34(4)

76
COLOUR ANALYSIS OF FRUITS USING MACHINE VISION SYSTEM FOR AUTOMATIC SORTING AND GRADING P. Sudhakara Rao, A. Gopal, R. Revathy and K. Meenakshi CEERI Chennai, CSIR Madras Complex, Taramani, Chennai - 600 113 ABSTRACT Food and other biological products are valued by their appearance. Appearance is a major factor in the judgment of quality and human eye has historically done this. The colour indicates parameters like ripeness, defects, etc. The quality decisions vary among the graders and often inconsistent. The adaptation of human eye to small changes in colour and the effect of the background on the perceived colour and colour intensity are the main sources of error. Hence, it is hard to provide precise guideline for visual inspection of fruits based on colour. In biological products, the light varies widely as a function of wavelength. These spectral variations provide a unique key to machine vision and image analysis. Machine vision technology offers the solution for all these problems. Of the many available colour models, HSI model provides a highly effective colour evaluation particularly for analyzing biological products. Using an RGB colour camera, it is also possible to obtain the colour information. However since the RGB colour model represents the image in three separate coordinates RED, GREEN and BLUE, it is not efficient for colour perception and image processing than compared to HSI model, where ‘H’ alone gives the colour perception. Grading of large number of fruits create a problem related to large variations among the population of fruits. This problem is addressed by using the tools provided by the statistical methods. The authors developed an on-line apple grading system, partially sponsored by the ministry of food processing, government of India, based on some of the most important external parameters including the fruit’s surface colour. The details including the experimental results and techniques of implementation of colour grading are presented in this paper. 1. INTRODUCTION Colour provides valuable information in estimating the maturity and examining the freshness of fruits & vegetables. Colour is one of the most significant criteria related to fruit quality. It indicates the parameters like ripeness, defects, etc.[1-5] The appearance of the fruit, affects the consumer acceptance and the value addition. Specially picked quality graders have usually performed this visual inspection. The quality decisions vary among the graders and are inconsistent. The adaptation of human eye to small changes in colour and the effect of the J. Instrum. Soc. India 34 (4) 284-291

Transcript of 34(4)

Page 1: 34(4)

COLOUR ANALYSIS OF FRUITS USING MACHINEVISION SYSTEM FOR AUTOMATIC SORTING

AND GRADING

P. Sudhakara Rao, A. Gopal, R. Revathy and K. MeenakshiCEERI Chennai, CSIR Madras Complex, Taramani, Chennai - 600 113

ABSTRACT

Food and other biological products are valued by their appearance. Appearance is amajor factor in the judgment of quality and human eye has historically done this. Thecolour indicates parameters like ripeness, defects, etc. The quality decisions vary amongthe graders and often inconsistent. The adaptation of human eye to small changes incolour and the effect of the background on the perceived colour and colour intensityare the main sources of error. Hence, it is hard to provide precise guideline for visualinspection of fruits based on colour. In biological products, the light varies widely as afunction of wavelength. These spectral variations provide a unique key to machine visionand image analysis. Machine vision technology offers the solution for all these problems.

Of the many available colour models, HSI model provides a highly effective colourevaluation particularly for analyzing biological products. Using an RGB colour camera, itis also possible to obtain the colour information. However since the RGB colour modelrepresents the image in three separate coordinates RED, GREEN and BLUE, it is notefficient for colour perception and image processing than compared to HSI model, where‘H’ alone gives the colour perception.

Grading of large number of fruits create a problem related to large variations among thepopulation of fruits. This problem is addressed by using the tools provided by thestatistical methods. The authors developed an on-line apple grading system, partiallysponsored by the ministry of food processing, government of India, based on some ofthe most important external parameters including the fruit’s surface colour. The detailsincluding the experimental results and techniques of implementation of colour gradingare presented in this paper.

1. INTRODUCTIONColour provides valuable information in estimating the maturity and examining the freshness

of fruits & vegetables. Colour is one of the most significant criteria related to fruit quality. Itindicates the parameters like ripeness, defects, etc.[1-5] The appearance of the fruit, affectsthe consumer acceptance and the value addition. Specially picked quality graders have usuallyperformed this visual inspection. The quality decisions vary among the graders and areinconsistent. The adaptation of human eye to small changes in colour and the effect of the

J. Instrum. Soc. India 34 (4) 284-291

Page 2: 34(4)

background on the perceived colour, type of illumination, viewing angle are some of the mainsources of error. Hence it is hard to provide precise guideline for visual inspection of fruitsbased on colour. Machine vision technology offers the solution for these problems.

The colour of an object is determined by wavelength of light reflected from its surface.In biological materials the light varies widely as a function of wavelength. These spectralvariations provide a unique key to machine vision and image analysis[6-7].

2. COLOUR MODELRed, Green and Blue are the primary colour components. They are additive, when adding

colored lights and subtractive when adding paint pigments. Although the process followed byhuman brain in perceiving colour is a psychological phenomenon that is not yet fullyunderstood, the physical nature of colour can be expressed on a formal basis supported byexperimental and theoretical results.

A colour model is adopted depending on specific application and the RGB model is mostcommonly used for colour monitors, colour video cameras etc. Similarly, CMY [cyan, magenta,yellow] model is used for colour printers and YIQ [Luminance, In phase, Quadrature] modelis used for TV broadcast. The colour models often used for image processing are RGB, YIQ,and HSI [Hue, Saturation and Intensity]. It has been experimentally found that HSI model ismost suitable for finding out the ripeness of fruits, vegetables, colour matching [7,8] samples,etc. Here, Hue is a colour attribute that describes a pure colour where as Saturation gives ameasure of the degree to which a pure colour is diluted by a white light. The Intensity isdecoupled from the colour information of the image and Hue & saturation components areintimately related to the way in which human beings perceive colour [8.9].

We adopted HSI model for sorting and grading of fruits by colour and developed aprototype for on-line sorting of Apples based on colour, size and shape. The basic imagecapturing system adopted by us is a colour CCD camera and a frame grabber card. Theyprovide the image in RGB model. The RGB model, after normalisation, is first converted intoHSI model using a set of converting equations.

For its cost effectiveness and usefulness it is necessary to have this grader system tofunction on-line and check each and every fruit under inspection. Hence the equipment shallhave a conveyer arrangement to transport the fruits in order to the imaging chamber. Thenumber of fruits captured in a single frame (instant) depends on the system design and itshardware. The image sent by the camera will be used to analyze for all the parameters underconsideration, more specific colour in the present case. The image will be analyzed by usingadvanced image processing algorithms to estimate the colour of image.

3. SYSTEM DEVELOPMENT FOR SORTING AND GRADING OF FRUITSBASED ON COLOUR VALUE - THE DESIGN PRINCIPLE

The setup as shown in Fig. 1 has a singulator that delivers one fruit after othersynchronously to the conveyor fruit stations. The conveyer assembly mechanism automaticallyorient apple to its stem and calyx, while moving the fruit to the illumination cell. The colourCCD camera is connected to a personal computer through a frame grabber card, whichconverts the image data in analogue form into digital form. The camera with its lens system

Colour analysis of fruits using machine vision system for automatic sorting and grading 285

Page 3: 34(4)

286 P. Sudhakara Rao, A. Gopal, R. Revathy and K. Meenakshi

Fig. 1. Line diagram of a Machine vision system for sorting and grading of fruits

has the field of view to cover 3 fruits on the conveyer. Hence three fruits are captured inevery single capture along with the machine background. The specially developed C++ systemsoftware package collects the scene for further processing and analysis.

We have experimentally found for all varieties that the Apple skin has negligible Blue colourcomponent in the RGB colour space. Hence all the parts of the prototype that are part of thecamera’s field of view are coloured with blue. This helped easy elimination of background tosegment the apples images from the captured scene.

The fruit continues to rotate around its stem and calyx axis while passing through thefield of view. Multiple image frames of a fruit are captured synchronously to the fruitsmovement using proximity switches, while it is passing through the field of view. Each captureis done by the frame grabber card on an occurrence of a trigger pulse on one of its digitalinput lines. Each scene has a new posture of each apple image in the scene. The designparameters like field of view, conveyor speed and apple rotation speed are such that fruit imagerepresent a view of the fruit’s surface in one of the six rotated positions, called postures,representing rotation in multiples of 60°. This scheme of image collection helps in analyzingthe fruit’s entire surface of the fruit.

Using image-processing techniques as outlined in this paper, this digital image aftersegmentation is analyzed to determine the fruits color. The basic image processing systemconsists of a Silicon Graphics works station, Frame grabber card (DT- 3154), colour CCDcamera and an Illumination system. PULNIX- 9700 progressive scan camera with resolutionof 768 x 494 pixels is used in the setup. A uniform diffused illumination system designed byus for the purpose, is used to capture clear images of the fruits.

The captured scene from the experimental setup has three images of apples including theimage of conveyor assembly in background. The scene data in RGB colour space is firstconverted into HSI colour space. By means of selective Hue thresholding three apples together

Page 4: 34(4)

in the image are first segmented. Further each apple image in the scene is separated from therest and grouped along with the respective apple image of different posture. Thus when theapple is passed out of the field of view, we obtain six images of the apple but of differentpostures, which form the inputs for the colour analysis.

4. IMAGE ANALYSIS USING RGB MODELThe histogram of a digital image with Gray levels in the range [0, i-1] is a discrete function,

ρ(rk) = n

k / n, Where r

k is the kth Gray level

nk is the number of pixels in the image with that Gray leveln is the total number of pixels in the image and k = O, 1, 2,...,i-1

ρ(rk) gives an estimate of the probability of occurrence of Gray level n

k. A plot of this

function for all values of k provides a global description of the appearance of an image. Althoughthe properties are global descriptions that say nothing specific about image content, the shapeof the histogram of an image gives us useful information about the possibility for contrastenhancement. For classification purposes we compare the test histogram with the referencehistogram and select the best match as the one whose histogram is most similar to that of thetest item.

For multi spectral imagery, the histogram of each component image can be displayed.The graphical representation of an image histogram is a plot of the percentage of image pixelsat each digital intensity value. The input data for the histogram consists of a table of numberscontaining the number of pixels at each intensity level. The first step in generating a graphicaldisplay of the histogram is to convert the distribution into a percentage distribution by dividingthe table entries by the number of pixels within the image. The second step involves allocationof the number of pixels to be used in the annotation area for presentation of the histogram.

5. COLOUR ANALYSIS OF FRUITS BASED ON HISTOGRAMIn case of colour image, in the RGB colour space, every individual colour component,

namely Red, Green and Blue has its histogram. Then, the percentage composition of everyindividual colour component, which an fruit possesses are to be evaluated. Using this percentagecomposition the level for a component can be set as a standard in classifying the apples basedon a particular colour orientation. For an Apple the higher percentage composition of the redcomponent was assigned the superior grade, the next lower composition the second grade &like wise the descending grades were assigned. This enabled the sorting of apples based onthe colour as a parameter.

6. STATISTICAL DATA BASED ON HISTOGRAM ANALYSISThe following are some of the statistical data that can be through histogram analysis:

Percentage of colour components

Percentage of Red = [r/(r + g + b )]* 100

where r = Σ (gray level * Number of pixels) for Red

g = Σ (gray level * Number of pixels) for Green

and b = Σ (gray level * Number of pixels) for Blue

Colour analysis of fruits using machine vision system for automatic sorting and grading 287

Page 5: 34(4)

288 P. Sudhakara Rao, A. Gopal, R. Revathy and K. Meenakshi

variance = Σ(y(x - x)2 / n) - [Σ(y(x - x) / n)]2

where x = gray level ranging from 0 - 255y = Number of pixels of corresponding gray level xx = mean of xn = Total number of pixels

standard deviation = √variance

Mean of colour components

mean of red = r / Number of pixelsmean of green = g / Number of pixelsmean of blue =b / Number of pixelsco - variance = (standard deviation / mean) x 100

Correlation factor

[Σ (xy) - (ΣxΣ y / n)] / √[Σx2 - (Σx)2 / n] [Σy2 - Σy)2 / n ]where x = gray level ranging from 0 - 255 for image 1

y = gray level ranging from 0 - 255 for image 2

7. EXPERIMENTATION RESULTSApples fruits were obtained from the local market. The fruits using human expertise are

classified into four different classes called GRADE-A, GRADE-B, GRADE-C and GRADE-Dbased on their Red colour component. Apples belongs to group A have more red percentagevalues while group D will have least red percentage values, each group having 10 Apples.The apple fruits are imaged using the experimental set up. The images are analysed and statisticaldata obtained and tabulated in Table - 1.

TABLE 1: Experimental results for the 4 GRADES of Apples

Grades Colour Percentage Mean Standard Co-variance CorrelationComponent Deviation Factor *

GRADE A RED 47.038 133.63 65.95 49.35 0.701GREEN 27.188 77.24 51.77 67.03 0.644BLUE 25.773 73.22 45.96 62.78 0.616

GRADE B RED 45.059 133.33 66.15 49.61 0.124GREEN 28.298 85.63 58.60 68.43 0.116BLUE 27.642 83.65 54.48 65.13 0.142

GRADE C RED 42.848 190.2 59.55 31.31 1. 000GREEN 28.699 127.39 80.13 62.90 1.000BLUE 28.451 126.29 74.70 59.15 1.000

GRADE D RED 35.422 140.91 67.50 45.32 0.296GREEN 32.557 136.87 82.07 59.96 0.381

BLUE 35.022 134.62 75.27 55.91 0.158

[* In this table the correlation factor is computed when Grade C apple is taken as test apple]

Page 6: 34(4)

The above properties of the pixel intensities are calculated using histograms. The meanis the measure of average brightness and the variance is a measure of contrast. The intensitymean and variance (or standard deviations) properties are used because of their relevance tothe appearance of an image. Correlation factor between two images gives the measure of howwell one image matches with the other. The above statistical analysis is useful only when weuse RGB.

8. DISADVANTAGES OF HISTOGRAM SPECIFICATION USING RGBThe pixel value in colour image is vector [RGB values], which implies the colour

equalization is three-dimensional (3D) process. If Histogram equalization is applied to each ofthese colors independently, changes in the relative percentage of red, green and blue for eachcolour vector may occur. Consequently, colour equalization must process all three componentsat once, which makes for the process of histogram analysis very complex.

The overall important aspect of colour image processing is that the algorithms developedshould be able to produce the least amount of detectable colour distortion; the processedquantities should be mapped as closely as possible to those that are perceptually important.This can be accomplished using HSI colour space.

9. IMAGE ANALYSIS USING HSIRGB data is first converted into HSI data. With image representation in the HSI domain,

the colour analysis was based on primarily the Hue value. Hue is a colour attribute that describesa pure colour, whereas saturation gives a measure of the degree to which a pure colour isdiluted by white light and finally intensity gives the effectiveness of the colour. The threedimensional RGB space is reduced to a one-dimensional ‘H’ Space for colour analysis. For adigitized colour image, the Hue histogram represented the colour components and the amountof that Hue in the image. Therefore, colour evaluation of fruits was achieved by analyzingHue histograms [6,7.8,11] . The advantage of using HSI model over RGB is as under:

The intensity component is decoupled from colour information Hue and Saturation components are intimately related to the way in which human

being perceive colour. Hue value is invariant to changes in light intensity.

These features make the HSI model an ideal tool for developing digital image processingalgorithms based on colour sensing properties of human visual system

10. METHODS OF ANALYSISThe Hue, Saturation and Intensity (HSI) domain was chosen to quantify Apple colour by

machine inspection [9], since this representation is closest to that of human perception anduse of the Hue domain allowed easy segmentation of the apple from the background. However,colour image information is sensed using combination of RGB. Various colour features canbe calculated from the RGB components by using linear or non-linear transformation.

Hue and Saturation are calculated from RGB[10,12] values by

H = (90O + tan-l [(2R - G - B)/[√3(G - B)]]+(180O if G < B ) x (255/360)

Colour analysis of fruits using machine vision system for automatic sorting and grading 289

Page 7: 34(4)

290 P. Sudhakara Rao, A. Gopal, R. Revathy and K. Meenakshi

S = 1- [min(R,G,B) / I] x (255/360)

A typical HUE histogram of different apple images is shown in Fig.(2).

11. CLASSIFICATION BASED ON HUE HISTOGRAM DISTRIBUTIONThe probability density of the Hue pertains to surface colour is maximum at the median

density pixel value. Hence the median density Hue value represents the overall distribution ofsurface colour of the apples. This method requires extensive training. The Meadin density ofHue histogram for a given class is taken as the basis for segregating the fruits belonging tothat grade. Hence during the training process 20 Apple samples belonging to a GRADE aretaken and meadian density of the combined histogram is calculated, thus obtaining a distinctmedian density value for each of the GRADE to build the reference table.

12. RESULT AND DISCUSSIONSWe have performed experiments using the developed experimental setup for sorting and

grading of fruits like apples based on colour as the parameter. We have obtained sample applefruits of different colours pertains to different grades with the help of Mis Himachal PradeshHorticultural Producers and Marketing Corporation, Shimla, India. Six different postures ofthe test apple are imaged and digitized by the computer system for colour analysis. We haveused HSI scheme for the colour classification of the Apples. Median density of Hue of theApple is used as the criteria for grading. We have assigned different grades depending on thevariation of red colour of the apples. In the Hue colour chart the red variation occurs between0 and 20 wherein 0 represent the 100% Red and as we go away from 0 the percentage of red

Fig. 2. Typical HUE histogram of Apple

Page 8: 34(4)

colour occurrence is decreasing. Above the Hue value of 20 the distribution of red colour isvery less. Following reference Table 2 is used for the Apples classification.

Table 2- Grading criteria for colour classification of apples.

Classification criteria basedGRADE on Hue value.

A 0-1

B 2-5

C 6-10

D 11-15

E 16-20

The system is trained 100 Apples, 5 sets of Apples each having 20 Apples presenting theparticular grade. It was found that changes in light intensity level during inspection affectedthe results of classification. Although in theory Hue should be independent of intensity of animage, Le lighting level, experiments showed that the Hue histogram of an apple shifted towardsthe green direction if the intensity is increased by changing the lens aperture or lighting level,and shifted towards red (left) if intensity decreased. Hence the intensity level is maintainedreasonable constant by maintaining the voltage level for the lamps. By representing mediandensity of Hue as a grading criterion, the image processing system achieved around 98 %accuracy in colour inspection of apples.

13. CONCLUSIONSMany commercial factors like pricing, identification of fruit by variety, grading etc., are

decided by many parameters like size, shape, colour, surface defects, etc. The adaptation ofhuman eye to small changes in colour and the effect of the background on the perceivedcolour and colour intensity are the main sources of error. Hence, it is hard to provide preciseguideline for visual inspection of fruits based on colour. In biological products, the light varieswidely as a function of visible wavelength region. Machine vision technology is applied toextract information about all these parameters using appropriate optics and imaging systems.Most of the image processing applications for parameter extraction deals with huge amountof data, hence, the algorithms developed are efficient and be optimized for speed to keep thesystem throughput as per the market demands. Colour representation in HSI provides anefficient scheme for colour discrimination. By representing median density of Hue as a gradingcriterion, the image processing system achieved around 98 % accuracy in colour inspectionof Apples. The image processing system incorporated in the prototype system also includeother quality parameter extraction such as shape and size. We are presently working on extendingthe system for incorporating external defect as one of the quality parameter in the overallclassification of the fruits. The machine vision system and methods developed are alsoapplicable for general use in colour processing and therefore could be used to inspect otherkinds of fruits and vegetables.

Colour analysis of fruits using machine vision system for automatic sorting and grading 291

Page 9: 34(4)

REFERENCES1. On encoding of arbitary geometric configurations - Freeman.H, IEEE Trans. Elect. Computers

EL-10, 260-268,1961.

2. Visual Pattern recognition by moment invariants, HU.MK, IRE, Trans. Info. Theory IT-8,179-187, 1962.

3. Automated Machine vision inspection of potatoes - Tao, Morrow, Heineman P.H. ASAE,No. 90-3531, 1990.

4. An instrument system for cereal grain classification using digital image analysis -Sapirstein,H.D, Neuman, M, Wright E.H. - J Cereal Sci, 6, 3-4,1987.

5. Corn quality evaluation with computer vision - Ding et, al. ASAE No. 90-3532, 1990.

6. Computer vision sorting by potatoes - Ph.D. Diss by Mc clure J.E., Pennsylvania S t a t e .Univ. 1988.

7. Machine vision for colour inspection of potatoes and apple -Tao Y. et. AI ASAE 38; 1555-1561, (1995a).

8. Grading of mushrooms using a machine vision system - Heinemann et. al - ASAE Vol. 37(5);1671-1677, 1994.

9. Machine vision inspection of golden delicious apples - Heinemann et al. ASAE Vol. 11(6)901-906, 1995.

10. Digital image processing - Gonzalez. R.C and Wintz - 1988.

11. Apple shape inspection with computer vision - Lee man et al. - Proc. Of IntI. Conf. On Sensorsfor non destructive testing - Measuring the quality of fresh fruits and vegetables -1997.

12. Gonzalez R. C. and Richard E. Woods, (1993), Digital image Processing, Addison - WesleyLongman Inc.

292 P. Sudhakara Rao, A. Gopal, R. Revathy and K. Meenakshi

Page 10: 34(4)

ON-LINE PROCESS CONTROL UNIT FOR JAGGERYMANUFACTURING INDUSTRY

S.T. Pawar and M.B. DongareDepartment of physics, Shivaji University, Kolhapur - 416004

ABSTRACT

Agriculture and agro-based industries fonn the backbone of Indian economy but theyare still adopting traditional methods. Jaggery industry is an imponam agroprocessingindustry in rural India. The major constraint in this industry is lack of standardisation inprocessing. In traditional method of jaggery manufacturing a so-called skilled personknown as “Gulvaya” plays a deciding role in clarification of sugarcane juice and inconfirming formation of ‘Kakavi’ (liquid jaggery) and jaggery. The presently developedmicrocontroller based on-line process control unit is field usable and can give audioindication of important parameters (pH and temperature) and display the parameters indecimal form on the LED numeric display. It has resulted in giving us a precise results,in order to produce a superior quality jaggery.

1. INTRODUCTIONIndia is the largest producer of sugarcane in the world occupying about 4.0 million-hectare

of land. The area in which sugarcane is grown in Maharashtra is 601 thousand hectares andis ] 1.1 % of the tota! area in which sugarcanes grow in India. The yield is 82394 kg/ha1. InIndia total sugarcane production is of 227.061ak tonnes. Out of total sugarcane production43.5% is used for jaggery and khandari2. About 10.3 million tonnes of jaggery is producedannually in India 7. The major constraint in jaggery industry is lack of standardisation inprocessing. In addition an unhygienic surrounding during manufacturing, packaging and storageare major problems.

2. EXPERIMENTAL DETAILSThe field experiment was conducted from Nov 2000 to March 2001. From our study

and field survey of 25 jaggery-manufacturing units in western part of Maharashtra, it hasbeen noted that, temperature and pH plays an important role in manufacturing process ofjaggery. Whatever may be the initial Brix of the cane juice the two important striking pointsare appearing at a fixed temperatures only. viz. Liquid jaggery (Kakavi) at 105°C and finalstriking stage (Golli S1age) at 118°C. Also it has been noted that initial pH of juice variesfrom 5.1 to 5.7. For clarification it is raised by the use of lime water to 5.9-7.00. Theneutrilisation is achieved by use of chemical clarificants. The observed range of pH after

J. Instrum. Soc. India 34 (4) 293-298

Page 11: 34(4)

294 S.T. Pawar and M.B. Dongare

neutrilisation is 4.8 to 5.4. These are the variations due to manual and approximate use ofclarificants. These leads to variation in quality of jaggery.

It is also found that for extra-special (Exia) and grade no. 1 jaggery the correspondingincreased and decreased pH values are to be 6.3 and 5.3 respectively.

Thus quality of jaggery is influenced by physical parameters like pH and temperature3.Hence there is a need of on-line field usable process control unit to be developed to meet therequirement offarmers to get superior quality jaggery. Some of the novel features of thedeveloped system are set point facility to meet region-wise variation in parameters and chumercontrol.

3. ON-LINE PROCESS CONTROL UNIT FOR JAGGERY MANUFACTURINGINDUSTRY

The developed microcomroller based on-line process control unit is designed specially tomeet the needs of fa;-mers. The system incorporates pH electrode and temperature sensor todisplay and give the audio indication of important stages in jaggery manufacturing process.The system is user triendly and does non-destructive measurement.

3.1 Hardware of the SystemThe developed system consists of (H2CI2) Calomel electrode, an instrumentation amplifier

(LM 321), temperature transducer (pt 100), constant current source formed by LM 324 andBC 557, ADC 0809 Board and microcontroller 89C51. The system can measure pH valuewith an accuracy of 0.1 of the reading. The system is energized by a highly regulated powersupply. The block diagram of the system is shown in Figure. 1.

Fig.I. Block diagram of microcontroller based on-line process control unit.

Page 12: 34(4)

3.2 pH ElectrodeTo obtain an accurate measurement of the emf developed at the electrode, the electronic

measuring circuit must have a high input impedance The developed emf is suitably amplifiedby the instrumentation amplifier before applying it to the ADC. When the combined electrodeis immersed in a solution, a potential is developed. This potential is actually very small of theorder of few millivolts.

3.3 PT -100 RTD SensorAs compared to Nickel and Copper, platinum has been found to be relatively linear within

the specified range. The platinum further has additional merits, which make it suitable for thepresent application. These merits are - I) high precision and accuracy 2) Ease of calibration3) high responsibility 4) fast response 5) Interchangeability with other resistance without anycompensation 6) Good performance in desired temperature range and 7) limited susceptibilityto contamination etc.5.Pt-100 is two terminal passive sensors. It has wide mt:asuring range -100°C to 600°C.

3.4 Instrumentation Amplifier (LM 321)Op-amp LM321 and opamp-LM324 ICs are connected in the non-inverting mode and

are used to amplify the millivolt signal generated by the pH electrode.

3.5 AID converter InterfacingADC 0809 is a monolithic CMOS device with 8-bit ADC. It uses successive approximation

as the conversion technique. ADC 0809 is a 8 - channel ADC for unipolar analog signals6.

3.6 MicrocontrollerMicrocontroller 89C51 is used to process the input data with setpoint values of temperature

and pH. The 89C51 have the following features

1. Eight bit CPU with accumulater A and B registers.

2. 16 bit program counter and data pointer.

3. 8 bit stack pointer.

4. Internal E2pROM of 4 K.

5. Internal RAM of128 bytes.

6. 32 I/O pins arranged as four 8-bit ports.

7. Two 16 bit timer/counter.

8. Full duplex serial data transmitter / receiver.

9. Control registers TCON, TMOD, SCON, PCON, IF and IE.

10. 2 External and 3 Internal interrupt sources

11. Oscillator and clock circuits.

According to the selected data, the appropriate control signals are generated and are appliedto the buzzer and motor drive unit 4.

On-line process control unit for jaggery manufacturing industry 295

Page 13: 34(4)

296 S.T. Pawar and M.B. Dongare

3.7. 7447 display drives / decoder and key board7447 is display driver for common anode type seven segment displays. Keyboard is used

to set the initial and final set points related to desired pH and temperature values.

4. SOFTWARE OF THE SYSTEMIn microcontroller based system, software design is a more demanding task than hardware

design. The software is written in assembly language of microcontroHer 89C51 to performthe following

1. Analog to digital conversion program2. Display program for pH and Temperature3. Data byte of pH and Temperature4. Delay program

5. CALIBRATION AND WORKINGUsing standard laboratory pH meter and thermometer carries out the calibration of the

unit. The pH electrode and temperature sensors are placed in the sugarcane juice-boiling pan.The power is supplied to the system by means of the stabilized IC regulated dc power supplies.

The analog outputs ITom the transducers, after suitable signal conditioning, amplificationand conversion are applied to the channels of the ADC 0809. Here channel 0 is used. ADC0809 convert’s analog inputs into digital outputs in hex ITom by executing the main program.The digital values are processed by microcontroller 89C51 and pH and temperature values forcane juice under process are displayed. The unit gives audio indication in accordance withpreset values of pH and temperature. It also controls the churner action in jaggery manufacturingprocess.

6. TESTING OF PROCESS CONTROL SYSTEMThe performance of the unit has been successfully tested at 10 jaggery-manufacturing

units in Kolhapur region. During test we have carried out the jaggery preparation by makinguse of process control device. Irrespective of soil type and sugarcane genotype we haverecorded the recovery and grade of the jaggery.

System Setting - I) pH was set at 6.3 for lime defection and at 5.3 for neutralization.

II) Temperature was set at 10S°C. for liquid jaggery(Kakavi)stage andat 118°C. for jaggery (Golli) stage. Table-l gives the result of 10jaggery-manufacturing units in Kolhpur region.

7. CONCLUSIONSThe observed results reveal that the microcontroller based on-line process control unit

for jaggery industry is sufficiently accurate in monitoring physical parameters of sugarcanejuice. The unit gives better result as compared to manua1 judgement by a person known as“Gulvaya” in jaggery manufacturing process. This unit is highly beneficial to the farmers indeciding the two important striking stages and it helps in optimum clarification so that a goodquality jaggery can be produced. From Table-l we found that the jaggery recovery rangesfrom 10-12 and grade of product is maintained between extra-special (exta) and grade-I.

Page 14: 34(4)

Table-1: Results of 10 jaggery-manufacturing units in Kolhapur region.

Jaggery Unit No Jaggery Recovery Jaggery Grade*

1 10 Exta2 11 13 12 Exta4 11 15 10 Exta6 12 17 11 18 10 Exta9 11 110 11 Exta

Grade of jaggery is recorded as per present jaggery grading method adopted in jaggerymarket. Any market does not follow scientific grading method. In market grading is doneby physical appearance of jaggery i. e. by checking test, color and hardness by knife(Granular size)

The grade numbers given by jaggery market are,1. Extra special grade (exta)2. Grade no-l3. Grade no-24. Grade no-35. Grade no-46. Grade no-5

The jaggery of the grade i.e. Exta and grade no-l fetches maximum price in market.

ACKNOWLEDGEMENTOne of authors (STP) is thankful to University Grants Commission, New Delhi for award

of teacher fellowship under FIP.

REFERENCES1. Damahe, B.A. 14(2000). Application of L T. in area of Agri. And Agrobased industry. Proc.

Seminar on INFOTECH.2. Patil, J.P. et. al. 1-3 (1996). Research bulletin on liquid jaggary MPKY. RES. PUB. NO. 17.3. Pawar, ST. et.al., 369 -374 (2001) Scientific studies on Jaggery Manufacturing process.

Co-operative Sugar Vol-32, No.5.4. Kenneth J. Ayala.54-60 (1996). The 8051 microcontroller Architecture, ogramming and

application. Second edition. Penram International publishing (India) .5. Raman K. Attri .et.al., 275-283(2000) Design Approach to use Pt RTD sensor. 1. Instrum.

Soc. India 30(4).6. B.Ram. (1995). Fundamentals of microprocessor and microcomputers. 4 th edition. Dhanapat

Rai Pub. Nai Sarak, Delhi.7. Dorge, S.K. I-II (1994). Proc. of National consultation meeting feb 27-28 : RS and JRS, Kolhapur.

On-line process control unit for jaggery manufacturing industry 297

Page 15: 34(4)

FABRICATION OF A PULSE-SHAPEDISCRIMINATION MODULE FOR NEUTRON-GAMMA

SEPARATION

Rakesh Kumar, Golda K.S, S. Venkatramanan, R.P. Singh,S.K. Datta and R.K. Bhowmik

Nuclear Science Centre, P.O. Box - 10502, New Delhi - 110067 INDIA.

ABSTRACT

A low cost high performance pulse shape discriminator (PSD) module for neutron gammadiscrimination has been developed for an array of eight neutron detectors. In this, zerocross technique (ZCT) has been exploited to achieve an optimum n-γ identification. Toreduce cable connection in the system, as well as the capacitance and delay effects, theconstant fraction discriminator (CFD) and PSD stages are compactly mounted in astandard single-width NIM module. To make the system still more compact two channelsare accommodated in a single NIM Module. The module accepts signal ranging from -200m V to-2V. The module is designed to handle fast rise time pulses of the orderof ~5ns. The module provides two NIM outputs, an n-γ logic signal and a crossovertiming output for n-γ discrimination in an external TAC.

1. INTRODUCTION

A detector array consisting of eight neutron detectors (BC50lA organic liquid scintillatorsl)is being built to study certain aspects of nuclear reactions e.g. the entrance channel effects,pre-equilibrium reaction component, level density parameters etc. and also for determinationof nuclear structure through neutron tagging. The liquid scintillators deployed for neutrondetection are also sensitive to gamma radiation. As the typical gamma output in a reaction isconsiderably higher than that of neutrons, it becomes necessary to distinguish between thetwo. Pulse shape discriminators (PSD) have been successfully used with neutron detectorsto discriminate neutron and gamma2-6.

There are two methods to do neutron-gamma (n-γ) discrimination; charge comparisonmethod2,3 and zero cross technique4-6. Here, the ZCT is used for n-γ discrimination. Thesingle width NIM module contains CFD and PSD; this makes the system compact.Results are presented for discriminating n-yrays with liquid scintillator BC50lA, which showsgood separation.

J. Instrum. Soc. India 34 (4) 298-308

Page 16: 34(4)

2. PRINCIPLE OF OPERATIONIn the zero cross-over method, the a node signal from the fast photomultiplier tube (PMT)

coupled to the scintillator is differentiated through a differentiating amplifier. When pulseswith different rise times undergo differentiation, one gets bipolar pulses with different zerocross over timings. Due to the difference in the interaction mechanism in the scintillator material,neutrons and gammas give pulses with different rise times and hence can be distinguishedand separated by using zero cross over method.

3. CIRCUIT DETAILSThe first prototype of the PSD is an integrated unit with CFD and P SD stages all

compactly mounted in a single width NIM module, as shown in the block diagram in Fig. 1.Therefore, in future we will be able to incorporate two (CFD-PSD) units in the same module.

Fig. 1. Block diagram of the CFD and PSD stages.

This unit, which utilizes basic ECL (Emitter-coupled logic) circuits, consists of a CFD and aPSD. A -2V supply for terminating ECL signals is locally generated with transistorized circuitfrom the -6V supply line. The input signal from the photomultiplier anode, which is fed througha LEMO connector on the back panel, is split into two - one going to the CFD stage and theother to the PSD stage. The front panel LEMO connections give outputs of CFD, Zero Cross,Strobe, PSD and also neutron gated PSD output. Provisions for monitoring threshold settingsare also there on the front panel.

Constant Fraction DiscriminatorThe CFD part uses standard constant fraction time technique by adding a delayed inverted

pulse to an attenuated input pulse. A dual ultra fast comparator (AD96687) is used as the

Fabrication of a Pulse-Shape Discrimination Module for Neutron-Gamma Separation 299

Page 17: 34(4)

300 Rakesh Kumar, Golda K.S, S. Venkatramanan, R.P. Singh, S.K. Datta and R.K. Bhowmik

discriminator, out of which one is used as Low Level Threshold (LL TH) adjuster and secondas a CFD. The LLTH voltage is generated from LM336 (-2.5V regulator) and a resistivenetwork. The range of threshold can be varied from -2mV to -200mV. The constant fractionis chosen as 0.2 and implemented by a resistive network. The input signal is delayed by usingaim LEMO (RG-174) cable that gives a fixed delay of 5 ns. The attenuated and delayed signalis then fed into a zero crossing detection circuit, which leads to constant fraction logic. Walkadjustment is provided through a front panel potentiometer along with a walk monitor point.The ECL AND gate (MC10105) is used to produce an AND signal of LLTH discriminatorand the CFD. After that an ECL monoshot (MClO198) is used to produce an ECL pulsecorresponding to the resultant of the AND pulse. The output is taken through a LEMOconnector on the front panel. Another CFD output is taken to generate the strobe in thePSD stage.

Pulse Shape Discriminator StagePMT’s split signal is integrated and differentiated in a fast shaping amplifier AD829 so

that the zero cross time of the output pulse depends on the fall time of the input signal. TheRC time constants are experimentally adjusted for best shape discrimination and resolutionand its value is fixed around l00ns each. To detect baseline zero cross over of the shapedpulse, an ultra fast comparator (AD96687) is used. This comparator gives a zero-cross pulse.A strobe of adjustable width and delay is generated with the CFD output using 2 monoshots(MC10198). One of the outputs from AD96687 is gated with strobe in such a way that ORgate output is a zero cross out put of both neutron and gamma and this can be used in anexternal Time-to-Amplitude Converter (TAC) to discriminate neutron and gamma. ThisNIM output is available through a front panel LEMO connection. The other output is gatedwith delayed and width adjusted strobe so that the gamma is eliminated and this logic gateoutput is further fed into an ECL monoshot for giving an ECL logic pulse. This pulse thengoes through an ECL to NIM converter, and a front panel LEMO connection gives thisNIM output.

4. TESTING AND CALIBRATION

The performance of the unit has been tested with a252Cf source using a 5" X 5" BC 501organic liquid scintillator. For the electronic set up (see Fig. 2), the anode signal from thePMT is given to the PSD module. The CFD output from the module is used as start in theTAC and the zero cross output is used as stop after giving necessary delay. The TAC outputis given to a 2k ADC (ORTEC-AD811). Dynode signal from PMT is amplified through EG&G571 spectroscopy amplifier and is given to another channel of the ADC. ADC data is collectedin a computer through CAMAC based data acquisition system. The efficiency of the moduleis measured in terms of its figure of merit (FOM), which is defined as the ratio of the peakseparation to the sum of Full Width at Half Maximum (FWHM) of two peaks. From the TACspectrum, given in Fig. 3, the FOM of the PSD is estimated. The FOM of our module is 2,which is comparable to that of the commercial PSD module CANBERRRA 2061A. The 2-Dspectrum of TAC vs. energy is plotted in Fig. 4, and shows clear separation between neutronand gamma.

Page 18: 34(4)

Fig. 2. Block diagram of a basic PSD electronic set up.

Fig. 3. Neutron-gamma time of flight (TAC) spectrum.

Fabrication of a Pulse-Shape Discrimination Module for Neutron-Gamma Separation 301

Page 19: 34(4)

Fig. 4. 2-Dimension spectrum (Pulse height vs. TAC)

5. CONCLUSIONThe performance of the modules designed by us show good n-y separation and their

FOM is comparable to those of commercially available modules. These low-cost, compactNIM modules demonstrate good n-y discrimination using neutron-detectors. Four PSD modules(each containing dual units) will be completed shortly.

ACKNOWLEDGEMENTThis investigation was sponsored through a Department of Science and Technology (DST)

project.

REFERENCES1. Bicron Corporation, USA, BC501A data sheet.

2. M. Moszynski et al., ‘Study of n-γ discrimination by digital charge comparison method for alarge volume liquid detector,’ Nucl. Instr. Meth. Phys. Res. A, 317, 262-272 (1992).

3. M. Moszynski et al., ‘Study of n-γ discrimination with NE213 and BC501A liquid scintillatorsof different size,’ Nucl. Instr. Meth. Phys. Res. A, 350, 226-234 (1994).

4. Gy Mathe et al., ‘Pulse shape discrimination method for particle identification,’ Nucl. Instr.Meth., 27, 10-12 (1964).

5. L.J. Heistek et al., ‘Pulse shape d iscriinination with a comparator circuit,’ Nucl. Instr. Meth.,80, 213-216 (1970).

6. J. Bialkowski et al., ‘A pulse-shape discriminator with high precision of neutron and gammaray selection at high counting rate,’ Nucl. Instr. Meth. Phys. Res. A, 275, 322-328 (1994).

302 Rakesh Kumar, Golda K.S, S. Venkatramanan, R.P. Singh, S.K. Datta and R.K. Bhowmik

Page 20: 34(4)

CLASSIFICATION OF FRUITS BASED ON SHAPEUSING IMAGE-PROCESSING TECHNIQUES

P. Sudhakara Rao, A. Gopal, S. Md. Iqbal,R. Revathy and K. Meenakshi

CEERI Chennai, CSIR Madras Complex, Taramani, Chennai - 600 113

ABSTRACT

The shape is one of the important visual quality parameters of fruits, vegetables, etc.Currently human sorters are employed to sort fruits based on shape. Shape is a feature,easily comprehended by human but difficult to quantify or define by computer. Most ofthe machine vision shape detection work has been done on industrial objects, whichhave more definite structure. Agricultural and biological products are unique in natureand the growing environment causing various boundary irregularities influences theirshapes. Image processing offers solution for sorting of fruits based on their shape.

Many researchers explored to establish relationship between object shape and itsboundary values in Fourier spectrum. The authors developed an on-line apple gradingsystem, partially sponsored by the Ministry of food processing, Government of India,based on some of the most important external parameters including the fruit’s shape.This paper presents some new approaches using correlation techniques, graphicalanalysis of radius and area signature, directional change of contour and boundaryFourier coefficients to extract the shape, particularly for fruits. The paper also discussessome of the experimental results.

1. INTRODUCTIONNon-destructive quality evaluation of food products is an important and very vital factor

in food/agricultural industry. Various parameters which define the quality of these products(eg., colour, shape, size, external defects, etc.,) are evaluated visually by human inspectors.Though rapid technological advances have taken place, the cost of image processing / machinevision techniques are still expensive and there is need for investigating less expensive processingtechniques. This should provide automatic grading process, based on shape as one of theparameters, to remove inconsistencies and reduce the likelihood of apples / fruits rejection atconsumer markets.

Shape is a feature, easily comprehended by humans but difficult to quantify or define bycomputer. Most of the machine vision shape detection work has been done on industrialobjects, which has more definite structure. Agricultural and biological products are unique innature that it is possible for a fruit, say apple to have a shape, which the vision system may

J. Instrum. Soc. India 34 (4) 227-239

Page 21: 34(4)

never have previously encountered. The growing environment, causing various boundaryirregularities, influences their shapes. Damage during harvesting and handling adds more kindsof shapes. Thus for apples, there is an infinite number of shapes from ‘well’ to ‘badly’ shapedapples [2]. Trained Inspectors judge apple shape according to how well it conforms to thecharacteristic shape of the variety. Four-shape categories are generally used - well formed,fairly well formed, slightly deformed and seriously deformed.

Various techniques for shape identification[1-8] and grading were adopted. These aregenerally summarized as

a. Boundary encoding technique for finding shape number.

b. Statistical analysis using moments, bending, energy, radius variation and fractures.

c. Structure analysis from geometry.

d. Spectrum analysis.

Although many such general techniques were documented in computer vision literature,the natural variability and diversity of biological materials create difficulties and practicalproblems[4]. The shape requirements are somewhat abstract and difficult to comprehend sincethere are no standard shapes available for comparison. These classes need to be quantifiedfor automatic grading, i.e., each shape classification should have a number or number rangeassociated with it.

In general two different kinds of methods[8]. Viz., Region based information and Boundaryinformation are adopted for Shape extraction / analysis of fruits. The first method is based ongeometric parameters while the second method involves computations of Fourier descriptors.

2. REGION-BASED INFORMATIONShape measurements are evaluated by calculating the coefficient of determination between

the length (major principal axis of inertia) and width (minor principal axis) of the fruitscalculated by the vision system and the polar and equatorial diameters measured manually.The relationship between the area obtained with the vision system and the product of the twomanual measurements has also been investigated. The vision system measured the length ofmajor and minor axis of the contour of random views of the fruits, and the manual data weretaken from pre-oriented fruits (“equatorial” and “polar” diameters). For this reason, theparameter “area” has higher correlation with manual measurements.

Shape evaluation is performed in each view [5] by determining the perimeter of the fruitin each view, the area enclosed by the perimeter, the convex hull of the perimeter and thearea enclosed by the convex hull. The computer uses the convex hull and the perimeter todetermine if a piece of the fruit is matched or mismatched. For example, in the case of sphericalfruit, the convex hull and the perimeter should be nearly the same. However, for a banana,the convex hull will enclose more area than the perimeter.

Sorting of fruits based on shape may also be done by calculating “Eccentricity (E)”, ofthe object fruit, which is defined as the maximum dimension divided by minimum dimensionof the object [8]. Another parameter that can be used as size/shape sorter is “Compactness”.

228 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 22: 34(4)

This is defined as the ratio, C = P2/4πππππ A; where P is the perimeter and A is the area of theobject. Alternatively, the inverse of compactness is called, “Circularity”. The Circularity is adimensionless number with a minimum value of ‘l’ for circles. The circularity is 1.27 for asquare, shows a high value for elongated objects.

3. BOUNDARY INFORMATIONThe Fourier descriptors[6,7] use only the boundary of the object and consider it as cyclic

waveforms, which provide a parameter description of the boundary of the trace. The Fourierdescriptors of an image are calculated by taking the FFT for the sequence considering “N”sample points on the contour of the image obtained after thresholding. Here “N” is made tobe numbers obtained as powers of ‘2’[1,2,6,8] i.e. 2,4,8,16 etc. The Fourier descriptors conveyuseful information about the boundary shape. In general, an image, which possess uniformharmonic values and also containing minimum number of harmonics is highly regular in shapeand its reconstruction can be obtained easily with lesser number of samples. The inverse Fouriertransform helps in reconstruction of the object based on the contour value of the Fourierdescriptors.

In general, global shape or the approximate shape of an object can be obtained back usingminimum number of harmonics say up to 4[8], but higher harmonics contribute to finer detailspertaining to sharp transitions.

4. SHAPE COMPUTATION METHODThe objective of this paper is to develop effective shape evaluation method that could be

used in machine vision systems for the purpose of grading apples. The shape of any irregularobject is unique in nature. The boundary information of an object determines the shape of anobject. The accuracy of shape definition depends on the accuracy of the boundary informationof an object. In a computer image, the boundary points are ‘pixels’. The boundary pixels canbe studied, compared and analyzed in many ways. However after shape analysis, the boundaryshould be re-traceable and the error should be kept to its minimum.

A new method “Improved Radius signature Analysis” is presented in this paper forthe shape analysis of fruits. In this method the centroid of the object is assumed to be invariantand appropriate aspect ratio applied.

5. EXPERIMENTAL SETUPThe experimental setup has a conveyer assembly to automatically orient apple fruit to

stem and calyx and then move the fruit to the illumination cell. It is experimentally found thatthe apple skin for all varieties has negligible blue component in the RGB colour space. Hencethe conveyor assembly, fruit holder, belt for friction drive that rotates the fruit are colouredwith dark blue. This will help easy elimination of background to extract the apple images fromthe captured scene. The system has a CCO camera to capture the scene that consists of fruit.Three fruits are captured in a single frame. The camera is connected to a Silicon GraphicsWorks station through the frame grabber card OT-3154 from Data Translation, Inc. Thegrabber card will convert the image data in analogue form into digital form, stored in the

Classification of fruits based on shape using Image-Processing Techniques 229

Page 23: 34(4)

memory and taken for applying suitable algorithm as outlined in this paper. The basic imageprocessing system consists of Silicon Graphics Works station, frame grabber card, CCOcamera and illumination system. PULNIX 9700 progressive scan colour digital camera withresolution of 768 x 494 pixels from PULNIX America, Inc. is used in the setup. A uniformdiffused illumination system is used to capture clear images of the fruits.

The captured scene from the experimental setup has three images of apples including theimage of conveyor assembly in background. The scene data in RGB colour space is convertedinto HSI colour space. By means of selective hue component elimination, the background istotally removed and three apples image alone is retained in the scene. The retained image datais used for extracting information like colour, size, shape etc. The image is converted intobinary form by thresholding and then the edge of each apple is detected by applying laplacianfilter for obtaining the contour of the image for each of the fruit. Then the contour is analyzedusing various techniques proposed in this paper for shape estimation. The shape informationthus obtained will be the basic input parameter for comparison and to assign appropriate gradefor each of the apples. Mathematical and the relevant analysis carried out in this paper assumethat the images are in the first quadrant.

6. IMPROVED RADIUS SIGNATURE ANALYSIS

Referring to the Fig.1 (a), the local radius corresponding to different points on theboundary profile of the image can be determined by first finding the centroid of the profile. If(x

k, y

k), k = 0,1,2,A, N be the N pixel points traced on the profile, then the centroid (x

c, y

c),

of the profile can be determined using Green’s Theorem based on area moment of inertiaas below.

Fig.1.

230 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 24: 34(4)

After finding the centroid, pixel points at equiangular intervals, say ‘N’, normalizedboundary points are obtained. The local radii corresponding to these points can then bedetermined as

and these values of Rk for N boundary points are plotted as shown in Fig. 1 (b).

Such a graph is made for a set of radii Rx for x = 0,1,2....N for a standard grade ofan apple and used as reference. The on-line current sample can now be analyzed for itsshape by measuring the boundary pixel points on it and finding out the new set of radiiRy for Y = 0,1,2...N, for the same number of equiangular intervals as taken in reference.The minimum radius may be taken as the starting data for both the reference and the samplefor plotting purpose.

The following three techniques are used for the analysisa) Correlation techniqueb) Graphical techniquec) Fourier technique

7. CORRELATION TECHNIQUEThe correlation co-efficient between the set of radii measured for same number of

equiangular intervals, for the reference as well as the sample apple, is obtained using KarlPearson’s correlation co-efficient method, as given below

Classification of fruits based on shape using Image-Processing Techniques 231

Page 25: 34(4)

r(x,y)=Cov (x,y)/(σσσσσxxσσσσσx)

where CF x and CF yare the standard deviation of the reference and the samplerespectively

Where Xav

& Yav

are mean of reference and sample data respectively, dx and dy aredeviations from assumed means of reference and sample data respectively.

The value of . ‘r’ gives an indication as to, how close the shape of the sample apple isassociated with that of reference apple. If r = ± 1, it means that there is a perfect relationbetween the set of data. If ‘r’ is nearer to zero, the set of data are uncorrelated.

The probable error (P.E.) for this set of data is obtained as

P.E. = 0.6745 (1- r2) / (√√√√√N ) (6)

Then the upper and the lower limits within which the coefficient of correlation in thepopulation” p” can be fixed as

p = r ± P.E. (7)

8. GRAPHICAL ANALYSISFor the set of points (k, Rk ), k = 1,2,A N = x, y a best fit polynomial curve of nth

degree can be obtained as below

Y = aO +a1x+a2x2 +a3x

3 + ΛΛΛΛΛ ΛΛΛΛΛ ΛΛΛΛΛ + anxn (8)

The normal equations for determining the constants aj’s (Where i = 0,1..n) are obtainedby the principle of Least Squares by minimizing the residual or error sum of squares [E],summation being extended over the given set of observations.

E = ∑∑∑∑∑ (y-ao - a

1x-a

2x2 - a

3x3 ΛΛΛΛΛ ΛΛΛΛΛ ΛΛΛΛΛ -a

nxn)2

(9)

The normal equations are given as

∂∂∂∂∂E / ∂∂∂∂∂ai = 0, (i = 0,1,2 ΛΛΛΛΛ ΛΛΛΛΛ n) (10)

=> ∑∑∑∑∑xi(y - a

0 - a

O - a

1x - a

2x2 - a

3x3 ΛΛΛΛΛ ΛΛΛΛΛ ΛΛΛΛΛ -a

nxn)= 0, (i = 0,1,2, ΛΛΛΛΛ ΛΛΛΛΛ ΛΛΛΛΛ) (11)

232 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 26: 34(4)

The constants aij are found out by solving equation (11).

The curve fitting can also be done by orthogonal polygonal method, which iscomputationally intensive.

For all practical purposes the co-efficient of ‘xn’ for degrees more than three are negligible,and hence it is enough that the curve fitting is done for a third degree. Hence the equation (8)reduces to

Y = aO + a1x + a2x2 +a3x

2

As the profile is continuously changing and closed, dividing the profile into known numberof segments and then fitting the curve for each segment can obtain a perfect fit. For an apple,the profile can be divided into four segments, as the change in the profile is more or lessuniform in the four quadrants. Such a graph can be drawn f(x) for the standard apple.

Similarly for the sample under analysis, the best-fit curve g(x) depicting its radius signaturecan be drawn as shown in Fig. 2.

The sample can then be analyzed by computing the ratio, (L2/A) where ‘L’ is the totallength of the curve and ‘A’ is the area below the curve. Analysis can also be done by computingradius of curvature at different points. The following equations may be used for finding outthese parameters.

The area below the curve under analysis is given as

A = ∫ ∫ ∫ ∫ ∫ y dx

Fig. 2 Best fit curve

The length of the curve is given as

The local maximum and local minimum for the curve under consideration exists at pointswhere dy / dx = 0. The radius of curvature (C) for the curve is given as

Classification of fruits based on shape using Image-Processing Techniques 233

Page 27: 34(4)

C = (1/R) = (d2y / dx2) / [I+(dy/dx)2]3/2

The bending (concavity and the convexity) of the curve is determined by the values ofd2y / dx2 = 0. The points of inflexion exist for the curve where d2 y / dx2 = 0 and d3y / dx3

= 0. For a regular object the number of maximum and minimum points and the points ofinflexions are minimum.

9. FOURIER TRANSFORMTreating the boundary signature as a one-dimensional digital signal, it can be translated

to Fourier domain as

Where, IF(u)I is the magnitude at harmonic frequency ‘u’ in the Fourier domain as shownin Fig.3. Here, only the magnitude is taken into consideration.

These harmonics will then represent the shape information of the object. The boundarynormalization and Fourier transformation will achieve significant information compression. Theharmonic components in the Fourier domain represent the magnitude of boundary frequencychanges in spatial domain R(k) of the radius boundary sequence. Specifically, F(O) will givethe average radius after normalization, F(1) will give the bending of an object, F(2) will givethe elongation of the object and so on.

10. RESULTS and DISCUSSIONSExperiments were performed using the experimental setup as described earlier. Apple fruits

of different sizes and shapes were obtained from MIs Himachal Pradesh Horticultural Producersand Marketing Corporation (HPMC), Shimla, India, one of the major apple dealers of the

Fig. 3. Fourier Domain

234 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 28: 34(4)

country. Four different postures of the test apple are imaged and digitized by the computersystem for shape analysis. Algorithms were developed to obtain consistent and accurate resultsfor shape classification of apples. For effective shape information extraction, a method ofharmonics multiplied by its magnitude F (h) * hm was established to provide an effectiveheuristic. It is determined as separator S. the higher the S, the more severity is the irregularityof the shape.

S = ∑∑∑∑∑F(h)*hm, m = 1,2,3

Of the many tests and experiments carried out by us with large number of apples, resultsof ten apples labeled as T1 ,T2 ... T1 0 are presented in Table-I. In this Table F1.. F10represent the first ten harmonics values of the Fourier descriptors and S represent the shapeseparator values.

Table-1 Harmonics data and shape separator values.

T1 T2 T3 T4 T5 T6 T7 T8 T9 T10

F0 91.785 88.8639 90.9192 82.0989 91.5184 91.7262 90.3222 96.1471 93.1194 89.7739

F1 23497.0 22749.0 23275.0 21017.0 23429.0 23482.0 23122.0 24614.0 23839.0 22982.0

F2 266.00 144.00 137.00 280.00 58.00 320.00 248.00 179.00 397.00 328.00

F3 296.00 418.00 626.00 532.00 494.00 205.00 614.00 262.00 935.00 810.00

F4 521.00 556.00 360.00 212.00 297.00 477.00 493.00 311.00 524.00 354.00

F5 211.00 167.00 327.00 219.00 265.00 245.00 381.00 403.00 183.00 399.00

F6 125.00 70.00 133.00 145.00 145.00 170.00 30.00 108.00 58.00 37.00

F7 173.00 176.00 154.00 88.00 157.00 54.00 101.00 111.00 64.00 186.00

F8 170.00 100.00 114.00 95.00 62.00 65.00 76.00 103.00 102.00 101.00

F9 66.00 53.00 45.00 36.00 25.00 32.00 29.00 33.00 29.00 77.00

F10 56.00 27.00 36.00 40.00 25.00 23.00 62.00 75.00 49.00 22.00

S 0.4802 0.3795 0.4167 0.3412 0.3416 0.3155 0.3824 0.4176 0.3642 0.4274

From the values above, it can be seen that test apples T4 and T5 have very close shapeseparator value, which can be confirmed from their gray level images as shown in Fig.4and Fig.5

The two apple’s T 4 and T5 boundary signature is drawn after normalization and it isseemed to be overlapping as shown in Fig.6

It can also be shown that the first 10 harmonics of the Fourier descriptors are same andthey can be as shown in Fig.7

Classification of fruits based on shape using Image-Processing Techniques 235

Page 29: 34(4)

Fig. 4 - Test Apple T4 Fig. 5 - Test Apple T5

Fig. 6. Boundary signatures of test apples T4 and T5

Fig. 7. Fourier descriptors of test apples T4 and T5

236 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 30: 34(4)

The same fact that the apples belong to the same category of shapes can be confirmedwith the correlation coefficient. The correlation between the above two images is found tobe 0.7783, that is more than a positive 0.5 correlation, which proves that the shapes are almostsimilar. The facts can be proved taking another example of T8 and T10 test apples, whichhave got a separator of 0.4176 and 0.4274 respectively. The correlation coefficient betweentheir radius signatures was also found to be 0.7564.

For example if we take apples of dissimilar shapes say T4 and T8 as shown in Fig. 8,then the Fourier transforms obtained in the graph are given in Fig. 9.

Fig. 8. Test apples T4 and T8

Fig. 9. Fourier descriptors of test apples T4 and T8

Classification of fruits based on shape using Image-Processing Techniques 237

Page 31: 34(4)

Fig. 10. Shape separator variation

The small variation in the Fourier descriptors in its first harmonic is enough to distinguishbetween the two apples, which have shapes like that of a circle and ellipse respectively. Thecorrelation coefficient techniques supports the same fact by giving the correlation as 0.4102,less than a +0.5, hence confirming that the above two apples belong to different categories.

In this manner the apples can be graded by setting the threshold values and ranges forthe shape separators. (Fig.10). The Thresholding and range setting can be done as shown inthe Table II in which, the values lying in the regions shown are categorized as one group.Then the groups formed are

Table II Classification of apples using shape separator values

Groups Shape separator range Apples belonging to the category

Group I 0.3 - 035 T4, T5, T6

Group II 0.35 - 0.4 T2, T7, T9

Group III 0.4 - 0.45 T3, T8, T10

Group IV 0.45 - 0.5 T1

11. CONCLUSIONThe new method proposed namely Improved Radius signature, have been discussed. This

method can be effectively used for comparing many samples against a reference shape forthe purpose of sorting and grading using correlation coefficient technique, Graphical analysisor Fourier transformation technique. These techniques are very useful for automatic grading

238 P. Sudhakara Rao, A. Gopal, S. Md. Iqbal, R. Revathy and K. Meenakshi

Page 32: 34(4)

of apples. In all these techniques, there is no limitation on the number of boundary points forcomparing the correlation coefficients. The main criteria adopted, is to develop algorithmsfor less expensive implementation. In image processing analysis, algorithms for computingdifferent parameters plays a vital role in achieving the desired results.

Presently the throughput of the system is two fruits per sec and the execution time forthe entire parameters colour, size and shape is taken care to meet the desired throughput. Asthe shape analysis is one of the many criteria to sort the fruits, considering all the parametersof interest, the algorithm’s processing time has large bearing on the system throughput beforethe object is sorted. One of the limiting factors being the processing time, sorting is beingimplemented on parallel lines with parallel hardware to meet the throughput demands.

REFERENCES1. Machine vision for colour inspection of potatoes and apple -Tao Y. et. al ASAE 38; 1555-

1561, (1995a).

2. A Fourier based separation techniques for Shape grading of potatoes using machine vision- Tao Y. et, al, ASAE 35:949-957, (1995b).

3. Grading of mushrooms using a machine vision system - Heinemann et. al - ASAE Vol. 37(5);1671-1677, 1994.

4. Shape feature extraction and classification of food material using computer vision -Gunasekaran. S and Ding.K- ASAE, Vol.37 (5): 1537-1545.

5. Machine vision inspection of golden delicious apples - Heinemann et al. ASAE Vol 11 (6)901-906, 1995.

6. Boundary estimation in complex imagery using Fourier descriptors - Jiang T and Merickel =Proc. of Intl. Cont. on Pattern Recognition 1: 187-190, 1988.

7. Digital image processing - Gonzalez. R.C and Wintz -1988.

8. Apple shape inspection with computer vision - Lee man et al. - Proc. of Intl. Cont. on Sensorsfor non destructive testing - Measuring the quality of fresh fruits and vegetables - 1997.

Classification of fruits based on shape using Image-Processing Techniques 239

Page 33: 34(4)

DIGITAL AND FAST PREDICTION OF MULTIPLEQUALITY PARAMETERS OF GND WATER

S.B. Kalyanaraman and G. GeethaPhysics Wing (DDE), Annamalai University, Annamalainagar - 608 002

ABSTRACT

Generally, among the various quality parameters of ground water, the ElectricalConductivity (EC) has significantly high correlation co-efficient ( > 0.75 to 0.99) withother parameters like Total Dissolved Solids (TDS), Total Hardness (TH), Chloride (Cl),Magnesium (Mg), Total Alkalinity (TA), Sulphate (SO

4)

’ Sodium (Na), Calcium (Ca),...

etc. Just by measuring the EC only it is possible for simultaneous prediction of otherparameters having very high correlation co-efficient values with EC.

1. INTRODUCTIONMany parameters characterising the quality of ground water are correlated to each other.

The linear regression equation of the type y = A + BX (where, A and B are regression co-efficients, X is the independent variable and Y is the dependent variable) is used and thecorrelation coefficient ‘r’ is found between the EC and other parameters [1,2]. It has beenestablished by many earlier workers that the EC has very high correlation with TDS (r ~ 0.9)and significantly high Correlation with some other parameters like Cl, TH, Mg, TA, etc. [3, 4,5 and 6]. Measurement of each parameter involves different instrumentation, procedures, andusually consumes more time [1].

Thus, an accurate measurement of EC alone, along with the known values of A and Bfor the parameters in the area of interest obtained through earlier measurement enables fastprediction of these parameters.

2. MATERIALS AND METHODSFifteen sites were selected in a village housing a sugar-cane distillery plant and the physico-

chemical parameters of the ground water were analysed by standard methods (APHA, 1989)[7] and given in Table - II as observed values.

3. STATISTICAL ANALYSIS

Correlation co-efficient and Regression co-efficientUsing sx package, the co-efficient r, A and B are evaluated for EC paired with other

parameters. The values obtained are presented in Table I.

J. Instrum. Soc. India 34 (4) 240-244

Page 34: 34(4)

Table - ICorrelation and Regression coefficient values for EC paired

with some water quality parameters

Water Quality Correlation Regression coefficientParameters coefficient

r A B

TDS 1.0000 0.001152 699.999Cl 0.9867 -110.793 229.834

Mg 0.8959 -30.9479 38.5TH 0.8941 -13.6307 314.836TA 0.8904 121.653 206.111

4. SYSTEM DESCRIPTIONA microcontroller (8051) based system is designed [8] for the ast and simuhaneous

prediction of some of the water quality parameters. The block diagram of the system is shownin figure 1.

Fig.1. Block diagram of digital, multiple prediction of GND water qualityparameter system

The data acquisition block comprises of a 1 MHz oscillator, Electrical conductivity cell(coll constant 1.0) current to voltage Converter, rectifier, op-amp buffer and an op-amp scalechanger. The output of this unit is adjusted to have values in the range 0 to 5 volts, so as tobe compatible with the Analog to digital converter (ADC 0809). The digital output of ADC isprocessed by the embedded J1C 8051 card with suitable ‘C’ program loaded in it.

4.1. Analog to digital converter (ADC 0809):Analog to digital conversion is obtained using add-on board consisting of ADC 0809

interfaced with programmable peripheral interface IC 8255.

4.2. Embedded Microcontroller Board:8051 CPU card [VPC-51] with 8KB RAM, 8KB EPROM, provided with RS232 serial

port is used to store and execute the program [9].

Digital and fast prediction of multiple quality parameters of GND water 241

Page 35: 34(4)

4.3. 16 x 2 Dot matrix Alphanumeric display cardVAPC - 013 model LCD display card is used to display the results. It has eight on board

keys to enter the data. Two rows each with 16 column LCD dot matrix is available to displaythe parameter name and the corresponding value.

5. WORKINGCalibration of the data acquisition system is first done, by initially immersing the

conductivity cell in the prepared standard solutions of KCI with mole &actions M = 0.01 andM = 0.001.

The cell is then thoroughly washed with distilled water and used for the measurement ofEC of the collected water samples. The regression co-efficient values of A and B Correspondingto the parameters obtained from preliminary studies in the given area are entered into theprogram in an order using the keys in the alphanumeric display card. The execution of theprogram results in the water quality parameters displayed one after the other in successionand in the given order.

The observed and predicted value of EC and other parameters for the water samplesunder study are presented in Table II.

6. CONCLUSIONThe developed microcontroller based system for fast prediction of water quality parameters

is highly sophisticated but simple to calibrate and use. The portability of the system enablesfield measurements leading to fast monitoring of these parameters.

ACKNOWLEDGEMENTThe authors are very much thankful to Vi-microsystems, Chennai for providing embedded

microcontroller, ADC and display cards and facilities to implement this work.

REFERENCES1. Draper and Smith. Applied Regression analysis (2nd edn) Jolm wiley and sons. New York

(1981).

2. Somasekhara Rao, K. “Correlations among water quality parameters of ground water of NuzvidTown and Nuzvid mandalam,” Ind. J. Env. Prot, 13(4), 261-266(1993).

3. Jain C.K. and Shanna M.K., “Regression analysis of ground water quality data of Sagar district,Madhya Pradesh”, Ind. 1. Env. Hlth., 42(4) 159-168 (2000).

4. Singanan A, Somasekhara Rao K. and Rambabu C., “A correlation study on physico-chemicalcharacteristic of ground water in Rameswaram Island”, Ind. 1. Env. Prot, 15(3), 213-217(1995).

5. Vengatachalam M.R. and Jebanesan A, “Correlations among water quality parameters forground waters in Chidambaram Town”, Ind. J. Env. Prot, 18(10), 734-738 (1998).

6. Garg V.K., Gupta R., Goel S., Taneja M. and Khurana B., “Assessment of underground drinkingwater quality in Eastern part of Hisar”, Ind. 1. Env. Prot., 20(6), 407-412 (2000).

7. APHA Standard methods for the examination of water and wastewater. 17th edition (1989)Washington, D.C.

8. Kenneth J. Ayala. The 8051 Microcontroller, Architecture, Program and Applications, PentramInternationals, India.

9. Technical manual - Intel 8051 - VI Microsystems Pvt Ltd. Chennai.

242 S.B. Kalyanaraman and G. Geetha

Page 36: 34(4)

Tabl

e -

II

The

obs

erve

d an

d pr

edic

ted

valu

es o

f so

me

GN

D w

ater

qua

lity

para

met

ers

in a

vill

age

situ

ated

wit

h a

Dis

tille

ry p

lant

Sl.

No.

EC

(mm

hos)

TD

S (m

g/I)

Obs

erve

d P

redi

cted

Cl (

mg/

I)

Obs

erve

d P

redi

cted

TH

(mg/

I)

Obs

erve

d P

redi

cted

TA (m

g/I)

Obs

erve

d P

redi

cted

Mg

(mg/

I)

Obs

erve

d P

redi

cted

11.

5510

8510

8522

624

5.45

380

474.

3747

044

1.13

2328

.79

21.

3997

397

319

320

8.68

464

423.

9948

040

8.15

2322

.62

31.

3191

791

719

319

0.29

376

398.

8038

839

1.66

2319

.54

41.

5610

9210

9224

824

7.75

576

477.

5144

844

3.19

2529

.17

51.

1781

981

916

715

8.11

376

354.

7336

036

2.80

1214

.14

62.

014

0014

0035

934

8.88

612

616.

0454

053

3.88

5046

.13

71.

9213

4413

4434

033

0.49

595

590.

8548

051

7.39

4843

.05

80.

9767

967

913

211

2.15

280

291.

7630

032

1.58

126.

43

91.

3393

193

118

519

4.89

350

405.

1034

039

5.78

720

.31

101.

1177

777

713

814

4.32

360

335.

8436

035

0.44

1911

.83

Digital and fast prediction of multiple quality parameters of GND water 243

Page 37: 34(4)

MEASUREMENT OF SQUARE OF REFLECTANCEUSING PRINCIPLE OF RATIOS

Om PrakashDivision of Standards, National Physical Laboratory, New Delhi – 110 012, India

ABSTRACT

A simple design of a spectrometer is suggested to measure the reflectance of planespecularly reflecting surfaces at normal / near normal incidence. Spectral energydistribution of reflectance of samples can be recorded by this instrument. The divisionof incident radiation beam into the sample and the reference beams facilitates to overcomethe difficulty arising due to source fluctuations. The performance of experiment in twoSteps helps effectively to obviate the effects because of (i) difference in path lengths ofthe sample and the reference beams, (ii) polarization produced by optical components,and (iii) stray light etc. The use of two integrating spheres attached with separatedetectors facilitates the setting of the beam splitters not so critical. The uncertaintywith the measurement is reduced as the square of the reflectance is measured.

I. INTRODUCTIONThe reflectance of optical components such as glass windows, filters and thin films

coatings etc. is an important parameter used to characterize them. The researchers have beenvery keen to measure the reflectance of these components at normal incidence, because atnormal / near normal incidence reflectance does not change significantly. Moreover, thereflectance value is nearly constant and free from polarization effects. But the measurementof reflectance at normal incidence is difficult, if it is not impossible. The reason for the sameis that for measurements in this situation, the radiation source and the detector are to be alignedin the same direction. A number of methods to measure the reflectance of samples of varyingnatures have been reported in the literature [1-5]. In the present article, a simple design of aspectrometer has been suggested. It is capable of measuring the reflectance at normal incidencenearly free from above mentioned effects etc.

2. WORKING PRINCIPLEThe schematic diagram of experimental set up is shown in Fig. 1. Here, O, is optics, C,

chopper; M, monochromator, S, sample; B1 and B2, beam splitters; S1 and S2, integratingspheres attached with screens P1 and P2, and detectors D1 and D2, respectively, L1 and L2,two similar Lock-in-Amplifiers; and R, ratio meter. A beam of monochromatic light from themonochromator M, is collected by the optics O and rendered parallel to reach the aperture A.

J. Instrum. Soc. India 34 (4) 244-248

Page 38: 34(4)

The size of this parallel beam is reduced in such a way so that it may pass through the openingof the chopper and is incident on beam splitter B1. At the surface of the beam splitter B1,the light beam is divided into two parts. A part beam is recorded at the sample surface. Thesample surface is oriented so that the incident rays fall upon it in the normal direction to itssurface. In this situation, the incident beam after reflection retraces its path back and afterpassing through the beam splitter B1, enters into the integrating sphere S1. On the otherhand, the part of the beam transmitted through beam splitter B1, is reflected from beam splitterB2, and enters into the integrating sphere S2. The screens P1 and P2 attached with theintegrating spheres S1 and S2, prevent the respective light beams falling directly on the detectorsD1 and D2 respectively. The light flux in the integrating spheres is converted into A.C. electricalsignals by the photo detectors, D1 and D2, because of chopping of the light beams and theseare proportional to the incident light fluxes in the spheres. The Lock-in-Amplifiers convertthe A.C. electrical signals from detectors D1 and D2, into D.C. signals and they are detectedand rationed by the ratio meter. The experiment can be performed in two Steps as discussedin the following paragraphs.

2.1. Step 1Let the intensity of the radiant flux coming out of aperture A is I. After reflection from

beam splitter B1 and sample S and transmitted through B1, the intensity flux I is reduced toa value of I.r.R1.t. Where, r and t, are the reflectance and transmittance of beam splitter B1;and R1 is the reflectance of sample S, to be measured. Similarly, the intensity of radiationflux transmitted through beam splitter B1, reflected from beam splitter B2, and reaching sphereS2 will be I.t.r. Here it has been supposed that B1 and B2 are identical beam splitters, so thattheir reflectances and transmittances may be taken to be the same. Let the radiation to electricalsignal conversion factor of detector D1 along with sphere S1 is F1(x) at wavelength x andthat of Detector D2 along with sphere S2 is F2(x). Let the ratio of electrical signals fromdetectors, D1 and D2 as detected through Lock-in-Amplifers L1 and L2 respectively, andrecorded by the ratio meter R, is K1, then one may write,

I.r.R1.t.F1(x) / I.t.r.F2(x) = K1

Or,

R1.F1(x) / F2 (x) = K1 (1)

2.2. Step 2In this step the position of integrating spheres S1 and S2 along with their respective

detectors are interchanged. Let the ratio of electrical signals from sample and reference beamchannels is K2, then one may write,

I.r.R1.t.F2(x) / I.t.r.F1 (x) = K2

Or,

R1. F2(x) / F1 (x) = K2 (2)

On multiplying (1) and (2), one may get,

R1. R1 = K1. K2

Measurement of square of reflectance using principle of ratios 245

Page 39: 34(4)

246 Om Prakash

It is evident from the above expression that the square of the reflectance of the samplecan be measured easily just by recording the ratio of the electrical signals due to referenceand sample beams in the two Steps of experiments as mentioned above.

3. OBSERVATIONSThe radiation source used in a tungsten halogen lamp. A variable frequency mechanical

chopper chops the light beam at an appropriate frequency. The reference signal to be fedinto the Lock-in-Amplifiers is derived from the chopper at the chopping frequency from builtin reference signal generator in it. The D.C. signals from the Lock-in-Amplifiers are fed intothe Ratio meter that records the ratio of these signals i.e. values of k1 and K2, in Step 1 andStep 2, respectively, in the experiment. The sample holder is in the form of a mirror mountthat has got provision to tilt the incident in the horizontal as well as vertical directions, and isheld in a mount with adjustable height.

In order to carryout the experiment nine samples of mirrors, optical flat, microscopeslide and quartz neutral density filter etc. were procured. Each sample was placed at theproper place in the sample holder and its surface was oriented in such a way that the incidentrays fall on it normally and are reflected back to retrace their path back and enter into thebeam splitter B1 and integrating sphere S1. The other portion of the light beam incident onbeam splitter B2 is reflected and enters into the integrating sphere S2. The central rays of thereflected beam (the sample beam) and the reference beam are adjusted to the same heights sothat they may enter the integrating spheres in Step 1 and Step 2, without any obstruction,when the integrating spheres are interchanged. The ratio meter readings, K1 and K2,corresponding to the positions of Step 1 and Step 2, were recorded for each sample. Thevalues of reflectance square and reflectance for each sample were calculated. Sampleidentification (ID), values of K1 and K2, (R1)2 and (R1) are given in Table 1. As is evidentfrom Table 1, the reflectance values measured for similar nature of samples are higher thanthe reported values. The reason for the same is that in the present case the reflectance has

Table 1. Sample ID, ratio meter readins, K1 and K2, respectively, K1. K2 and Reflectance, R1.

Sl. Sample ID Ratio meter Ratio meter K1, K2 Reflectancereading reading

K1 K2 (R1)2 R1

1. Front coated Gold mirror 0.5078 0.4565 0.23181 0.4814

2. Back coated Aluminum mirror 0.5078 0.4420 0.22444 0.4737

3. Front coated Aluminium mirror 0.5161 0.4347 0.22434 0.4736

4. Optical flat small 0.3348 0.0735 0.02460 0.1568

5. Transmission Grating 0.3593 0.0735 0.02640 0.1624

6. Microscope slide 0.2941 0.0629 0.01849 0.1359

7. Sapphire window 0.3174 0.0597 0.01894 0.1376

8. Fused Quartz window (Plate) 0.3733 0.0928 0.03464 0.1861

9. ND Filter Ealing, 30% T integrated 0.4794 0.1151 0.05517 0.2348

Page 40: 34(4)

been measured for the integrated light whereas the reported values are spectral ones at selectedwavelength bands. The reproducibility of the observations of K1 and K2, was fairly good, asis evident from Table 2.

4. DISCUSSIONReferring to the Fig. 1, it is evident that the beam splitters are set nearly at 45 degrees to

the direction of incident radiation. Also, the sample surface, whose reflectance is to bedetermined can be set at normal / near normal incidence. The settings of both of the above

Table 2. Reproducibility of observations of ratio meter i.e. K1 and K2. Run No., values of K1and K2, for the sample of front-coated gold mirror.

Run No. Value of K1 Value of K2

1 0.5078 0.4565

2. 0.5078 0.4565

3. 0.5078 0.4565

4. 0.5078 0.4565

5. 0.5078 0.4565

Fig.1. Schematic ray diagram of experimental set up for measurement of square ofreflectance at normal incidence.

Measurement of square of reflectance using principle of ratios 247

Page 41: 34(4)

components are not critical because of use of integrating spheres. Though in equation (1)and (2), the terms for effects of polarization and differences in path lengths of the beamshave not been included, even then it is easy to show that in rationing operation in two stepsof experiments, their effects are neutralized. Evidently, the expression for R1. R1 does notinclude any parameters such as beam intensity I; reflectances, r; transmittances, t, of beamsplitters; and the response functions of the detectors, F1(x) and F2(x), etc. and is onlydependent on the ratios of electrical signals due to the sample and the reference beams in thetwo Steps of the experiment. The uncertainty in the measurements of reflectance is furtherreduced because of measurement of the value of R1.R1, i.e. the square of reflectance of thesample. The signal to noise ratio in the measurements is appreciably increased because, ofprocessing electronics. The measurements can be performed in the ambient light, as only thechopped radiation is detected by the lock-in-amplifiers at the chopping frequency of beams.Only the detector signal corresponding to the chopping frequency of beams are amplified anddetected, the other frequency signals such as noise etc. are rejected by the lock-in-amplifiers.A PC can also be used to monitor the measurement of the monochromator drive. A softwareprogramme may be developed for this purpose and also to perform other functions such asto record the observations of ratio meter and to acquiring and storing data for future use.The uncertainty with the reflectance measurement results is about 5(10-5). This shows theusefulness of this method.

REFERENCES1. H. Takahashi, M. Kimura and R. Sano, Opt. & Las. Technol. 21 (1989) 39.

2. P. Wu, P. Gu and J. Tang, Appl. Opt. 33 (1994), 1975.

3. J. Stone and L.W. Stultz, Appl. Opt. 29 (1990) 583.

4. G. Bader, P.V. Ashrit, S. Eloualik, F.E. Girouard and Vo-Van Troung, Rev. Sci. Instrum.62 (1991) 2398.

5. S. Gil, G.A. Clarke, L.Mc. Garry and C.E. Waltham, Appl. Opt. 34 (1995) 695.

248 Om Prakash

Page 42: 34(4)

MICRO-CONTROLLER BASED MONITOR FORARC WELDING ANALYSIS

P. Kavitha*, T. Balasubramanian* and S. Manoharan***Department of Physics, Regional Engineering College. Trichirappalli - 620 015

**Welding Research Institute, BHEL. Trichirappalli - 620 014

ABSTRACT

Welding is the accepted joining process for various metals all over the world. Arc weldingis employed for joining of thick plates and pressure vessels. Various welding equipmentmanufactures are manufacturing new models of welding power sources every year in alldeveloped countries. Presently there is a need to identify the quality of weldingproduced by each of these different welding power sources. Hence, the parametersaffecting the arc welding quality such as arc voltage and arc current have to be monitoredand analyzed thoroughly. In order study these parameters, a micro controller basedmonitor for arc welding system has been developed based on today’s electronicstechnology. Embedded systems are used for this purpose, necessary hardware andsoftware were assembled, welding parameters were monitored and the process signalswere analyzed. With the developed monitor and software one can identify and choosethe best welding power source for given application among the variety available in themarket.

1. INTRODUCTIONMonitoring of different types of welding and control of the current using micro-controller

has been explored recently [l]. An arc-welding monitor using 8085 microprocessor-basedsystem has been developed and reported earlier [2]. Such microprocessor based weldingmonitors are extremely used for testing of various commercial welding power sourcesmanufactured in India [3]. Research works are also carried out using those welding monitorswith micro-controllers [4].

During arc welding, the arc voltage and current vary depending upon the various setparameters. Variation of these parameters affect the arc welding quality. The arc voltage andcurrent behavior have to be studied for effective parameter estimation. Thus, the arc voltageand current are monitored during the welding process through a signal conditioner employedin the setup.

The signal conditioner is used to reduce the high voltage and current levels to the desiredlevel of 0-5 volts. Signals are acquired and stored in the micro-controller. The micro-controllerprogram operates through the front-end interface, which is a visual basic program. This enables

J. Instrum. Soc. India 34 (4) 249-256

Page 43: 34(4)

250 P. Kavitha, T. Balasubramanian and S. Manoharan

one to transfer the data from micro-controller to a default directory in order to compute andplot the parametric graphs for the captured data. From the graphs, the arc welding stabilityand quality of welding can be analyzed.

2. HARDWARE DESCRIPTIONThe hardware that acquires the voltage and current signals during the welding process is

shown in Fig. 1. Intel 8051 is a versatile micro controller employed in the control designs[5]. The current and voltage signals are processed by an analog signal conditioner, andtransferred to an 8-bit micro-controller to be analyzed by the PC program.

3. DATA ACQUISITIONThe circuit diagram used to process the arc voltage and current is shown in Fig.2.

Voltage AttenuatorThis unit comprises of five relays, which are controlled by the micro controller through

8051 Interface. Using software, this unit can be made to operate each relay separately.

Current AttenuatorThe welding current falls in the range of 40 to 400 amperes. The current from the welding

process is passed through a shunt. The shunt used here has a rating of 1000A-75mV. Currentswith in the range (l000A) are converted to a standard value of 75mV.

Figure 1. Block diagram of Micro-controller Based Monitor for Arc Welding analysis

Page 44: 34(4)

Fig

ure

2.

Cir

cuit

Dia

gram

of

Ana

log

Sign

al c

ondi

tion

er

Micro-controller based monitor for arc welding analysis 251

Page 45: 34(4)

ClipperA clipper circuit is used to clip off a certain portion of the weld voltage to obtain a desired

output voltage.

Low Pass FilterThe third order low pass Butterworth filter is used ‘to provide a flat frequency response:

Thi: cutoff frequency of the filter is 1 KHz. The arc welding voltage has a large amount ofnoise in the higher frequency, Hence a low pass filter is used to suppress the highfrequency components.

ComparatorThe comparator used here is for indication. Its output is connected to a LED. The LED

glows if an over range value occurs. The voltage can then be reduced further to the compatiblelevel.

Multiplexer (HI508)HI508 is a single 8 input differential 4 channel CMOS analog multiplexer. This chip includes

an array of eight analog switches, a digital decode circuit for channel selection, a voltagereference for logic thresholds, and an ENABLE input for device selection when severalmultiplexers are present. In this experimental setup, multiplexer is used to select the voltageand current alternatively.

Analog to Digital Converter (AD574A)The analog signal conditioner ADC is set to 12-bit data conversion mode in the present

setup. The CS and Ao pins are grounded. CE and R/C pin are used to initiate the data readand conversion operations. The analog input range is limited to 10V and the 12-bit digital outputis sent to the 8051 micro controller.

Micro-Controller 8051The micro-controller is mainly employed for following three purposes To read and store the data. To select the relay circuit. To transfer the stored data to host computer.

The flow diagram of the 8051 micro-controller is shown in Fig. 3. A high-level programdeveloped in Visual Basic is accessed for this Micro-controller.

4. INTERFACING THE PC AND MICRO-CONTROLLER4.1 The Hardware

The 26 Pin parallel port is used to interface the micro-controller and the analog signalconditioner. The signals are processed and stored in the micro-controller. The RS- 232C serialport is used to interface the computer and micro-controller. Data acquisition software programhas been developed to acquire the data that will be stored in the Pc.

4.2 The Software Descriptiona. Main Form

The main form consist of menus such as File, Settings, Acquire, graph, Help and Exit.

252 P. Kavitha, T. Balasubramanian and S. Manoharan

Page 46: 34(4)

The file menu has the menu items such as default directory, load settings, save settingsand Exit. This section is used to store the data in the default directory. The settings menu isused to select the relay, samples, the COM port, and mode. The setting menu window isshown in Fig. 4.

The acquire menu has menu items such as set parameters, capture, and transfer data.The data acquisition window is shown in Fig. 5.

The transfer menu is used to send the transfer command to the master controller. TheMSCOMM control is used to capture the data and store it in a separate file.

Figure 3. Flow Diagram of 8051 Micro-controller Program

Micro-controller based monitor for arc welding analysis 253

Page 47: 34(4)

b. Graph FormThe X and Y-axis are automatically plotted based on the samples per second settings.

The peak average current, average current, spatter index and ignition index are evaluated fromthe graph. The graph window is sown in Fig. 6.

Figure 5. Data Acquiring window

Figure 4. Parameter Setting window

254 P. Kavitha, T. Balasubramanian and S. Manoharan

Page 48: 34(4)

5. FORMULAE

Figure 6. Graph window

Micro-controller based monitor for arc welding analysis 255

Page 49: 34(4)

6. CONCLUSIONAnalyzing the arc welding with the help of computers is an added advantage because of

its speed, accuracy and efficiency. The system has been developed with a variety of usefulfeatures. A software program is used to analyze the arc welding with GUI capabilities. Theaverage current, peak average current, spatter index, ignition index are calculated. From thecalculated value, the stability and the quality of welding have been analyzed. With the developedmonitor and software one can identify and choose the best welding power source for givenapplication from among the wide variety available in the market.

REFERENCES1. V.R. Samule, K. Asok kumar and S. Manoharan, ‘Emerging trends in Microprocessor based

monitoring and control system for welding applications-developments in WRI’, WRI Journal,Vol. 20, No 3. pp 81-88

2. S. Manoharan, VR. Samuel et.al, ‘Microprocessor based on line monitor for arc welding’ bypublished in WRI Journal, Vol. 13, no.2

3. K. Padmanaban and S. Manoharan, ‘Selection of power source for narrow gap SA W’, presentedin National welding seminar held at Bangalore in 1997.

4. S. Manoharan, K.L. Rohira, A. Raja and Dr. S. Palani, ‘Evaluation of welding power source’by published in India Welding Journal, April-June 1991.

5. Hand book on ‘Intel 8-bit Embedded Controller’.

256 P. Kavitha, T. Balasubramanian and S. Manoharan

Page 50: 34(4)

MICRO CONTROLLER BASED CONDUCTIVITYMETER

P. Bhaskar*, Parvathi C.S.*, Nagabhushana Katte**,K. Nagabhushan Raju** and K. Malakondaiah**

*Department of Instrumentation Technology, P. G. Centre, Yarigera, Raichur - 584 133,Karnataka State, India

**Department of Instrumentation, Sri Krishnadevaraya University, Anantapur - 515 003.(A.P.) India

ABSTRACT

A microcontroller based conductivity meter has been designed and fabricated. The useof microcontroller (8031) has made the instrument compact and versatile. A 41/2 digitAID converter is used to improve the accuracy and resolution of the system. Thetemperature measurement is also provided in the system since the conductivity of thesolution varies with temperature. The paper deals with the hardware and software featuresof the system.

1. INTRODUCTIONElectrolytic conductivity is a measure of the ability of a solution to carry an electrical

current1. Solution of electrolytes conducts an electrical current by the migration of ions underthe influence of electric field. The conductance of a solution is the reciprocal of the electricalresistance. It is expressed in mhos. Conductance C is directly proportional to the cross sectionalarea ‘A’ and inversely proportional to the length’ I’ of a uniform conductor.

Thus, C α A/l = K A/l

Where, K is a proportionality constant called specific conductance or conductivity. ‘A’and ‘I’ are numerically equal, the specific conductance becomes the conductance and henceit can be measured in terms of conductance.

2. INSTRUMENTATION

2.1. PrincipleThe operational amplifier in inverting configuration is used for measurement of conductivity

of an electrolytic solution2. It consists of an input resistance Ri and a feed back resistanceRf. Here the conductivity cell is used as Ri. The output voltage of the op-amp is given by

VO = -(Rt/Ri) * ViRi = -(Rt/Vo) * Vi

J. Instrum. Soc. India 34 (4) 257-267

Page 51: 34(4)

258 P. Bhaskar, Parvathi C.S., Nagabhushana Katte, K. Nagabhushan Raju and K. Malakondaiah

When Vi = lV, and Rf = 1000 Ohms

K = C = 1/Ri

K = VO/1000

K = VOm mhos

So, the inverting amplifier is designed such that the reciprocal of the input resistance isequal to the output voltage, which is then acquired by the microcontroller through ADC andfinally it is displayed on LCD module. Simultaneously the measurement of temperature is alsoincorporated in the system.

2.2. Hardware DetailsThe block diagram of the system is shown in the Figure. 1. The functional blocks of the

system are given below.

Figure. 1. Block Diagram of Microcontroller based Conductivity Meter

(a) Conductivity Cell(b) Inverting Amplifier(c) A.C. Source(d) Precision Rectifier(e) AID Converter(f) Temperature Measuring Circuit(g) Microcontroller Card with LCD display

The complete schematic diagram of microcontroller based conductivity meter is shownin Figure. 2. The salient features of the individual blocks of the system follow.

Page 52: 34(4)

Fig

ure.

2. S

chem

atic

Dia

gram

of

Mic

roco

ntro

ller

bas

ed C

ondu

ctiv

ity

Met

er

Micro controller based conductivity meter 259

Page 53: 34(4)

260 P. Bhaskar, Parvathi C.S., Nagabhushana Katte, K. Nagabhushan Raju and K. Malakondaiah

(a) Conductivity CellThe Conductivity cell consists of a pair of electrodes that are firmly located in a constant

geometry and which are immersed in an electrolytic solution whose conductivity is to bemeasured. The cell used in the present study consists of two platinum electrodes of 1cm2

cross sectional area that are separated by a distance of 1 cm. The cell constant of the cellused in the present study is 1.01.

(b) Inverting AmplifierThe inverting amplifier containing the conductivity cell as input resistance (Ri) is designed

in such a way that the reciprocal of the input resistance is equal to the output voltage. Thenthe conductance of a solution becomes the output voltage of the operational amplifier in termsof milli-mhos. As the output of operational amplifier is ac, it cannot be directly applied toanalog to digital converter. Hence, a precision rectifier and RC filter convert the signal intodc. The precision rectifier is constructed using two operational amplifiers A2 and A3.

(c) A.C. SourceA.C source consists of function generator, buffer and transformer. A function generator3

is designed using ICL8038 to generate sinusoidal signal for the excitation of the cell. Sinceit’s excellent amplitude and frequency stability it has been chosen in the present study. Theamplitude and frequency of generator is exactly adjusted to 1 V and 1 KHz respectively.

To avoid the loading effects, the sinusoidal signal from the function generator is appliedto cell through an isolation transformer4. As the output signal of the function generator cannotbe given to transformer directly, an emitter follower is employed to drive the transformer.

(d) AID ConverterThe output of the inverting amplifier is given to the 41/2 digit dual slope AID converter

ICL 71359 which is interfaced to the Microcontroller. Figure 3 shows an interface betweenan AID converter with micro controller through 8255. Port-A of 8255 is used to read thedigit strobes and BCD data from AID converter. The output of the AID converter is inmultiplexed BCD form. The BCD data is available on B 1, B2, B4, and B8 lines of the converter.Logic High on one of the digit strobe lines D5, D4, D3, D2, and Dl indicates the presence ofBCD code for corresponding digit on data lines. The data from this AID converter are readby polling the bit corresponding to a strobe line (until that bit goes high) and store the data inreserved memory locations for future reference. After reading over the BCD code for onedigit, the bit corresponding to the strobe line for the next digit is poled until that bit goes Highand stored in the next memory location. The process is repeated until the data for all the5-digits are acquired.

(e) Temperature Measuring CircuitAccurate conductivity measurement depends on temperature compensation5. Hence, an

integrated circuit temperature sensor LM335 is used in the present study to measure temperatureof the solution6. The LM335 is Kelvin’s sensor i.e., its output is 2.73±V/Oc. In order to getthe temperature in degree Celsius, the voltage 2.73V must be subtracted from the output of

Page 54: 34(4)

Fig

ure.

3. S

chem

atic

Dia

gram

of

Mic

roco

ntro

ller

boa

rd

Micro controller based conductivity meter 261

Page 55: 34(4)

262 P. Bhaskar, Parvathi C.S., Nagabhushana Katte, K. Nagabhushan Raju and K. Malakondaiah

the sensor. Operational amplifier produces a reference voltage of 2.73V, which is subtractedfrom the output of the sensor using an operational amplifier in differential mode. The differentialamplifier produces a change 10m V/OC. The output of differential amplifier is given to the41/2 digit AID converter through a relay.

(f) Microcontroller Card with LCD DisplayThe schematic diagram of the micro controller board designed and fabricated by the authors

for dedicated applications is shown in the Figure 3. It contains all the features required forit to function as stand alone system7-8. The board contains the following features.

CPU 8031 (12 MHz)

Program Memory 2764 (8K) 0000H - 1FFFH

Data Memory 6264 (8K) 2000H - 3FFFH

Parallel I/O 8255 4000H - 4003H

Serial Port One (1200 dB)

The role assigned to the microcontroller in the present is to acquire the data from AIDconverter, process it and display the same on LCD display.

The buffered data lines (D0-D7) of micro controller are connected to the data lines ofmemory chips i.e. EPROM and RAM. Since both EPROM (2764) and RAM (6264) requires13 address lines to address over 8K bytes (Registers). Hence the address lines A0 to A12 of8031 micro controller are connected to the address lines of EPROM and RAM respectively.The remaining address lines A13 to A15 are used for selecting the chips with the help of a 3to 8 decoder i.e.74LS138 IC.

The RD and PSEN of the microcontroller are combined with the help of an AND gate(74LS08), the output of the AND gate is connected to the output enable (OE) of EPROM.This arrangement enables the data to be read from the EPROM. Sending the address to EPROMcan access the contents of the memory at any location. Since, the read and write operationswill take place in the RAM, the RD and WR signals are connected to the OE and WE pins ofthe chip respectively. Read and write operations are low active Intel 8255 is a general-purposeprogrammable I/O device designed for simple input/output operations. It has 24 I/O pins thatare divided into three ports of 8-bits each viz., Port-A (PA0-PA7), Port-B (PB0-PB7) andPort-C (PC0-PC7). Port-C can be divided into two four bit ports (PC0-PC3) and (PC4-PC7).

Since there are three ports, two address lines are sufficient to select all the ports. Theaddress lines A0, A1 of the micro controller are connected to the A0, A1 of 8255, which provideport selection. The RD and WR pins of the microcontroller are connected to RD and WR of8255 to synchronize its read/write operations with the micro controller. The data lines Do toD7 of microcontroller are connected to 8 data lines of 8255 for transmitting or receiving thedata from the microcontroller. The CS of the 8255 is connected to the Y2 of the decoder(74LS138). The remaining 16 address lines A0, A1 and A13, A14, A15 are used for determiningthe selection of a port or control register of 8255. The contents of the Control Register arecalled ‘control word’, which specifies an I/O function for each port. Writing a control word

Page 56: 34(4)

when both Ao and Ai are at logic high can access this register. This register is not accessiblefor a read operation.

The addresses of EPROM, RAM, and 8255 are given below.

EPROM - 0000H to 1FFFH (8K bytes)

RAM - 2000H to 3FFFH (8K bytes)

8255 - 4000H Port-A

4001H Port-B

4002H Port-C

4003H Control Register

Two ICs 1488 and 1489 are also incorporated in the microcontroller board for serialcommunication with the host computer. This enables the computer to read the conductivityand temperature data of the sample and store it in the specified file for further analysis. Thisfacility makes the instrument suitable for research applications.

The interfacing of the Liquid Crystal display module with the microcontroller through8255 helps in making the instrument portable. The data bus D0-D7 of display module10 isconnected to PB0-PB7 of port-B of 8255 and the control pins RS and E are connected to thePC0 and PC1 pins of 8255 respectively. The command word is sent to the command registerby making RS LOW through PC1 and the data can be sent to the display RAM by making RSHIGH through PC1. A HIGH pulse of about 40 microseconds duration is sent to the pin E ofthe microcontroller to fetch/send the data/command word to the display, which can be donethrough the pin PC0 of 8255. The controller pin RJW is connected to the ground. The secondpin of the display is connected to the variable terminal of the potentiometer, which varies thecontrast of the display. After measuring the conductivity and temperature of the solution themicrocontroller displays them on the LCD module. The photographs of the system are shownin Figure. 5.

3. SYSTEM WORKING

Since the conductivity cell acts as input resistance (Ri) of the inverting amplifier, theoutput voltage of the amplifier represents the conductivity in milli mhos of the solution undertest. The input of the inverting amplifier is 1 V /l KHz. sinusoidal signal. The precision rectifierand RC filter are used to convert sinusoidal output voltage of the inverting amplifier into puredc voltage. Then it is given to the AID converter through double pole single throughelectromechanical relay, which helps in switching between two measuring parametersi.e., temperature and conductivity. The AID converter provides the digital data correspondingto the parameter under measurement. The same is read by the micro controller throughport-A of Programmable Peripheral Interface device 8255. The port bit P1.0 of micro controlleris used to select one of the two measuring parameters. The micro controller sends HIGHon P1.0 to select the conductivity input, and then acquires it through AID converterand displays the same on the liquid crystal display. Then the microcontroller selects thetemperature by sending LOW on P1.0, and then it acquires and displays the same onliquid crystal display.

Micro controller based conductivity meter 263

Page 57: 34(4)

264 P. Bhaskar, Parvathi C.S., Nagabhushana Katte, K. Nagabhushan Raju and K. Malakondaiah

Figure 4. Flowchart for Microcontroller based Conductivity Meter

Page 58: 34(4)

4. SOFTWARE DETAILSThe role ofthe software in the present study is given below.

(a) To initialize the ports of 8255.

(b) To initialize the liquid crystal display.

(c) To scan the keyboard

(d) To read digital data through AID converter corresponding to the key pressed,process it and displays the same on the liquid crystal display.

The flowchart of the program is presented in the Figure 4. The detailed program is writtenin the microcontroller assembly language.

5. RESULTS AND CONCLUSIONSThe micro controller based conductivity meter is designed and fabricated. The system is

calibrated and its performance is tested with the standard resistors of 500Ω, 1.0 KΩ and10.0 KΩ and some standard solutions11. The results are tabulated in Table-I, which are ingood agreement with the literature values.

Figure.5. Photographs of the Microcontroller based Conductivity Meter

Micro controller based conductivity meter 265

Page 59: 34(4)

REFERENCES1. G. Chatwal and Anand. Instrumental Methods of Chemical Analysis, Himalaya Publishing

House (1998).

2. R.S. Khandpur, Handbook of Analytical Instruments, Tata Mc Graw-Hill Publishing Ltd. (2002)

3. R.F. Coughlin and F.F.Driscoll, Operational Amplifiers and Linear Integrated Circuits, PrenticeHall Inc. (1987).

4. Hobart H.Willard, Lynne L.Merrit, John A.Dean and Frank A.Settle, Instrumental Methods ofAnalysis, CBS Publishers & Distributors (1986).

5. Douglas M.Conidine, Process/Industrial Instruments & Controls Handbook, McGraw HillInternational Editions, 4th edition (1993).

6. J. Michael Jacob, Industrial Control Electronics, Prentice Hall (1988).

7. MCS51 Users Manual, Intel Corporation (1981).

8. R.S. Khandpur, Handbook on Microcomputers, Tata McGraw-Hill Publishing Company Ltd.(1989).

9. Intersil Data book (1988).

10. Oriole’s LCD Module - Series User’s Manual (1996).

11. Samuel Glasstone, Electrochemistry, East-West Press Private Limited (1942).

Table - I

S.No Resistor / Solution Specific SpecificSolution Normality Conductance Conductance

at 25°C at 25°C(10-3mhos Cm-1) (10-3mhos Cm-1)

Present Study Literature Values11

1. 500 Ω 1.9991 2.0000

2. 1000 Ω 0.9994 1. 0000

3. 1O.OK Ω 0.0993 0.1000

4. KCI 0.1N 12.849 12.856

0.01N 1.4090 1.4087

5. AgNO3 0.1N 10.865 10.878

0.01N 12.479 12.476

6. NaI 0.1N 10.873 10.878

0.01N 1.1916 1.1924

266 P. Bhaskar, Parvathi C.S., Nagabhushana Katte, K. Nagabhushan Raju and K. Malakondaiah

Page 60: 34(4)

COMPUTER BASED DC MICROMOTOR SPEEDCONTROL SYSTEM

P. Bhaskar*, Parvathi C.S.*, L. Shrimanth Sudheer* and A.B. Kulkarni***Department of Instrumentation Technology, P.G. Centre, Yaragera, Raichur - 584 133

**Department of Applied Electronics, Gulbarga University, Gulbarga - 585 106

ABSTRACT

A computer based DC micromotor speed control system has been designed andfabricated. In the present study, a Data Acquisition Card has been designed for themeasurement and control of DC micromotor (Minimotor 2230U 015S - 8400 rpm -FaulhaberDC motors, Switzerland). The comparative study of two controllers (PID & ImprovedPID) have been discussed for step response and load variations (magnetic brake). It isobserved that the improved PID controller has better time response in terms of shorterrise time, settling time, no overshoots, undershoots and zero steady state errors thansimple PID controller.

1. INTRODUCTIONThe DC motor speed is one of the important process parameter that is to be monitored

and controlled in Industry. No doubt, several investigators1-4 have designed and fabricatedDC motor speed controllers. But the attempts to control micromotors are rather scarce inspite of their several industrial applications. The DC motors are widely used in the variablespeed applications due to the ease of speed control. In closed-loop system, the speed can bemaintained constant by adjusting the motor terminal voltage5. Smaller DC motors operate atlower voltages, which make them easier to interface with control electronics6.

2. PRINCIPLEThe Fig.1 shows the block diagram of computer based DC micro motor speed control

system. The computer measures the speed of DC micromotor through optical encoder,frequency to voltage (F/V) converter and analog to digital (A/D) converter. The computerthen displays the speed of DC motor on the monitor. After measuring speed of the DC motor,the computer compares it with the set-point speed; the difference is called as error, which isapplied to one of the control algorithms (Proportional + Integral + Derivative (PID) or ImprovedPID). The PID controller gives a good transient as well as steady-state control. It offersrapid proportional response to error, while having an automatic reset from the integral part toeliminate residual error. The derivative section stabilizes the controller and allows it to respondto the rapid changes or transients in error. Hence, no oscillations are found in the PID

J. Instrum. Soc. India 34 (4) 267-276

Page 61: 34(4)

268 P. Bhaskar, Parvathi C.S., L. Shrimanth Sudheer and A.B. Kulkarni

response. The output of these algorithms is a digital value, which is fed to digital to analog(D/A) converter. The output of D/A converter is analog voltage, which is proportional to theerror. The output of D/ A converter is given to the DC motor through buffer. The latter providesthe required current for the DC micromotor. The output of the D/A converter is continuouslyvaried until the speed of the DC motor attains the desired speed.

Fig. 1 Block Diagram of Computer based DC micromotor speed control System.

Fig. 2 Schematic diagram of the speed control of DC Micromotor.

Page 62: 34(4)

3. HARDWARE DETAILSThe Fig.2 shows the complete schematic diagram of the system. It consists of the

following elements.

a. DC Micromotorb. Optical Encoder and F/V converterc. Data Acquisition Cardd. Personal Computere. D/ A converterf. Actuator

a. DC Micromotor

The DC micromotor (2230U 015S) used in this application is a product from FaulhaberDC motors, Minimotor SA 6980 Croglio, Switzerland7. Minimotor products are based on thepatented self-supporting skew wound coil technology. The main features of these micromotorsare: less weight, low power, high speed etc. The lifetime of these motors vary from a fewhundred hours to more than 10,000 hours.

The specifications of DC micromotor (2230U 0158) are :

* Nominal voltage : 15 V

* Output power : 2.63 W

* No load speed : 8400 rpm

* No load current : 0.007 A

* Operating temperature range : -30°C to 85°C

* Commutation : Precious metal

* Magnetic material : Al NiCo

* Weight : 50 gms

b. Optical Encoder and F/V ConverterThe optical encoder is a transducer that is connected to the shaft of the DC micromotor,

which converts the speed of the motor into corresponding frequency. The optical encoderused in this application produces 12 pulses for one revolution. IC LM2907 (F/V converter)converts these TTL compatible pulses into the corresponding voltage. The output voltage ofF/V converter is directly proportional to the speed of the DC motor.

c. Data Acquisition Card

A complete Data Acquisition Card has been designed and fabricated by the authorsindigenously for the present study. The card is designed as a multipurpose card. The cardcan be used not only for measurement but also for controlling the parameter being measured.The card mainly consists of a high speed Analog to Digital (A/D) converter AD1674,Programmable Peripheral Interface (PPI) 8255 and Programmable Interval Timer (PIT) 8254.These three devices are interfaced to computer at different addresses through the I/O slot.

Computer based DC micromotor speed control system 269

Page 63: 34(4)

270 P. Bhaskar, Parvathi C.S., L. Shrimanth Sudheer and A.B. Kulkarni

Interfacing of Analog to Digital Converter AD1674 with Computer:Interfacing of AID converter ADI674 with the computer is presented in Fig.3. The AID

converter is interfaced to the computer through its I/O slot. In computer all the address, dataand control lines are terminated in the I/O slot and these lines are essential for interfacing anyperipheral device to the CPU of the computer8. The address lines (A0 - A9), control lines(AEN, RD, WR, RESET) and data lines (DO - D7) are brought onto the card through twolatches 74LS573 (lCI & IC2) and one bi-directional buffer 74LS245 (IC3). The latches andbuffer are used to protect the address, control and data lines of the CPU. If any short circuitor spark is generated in the external circuit these latches are destroyed and CPU is protected.

Since the I/O slot (ISA) of the computer has only 8-bit data lines, to interface 12 bitAID converter two 8-bit latches 74LS573 (IC4 & IC5) are used as shown in the circuit(Fig. 3). The data lines (DO - D7) are connected to latch (IC4) and remaining data lines (D8-D II) are connected to another latch (IC5). The pins R/C and STATUS of analog-to-digitalconverter are connected to the PCO and PC7 of programmable peripheral interface (PPI) 8255

Fig. 3. Schematic diagram of Data Acquisition Card.

respectively to start conversion and to check whether conversion is completed or not. TheSTATUS pin is also connected to the STROBE pins of two latches IC4 & IC5 which strobesthe 12 bit digital data of AID converter onto the outputs of these latches after convertinganalog voltage into digital.

A 3 to 8 decoder 74LS 138 (IC6) is used to select these latches and other peripheraldevices present in the card. A dip switch logic has been provided in the card to change thebase address of the card. This provision enables the user to insert more than one card insame computer. This logic is designed with 8-bit digital comparator 74LS688.

Page 64: 34(4)

This comparator consists of 8 numbers of two inputs NAND gates. One set of inputsPO, PI, P2, P3, P4, P5, P6 & P7 are connected to the buffered lines A4, A5, A6, A7, A8,A9, AEN and +5V of the computer. Another set of inputs QO, Ql, Q2, Q3, Q4, Q5, Q6 andQ7 are connected to the dip switches. When the data on the dip switches match with thedata on the address lines, the comparator output goes low. This signal is used to enable thebi-directional data bus buffer IC3 74LS245 and the decoder IC6 74LS 138. This signal actsas a board selecting signal. Hence, the status of the switches is very important to providespecific address to any peripheral device. For example, the following dip switches status givesthe base address of the board as 0300H.

SWI SW2 SW3 SW4 SW5 SW6 SW7 SW8 AddressON ON ON ON ON OFF ON OFF 0300H to 030FH

Similarly, a programmable peripheral interface 8255 and programmable interval timer 8254have been interfaced to the computer in the following addresses.

8254 : 030CH to 030FH8255 : 0308H to 030BH

d. Personal ComputerIn the present study a personal computer with the following features is employed for

controlling the speed of DC micromotor.

* Pentium-I, 166 MHz Intel Microprocessor* 16 MB RAM* 1.2 GB Hard disk drive* 1.44 MB Floppy disk drive* Two Serial Ports* One Parallel Port* Six ISA slots to connect I/O, AID, D/ A converter cards etc.

e. D/A converterIn the present study a 12 bit D/A converter AD7541 is used. As shown in Fig.2, the D/

A converter is interfaced to the data acquisition card through port-A and port-B of 8255. Theoutput of PID equation is digital data, which is converted into analog voltage by D/ A converter.The output of D/ A converter is given to the motor through an actuator.

f. ActuatorThe voltage from D/A converter cannot drive the motor directly due to mismatch of power.

Hence, a voltage follower with Darlington pair is added to the output of D/ A converter as anactuator to drive the motor.

4. WORKING OF THE SYSTEMThe main objective of our work is to measure the real time speed of DC micromotor and

to control it for the desired speed. The shaft of DC micromotor is connected to the opticalencoder (transducer), which converts the speed of DC motor into a train of TTL-compatible

Computer based DC micromotor speed control system 271

Page 65: 34(4)

272 P. Bhaskar, Parvathi C.S., L. Shrimanth Sudheer and A.B. Kulkarni

pulses. This train of pulses or frequency is not directly accessible by the data acquisition card,since this card needs the signal to be in analog voltage form. Hence, this frequency is convertedinto voltage by Frequency to Voltage (FN) converter using IC LM 2907. The analog voltage,which is proportional to the speed, is accessed by computer through AID converter. Thisvoltage is converted into corresponding frequency by the equation f=a

1v+a

0. This equation is

fitted by least square curve fitting, where f is the frequency of the signal generated fromoptical encoder and v is the output voltage of FN converter. a

l = 80.156 and a

o=0.04276 are

the slope and intercept of the characteristic line of DC micromotor respectively. This frequencyis converted into speed by the equation:

Speed = (Frequency * 60 seconds) * 1/p rpm.Speed = (Frequency * 5) rpm

where p = number of pulses for one revolution. For the optical encoder used, 12 pulsesare generated for one revolution. This is the measured DC motor speed, which is comparedwith the set value (1000 rpm) to get error i.e., error = set value - measured value. Thencomputer solves the proportional+integral+derivative (PID) equations. To enable the computerto implement PID and improved PID control algorithms, the continuous differential equationsare converted into discrete difference equations9-11. The velocity algorithms for PI and PIDare given by the following equations:

Vn = Vn-1 + Kp(en - en-1) + Ki enT ....... (i)

Vn = Vn-1 + Kp(en - en-1) + Kien T + Kd/T [(en - 2en-1 + en-2)] ....... (ii)

The above equation (ii) is a standard PID difference equation which can be modified intoan improved PID difference equation by using Trapezoidal rule and Interpolation techniquewhich is given by the following equation.

Vn = Vn-1+Kp(en-en-1)+Ki (en + en-1)/(2*T)+Kd/(6*T) [(en + 2*en_1- 6*en_2 + 2*en_3+2*en-4]

....... (iii)

At any instant of time, the current value of the PID output V n is calculated based on theprevious value of the PID output V

n-1, current error en, previous error e

n-1, previous to previous

errors en-2

, en-3

, en-4

, the cycle time T and weighing constants(Kp, K

i, K

d). The output of control

program is a digital value, which is fed to the D/A converter, which converts the digital datainto corresponding analog voltage. Since the D/A converter cannot drive the motor due tomismatch of power, voltage follower with Darlington pair is added as an actuator to drivethe motor.

5. SOFTWARE DETAILSThe software is written in C language. The flow chart (Fig. 4) of the program shows

the DC motor speed measurement and control. The software first acquires the speed of theDC motor, finds error by comparing current speed with set-point speed, solves PID equationsand finally outputs control signals to the actuator. The software makes the system user friendly.It provides on-line tuning of PID parameters. The DC motor speed data can be stored in thefile name given by the user at the beginning of the program or at any time. The software also

Page 66: 34(4)

Fig. 4. Flow chart of the program

Computer based DC micromotor speed control system 273

Page 67: 34(4)

274 P. Bhaskar, Parvathi C.S., L. Shrimanth Sudheer and A.B. Kulkarni

provides on-line variation of set-point, which facilitates the system to study the step variationresponse. The loop time of the system is about 1 sec.

6. EXPERIMENTAL RESULTS AND DISCUSSIONS:The experimental results of DC micromotor (2230U 015S) are discussed here for desired

speed and load variations.

For Desired SpeedFig.5 shows the response of DC motor for simple PID controller and improved PID

controller for a desired speed of 1000 rpm. The settling time is 30 seconds and 20 secondsfor simple PID controller and improved PID controller respectively.

Fig. 5. Step Response of DC motor for PID Controller and Improved PID Controller fora rated speed of 1000rpm.

Page 68: 34(4)

For Load VariationsFor the DC motor, the load is applied in the form of magnetic brake. The magnetic brake

works by means of an aluminum disk, which is mounted on the motor shaft. When the diskis rotated between the poles of magnet, eddy currents form on the disc, producing the effectof a frictional load. When the motor is running at a rated speed of 1000 rpm, the load isapplied. Fig.6 (a), and (b) shows the response of DC motor for improved PID controller andsimple PID controller respectively. From the graphs the following points are observed.

Fig. 6 Load variations Response of DC motor for PID Controller Improved PIDController for a rated speed of 1000rpm.

Computer based DC micromotor speed control system 275

Page 69: 34(4)

When the load is applied, the undershoot speed is 725 rpm and 700 rpm and when theload is removed, overshoot speed is 1250 rpm, and 1300 rpm for improved PID controllerand simple PID controller respectively, with the corresponding settling times of 10 secondsand 12 seconds respectively.

The improved PID controller response is better than the simple PID controller responsein terms of shorter rise-time, shorter settling-time, no overshoots and undershoots. The systemis quite successful in measuring and controlling the speed of the DC micromotor.

REFERENCES1. Ramakant A. Gayakwad, Op-Amps and Linear Integrated Circuits, PHI, 3rd ed., 2000

2. Douglas V. Hall, Microprocessor Interfacing and Programming, McGraw-Hill, 1998.

3. Schuler, McNamee, Modem Industrial Electronics, McGraw-Hill International Editions, 1993

4. Michael Jacob .J, Power Electronics: Principles and Applications, Delmar-ThomsonLeaming, 2002

5. Sen P.C., Principles of Electric Machines and Power Electronics, John Wiley and Sons,2nd ed., 2001.

6.. Christopher T. Killian, Modem Control Technology, West Publishing company, Minnespolis/St. Paul, 1996.

7. Faulhaber DC Motors, “MINIMOTOR”, Minimotor SA 6980 Croglio, Switzerland, 1999-2000,pp 53.

8. Lewis C. Eggebrecht, Interfacing to the IBM Personal Computer, Howard W, Sams & Co.,Inc., 1983.

9. Krishna Kant, Computer-Based Industrial Control, Prentice-Hall of India Pvt. Ltd, New Delhi,2002.

10. Michael Jacob M., Industrial Control Electronics - Applications and Design, Prentice Hall,England Cliffs, (1988).

11. Benjamin. C, Kuo, Digital Control Systems, Holt-Saunders International Editions.

12. Liptak .B.G, “Instrument Engineers’ Handbook - Process Control”, Butterworth HeinemannLtd., Oxford, 1995.

276 P. Bhaskar, Parvathi C.S., L. Shrimanth Sudheer and A.B. Kulkarni

Page 70: 34(4)

MICROSENSOR SYSTEM FOR MEASUREMENT OFATMOSPHERIC TURBULENCE RELEVANT TO

OPTICAL PROPAGATION

ANIL K. RAZDANScientist, G-FAST, PI-Metcalfe House, Defence Science Centre, Delhi - 110 054

ABSTRACT

In order to model laser beam propagation (or optical propagation in general) through aturbulent atmosphere, one has to have knowledge about the strength of turbulence which thelaser beam is encountering. The refractive index structure constant (Cn) is the single mostimportant parameter which is measure of the atmospheric turbulence strength. A microsensorsystem has been designed and fabricated for measurement of the turbulence parameter (Cn).The instrument uses a pair of fine platinum wires (developed inhouse; wire diameter~4-6 µm) as sensors for measurement of small (< 0.05 K) and fast (>50 Hz) temperaturedifference (DT) fluctuations existing in the atmosphere. This paper will give the details aboutthe design, fabrication, test and calibration of the instrument for its use to measure Cn. Theinstrument was used for long hours to measure Cn in the open atmosphere and study its diurnalvariation over a 24 hour period under various atmospheric conditions covering weak, moderateand strong turbulence conditions. The results of these measurements are presented.

1. INTRODUCTIONAir movements are characterized by disordered variations of both the magnitude and the

direction of the velocity at any point. The result is vigorous mixing. Such motion is calledturbulent as distinct from laminar motion in which mixing does not occur and the velocity ata given point is either constant or varies in a regular fashion[1,2]. Turbulent air motion representsa set of eddies of various sizes from extremely large with a characteristic scale Lo (outerscale of turbulence) to extremely small with a scale lO (called inner scale of turbulence). Theouter scale Lo varies in size from tenths of a meter to a few meters near ground level, whilethe inner scale lO is of the order of one millimeter near the ground level.

As a result of atmospheric turbulence the density of air fluctuates which results in thechange of atmospheric index thereby affecting the angular spectrum of the optical wavefrontpropagating through the atmosphere. The temperature variations are mainly responsible forthe density in homogeneities in the atmosphere because the pressure fluctuations are dissipatedvery quickly (they travel with speed of sound).

J. Instrum. Soc. India 34 (4) 277-283

Page 71: 34(4)

278 Anil K. Razdan

Ideally an instrument for measuring these temperature variations should be capable ofmeasuring temperature at a point in the atmosphere without itself altering the temperature.Directly measuring the index of retraction by optical interferometry seems attractive at first,because no heat exchange is required to make the measurement. Unfortunately measuringsmall volumes of air involves the proximity of large pieces of glass which disturb the air flow.An alternative approach involves the temperature measurement of a solid in near equilibriumwith the surrounding air. A study of the performance of a resistance thermometer indicatesthat this means of air temperature measurement is sufficiently close to the ideal to be useful.

2. CONSTRUCTION OF PROBESFine platinum wires with diameters in the range of 4-6 µm and resistance 4-8 Ω/mm are

used as temperature sensors. These wires are produced using the wollaston process[3]. Inthis process a relatively thicker Platinum wire is embedded in a Silver rod along its axis. Thiscomposite piece is then drawn by conventional method into a wire till the Platinum wire reachesits desired thickness. The Silver clad Platinum wire or wollaston wire is easy to handle. Arectangular hole of (0.5 x 0.6) cm2 is cut in the center of a circular mica disc of ~ 4.0 cmdiameter. A series of holes (0.5 mm diameter) are drilled on two edges and along the lengthof the rectangular hole. Five holes are along one edge and six along the opposite edge. Holesof same dimensions are drilled on either side along its breadth. Two Aluminium discs havingcentral rectangular hole (1 x 1.3) cm2 are pasted concentrically on the two sides of the circularmica piece. Wollaston wire is then threaded in a zigzag way through the pin holes of the micasheet and its two ends are respectively soldered by Indium to two Aluminium electrodes. Theholes are covered with thin layer of araldite to give strength to the wire at these points. Thecomplete probe is then dipped in dilute nitric acid bath, the electrodes and araldite beingprotected with paraffin wax. When the silver is completely dissolved in the acid, the paraffinwax is removed carefully first in hot water and then in petroleum ether. The sensor is nowready. A sketch of the finished sensor is shown in Fig. (1).

Fig. 1. Diagram of temperature sensor (mounted)

Page 72: 34(4)

3. MEASUREMENT PRINCIPLE:Figure 2. shows the block diagram of the measurement system. The Platinum wire probes

P1 and P2 are used in pairs separated by a distance of a few centimeters. Taking the root meansquare (R.M.S) value of the temperature difference between the two probes separated by adistance ‘r’ (which is proportional to the voltage difference across the points A and D; thecalibration constant for conversion of temperature difference into voltage is obtained by thecalibration procedure described later), we calculate the value of the temperature structureconstant (C

T) by using the relation[4] :

CT

2 = [< (T2 - T

1)2>] r-2/3 (1)

⇒ CT = [<(T

2-T

1)2>]l/2 r-1/3 = D

Tl/2 r-1/3 (2)

where < — > is the ensemble average, DT is the temperature structure function, r is the

separation between the probes and T1 and T2 are the individual temperatures measuredby the probes. According to the Kolmogoroff theory, C

T2 is independent of r. It is related to

the refractive index structure parameter (Cn) as follows[4] :

Cn

2 = [79 P/T2 x 10-6]2 CT

2 (3)

Cn = [79 P/T2 x 10-6] C

T(4)

Fig. 2. Block diagram of Cn measurement systemP

1, P

2 : Microthermal probes; R

f : Fixed resistance; R

v : Variable resistance

Microsensor system for measurement of atmospheric turbulence relevant to optical propagation 279

Page 73: 34(4)

280 Anil K. Razdan

where P is the pressure in millibars and T is the temperature in K. The advantage ofusing the probes in pairs is that structure function measurement acts as a spatial filter,discriminating against irregularities with scales larger than” r”. In this way it is possible toobserve directly those irregularities that are most important in their optical effects.

4. DESCRIPTION OF INSTRUMENTThe two probes P

1 and P

2 constitute two arms (arm 1 and arm 2) of a wheatstone bridge

(see fig. 2). The third arm is a fixed resistance Rf and the fourth arm is a variable resistanceR

v. The cold resistances of the two probes were determined to be 204.0 ohms and 154.0

ohms after measuring the resistance of the probes at a known constant temperature using aconstant temperature bath. A battery serves as a source of current to the bridge circuit. Thecurrent flow is restricted to a value so that the self heating of Platinum wire due to currentflow is negligible. Initially the two probes are kept at zero separation and the bridge is balancedby the adjustable resistance of the fourth arm. Thereafter the two probes are kept at a knowndistance apart in the region under investigation. The output between the points A and D issuitably amplified and after passing through a high pass filter is fed to an RMS unit givingRMS value of the temperature difference fluctuations (the amplified output can be expressedin terms of the temperature difference between the probes by using the calibration constantobtained by the calibration procedure described later). CT is then calculated according toequation (2). A sample of a chart recorder output of temperature fluctuations is shown inFigs. 3 (a), (b), (c). This is repeated for different separations between the two probes and an

Fig. 3. Samples of temperature difference measurements made on a clear sunny day withsensors 1.5 m above the ground and speed 3.6 cm apart.

a) at 0545 hrs b) at 1020 hrs c) at 1300 hrs

Page 74: 34(4)

average value of CT is obtained. The average atmospheric pressure and the average temperatureare simultaneously measured and Cn is calculated using equation (4).

5. CALIBRATION PROCEDUREReferring to Fig. (2) we see that by changing the measuring current slightly by introducing

a small change in the resistance (∆R) in arm 1 or arm 2 of the wheatstone bridge it is possibleto change the potential difference between the points A and D by an amount ∆V that wouldoccur through a temperature difference of ∆T between the two probes. The procedure adoptedfor calibration is as follows.

The two probes P1 and P2 are kept at zero separation (at a known constant temperature)and the bridge is balanced by. the adjustable resistance of the fourth arm. The probes P1 andP2 are next replaced by precision potentiometers PTI and PT2 introducing the equivalentresistance of probes P1 and P2 respectively at that particular temperature. The potentiometerresistance is adjusted to produce step changes in the amplifier output corresponding to -2°C-1°C, +1°C and 2°C temperature difference between the probes P1 and P2. The change inresistance to be introduced for 1°C change in temperature at a particular probe is equal toαRO where RO is the cold resistance of the probe and ‘α’ is the temperature coefficient ofresistance for Platinum (0.00392Ω/Ω/OC). Thus a calibration constant is obtained relating theAmplifier output voltage and the temperature difference existing between the two probes.

6. RESULTS AND DISCUSSIONThe instrument was used to study the diurnal variation of refractive index structure

parameter (Cn) in the atmosphere. Samples of temperature difference measurements made ona clear sunny day with sensors 1.5 m above the ground, spaced 3.6 cm apart are shown inFig. (3) a, b, c. The wind velocity during the measurements was recorded as 0.7 m/s (min)to 3.7 m/s (max). The average pressure and temperature were also measured simultaneouslyalongwith the temperature fluctuation measurements & Cn estimated according to relation (4).The calibration factor in this case was found to be 170 mV = 0.5°C. The measurementswere repeated for different separations between the sensors viz. r = 1.6 cm, 3.6 em, 9.6 cm,16.00 cm and average value of CT was calculated. The average temperature and averagerelative humidity (R.H) during these three sample measurements were recorded to be 33°C,49%; 36.2°C, 30% and 39°C, 37% respectively. Table (1) gives the value of CT & thecorresponding value of CT as a function of time. Mean pressure values are also given. Fig.(4) illustrates a typical example of the diurnal variation of Cn with probes 3.6 cm apart and ata height of 1.5 m above the ground. The diurnal cycle of Cn in Fig. (4) is typical of itsbehaviour in the surface layer of the earth (approx. the lowest 0-100 m). The intensity of thetemperature fluctuations and hence Cn is quite small at night and in the early morning hoursuntil solar heating of the ground becomes significant enough to initiate the convective instability.This instability generates temperature fluctuations at a given height in the following manner.The sun heated surface materials warm a thin layer of the air above. If air parcels from thelayer become displaced upward, they find themselves warmer and hence less dense and morebuoyant than the ambient air; therefore, they continue to accelerate upward. It is the mixingof these hot rising air parcels with cool descending air parcels that produces the observedtemperature irregularities. This process continues, causing Cn to increase until the solar heatingsubsides producing a consequent drop in Cn as seen in Fig. 4 around 1200 hrs.

Microsensor system for measurement of atmospheric turbulence relevant to optical propagation 281

Page 75: 34(4)

282 Anil K. Razdan

Fig. 4. Plot showing diurnal variation of Cn on a typical hot summer day

Table 1 : Recorded values of mean pressure and temperature and estimatedvalues of CT and Cn

Time CT Mean Pressure Mean Temperature Cn (x10-6)(hrs) (m-1/3) (mb) (K) (m-1/3)

510 0.287 1008 306 0.242538 0.40 1008 306.3 0.336545 0.627 1008 306.5 0.528615 0.640 1008 306 0.528710 0.661 1008 306 0.544815 0.918 1008 309.7 0.757915 0.966 1008 309 0.8011025 1.234 1008 309.2 1.0231100 1.557 1008 309.2 1.291210 2.811 1008 311 2.31300 2.144 1008 312 1.7451410 1.837 1008 312 1.4951500 1.7055 1008 311.5 1.4581600 1.467 1008 312.5 1.191905 1.404 1008 310.5 1.1532000 0.920 1008 309 0.7592210 0.779 1008 308.75 0.6432258 0.561 1008 308.8 0.4652330 0.223 1008 308.5 0.184

Page 76: 34(4)

ACKNOWLEDGEMENTSThe author wishes to express his gratitude to Sh. K V S S Prasad Rao, CC R&D (Tech)

& DS, DRDO HQRS. and Sh. K S Jindal, Du-ector, LASTEC, for their kind permission topublish this work.

REFERENCES1. J.W. Strohbehn, “Laser Beam propagation in the Atmosphere” (Topics in Applied Physics;

Vol. 25, Springer Verlag, New York (1978).

2. V.E. Zuev, “Laser Beams in the Atmosphere” , Plenum (1982).

3. Kartar Singh, et al “Fabrication of Microphone Grids” Internal report DSL Report no. 3/70,Defence Science Centre, Metcalfe House, Delhi-54

4. R. S. Lawrence. et. al., “Measurements of atmospheric turbulence relevant to opticalpropagation” J. Opt. Soc. Am. 60, 826-830 (1970).

Microsensor system for measurement of atmospheric turbulence relevant to optical propagation 283