Adaptive bias voltage driving technique of uncooled infrared focal plane array

Adaptive bias voltage driving technique of uncooled infrared focal plane array

G Model IJLEO-53076; No. of Pages 4 ARTICLE IN PRESS Optik xxx (2013) xxx–xxx Contents lists available at SciVerse ScienceDirect Optik journal home...

563KB Sizes 6 Downloads 164 Views

G Model IJLEO-53076; No. of Pages 4

ARTICLE IN PRESS Optik xxx (2013) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Optik journal homepage: www.elsevier.de/ijleo

Adaptive bias voltage driving technique of uncooled infrared focal plane array Xiubao Sui ∗ , Qian Chen, Guohua Gu Jiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, NUST, Nanjing 210094, China

a r t i c l e

i n f o

Article history: Received 16 August 2012 Accepted 2 January 2013 Keywords: IRPFA Adaptive bias voltage Response dynamic range of IRFPA Scene temperature range

a b s t r a c t In order to have the ability of detecting the infrared scene with different temperature range, the bias voltage of uncooled infrared focal plane array (IRFPA) is usually fixed. While, under this condition two situations may occur. One is that the dynamic range of the scene is far bigger than IRFPA, and then the response of IRFPA is not the real scene gray. The response of IRFPA at some scene temperature may be cut off. The other is that the dynamic range of the scene is far smaller than IRFPA, and then most response dynamic range is wasted which reduces the ability of distinguishing details. In order to solve this problem, we brought forward the adaptive bias voltage driving technique of uncooled IRFPA in this manuscript. We computed the real scene temperature range according to the response of IRFPA and current bias voltage, set the adjustable threshold of bias voltage and computed the matching bias voltage. Adopting the technique, we can guarantee that the response dynamic range is accordant with the real scene continuously. The theory derivation and experiment demonstrate that the technique is effective in making the scene temperature dynamic rang and response dynamic range of IRFPA be consistent. And the technique can increase the IR image detai1 resolution by a big margin. © 2013 Elsevier GmbH. All rights reserved.

1. Introduction Uncooled infrared focal plane array (IRFPA) usually has an adjustable response dynamic range and it can be set by its own bias voltage. But, most of the thermal imagers do not make use of the characteristic of the uncooled IRFPA. In order to have the ability of detecting the infrared scene with different temperature range, the infrared (IR) thermal imaging system is often set to a relatively large dynamic range. This approach dose not fully plays the role of the IRFPA. For instance, the nominal detecting range of the DL700 uncooled thermal imaging system produced by the DALI-TECH of China is 20 ◦ C–400 ◦ C [1]. The 320 × 240 uncooled focal plane detector of ULIS cooperation has a detecting range of −10 ◦ C to +50 ◦ C. However, the real target scene can hardly cover the whole temperature dynamic range of either DL700 or the 320 × 240 uncooled focal plane detector. It would be a waste to the dynamic range of both the focal plane detector and the analog to digital conversion (AD) chip if the thermal imaging system cannot adjust its temperature dynamic range accordingly to the change of the temperature of the target scene. As known, the bigger dynamic temperature range the thermal imaging system can detect, the lower temperature resolution the images have. So, the approach of fixing bias voltage of IRFPA will directly result in a low temperature resolution of the IR images. Although some software algorithms which attempt to stretch the IR images by means of combining the grade

∗ Corresponding author. E-mail address: [email protected] (X. Sui).

levels has been developed, such as the histogram equalization algorithm based on Gray-level Redundancy [2,3], the performance of output IR images is visually improved compared to the original one when the algorithm is applied. Actually, the software methods do not increase the image information. The temperature details have lost before the histogram equalization algorithm is performed. The effect of the algorithm is only to stretch the IR images to the whole gray range ponderously. No more information is got except for the visual improvement. Also, the algorithm cannot improve the temperature resolution of the IR images. For the 320 × 240 uncooled focal plane detector of ULIS cooperation, the VFID and VEBASAGE parameters can adjust the relative dynamic range and absolute response temperature of the detector, this makes the IRFPA detector can adaptively adjust the two offset voltages denoted as VFID and VEBASAGE [4], so that it becomes possible to fit the temperature response range of the detector with the temperature range of the infrared scene. In this manuscript, we will introduce the adaptive bias voltage technique by the 320 × 240 uncooled focal plane detector. The remainder of this manuscript is organized as follows. In Section 2, the adaptive bias voltage driving technique is presented in detail; in Section 3, the performance of the proposed technique is evaluated; the conclusions is finally summarized in Section 4. 2. Technique description Adaptive digital technique of IRFPA [5] is adaptive to change the AD chip bias voltage to ensure that the analog signal input to the AD chip can be “full-scale digital output” according to the temperature

0030-4026/$ – see front matter © 2013 Elsevier GmbH. All rights reserved. http://dx.doi.org/10.1016/j.ijleo.2013.03.026

Please cite this article in press as: X. Sui, et al., Adaptive bias voltage driving technique of uncooled infrared focal plane array, Optik Int. J. Light Electron Opt. (2013), http://dx.doi.org/10.1016/j.ijleo.2013.03.026

G Model

ARTICLE IN PRESS

IJLEO-53076; No. of Pages 4

X. Sui et al. / Optik xxx (2013) xxx–xxx

2

Digital

the factors above will introduce non-uniformity to IRFPA. The video output non-uniformity of the IRFPA is caused by comprehensive factors such as IRFPA readout circuit, semiconductor characterization and amplifying circuit [6,7], and we cannot eliminate them completely using non-uniformity correction algorithm. So, shying away the non-uniformity is a good way. The non-uniformity of different pixels in the IRFPA will result that field-effect-tube (FET) threshold voltage, Vcons and Cint are different, namely, the parameters in Eq. (1) will be different for different pixels. So, the dynamic range of IRFPA needs to be adjusted on the basis of the typical dynamic range, i.e.:

driving

signal Analog IR

bias Digital

voltage Signal

FP A

AD

Temperature control system

FPGA

9240

signal output

filter

Analog

output

signal Bias voltage 1

Bias voltage 2

Serial D/A Chip

Fig. 1. Adaptive bias voltage driving technique block diagram.

range of the infrared scene. The focal plane detector also has a response dynamic range to the infrared scene. So if we can adaptively adjust the dynamic range of the detector, its performance will be given full play. Reference [4] shows that the mathematics model of 320 × 240 uncooled IRFPA of the ULIS Company is: Vsamp = Vcons −

+

1 Cint



(VEBASAGE − VTH EBASAGE ) × tint Rb

(VFID − VTH FID )C × R(Ts )˛(kpv − G)c1











exp −˛

− exp −˛ KT (0) P0 + Tcons − Ts



KT (0) P0 + Tcons − Ts + c1



kpv − G tint C



(1)

where, Vcons is the positive terminal voltage of the amplifier in the readout circuit (ROIC); Cint is the integration capacitor of the; tint is the integration time of ROIC; Rb is the blind resistance value; C is the heat capacity of the pixel; VTH EBASAGE is the threshold voltage of the FET whose gate voltage is VEBASAGE ; VTH FID is the threshold voltage of the field effect tube (FET) whose gate voltage is VFID ; Ts is substrate temperature of microbolometer; R(Ts ) is the resistance value of the pixel with its temperature; ˛ is the temperature coefficient of the pixel; kpv , c , KT(0) and Tcons are different constants separately; G is the equivalent thermal conductivity of the pixel to environment; P0 is power amplitude of modulated infrared light; VFID and VEBASAGE are the bias voltage of the detector. VFID and VEBASAGE determine the dynamic range of the analog output signal and the direct-current voltage of output signal respectively. So if we adjust these two voltages according to the temperature change range of the infrared scene, we can ensure the maximum utilization of the IRFPA’s temperature dynamic range and improve temperature resolution of the thermal imager. In the implementation process of adaptive bias voltage driving technique, for each pixel, the other parameters in Eq. (1) keep unchanged except for VFID and VEBASAGE . Bias voltage VFID and VEBASAGE supported to the IRFPA are offered by the serial digitalto-analog (D/A) chip in Fig. 1. The same serial D/A chip also offers bias voltages to AD chip (AD9240). In order to avoid the interference of blind pixels, we need exclude the blind pixels before adjusting the dynamic range of IRFPA, namely, all pixels that involved in calculating are the ones which are not blind pixels. In addition to blind pixels, the non-uniformity of IRFPA is another factor that we need to consider before adjusting the dynamic range of IRFPA. Because of the non-uniformity of semiconductor material (for example, the non-uniformity of concentration of impurities, crystal defects and the non-uniformity of internal structure, etc.), mask error, defect and craft limitations, the output amplitude of every non-blind pixel is not the same when the IRFPA faces a uniform blackbody. And then

Vvariable = (1 − knonuniformity )(Vsortie

max

Vverify

max

= (1 − knonuniformity )Vsortie

max

Vverify

min

= (1 − knonuniformity )Vsortie

min

− Vsortie

min )

(2) (3) (4)

where Vvariable is the effective dynamic range of IRFPA in the actual calculation; Vverify max and Vverify min are the maximum and minimum output values of IRFPA separately; knonuniformity is the largest non-uniformity coefficient of IRFPA; Vsortie max is the maximum output voltage value of IRFPA; Vsortie min is the minimum output voltage value of IRFPA. When we compute the dynamic range of IRFPA by the parameters provided by ULIS corporation, Eq. (2) can ensure that the output voltage value will not overflow or be cut off because of the non-uniformity. When the effective dynamic range of IRFPA is fixed and the blind pixels are excluded, the following steps should be carried out: (1) calculate the maximum temperature Tview max and minimum temperature Tview min in the infrared scene according to Eq. (1), the parameters provided by ULIS corporation, the parameters of infrared optical system, the output voltage of IRFPA and VEBASAGE , VFID and Cint . (2) Calculate the maximum temperature Tmax and minimum temperature Tmin that IRFPA can detect according to Eqs. (1), (3), (4) and VEBASAGE , VFID which are adopted in the current system. In the calculation process, Tmax is corresponds to Tview max in Eq. (3) and Tmin is corresponds to Tview min in Eq. (4). (3) Compare Tview max , Tmax , Tview min and Tmin , separately. If



Tmax − 2T ≤ Tview Tmin + T ≤ Tview

max

min

≤ Tmax − T

≤ Tmin + 2T

(5)

is satisfied, we need not adjust the values of VEBASAGE and VFID , because Eq. (5) indicates that the dynamic range of IRFPA is suitable. T in Eq. (5) is the threshold temperature need to be manually set and should be set according to the noise level of the circuit system of thermal imager. If we do not set threshold temperature T, similar with “adaptive digital technique of IRFPA” [5], the system noise will result that Tview max is equal to Tmax at some time. Even if the maximum scene temperature Tview max may be bigger than Tmax , Tmax is the maximum temperature that can be calculated according to VEBASAGE and VFID and so the maximum Tview max cannot exceed Tmax . The system noise will result that Tmax is smaller to Tmax at other time. At last, the system noise will result that the thermal imager adjusts VEBASAGE and VFID constantly and the infrared images do not keep stable. If Eq. (5) cannot be satisfied, it illustrates that the dynamic range of IRFPA may be much smaller or much bigger than the dynamic range of actual infrared scene. At this situation, VEBASAGE and VFID should be adjusted and satisfy Eq. (6):



Tview

max

= Tmax − 1.5T

Tview

min

= Tmin + 1.5T

(6)

When we adjust Tview max and Tview min according to Eq. (6), Eq. (5) can be satisfied and also the scene temperature Tview max and Tview min are set in the middle area of the dynamic range of IRFPA.

Please cite this article in press as: X. Sui, et al., Adaptive bias voltage driving technique of uncooled infrared focal plane array, Optik Int. J. Light Electron Opt. (2013), http://dx.doi.org/10.1016/j.ijleo.2013.03.026

G Model

ARTICLE IN PRESS

IJLEO-53076; No. of Pages 4

X. Sui et al. / Optik xxx (2013) xxx–xxx

3

Fig. 2. The thermal imager performance testing system from CI Company.

So, Eq. (6) can make that VEBASAGE and VFID need not be adjusted frequently when the scene temperature dynamic range has small changes. This will keep the infrared images stable for a long time. If the dynamic range of IRFPA which has been set is not enough to the scene temperature dynamic range, Tview max and Tview min which are computed according to VEBASAGE and VFID are not the accurate peak values of infrared scene and the accurate Tview max and Tview min should be fixed again. The method is to set VEBASAGE and VFID as the value that the test report of IRFPA provides. Regulate integration time and frame time of IRFPA to make IRFPA has a normal analog output when facing a uniform reference with environment temperature. Compute Tmax and Tmin that IRFPA can detect according to Eqs. (1), (3) and (4) and then get Tview max and Tview min of the infrared scene. If Eq. (7) is established,



Tview

max

< Tmax

Tview

min

> Tmin

(7)

Tview max and Tview min are the true reflections of the infrared scene temperature peak value and then record this value. If Eq. (7) cannot be established, we should regulate VEBASAGE and VFID according to Reference [4] until Eq. (7) can be established and then record Tview max and Tview min . (4) Combine Tview max , Tview min with Eq. (6) and the transformational equation of Eq. (1) [4], then we can get:

⎧ ⎪ ⎪ ⎪ ⎪ Vverify ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ Vverify



max

min

1 = Vcons − Cint

lP01 + mp01 ×



1 = Vcons − Cint

lP01 + mp01 ×

exp(np01 (Tview

max

4

c2 + c2 (Tview exp(np01 (Tview

min

max

+ 1.5T )

4

max

4

+ 1.5T ) + np02 )

c2 + c2 (Tview

4

2

2

Ts ) np01 = −(1/4)˛KT (0) As K × (2a/f  ) × ˇp2 /(ˇp − ˇ) − ˛c2 (kpv − G/C)tint ; np02 = −˛(Tcons − Ts ) − ˛c2 (kpv − G/C)tint ; , a, f , ˇp , ˇ, As are all optical parameters; K and c2 are constants. Solve Eq. (8), we can get adaptive VEBASAGE and VFID according to infrared scene temperature. Adaptive bias voltage driving technique is achieved when VEBASAGE and VFID are set by serial D/A chip. 3. Technique performance verification The adaptive bias voltage driving technique improves the temperature resolution of the infrared images, namely, the MRTD of the infrared images can be improved. The histogram equalization algorithm based on Gray-level Redundancy only processes the output digital signals, increases no original infrared data

min

− 1.5T )



4

− 1.5T ) + np02 ) − exp(np01 (TE min − 1.5T ) + np02 )

−(1/4)˛KT (0) As K × (2a/f  ) × ˇp2 /(ˇp − ˇ) ; np02 = −˛(Tcons − 2

theoretically. Of course it cannot improve the MRTD of the infrared detectors. In this section, non-uniformity correction algorithm, histogram equalization algorithm based on Gray-level Redundancy after nonuniformity correction algorithm and adaptive bias voltage driving technique after non-uniformity correction algorithm will separately process the infrared images. The processed images will be tested by the thermal imager performance testing system made by CI Company from Israel. We use the method to indicate objectively that adaptive bias voltage driving technique improves the MRTD of the infrared images apparently. And next, we will compare the images processed by the techniques above to offer subjective verification. The infrared detector performance testing system is showed in Fig. 2. The hardware platform includes CI testing instrument, optical

+ 1.5T ) + np02 ) − exp(np01 (Tview

lP01 = (VEBASAGE − VTH EBASAGE ) × tint /Rb; where: mp01 = (VFID − VTH FID )C/R(Ts )˛(kpv − G); np01 = 2

Fig. 3. Images processed by different techniques. (a) Processed by non-uniformity correction algorithm only, (b) processed by histogram equalization algorithm based on Gray-level Redundancy after non-uniformity correction algorithm, (c) processed by adaptive bias voltage driving technique after non-uniformity correction algorithm.



(8)

4

bench, collimator, trestle, high performance computer, video connector, radiation source controller, area blackbody, thermal imager and four rod target. The principle is: the thermal imager aims at the four rod target of testing system, which imaged by the thermal imager. The video signal will be sent to the computer of the testing system and be analyzed and calculated. The MRTD at different frequencies will finally be computed with the temperature differences between the four rod target and background. The average values of five times measuring the MRTD with different techniques are listed in Table 1. Comparing the data, the MRTD value is almost the same processed by histogram equalization algorithm based on Gray-level Redundancy and only non-uniformity correction algorithm at the same spatial frequency. Although there is a difference of 0.01 K, it is not apparent. And the difference of 0.01 K is within the system noise error range. So, we think that the histogram equalization algorithm based on Gray-level Redundancy cannot increase the temperature resolution of images. When we compare the MRTD value between the histogram equalization algorithm based on Gray-level Redundancy and

Please cite this article in press as: X. Sui, et al., Adaptive bias voltage driving technique of uncooled infrared focal plane array, Optik Int. J. Light Electron Opt. (2013), http://dx.doi.org/10.1016/j.ijleo.2013.03.026

G Model IJLEO-53076; No. of Pages 4

ARTICLE IN PRESS X. Sui et al. / Optik xxx (2013) xxx–xxx

4 Table 1 MRTD values of infrared images process by different techniques. Infrared images processed by different techniques

Non-uniformity correction algorithm Histogram equalization algorithm based on Gray-level Redundancy after non-uniformity correction algorithm Adaptive bias voltage driving technique after non-uniformity correction algorithm

MRTD

0.15 cy/mRad

0.33 cy/mRad

0.81 cy/mRad

1.33 cy/mRad

0.23 K

0.289 K

0.577 K

0.869 K

0.22 K

0.288 K

0.577 K

0.868 K

0.13 K

0.189 K

0.458 K

0.749 K

adaptive bias voltage driving technique, we can find that the MRTD of the images processed by adaptive bias voltage technique is much lower than the histogram equalization algorithm based on Graylevel Redundancy. So we can get a conclusion that the adaptive bias voltage driving technique can increase the temperature resolution of thermal imager. The conclusion is consistent with the theoretical analysis. Fig. 3 are the infrared images processed by the three techniques mentioned above. Compare the images we can find that visually the image quality of Fig. 3(c) is worse than Fig. 3(b). The result indicates that histogram equalization algorithm based on Gray-level Redundancy can improve visual effect. But compare Fig. 3(b) and (c), it is obvious that the hair detail in Fig. 3(c) is better than in Fig. 3(b) and it indicates that the adaptive bias voltage driving technique improve the temperature resolution of images. The objective data and subjective images are consistent. 4. Conclusion Aiming at the characteristics that the temperature resolution of thermal imager is not high, we analyzed the imaging theory of thermal imager. Based on the imaging theory and the mathematics model of IRFPA, we brought forward the adaptive bias voltage driving technique in this manuscript. The technique can adjust the temperature dynamic range of IRFPA adaptively according to infrared scene temperature range and maximize the temperature resolution performance of IRFPA. The test result indicates that the

technique is effective in improving the temperature resolution of infrared images. Acknowledgments The work was supported by Jiangsu Natural Science Foundation (BK2011698); Specialized Research Fund for the Doctoral Program of Higher Education of China (20113219120017); China Postdoctoral Science Foundation (20110491424, 2012T50478); Jiangsu Planned Projects for Postdoctoral Research Funds (1101081C); NUST Research Funding (2011XQTR01, 2011YBXM74). References [1] The DL700C Operating Instructions, Zhejiang Dali Technology Co., Ltd., China, 2004. [2] Zhang Xiao, DongYan Xue, ZhaoWen Juan, Digital Image Processing Technology, Metallurgical Industry Press, Beijing, 2005, pp. 100–120. [3] Zhu Xiu Chang, Liu Feng, Hu dong, Digital image processing and image communication, Beijing University of Posts and Telecommunications Press, Beijing, 2002, pp. 156–180. [4] Xiubao Sui, Research of uncooled staring thermal imager imaging theory and key technology, Doctoral Dissertation, Nanjing University of Technology and Engineering, 2009. [5] Xiubao Sui, Qian Chen, Lu. Honghong, Adaptive digital technique of IR FPA, Acta Armamentarii 29 (5) (2008) 548–551. [6] Liu Zhi Cai, Li Zhi Guang, Review of the infrared camera image processing technology, Infrared Technol. 22 (6) (2000) 27–32. [7] Sui Xiu Bao, Chen Qian, Gu Guohua, Research of the environmental temperature’s effect on infrared image non-uniformity, Acta Photon. Sin. 37 (12) (2008) 2572–2575.

Please cite this article in press as: X. Sui, et al., Adaptive bias voltage driving technique of uncooled infrared focal plane array, Optik Int. J. Light Electron Opt. (2013), http://dx.doi.org/10.1016/j.ijleo.2013.03.026