in this study. As this noise has a relationship that is inversely proportional to the frequency, it is proposed to spread the spectrum using a dither signal as a solution. The spectrum is spread by allowing a dither signal into the scan circuit, which injects an electron beam into the sample. In this way, the level of the contact noise is minimized and removed from the image. Funding This work was supported by the Seoul R&BD Program (Grant No. 10583). References 1 Tanaka K , Nakamura Y , Matsui K . Restoration of grey level picture with less blur using local information, Visual Communication and Image Processing'92 , SPIE , 1992 , vol. 1818 (pg. 1408 - 1413 ) 10.1117/12.131412 OpenURL Placeholder Text WorldCat Crossref 2 Kim J G , Doo I H , Kyung J Y . Performance of digital matched filter for band-limited white noise , IEEK Conf. , 1992 , vol. 15 (pg. 53 - 56 ) OpenURL Placeholder Text WorldCat 3 Jung Y G , Na S H , Lim Y C , Yang S H . Reduction of audible switching noise in induction motor drives using random position PWM , IEE Proc. Electric Power Appl. , 2002 , vol. 149 (pg. 195 - 202 ) 10.1049/ip-epa:20020244 Google Scholar Crossref Search ADS WorldCat Crossref 4 Cichowski A , Nieznanski J , Wojewodka A . Shaping the SPL spectra of the acoustic noise emitted by inverter fed induction motors , 2003 Proceedings on IEEE IECON'03, Roanoke, Virginia, USA, November (pg. 2923 - 2928 ) 5 Hsu F , Giordano A A . Digital whitening techniques for improving spread spectrum communication performance in the presence of narrow band jamming and interference , 1978 IEEE Transactions on Communication (COM-26) (pg. 209 - 216 ) February 6 Jung K O , Kim D H . A dither signal imposition to enhance an image in a scanning electron microscope. , Microelectron. Eng. , 2011 , vol. 88 (pg. 2601 - 2603 ) Google Scholar Crossref Search ADS WorldCat 7 Danimoto S . Op-amp practical skills , 1988 SEUN Ltd., pp. 52–60 (Korean translated). OpenURL Placeholder Text WorldCat 8 Utlant F . Spread spectrum principles and possible application to spread spectrum utilization and allocation. , IEEE Commun. Soc. Mag. , 1978 (pg. 21 - 30 ) 10.1109/MCOM.1978.1089761 OpenURL Placeholder Text WorldCat Crossref 9 Trzynadlowski A M , Blaabjerg F , Prdersen J K , Kirlin R L , Legowski S . Random pulse width modulation techniques for converter-fed drive system – a review. , IEEE Trans. Ind. Appl. , 1994 , vol. 30 (pg. 1166 - 1174 ) 10.1109/28.315226 Google Scholar Crossref Search ADS WorldCat Crossref 10 Kim D H , Kim S J , Oh S K . Image improvement with a modified scanning waves and noise reduction in a scanning electron microscope , Nucl. Instrum. Meth. Phys. Res. A , 2010 , vol. 620 (pg. 112 - 120 ) 10.1016/j.nima.2010.03.134 Google Scholar Crossref Search ADS WorldCat Crossref 11 Bae C H , Ryu J H , Lee K W . Suppression of harmonic spikes in switching converter output using dithered sigma–delta modulation , IEEE Trans. Ind. Appl. , 2002 , vol. 38 (pg. 159 - 166 ) 10.1109/28.980370 Google Scholar Crossref Search ADS WorldCat Crossref 12 Trzynadlowski A M , Borisov K , Li Y , Qin L . A novel random PWM technique with low computational overhead and constant sampling frequency for high-volume, low-cost applications , IEEE Trans. Power Electron. , 2005 , vol. 20 (pg. 116 - 122 ) 10.1109/TPEL.2004.839824 Google Scholar Crossref Search ADS WorldCat Crossref © The Author 2011. Published by Oxford University Press [on behalf of Japanese Society of Microscopy]. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
TI - Scanning electron microscope image enhancement using spread spectrum through dither signal imposition
JO - Journal of Electron Microscopy
DO - 10.1093/jmicro/dfr074
DA - 2011-12-01
UR - https://www.deepdyve.com/lp/oxford-university-press/scanning-electron-microscope-image-enhancement-using-spread-spectrum-vPC3EC6280
SP - 367
EP - 373
VL - 60
IS - 6
DP - DeepDyve
ER -