This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 1
paper id: 1570507789

Milk Tester: Simultaneous Detection of Fat Content

and Adulteration
Moupali Chakraborty, Student Member, IEEE and Karabi Biswas, Member, IEEE

Abstract—This work reports an integrated milk testing instru- specific brand as calibrating solution), and is easily available
ment for the simultaneous detection of fat content and adulterants for calibration.
in milk. The prototype consists of a small temperature control In this work, we have also tried to explain the frequency
system and an integrated signal conditioning circuit for fat and
adulteration detection. The calibration methodology of the instru- selection criteria from the undergoing chemical kinetics be-
ment with respect to a proposed standard solution is illustrated havior of the sensor. The complex impedance plot shows that
in detail. Different modules of the prototype are explained with at a particular frequency, a gradual transition takes place from
the analysis of the error that occurred in different segments. diffusion to the charge-transfer zone. An electrical equivalent
Power consumed by the modules is measured to improve the circuit is proposed using MEISP software from the experimen-
design of the instrument for field work. The proposed instrument
is also compared with the commercially available milk testing tal data, to validate the above.
instruments to understand the market challenges. In addition to The paper is organized as follows: apart from the introduc-
this, a study is carried out to understand the chemical kinetics tion, the paper has six more sections. The details of the devel-
behavior of the sensors in milk for the selection of operating oped prototype is elaborated in section-II-IV, which includes
frequency. the block diagram of the complete instrument, the temperature
Index Terms—Milk testing instrument, milk fat, adulteration, control system, integrated signal transduction circuit, packag-
lipase, PMMA, phase angle, signal conditioning circuit, calibra- ing, calibration, and error analysis of the individual module.
tion, error analysis, chemical kinetics.
Section-V provides a flow chart for measurement steps. A
comparison of the proposed prototype with the commercial
I. I NTRODUCTION milk testers is also included in this section. In section-VI,

M ILK is one of the staple foods which is consumed all the underlying chemical kinetics of the measuring system is
around the world for its nutritional value [1]. But at elaborated; and the criteria for selecting a particular frequency
the same time, milk is one of the most vulnerable food which for measurement are explained using the complex impedance
can be adulterated very easily. Typically, adulteration is done plot. Section-VII provides the conclusion.
to increase the volume of the milk or to increase the shelf
life. It has become one of the serious issues in most of the II. P ROTOTYPE DEVELOPMENT
developing countries, and also to some advanced countries as The complete block diagram of the proposed milk testing
well. Unfortunately, a user-friendly, low-cost device, which equipment is shown in Fig. 1. The prototype contains a small
can measure the fat content of milk along with the adulteration temperature control system and a suitable signal conditioning
detector is not available in the market. circuit, for the detection of fat content along with adulteration
In this context, this paper presents an integrated instrument in milk. Lipase + PMMA coated sensor is used for fat content
to detect fat content along with five common adulterants determination [3], [4] and the PMMA coated sensor detects the
in milk: and is an extension of the earlier research works presence of adulterants in milk [2]. The key parameter is the
by the authors [2]–[4]. A novel lipase immobilized PMMA phase angle of the impedance, which changes when the sensors
(Polymethyl Methacrylate) coated sensor [3], [4] was used to are dipped into the milk sample [2]–[4]. In the proposed
detect fat %, whereas, only PMMA coated sensor was used instrument, this change in phase angle is converted to a voltage
for adulteration detection [2]. However, in both the cases, an by an integrated signal conditioning circuit, where a switch
external heating system was used as it has been found that the
most repeatable results are obtained when the milk samples are
kept at a temperature around 42 C [2]–[4]. Hence, we need an
integrated heating system to make a portable instrument for
field testing. Therefore, a small temperature control system,
and an optimized signal transduction circuit are developed
to make a complete milk testing equipment. In addition to
this, a calibrating solution is proposed, which is a standard
solution (previously we used a particular type of milk of a
The authors are with the Electrical Engineering Department, Indian
Institute of Technology Kharagpur, Kharagpur 721302, India (e-mail:
[email protected]; [email protected]).
Research support from DST, Sanction No. EEQ / 2016 / 000540 Fig. 1: Block diagram of the proposed instrument

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 2
paper id: 1570507789

(a) (b)
Fig. 2: Temperature versus time response for: (a) 12 V heater (a) (b)
(b) 230 V heater
Fig. 3: Proposed instrument: (a) front view (b) top view

is used to choose the sensor (fat detection or adulteration

detection). Different modules of the system are discussed in the desired range. The PID control mechanism is used to
below. control the temperature. To show the set point and the process
value, a three-digit LED display is attached. A thermocouple
[5] is connected externally to read the temperature of the milk
A. Temperature control System sample.
In previous work by the authors [3], it was shown that, at
around 42 C temperature, sensors give most repeatable results B. Signal conditioning circuit
and an external temperature bath was used to keep the milk The signal conditioning circuit consists of IC XR2206,
samples at this particular temperature. Here, the temperature which generates 500 Hz and 5.6 kHz [2]. Both 500 Hz and
control system is an integrated part of the proposed instrument. 5.6 kHz excitation are needed for adulteration detection [2]
Different parts of the system are as follows: and only 500 Hz for fat detection [3]. A DPDT (double-
1) Power supply: A 230 V ac power supply is used to pole, double-throw) switch (S1 in Fig. 1) is used to select
operate the digital temperature controller and the heater of the frequency. The phase detection (PD) part of the circuit
the instrument. Voltage regulator IC 7812 generates 12 V dc, is described in [2] and [3]. For the proposed instrument, one
which is required for the signal conditioning circuit of the more DPDT switch (S2 in Fig. 1) is added with this PD to
prototype. select fat or adulteration detection. The output of PD is fed to
2) Container: A stainless steel cylindrical container, having the ATmega328P microcontroller through an RC low pass filter
a diameter of 5 cm and a length of 5.5 cm is used for the milk [2], [3]. Maximum phase shift obtained by the PD module is
sample. The container has a steel lid with two holes, one for 90 and the maximum input voltage to the microcontroller
temperature sensor (commercial sensor is used) insertion and is 5 V. Therefore, the filter circuit is tuned in such a way that
another for milk sensor. 5 V corresponds to 90 phase shift and 0 V is for 0 phase
3) Heater: A 12 V heater is first considered as the signal shift. A 16 ⇥ 2 liquid crystal display (LCD) is connected with
conditioning circuit of this instrument works at 12 V supply. the microcontroller to display the result.
Nichrome wire with 38 gauge size is used to make the 12
V heater. Measured resistance of the wire was 4 ⌦ and the
C. Packaging
current drawn by the heater was 1 A. Using this 12 V heater,
a temperature (22 C - 50 C) versus time plot for the milk The proposed instrument has two casings, made by cloth
sample (in the container) is shown in Fig. 2 (a). Linear fit of acrylic sheet. One for the temperature control system (A) and
the data is also plotted in the same graph and the slope is another for the signal conditioning circuit (B). The complete
0.02642 C/s. It takes 12 min to reach 42 C from 22 C. instrument is shown in Fig 3. Part A has an SPST (single-pole,
To overcome the problem of more heating time, a heater of single-throw) switch in the backside of the instrument to make
Nichrome wire with 39 gauge size is made, which works at the heater ON. Part B has two external DPDT switches (S1
230 V power supply. The wire is 2.3 m long, having resistance and S2 ). S1 switch has two positions, ‘500 Hz’ and ‘5.6 kHz’
470 ⌦, and wounded on mica sheet. Current drawn by the for choosing the frequency. Similarly, S2 switch has ‘FAT’ and
heater is 0.34 A. Data plot of temperature (22 C - 50 C) ‘ADUL’ positions for choosing the corresponding sensor.
versus time for milk sample with this heater is shown in Fig.
2 (b). Linear fit of the data in the range of 24 C - 50 C is III. I NSTRUMENT CALIBRATION
also plotted and the calculated slope is 0.14 C/s. It takes 3.6 The proposed instrument requires day-to-day calibration.
min to reach 42 C from 22 C. Therefore, a trade off is to The calibration process for fat and adulteration against a pure
be made between the heating time and the power supply. milk sample was one of the main problems to make a user-
4) Digital Temperature Controller: A commercial model friendly instrument. Every day, the availability of pure milk is
of TC513 digital temperature controller, having a size of 10 difficult. Moreover, if the milk is not pure, whole calibration
cm ⇥ 4.8 cm ⇥ 4.8 cm, is used to maintain the temperature as well as the adulteration detection process will be erroneous.

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 3
paper id: 1570507789

(a) (b)
Fig. 4: Plot to show the (a) effect of temperature and (b) (a) (b)
reusability of the standard solution
Fig. 5: Calibration time: sensor not in use for (a) 24 hours (b)
1 hour
Therefore, to detect the adulterants in milk by the proposed
instrument, a standard solution is made, which is similar to pure milk with respect to the standard solution and a minimum
the standard solution as used for fat detection process [3]. phase angle difference is obtained as 4.5 . For the adulterated
milk samples (adulterants are: detergent, ammonium sulphate,
A. Standard solution for calibration neutralizer, and common salt) the average phase angle is 29
A reference solution is prepared by adding 0.5 g of sodium and depends on the concentration of adulterants. Considering
chloride (NaCl) in 50 ml of distilled water at room tempera- this difference in phase angle, a calibration method is set
ture, as described in [3]. in the microcontroller for the proposed instrument to detect
1) Effect of temperature: Temperature effect of the standard adulterated milk.
solution is studied by varying the temperature between 16 C 1) Methodology:
to 70 C. In this range, magnitude and phase angle of the • The sensor is dipped in standard solution and the mea-
sensor impedance is plotted in Fig. 4 (a). Linear fit of the data sured voltage Vs is stored in the microcontroller. Time of
is also plotted in the same figure. It is found that the phase dipping is discussed in ‘calibration time’ subsection.
angle varies 0.011 / C whereas, magnitude varies 0.428 ⌦/ C. • 90 phase difference is equivalent to 5 V. Therefore,
As the variation of phase angle is negligible with temperature, 4.5 phase difference is equivalent to 0.25 V.
the standard solution is used at room temperature (26 C to • The reference voltage VR , to detect the adulterated milk
30 C). sample, is calculated as VR = Vs - 0.25 V.
2) Reusability: The shelf life of the standard solution is • For an unknown milk sample, measure the voltage VU .
more than one year. The solution prepared in July 2018 is If VU > VR : adulterated milk.
still (July 2019) in use for calibration. The solution level is If VU <= VR : pure milk.
marked in the beaker and the beaker is covered with aluminum Experimentally it has been seen that, phase angle variation in
foil. Whenever the solution level is below the marked point, standard solution for consecutive two days is within the range
distilled water is added to reach the level. Phase angle plot of 1 . Maximum variation of 1 can introduce around 22%
for the same standard solution in the period from July’18 to ( 4.5
1
⇥ 100) error in measurement. Thus, the error is effective
June’19 is shown in Fig. 4 (b). if the system is not calibrated every day.
From this experimentally obtained figure, it can be noticed 2) Calibration time: When the sensor is used for the first
that, in the frequency range of 20 Hz to 2 MHz, the phase time or not in use for a time gap of 3 hours or more, the sensor
angle at a particular frequency remains similar (with some needs to be dipped in the standard solution for 6 min. Several
deviations) over a period of one year. Therefore, the solution experiments are performed to record the phase angle versus
can be reused for a long time without any degradation and no time response of the sensor. One of the results is plotted in
need for preparing a day-to-day solution. Fig. 5 (a) for a sensor that is not in use for 24 hours. Data is
taken for 18 min duration in the interval of 10 s. Fig. 5 (a)
B. Calibration of the instrument indicates that after 6 min, the phase angle deviates ± 0.25
from the final value.
Sodium Chloride breaks down into N a+ and Cl ions in
Secondly, when the sensor is not in use for half an hour
water. Experimentally it has been seen that 0.5g/50ml of NaCl
or more, then the dipping time should be 1 min. One of the
solution has more ionic concentration than 50 ml of pure
results is shown in Fig. 5 (b), where the sensor is not in use
milk1 . Therefore, the absolute value of the phase angle is more
for 1 hour. Data is taken for 1 min duration in the interval of
in this standard solution. Difference in phase angle between
1 s. It can be seen from the figure that, the phase angle settles
the standard solution and the milk samples are used for the
at a final value after 50 s time.
calibration. At 5.6 kHz, the average phase angle reading mea-
sured for NaCl solution is 28 and that of unadulterated milk C. Lifetime, reproducibility, and repeatability of sensors
is 21 . Around 400 sets of experiments are carried out for
To find out the lifetime, repeatability, and reproducibility of
1 Milk which does not contain any adulterants in it is referred as pure milk lipase + PMMA and PMMA coated sensors, phase angle data,

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 4
paper id: 1570507789
TABLE I: RSD % of reproducibility and repeatability of lipase
+ PMMA and PMMA coated sensors
RSD % of lipase + PMMA RSD % of PMMA
Parameters
coated sensor coated sensor
Reproducibility 1.4 1.6
Repeatability 8.4 2.31
Repeatability 0.75 for LP1 0.42 for P1
Repeatability 0.74 for LP2 0.42 for P2
Repeatability 0.72 for LP3 0.41 for P3
Repeatability 0.77 for LP4 0.41 for P4

Fig. 6: Lifetime measurement of lipase + PMMA coated sensor

Fig. 8: Temperature versus time response for temperature

controller with 230 V heater at 35 C set point

(a) (b)
coated sensor. This is because of the short lifetime of lipase
Fig. 7: (a) Reproducibility, and (b) repeatability of lipase + + PMMA coated sensor compared to PMMA coated sensor.
PMMA, and PMMA coated sensors However, in a continuous dipping of 20 times, both of them
are repeatable with phase angle RSD of 0.7% for lipase +
PMMA (LP1 to LP4), and 0.4% for PMMA (P1 to P4) coated
collected at 500 Hz, are plotted accordingly.
sensors (viz. Table. I).
1) Lifetime: Experimentally it has been seen that, after
400 dipping of the lipase + PMMA coated sensor, erroneous
IV. E RROR A NALYSIS AND STUDY OF POWER
readings occur and the surface of the sensor also gets slight
CONSUMPTION BY THE INSTRUMENT
blackish. Phase angle versus number of dipping of the sensor
in NaCl solution, along with polynomial fit is shown in Fig. A. Error analysis of temperature control system
6. It can be seen that the deviation of phase angle from the 1) Heater: The heater is covered with stainless steel plate
fitted line increases after 400 dipping. Therefore, the lifetime and the container is also made from the same metal. Stainless
of the lipase + PMMA coated sensor is around 400 dipping. steel is the alloy of chromium-nickel steel, which has the
Adopting the same procedure, it has been found that the thermal conductivity of 0.148 - 0.163 W/cm C within 100
PMMA coated sensor can withstand 1000 dipping [2]. The C [6]. When the heater is not in use, initially it takes around
reduced lifetime of the lipase + PMMA coated sensor is due 40 s for the rise of 1 C. However, when the heater is heating
to the presence of the biological element in the coating. continuously, then it takes around 7-8 s to rise 1 C.
2) Reproducibility: To test the reproducibility of the sen- 2) 12 V power supply: The generated 12 V supply gives
sors, phase angle data for four lipase + PMMA coated sensors the maximum value of 12.3 V and the ripple of around 100
(LP1 to LP4), and four PMMA coated sensors (P1 to P4) mV. Ripple introduces an error of 0.57 %.
are considered. With 20 sets of dipping, phase angle data 3) Digital temperature controller: For the temperature con-
collected at 500 Hz frequency, are compared for four sensors trol system, if the desired temperature is set at 42 C, due to
and shown in Fig. 7 (a). Relative standard deviation (RSD % the thermal conductivity of the steel container, the temperature
= Standarddeviation
mean ⇥ 100) of the reproducibility is tabulated rises to 54 C during the 1 min test time. Therefore it
in Table. I. Analyzing Fig. 7 (a) and Table. I, it is apparent introduces an error of 12 C. Hence, the set point is kept
that both the sensors are reproducible. at 35 C to mitigate the overheating problem. As the heater
3) Repeatability: Phase angle data plot of both lipase + is covered with thick stainless steel plate, and the container is
PMMA coated and PMMA coated sensors, are calibrated also made of steel, time is required to transfer the heat from
consecutively for 15 days, and shown in Fig. 7 (b). Number of the nichrome wire to milk. Thus, after 35 C, when there is
dipping of the lipase + PMMA coated sensor is 425 and that of no current through the wire, then also the heat transfers to the
PMMA coated sensor is 386 within these 15 days. RSD % of milk sample and the temperature increases up to 42 C, and
the repeatability is tabulated in Table. I for both the sensors. remains at that temperature during the 1 min test time.
It can be seen from Fig. 7 (b) and Table. I, PMMA coated Temperature versus time data plot of the heating system at
sensor gives more repeatable results than lipase + PMMA 35 C set point is shown in the Fig. 8, where the temperature

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 5
paper id: 1570507789

settles down at 47 C from 27 C. As it was mentioned

in the previous work that temperature of 40 C to 50 C
is the optimum temperature for detection [3], this error in
temperature is not effective for the detection system.

B. Error analysis of signal conditioning circuit

The generated frequency from IC XR2206 varies between
498 to 508 Hz in case of 500 Hz, and for 5.6 kHz, it
varies between 5.56 to 5.67 kHz. Therefore, these frequency
generations introduce an error of 1.6 % for 500 Hz and 1.25
% for 5.6 kHz. The effect of these errors on detection is as
stated below.
• In the case of fat detecting sensor, for the frequency
change from 500 Hz to 510 Hz, phase varies around
0.07 .
• In the case of adulteration detecting sensor, for the
frequency change from 5.6 kHz to 5.7 kHz, phase angle
varies around 0.1 .
Phase angle variation of 0.1 is negligible compared to 4.5
for the adulteration detecting sensor (see III-D), and to 8 [3]
for the fat detecting sensor. Therefore, these errors are not ef-
fective. Error analysis for phase detection and microcontroller
has already been discussed in the previous work [3].

C. Power consumption by the instrument

The instrument presently works at 230 V supply. Applica- Fig. 9: Flow chart to detect the fat content and presence of
tion wise this instrument should be used at milk collection adulterants in milk by the proposed instrument
point or at some remote area where the power supply may
not be easily available. Therefore, the power consumed by the
1) Safety: Standard safety rules must be followed by the
instrument is calculated to improve the design of the circuit
user to protect the instrument against misuse.
to make it suitable for field work.
• The thermocouple should be inserted in the container
1) Temperature control system:
at the time of measuring. Otherwise, in the absence of
• The heater unit draws 0.34 A current from 230 V supply,
control action, the heater will be overheated and the
i.e. the consumption of power is 78.2 W. The 12 V heater
Nichrome wire may get broken.
consumes 12 W power. Therefore, a trade-off is needed
• The temperature set point of the controller should be
between the power consumption and the time of heating.
below 40 C.
• Digital temperature controller with its display unit con-
sumes 5 W power [7]. 2) Operational flow chart: The operating instructions of the
• Current consumption by IC 7812 for 12 V is 5 mA.
instrument are represented by the flow chart as shown in the
Therefore, power consumed is 0.06 W. Fig. 9. Oval, rectangular, rhombus, and parallelogram shapes
are used as per the standard norms of flow chart representation
The total current consumed by the signal conditioning circuit
[8]. The input and output of the instrument are shown by oval
is measured as 80 mA, i.e., power consumption is around 0.96
shape, whereas the rectangular and rhombus shapes show the
W. Including the power consumption of the heating system,
intermediate processes and decisions in LCD. Parallelograms
total power consumed by the instrument is 78.2 + 5 + 0.06
are used for user input from keypad.
+ 0.96 = 84.22 W. Main part of power consumption is due
3) Maintenance: Both the sensors are washed with water
to the heater. Therefore, further research work needs to be
and wiped with tissue paper after each dipping to remove
carried out to design a fast heating system, which consumes
the suspended particles on the surface. An automatic washing
lower power to make the instrument battery operated.
technique is under study. Wash the container with water before
measuring the milk sample.
V. U SER M ANUAL AND A C OMPARATIVE STUDY WITH THE
COMMERCIAL MILK TESTERS
B. Validation with the standard milk tester
A. Manual The instrument is under test in National Dairy Research
Complete instructions to use the milk tester along with the Institute (NDRI), Kalyani, India, where the cows of differ-
safety and maintenance process are described below. ent breeds/origins like ‘Red Sindhi’, ‘Jersey’, ‘Tharparkar’,

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 6
paper id: 1570507789
TABLE II: Validation of the proposed instrument with a standard milk tester
Validation with PM Validation with ADM and PM
Fat %
Cow Output of Voltage (V) in proposed Quantity of adulterants Output of Voltage (V) in proposed
(measured by
No. MilkoScreen instrument and output in 50 ml milk MilkoScreen instrument and output
MilkoScreen)
C1 5.25 Normal 1.79 (PM) 0.2 g N aCl Normal 2.11 (ADM)
C2 6.14 Normal 1.48 (PM) 0.2 g N aOH Abnormal 2.04 (ADM)
C3 4.63 Normal 1.42 (PM) 0.2 g Detergent Abnormal 2.02 (ADM)
Ammonium
C4 4.91 Normal 1.7 (PM) 0.2 g (N H4 )2 SO4 2.08 (ADM)
Sulphate
C5 3.61 Normal 1.69 (PM) 0.2 g N aHCO3 Abnormal 2.04 (ADM)
C6 3.06 Normal 1.63 (PM) No mixing Normal 1.46 (PM)
C7 4.64 Normal 1.49 (PM) No mixing Normal 1.42 (PM)
C8 4.41 Normal 1.74 (PM) No mixing Normal 1.57 (PM)
C9 7.35 Normal 1.76 (PM) No mixing Normal 1.55 (PM)
C10 6.8 Normal 1.56 (PM) No mixing Normal 1.5 (PM)
PM: Pure Milk ADM: Adulterated Milk

‘Holstein Friesian’, and unknown breed, named ‘Desi’, are VI. F REQUENCY SELECTION OF THE MEASUREMENT
available and numbered accordingly to identify the cows. The lipase + PMMA coated sensor in [3], [4], and the
Preliminary test results of the proposed instrument for 10 milk PMMA coated sensors in [2] behave as an electrochemical
samples (from C1 to C10), freshly collected from cows with system when dipped in milk. At a particular frequency of 500
different breeds, are given in Table II. NDRI has a commer- Hz, the lipase + PMMA sensor gives different phase angles in
cial milk analyzer ‘MilkoScreen’ from the German Indifoss different fat content milk [3], [4]. Similarly, the PMMA coated
company [9]. All the freshly collected milk samples are tested sensor detects adulteration at 500 Hz for milk with high fat
in the proposed instrument, as well as in MilkoScreen. percent and at 5.6 kHz for low fat. We have tried to explain
The methodology of the proposed instrument is discussed the detectability of the sensors at different frequencies from
in section III-B-1, where Vref is required to detect the adul- the chemical kinetics behavior of the detection system.
terants in milk. For the validation of these 10 samples with
MilkoScreen, the Vref calculated as 2.01 V. From Table II, it
is apparent that the voltage output of all the samples (C1 to A. Chemical kinetics behavior of lipase + PMMA coated
C10) are below Vref , and thus they are detected as pure milk sensor
(PM) by our instrument. These 10 samples are also detected A sinusoidal voltage is applied at the two terminals of the
as PM in MilkoScreen along with the fat % of each sample, lipase + PMMA coated sensor and the phase and magnitude
as tabulated in Table II. of the sensor impedance are determined. As explained in [24],
For further validation of the instrument, five adulterants if the measurements are made over a wide frequency range,
(N aCl, N aOH, detergent, (N H4 )2 SO4 , and N aHCO3 ) at then different physical processes may be separated through
0.2% concentration are mixed with five fresh samples of NDRI their different time constants. Therefore, the equivalent circuit
and rest of the five samples are kept as it is. This time, is modeled in MEISP software at the frequency range of 20
Vref calculated as 1.96 V. Results of both the instruments are Hz to 5 kHz [4]. This frequency range is chosen because of
tabulated in the right part of Table II. All the adulterants are 500 Hz (frequency for fat detection) lies within this range.
detected as adulterated milk (ADM) in both the instruments, An illustrative view of the electrical model of the lipase +
except NaCl, which could not be detected in MilkoScreen. It is PMMA coated sensor at the frequency range of 20 Hz to 5
worth to be noted that, though the milk samples are of different kHz is shown in Fig. 10, where the model of [4] is explained
fat content (3.06% to 7.35%), the proposed instrument can elaborately.
measure the adulteration in milk successfully. Thus, the initial • Impedance ZM of the model (viz. Fig. 10) contains liquid
verification of the designed instrument against a standard medium (electrolyte) resistance (Rm ) and fat globule’s
instrument worked well. Further research and data analysis resistance (Rser ) and capacitance (Cser )
of the proposed system is under study. • According to [25], effects from the diffusion of ions
through the layer, predominate at the liquid-film bound-
C. Comparative study with the commercial milk testers ary. Surface of the lipase + PMMA coated sensor is
Commercially available milk testing instruments are mostly porous in nature as shown in Fig. 11. Therefore, in this
for the determination of fat/SNF content in milk. The per- model, impedance at the liquid-film boundary is denoted
formance of the proposed instrument is compared with some by ZLF (viz. Fig. 10), which includes coating resistance,
commercially available milk testing instruments and tabulated capacitance (RI and CI ), and diffusion resistance, capac-
in Table III. Specifications and price of the instruments along itance (Rd and Cd ).
with the method of testing and the parameters of detection are • The charge transfer and double layer effect are predom-
mentioned in this table. Analyzing Table III, it is clear that, inated at the film-metal boundary [25], therefore, the
the proposed milk tester can detect fat and adulterants of the impedance ZF M of the model contains the charge transfer
milk in an inexpensive way. resistance (Rct ) and double layer capacitance (Cdl ).

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 7
paper id: 1570507789
TABLE III: A comparative study of the proposed instrument with the commercially available milk testing instruments
Specifications
Ref- Name of the (voltage / power
Method of detection Detecting parameters Price
erence instrument consumption /
weight / dimension)
230 V ac, 12 kg, INR
[10] Gerber centrifuge Chemical analysis Fat: 0 - 10%
diameter: 350 mm 10,000
12V dc, 6 A, 72 W,
[11], INR
Milko tester 16 kg, 230 mm ⇥ Scattering of light Fat: 0 - 13%
[12] 30,000
310 mm ⇥ 530 mm
MilkoScan Mars 12 V dc, 5A, 60 W, Fourier transform
[13], 0-48 % fat, 0-6 % protein, 0-50 % total solids, INR
from Foss Electric 10.5 kg, 345 mm ⇥ infrared spectrometer
[14] 0-12 % SNF, 0-6 % lactose 5,00,000
co., USA 280 mm ⇥ 285 mm (FTIR) analysis
Lacto star automatic
230 V ac, 180 VA,
[15], milk analyzer from Thermo-optical 0-35 % fat, 0-10 % protein, 0-5 % minerals, 0-15 INR
15.5 kg, 440 mm ⇥
[16] Funke Gerber, method % SNF, 0-10 % lactose 4,00,000
440 mm ⇥ 200 mm
Germany
Lactoscope FTIR 110 V - 240 V ac,
[14], Fourier transform
Advanced Mid-IR 650 VA, 86 kg, 450 0-55 % fat, 0-25 % protein, 0-60 % total solids, INR
[17], infrared spectrometer
Dairy Analyzer from mm ⇥ 2300 mm ⇥ 0-25 % carbohydrates 5,00,000
[18] (FTIR) analysis
Delta instrument 500 mm
100 V - 240 V ac, 50 Fat: 1.5-13.5 %, SNF: 7-12 %, Protein: 3-4.5 %
Yama milk analyzer
W, 4.5 kg, 380 mm Mid infrared (MIR) INR
[19] YAMA MIR 350 urea, sugar, salt and detergent: 2g/l
⇥ 310 mm ⇥ 285 technology 4,00,000
mm added water: 10 %
Fat, SNF, protein
MilkoScreen from 230 V ac, 8 kg, 260 Fourier transform
Adulterations : added Water: 20 %, urea: 2.5 g/l, INR
[9], [20] German Indifoss mm ⇥ 190 mm ⇥ infrared spectrometer
melamine, ammonium Sulphate: 1 g/l, and 5,00,000
company 250 mm (FTIR) analysis
sucrose: 7 g/l
Not
[21], Near Infra Red (NIR)
Laboratory prototype Not mentioned Added water in milk men-
[22] spectroscopy
tioned
Ksheer scanner from
CSIR-CEERI, Pilani, 230 V ac, 40 W, 14 Urea: 1 g/l, common salt, ammonium sulphate: 2
[20], Multivariate data INR
and commercialized kg, 300 mm ⇥ 150 g/l, detergent, caustic soda, sodium bi-carbonate:
[23] analysis 2,00,000
to EMAT by REIL mm ⇥ 350 mm 3 g/l, and added water: 70 %
company, India
Low fat milk (LFM): (<= 5 % fat),
High fat milk (HFM): (> 5 % & <= 12 % fat),
Very high fat milk (VHFM): (> 12 % fat) INR
230 V ac, 83 W, 5 Change of phase (upto 20% fat is measured) 10,000
Proposed instrument kg, 210 mm ⇥ 120 angle of the sensor Common salt (NaCl): 1 g/l (compo-
mm ⇥ 210 mm impedance Ammonium Sulphate: 1 g/l nent
Detergent: 2 g/l cost)
Sodium bicarbonate: 2 g/l
Sodium hydroxide: 2 g/l

A sinusoidal excitation voltage with an angular frequency of can be observed that the measuring process exhibits both the
! rad/s is applied to the two terminals of the sensor. The milk charge transfer and diffusion kinetics.
medium is consisting of ions. When the sensor is dipped in that In a spatially-restricted medium such as the porous film, a
medium, ions are injected into the film, due to the generation characteristic frequency separates these two distinct regions.
of ionic concentration wave C (x, t) by the applied voltage [25]. This frequency is given as [25],
The concentration wave depends on the concentration gradient D
between the liquid-film boundary and film-metal boundary, as !d = 2 ; (2)
L
mentioned in Fig. 10, and denoted by position ‘x’. According where, L is the thickness of the film. As D and L are constants,
to Fick’s first law [26], the rate of diffusion (J) depends on the !d is also constant for a particular system. By analyzing the
concentration gradient, where D is the diffusion coefficient. data of Fig. 12 (a), it can be seen that, at around 500 Hz,
@C a gradual transition from diffusion to charge transfer takes
J= D (1) place in all the milk samples. Therefore, the characteristics
@x
frequency is around 500 Hz. Fig. 12 (b) gives the zoomed
Complex impedance plot of the experimental data for three view of the transition part. It shows that, with the increase of
milk samples with the fat content of 3%, 6%, and 12%, in the fat content in milk, this transition point shifts towards left
the frequency range of 20 Hz to 5 kHz, is shown in Fig. 12 of the complex impedance plot i.e. towards the high frequency
(a). A straight line of slope 45 in the Nyquist plot represents region. This is because of the increase in diffusion rate (J)
the diffusion region of the electrochemical system [27], [28]. with the increase of concentration gradient at the solution-film
Whereas, the half semi-circle in the Nyquist plot indicates the interface (Fick’s law). The transition point occurs early as the
charge transfer region [27], [28]. Hence, from Fig. 12 (a) it ion concentration increases (i.e. the increase of fat content).

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 8
paper id: 1570507789

(a) (b)
Fig. 12: (a) Complex impedance plot for three milk samples
with fat content of 3%, 6%, and 12%, in the frequency range
of 20 Hz to 5 kHz (b) zoomed view of the transition region
Fig. 10: Equivalent electrical circuit model of the sensing
system used for fat detection

Fig. 13: Complex impedance plot for pure and adulterated milk
at the frequency range of 20 Hz to 2 MHz
Fig. 11: Porous surface of lipase + PMMA coated sensor [4]

are integrated with this prototype: 1) a small temperature

Thus, a clear difference in impedance can be noticed between control system, which is developed to keep the milk at a
the different fat content milk samples at this point. warm temperature of 42 C, and 2) an optimized signal
conditioning circuit to read the responses of two sensors.
B. Chemical kinetics behavior of PMMA coated sensor Lipase + PMMA coated sensor for fat detection, and only
The same behavior of lipase + PMMA coated sensor is PMMA coated sensor for adulteration detection. Lifetime,
also observed for PMMA coated sensor. The Nyquist plot, reproducibility, and repeatability analysis of both the sensors
in the range of 20 Hz to 2 MHz, for the PMMA coated are studied in detail.
sensor in pure and adulterated milk is shown in Fig. 13. A calibration methodology with respect to a standard solu-
Adulterated milk samples are prepared by separate mixing of tion is proposed for this prototype to detect the adulteration in
five common adulterants, having a concentration of 0.2 % (2 milk. Error analysis has been carried out for the individual
g/l), with the pure milk (PM). These are ammonium sulphate module of the instrument. Power consumed by the instru-
(AS), sodium chloride (NaCl), sodium hydroxide (NaOH), ment is 83 W, where the heater unit itself consumes 78 W.
sodium bi carbonate ((NaHCO3 ), and detergent (Det). LFM Therefore, further research is required to minimize the power
milk, having fat content of 3 % are used. It is worth to consumption of the heater.
mention that, experimentally it has been proved, the minimum A flow chart is included to illustrate the operation of the
detectable amount of these five adulterants are 1 g/l for NaCl, milk tester. Working of the proposed instrument is validated
and AS, and 2 g/l for detergent, NaOH, and NaHCO3 , when with a standard instrument and it is also compared with the
the standard solution is used for the calibration. commercially available milk testers to understand the market
In the Fig. 13, a transition is observed from diffusion challenges. Research is carrying on to detect other adulterants
to charge transfer region for the PMMA coated sensor as like urea, formalin, boric acid and sugar in milk along with
observed for lipase + PMMA coated sensor. This gradual the detection of SNF, protein and lactose content.
transition occurred around 5.6 kHz (viz. Fig. 13). However, The work further investigates the issue of a particular
a clear difference can also be observed in diffusion region due frequency for the detection of fat and adulteration in milk.
to the high ionic concentration of the adulterated milk. Complex impedance plot of the lipase + PMMA coated sensor
shows that a gradual transition occurs from diffusion control
VII. C ONCLUSION to charge transfer control region at around 500 Hz frequency.
In this work, a simple and low-cost milk tester is proposed Similar behavior is also observed for PMMA coated sensor at
to detect the fat and adulteration in milk. Two main parts 5.6 kHz.

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIM.2019.2955796, IEEE
Transactions on Instrumentation and Measurement
IEEE TRANSACTION ON INSTRUMENTATION AND MEASUREMENT 9
paper id: 1570507789

ACKNOWLEDGMENT [26] J. Lowney and R. Larrabee, “The use of Fick’s law in modeling diffusion
processes,” IEEE Transactions on Electron Devices, vol. 27, no. 9, pp.
The authors gratefully acknowledge the laboratory facility 1795–1798, 1980.
of National Dairy Research Institute (NDRI), Kalyani, India [27] E. Barsoukov and J. R. Macdonald, Impedance Spectroscopy Theory,
Experiment, and Applications. Wiley Interscience, Hoboken, New
for validation of the instrument. Jersey, 2005.
[28] M. Grossi and B. Ricco, “Electrical impedance spectroscopy (EIS) for
biological analysis and food characterization: a review,” Journal of
R EFERENCES Sensors and Sensor Systems,, vol. 6, pp. 303–325, 2017.
[1] M. Guetouache, B. Guessas, and S. Medjekal, “Composition and nu-
tritional value of raw milk,” Biological Sciences and Pharmaceutical Moupali Chakraborty received her B.Tech degree
Research, vol. 2, no. 10, pp. 115–122, December 2014. in Electronics & Instrumentation Engineering from
[2] M. Chakraborty and K. Biswas, “Limit of detection for five common Techno India College of Technology in 2010 and
adulterants in milk: A study with different fat percent,” IEEE Sensors M.Tech degree in Instrumentation Engineering from
Journal, vol. 18, no. 6, pp. 2395–2403, March 2018. NIT Agartala in 2015. Since July 2015, she is a
[3] M. Chakraborty and K. Biswas, “Hardware platform to detect fat Research Scholar in the Department of Electrical
percent in milk using a lipase immobilized PMMA-coated sensor,” IEEE Engineering at IIT Kharagpur, India and working
Transactions on Instrumentation and Measurement, vol. 68, no. 11, pp. in the area of sensor design and development of
4526–4534, November 2019. instrumentation system.
[4] M. Chakraborty and K. Biswas, “Electrical model for lipase immobilized
PMMA coated sensor to detect fat content in milk,” in Proc. IEEE Int.
Instrum. Meas. Tech. Conf. I2MTC, May 2019.
[5] O. D. Ernest, “Measurement systems: Application and design,” McGraw-
Hill book Co., Singapore, 1990. Karabi Biswas (M’09) received her B.Tech. and
[6] S. Shelton, “Thermal conductivity of some irons and steels over the Ph.D. degree from Indian Institute of Technology,
temperature range 100 to 500 C,” Bureau of Standards Journal of Kharagpur, India, in 1992 and 2007 respectively, and
research, vol. 12, no. 4/6, pp. 441–450, 1934. M.Tech. degree from Jadavpur University, Kolkata,
[7] http://www.selectautomation.net/downloads/TC513-temprature- India, in 2000. She is presently an Associate Pro-
controllers.pdf, accessed: 2019-11-16. fessor in the Department of Electrical Engineering,
[8] N. Chapin, “Flowcharting with the ANSI standard: A tutorial,” ACM IIT Kharapur, Kharapur, India. Her research interests
Computing Surveys (CSUR), vol. 2, no. 2, pp. 119–146, 1970. include study of fractional order systems, sensor
[9] http://www.milkoscreen.com, accessed on: 2019-11-16. design and development of instrumentation system.
[10] Manual of methods of analysis of foods and milk products. Food safety
and standards authority of India, Ministry of health and family welfare,
Government of India, 2015.
[11] R. E. Ginn, “The milko-tester and problems related to its use,” Journal
of Milk and Food Technology, vol. 37, no. 4, pp. 218–221, 1974.
[12] http://www.reiljp.com/pdf/emt.pdf, accessed: 2019-11-16.
[13] https://www.fossanalytics.com/en-in/products/milkoscan-mars,
accessed: 2019-11-16.
[14] J. Lynch, D. Barbano, M. Schweisthal, and J. Fleming, “Precalibration
evaluation procedures for mid-infrared milk analyzers,” Journal of dairy
science, vol. 89, no. 7, pp. 2761–2774, 2006.
[15] https://uk.vwr.com/store/product/8563349/instrument-for-milk-analysis-
lactostar, accessed: 2019-11-16.
[16] http://www.bennyimpex.in/, accessed: 2019-11-16.
[17] https://www.perten.com/Products/LactoScope-FTIR/, accessed on: 2019-
11-16.
[18] D. Instruments, “An advanced instruments company, lactoscope FTIR
user’s guide. milk and liquid dairy product analyzer,” Netherlands: Delta
Instruments, 2007.
[19] http://www.everestinstruments.in/products/milk-analyzers/yama-mir-
350, accessed on: 2019-11-16.
[20] R. Sharma, B. Mann, K. Satya, and D. K. Nanda, “Quick tests for check-
ing adulteration in dairy products,” in Enabling Extension Functionaries
to Address Field Level Problems in Animal Husbandry in compendium
of All India Animal Husbandry Officers Workshop-2015. by: Dr. AK
Srivastava Director & Vice Chancellor ICAR-NDRI, Karnal, 2015, p. 89.
[21] L. de Souza Ribeiro, F. A. Gentilin, J. A. de França, A. L. d. S. M.
Felı́cio, and M. B. d. M. França, “Development of a hardware platform
for detection of milk adulteration based on Near-Infrared diffuse reflec-
tion,” IEEE Transactions on Instrumentation and Measurement, vol. 65,
no. 7, pp. 1698–1706, 2016.
[22] L. da Silva Dias, J. C. da Silva Junior, A. L. d. S. M. Felı́cio, and
J. A. de França, “A NIR photometer prototype with integrating sphere
for the detection of added water in raw milk,” IEEE Transactions on
Instrumentation and Measurement, no. 99, pp. 1–8, 2018.
[23] G. Sahni, “Council of scientific & industrial research: Creating scientific
knowledgebase–delivering technology,” 2018.
[24] C. Ho, I. Raistrick, and R. Huggins, “Application of AC techniques to
the study of lithium diffusion in tungsten trioxide thin films,” Journal
of The Electrochemical Society, vol. 127, no. 2, pp. 343–350, 1980.
[25] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, and P. R. Bueno,
“Theoretical models for ac impedance of finite diffusion layers exhibit-
ing low frequency dispersion,” Journal of Electroanalytical Chemistry,
vol. 475, no. 2, pp. 152–163, 1999.

0018-9456 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.