Proc.

of the IEEE International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA)
25-27 November 2014, Kuala Lumpur, Malaysia

Comparative Investigation Using GaAs(950nm),

GaAIAs (940nm) and InGaAsP (1450nm) Sensors for
Development of Non-Invasive Optical Blood Glucose
Measurement System
Nina Korlina Madzhi, Sarah Addyani Shamsuddin, Mohd Firdaus Abdullah
Faculty of Electrical Engineering
Universiti Teknologi Mara
40450 Shah Alam, Selangor, Malaysia
nina6875@yahoo.com
sarah.addyani@gmail.com

Abstract—Previously, many researches had been done on

non-invasive using near-infrared sensing. Sia [1] had
investigated near-infrared sensing using signal penetrating Index Terms—Infrared, Near-infrared, Non-invasive, Blood
Glucose.
finger method. However, by using finger penetration, there
are no results obtained. He only obtained signal using I. INTRODUCTION
glucose concentration. Therefore the objectives of this
research are to investigate the performance of three Based on the data by International Diabetics Federation
different wavelength of sensors; infrared 940nm and (IDF) [2] in year 2010, there were 284.6 millions of diabetes
950nm and also near-infrared 1450nm. Sensor that gave patients globally. IDF estimated that the total of people around
the best output had been chosen to design a non-invasive the world with diabetes, will doubled with increment of 54% in
optical blood glucose measurement device. Generally, the year 2030. The number is very large and people should be
overall system consists of three parts including sensor aware of this matter. Figure 1.1 below has shown the global
part, signal conditioning circuit, and also numerical projections for number of people with diabetes in year 2010
display. The initial design tested by considering initial and 2030 in whole wide world. Moreover, having diabetes also
voltage 1.616V to 1.68V which referred to previous can lead one to blindness, kidney failure, heart attack, stroke
research by Sia [1] as the output of the sensor. Then and amputation. Yearly, there are four million people died
proceed by using test tube which contains various because of diabetes.
percentage of glucose concentration. The same methods
had been used to the human samples fingers instead of test
tube. From the experiment, output graph of the 950nm
shows more consistent pattern compared to the 940nm.
950nm also has a larger range scale for voltage which
from 5.016V to 5.4633V compare to the 940nm voltage
range scale which from 5.0327V to 5.4201V. Further test
on human finger had been done by using 950nm infrared
but the output voltages were too small. The performance
of the measurement can be improved by controlling the Figure 1.1 IDF Regions and Global Projections for Number of People with
surrounding condition and fixed the path length between Diabetes (20-79 years), 2010-2030 (Picture courtesy of IDF Diabetes Atlas [2]
transmitter and receiver. Test using test tube showed that
the near infrared and infrared were capable to predict To date, people in the entire world mostly live in
different glucose concentration. By using circuit designed, unhealthy daily routines and lifestyles which can lead to
it can be seen that the voltage reading became higher illnesses. Based on research by M. Mafauzy, Z. Hussein and
compared to before meal. Therefore, it can be concluded S. P. Chan [3], quality of life evaluation showed that about
that the circuit design functions accordingly and also the one third of diabetis patients in Malaysia have poor quality of
non-invasive. During human sample test, increment life. For example, dietary is one of the caused that can
pattern can be seen from fasting to non-fasting condition increase people’s blood glucose reading which also known as
but the main effect is all samples have different fingers’ diabetes. Dietary including eating time, type and amount of
diameter which each of user needs to be calibrated. the food especially carbohydrates which give large effect to

978-1-4799-0841-3/14/$31.00 ©2014 IEEE

the blood glucose level. However, glucose is also important The output voltage for each stage of signal conditioning
since it is the main carrier of energy in the human organism circuit had been tested to make sure that the results are
[4-6]. The most important factor is that the amount of food according to the expected results as in the calculation.With the
that we take should be suitable with our daily activities. circuit design, process had been preceded with the assembly
Besides, lack of exercise and physical activities plus stress can on the breadboard. Troubleshoot had been done to make sure
also increase the level of glucose. It is better if people can that the output results were similar to the results in the
check their blood glucose level once in a while because simulation. Then the data from the simulation had been
knowing their health condition can make people appreciate collected. PCB for overall of the system had been designed
more with their lives. People also can make adjustment to and fabricated. The system had been tested in two main tests
improve their lifestyle and control meals they eat. For diabetes which are by varying concentration of glucose solution and
patients, it is important for them to regularly test and record also with the human finger as samples. To verify and support
the result to see the improvement of their health by changing the output result, the human samples also had been tested with
their lifestyle. Hence, the existence of the portable blood commercial finger-prick method. The output results had been
glucose self check device is very helpful to the patients. compared as the verification. Programming for the PIC had
been developed as the numerical output display of the system.
II. METHODOLOGY The program had been done to make sure that the output can
display the status of the output reading if either with low,
medium and high blood sugar. All the experiments had been
done by testing with both by varying concentrations of
glucose solution and also by testing on the real human finger
as samples in the experiments. All data had been collected and
then analysed. Lastly, by completing all the steps and all the
data had been analysed, the report was written and
documentation had been done.
III. RESULT AND DISCUSSION

TABLE I Result of experiment using test tube

Glucose Vo Photodiode Vo Photodiode
Concentration (%) (V) (1450nm) (V) (940nm)
100 0.938 0.4831
90 0.937 0.4745
80 0.935 0.4668
70 0.934 0.4458
60 0.933 0.4499
50 0.933 0.4515
40 0.933 0.4477
30 0.933 0.4544
20 0.933 0.4541
10 0.932 0.4367

Figure 2.1 Flow chart of the overall process in construct this project.

Figure 2.1 shows the flow chart of the overall

process required in this project. Starting by doing detail
research and literature review related to this topic. From the
literature review, the findings of each research had been
analyzed which lead to research problems. Besides that,
advantages and disadvantages of the method used in each
research had been identified. With the research and literature
that had been done, the focus in the topic can be smaller and Figure 3.1 Comparison between LED 1450nm and LED 940nm output
more specific and objectives of the project then were created.
As the next step, circuit had been designed. Designing started
Since the output readings of both transmitters are not
by theoretical calculation. Generally, the system need three
affected by different percentage of glucose concentration,
parts which are sensing part, signal conditioning circuit and
circuit design had been adjusted from the initial design. These
Programmable Integrated Circuit (PIC) as the display.
show that the output voltage from the photodiode too small
Theoretically, the circuit had been designed and voltage
compared to the initial input voltage that had been considered
output for each step had been calculated and considered. From
while designing the circuit. The stage of amplification had
that, all the components required can be listed.After that, the
been amplified higher to make sure that the voltage readings
simulation for the preliminary designed circuit had been done.
are much higher than the previous output. Linearization part 7 56 163 0.1460 0.2080 29.81
had been adjusted as suitable to the output reading as long as 8 57 158 0.1335 0.1815 26.45
9 63 169 0.2360 0.2560 7.81
the final output voltage are not exceed than 5V.

TABLE II Glucose Reading for 940nm and 950nm

Infrared Wavelength
%
940nm 950nm
100 5.4201 5.4633
90 5.3396 5.2812
80 5.2489 5.2225
70 5.2045 5.2138
60 5.276 5.2064
50 5.1891 5.182
40 5.1625 5.147
30 5.151 5.128
20 5.1345 5.087 Figure 3.3 Fasting versus Non-Fasting
10 5.0771 5.053
0 5.0327 5.016
From TABLE III and Figure 3.3 above shows that
there are increments in voltage reading after having meal
compared to during fasting. Even though the output voltage is
small values, but then the pattern of increments still can be
seen. The pattern of effect after taking meals also can be
accepted since after having meals, the blood glucose reading
for each volunteer are increasing.

A. COMPARISON ANALYSIS BETWEEN INFRARED AND

NEAR INFRARED SENSOR TESTING
Figure 3.2 Graph Output Voltage of 940nm and 950nm versus Glucose
Concentration. Light emitting diode, LED is a complex
Figure 3.2 shows, the different voltage output semiconductor coverts electrical signal into corresponding
reading from two IR wavelengths which 940nm and 950nm. light signal. The advantages of using LED are small in size,
From the graph, the 950nm looks more consistent compare to high radiance which emit lots of light in a small area, high in
the 940nm. 950nm also has a larger range scale for voltage reliability and can be modulated (turned on and off) at high
which from 5.016V to 5.4633V compare to the 940nm voltage speed [7].Based on the application that needs to be used, there
range scale which from 5.0327V to 5.4201V. The percentage are several LEDs’ performance characteristics that can be
increase between 940nm and 950nm is 5.99%.Based on test of determined such as peak wavelength, spectral width, emission
varying percentage of glucose concentration, it shows that the pattern, power and speed. The peak wavelength is the
950nm infrared give better and more stable output, further test wavelength where the source emits most power. Spectral
on human finger had been done by using 950nm infrared as width is the range of wavelength that light is emitted in
the transmitter of the device. The test had been done using practise which also known as spectral width of the source [7].
human finger as the sample. The finger is placed between There are two types of LED structures which are edge emitters
phototransistor and photodiode. The output reading is and surface emitters. The differences of these two types are
measured by using multimeter at the last stage of the signal edge emitter LED is more complex and expensive. The
conditioning circuit. emitting spot also small resulting high output power. While
surface emitter LED have simple structure and comparatively
inexpensive. It offers low to moderate output power levels and
low to moderate operating speed [7].LED optical output is
approximately proportional to drive current. Temperature is
TABLE III Human Sample Reading one of the factors that affect the optical output. Figure 3.4
Non- Percentage shows the detail of the typical behaviour of LED.
Weight Height Fasting
No. Fasting Difference
(kg) (cm) (V)
(V) (%)
1 96 171 0.13096 0.1710 23.42
2 91 175 0.1340 0.1590 15.72
3 55 160 0.2014 0.3056 34.10
4 73 174 0.1004 0.1300 22.77
5 67.6 173 0.1027 0.1202 14.55
6 64.5 170 0.1820 0.1940 6.19
 voltage of the transistor. Equation below shows the general
form of the calculation [7].

ILED = VPOWER – VLED – VSAT/R3 …(2)

Where:
VPOWER = DC power supply voltage
VLED = forward voltage drop of the LED
VSAT = drive transistor saturation voltage
R3 = series LED current limiting resistor
ILED = peak LED current
Figure 3.4 Optical Outputs vs. Current in LED (Picture courtesy from fiber-
optics.info [53]). Experiment and testing had been done in controlled
environment since LED is sensitive to ambient light condition
LEDs have voltage drop about 1.1 to 1.5 Volts. [7]. All experiments had been done in same laboratory with
Shorter wavelength diodes have largest voltage drops [7] of fixed position of devices in the experiment setup.
the LED. For all LEDs used in this work, voltage drop of LED
940nm is 1.35V, LED 950nm is 1.3V while LED1450nm is
B. REFINEMENT OF NEAR- INFRARED TESTING
1.2V. From that, the pattern can be seen where as the
wavelength increases, the voltage drop decreases. This is due
As the same signal conditioning circuit used with the
to the bandgap energy, Eg which higher wavelength have
near-infrared, the result of output voltage during human finger
smaller energy gap and resulting smaller voltage drop.
test shows no changes. The circuit was redesigned for the
near-infrared testing. And did the re-test to test either the
Eg = hc/l = 1240eV-nm/l …(1)
sensor can give better output or not. The linearization circuit
had been removed from the design since output from low pass
Where:
filter was not exceeding 5V.
h = Plank’s constant = 4.13 x 10-15 eV.s
c = speed of light = 2.998 x 108 m/s
1) Test Tube Experimental
l = wavelength in nm
Experiment had been done using the difference of
From equation above, the energy gap of LED based
glucose concentration and the output readings from the test are
on its emission wavelength had been predicted as in TABLE
as below:
IV.
TABLE V Output Voltage of 1450nm Vary Glucose Concentration
TABLE IV Common Light Emitter Materials and Characteristics. Test
Glucose Glucose Output
Material Formula Energy Gap Wavelength Percentage Concentration Voltage
Gallium (%) (mg/dL) (V)
GaP 2.24 eV 550 nm
Phosphide 100 200 3.4174
Aluminum 90 180 3.2009
AIAs 2.09 eV 590 nm
Arsenide 80 160 2.9830
Gallium 70 140 2.8120
GaAs 1.42 eV 870 nm
Arsenide 60 120 2.4763
Indium 50 100 2.2880
InP 1.33 Ev 930 nm
Phosphide 40 80 2.1062
Aluminum- 30 60 1.5711
Gallium AIGaAs 1.42-1.61 Ev 770-870 nm 20 40 1.1660
Arsenide 10 20 0.8730
Indium- 0 0 0.6000
Gallium- 1100-
InGaAsP 0.74-1.13 Ev
Arsenide- 1670nm
Phosphide
Figure 3.5 shows the output pattern of the output
voltage from the test of different glucose concentrations.
From TABLE IV, shows that there are variation of
Graph shows that the voltage is nearly direct proportional to
material used in LED. In this work, three variation of
the percentage of glucose concentrations. From the pattern, it
materials had been used which are GaAlAs (940nm), GaAs
shows that when the percentage of glucose concentration is
(950nm) and InGaAsP (1450nm). Voltage drop which also
high, voltage output also high and vice versa.
called as forward voltage is indicated the energy gap that
corresponds to the energy of the emitted photons. LED current
can be calculated from the known voltage drop and saturation
 TABLE VI Overall Results of Near-Infrared Human Testing.
Output Voltage vs. Percentage of
Glucose Concentration
Output Voltage (V

Blood Glucose Reading

Before Meal After Meal
5 Commercial Present Commercial Present
Meter Designed (V) Meter Designed (V)
(mmol/L) (mmol/L)
0
5.8 1.7111 1.7069 7.1 1.7283 1.7191
0 20 40 60 80 100 5.6 1.6412 1.6352 5.0 1.6571 1.6790
Percentage of Glucose Concentrations (%) 5.5 1.6407 1.6547 5.9 1.6510 1.6549
5.4 1.6521 1.6466 6.3 1.6760 1.6677
5.4 1.6659 1.6493 5.2 1.6615 1.6570
Figure 3.5 Graph of Output Voltage vs. Percentage of Glucose Concentration 4.9 1.7056 1.6776 5.7 1.6509 1.6596
4.9 1.6622 1.6570 5.5 1.7238 1.6583
4.8 1.3762 1.6644 5.4 1.7123 1.6929
Based on Diabetes Research and Wellness 4.7 1.6712 1.6573 6.7 1.6436 1.6522
Foundation [8] shows that random blood sugar in normal 4.7 1.6754 1.6893 5.8 1.6828 1.6900
reading is from 80 to 139mg/dL and one will consider have 4.6 1.6827 1.6976 6.0 1.7107 1.6970
diabetes when the glucose reading is 200mg/dL and above. 4.6 1.6596 1.6421 5.4 1.6496 1.3501
4.6 1.6696 1.6419 6.1 1.6973 1.3669
From the data that had been collected, by comparing output
4.5 1.6620 1.6544 4.3 1.6634 1.3524
voltage with the glucose concentration, the results are 4.3 1.6843 1.6988 5.0 1.7025 1.3833
successful in getting the relationship between the output 4.3 1.6888 1.6717 4.7 1.7345 1.7266
voltage readings with the output display of PIC. 3.9 1.6848 1.6907 5.2 1.7239 1.7819

2) Human Sample Final Testing

From TABLE VI, it can be seen that all the human
As all the test and experiments had been done samples taken are in normal range of blood glucose reading.
previously, the most suitable sensing part of the system is This is due to the fact that the samples are taken from healthy
using LED1450E pairing with photodiode FGA01FC. In students at the age of 23 years old with no diabetic condition.
previous test also try and error of the signal condition circuit The testing is only to prove that the circuit designed
had been done in order to give the best and most accurate functioning well with verification using commercial glucose
output results. After the finalized all the components and meter. However, from the observation of the output readings,
circuit construction, the actual testing had been done with 17 the output voltages are not consistent compared to the
volunteer as the human sample. In this test, another factor also commercial meter. This may occurred because of the different
had been taken. The test consists of body mass index, blood finger diameter size of each person that effected the signal
pressure and also their heart beat. These elements had been penetration through different person. But then, increment
observed to see if there are any effects to the output pattern still can be seen from fasting to non-fasting condition
reading.Test also had been done using the present design. for every sample. The obvious effect is all samples have
Reading had been taken twice to make sure the repeatability of different fingers’ diameter which each of user needs to be
the system. Lastly, volunteer also had to do the blood glucose calibrated.
test using commercial meter with the finger pricking method
as verification. The commercial reading had been taken to test C. PIC OUTPUT DISPLAY
the performance of the present designed. Test of both blood
glucose meters had been done in two different conditions When the development of sensory part is complete,
which are during fasting and non-fasting. Volunteer had been several experiment is conducted to determine whether the
asked to fast for at least eight hours. At 10 am, the first test instrument that have been develop was able to process the
had been done and the data obtained considered as during electrical signal given by photodiode and display it on LCD.
fasting. After that, they had their breakfast which the meals The experiment was implementing by using led with
also had been controlled. All of them had been given same wavelength 1450nm as transmitter and InGaAs photodiode as
meals which each of them had one wrapped nasi lemak and receiver to measure glucose concentration. The suitable
250ml of sugar added fruit juice. After two hours, second voltage range needed to interface with microcontroller is
reading had been taken as the non-fasting condition.Results between 0V to 5V. Therefore the photodiode output voltage
obtained had been taken and as in Table 25. From the data that was linearized into the suitable range. By using the same
had been collected, by comparing output voltage with the configuration of linearization and summing amplifier that has
glucose concentration, the results are successful in getting the been discussed in previous chapter, the photodiode output
relationship between the output voltage reading to the output voltage are able to converted into a suitable range to operate
display of PIC where had been programmed. with microcontroller. Figure 3.6 show the result from display
panel that display the result from the experiment after be
processed by signal conditioning circuit.
 fasting to non-fasting condition for every sample. The obvious
effect is all samples have different fingers’ diameter which
each of user needs to be calibrated. So it suggested in future
work to add calibration knob or reset button so that for every
testing with different patient as sample, the reading will starts
at ‘0’.
Besides, the instrument based on self-monitoring
glucose level still can be enhanced by adding trigged alarm.
Different type of triggered alarm that produces different type
Figure 3.6 Output Reading Display.
of sound can be used to indicate which glucose level that has
been measured. This method was recommended to overcome
IV. CONCLUSION the problem to those have blurred vision especially older
people. The current instrument need to be power up by using
The first objective had been achieved by comparing two direct current supplies with needed voltage 18V. As
the performance of infrared and near-infrared as the sensing recommendation, the future instrument can be powered by
device of non-invasive blood glucose. The instrument based using lithium battery so that the instrument can be carry out
on self-monitoring glucose level was developed as and the easily and can be use anytime and anywhere.
experimental processes had been done well. Signal conditional
circuit was able to filter noise produced by near infrared ACKNOWLEDGMENT
sensor by blocking high frequency. The benefit from noise The authors would like to express heartily gratitude to project
filtering is better glucose signal can be measured and accurate supervisor for the guidance and enthusiasm given throughout
reading can be obtain. Besides that, signal conditioning circuit the progress of this project and to RMI UiTM for their
was able to amplify the input voltage from near infrared financial support (grant: 600-RMI/DANA 5/3/RIF
sensor into suitable value to be observed. The output from (87/2012))and Faculty of Electrical Engineering for the
signal conditioning circuit are also successful be linearize facilities provided.
from 0V to 5V. From that voltage range, the signal
conditioning circuit was able to interface with microcontroller. REFERENCES
The input from signal conditioning circuit was able to be [1] D. Sia, “Design of a Near-Infrared Device for the Study of
converted to digital value. By displaying the digital value on Glucose Concentration Measurements,” B. Eng. Thesis, Dept.
display panel, people will be able to determine their glucose Elect. And Biomedical Eng., McMaster Univ., Hamilton,
level which leading them to better lifestyle.It can be concluded Ontario, Canada, 2010.
that the circuit design can functions accordingly and also non- [2] International Diabetes Federation. (2009). IDF Diabetes Atlas
invasive. During human sample test, increment pattern can be (4thed.) [Online] Available: http://www.idf.org/diabetesatlas/
seen from fasting to non-fasting condition but the main effect [3] M. Fafauzy, Z. Hussein, and S. P. Chan, “The Status of Diabetes
is all samples have different fingers’ diameter which each of Control in Malaysia: Results of DiabCare 2008”, Med. J.
user needs to be calibrated. The calibration also may due to Malaysia, vol. 66, No. 3, Aug. 2011
different people have varying amounts of protein, fats and [4] Nordqvist. (2013, February 19). What Are Carbohydrates? What
water. Besides, doing measurement on the skin tissues, the Is Glucose?. Medical News Today [Online]. Available:
absorbance spectra can be affected by other elements http://www.medicalnewstoday.com/articles/161547.
contained in blood such as water, albumin, globulin and also [5] Gwen, “What is Diabetes?,” in Diabetes Research & Wellness
environmental factors. However environmental lighting, Foundation What is DIABETES?, vol. 3, Hampshire, United
temperature and humidification had been controlled by doing Kingdom, 2013.
all the experimental process in the same room. [6] C. E. Ferrante do Amaral and B. Wolf, “Current Development
in Non.invasive Glucose Monitoring,” Medical Engineering &
V. RECOMMENDATION FOR FUTURE WORK Physics, vol. 30, pp 541-549, 2008.
From the observation of the output readings, the [7] Emcore Corporation, “Light-emitting Diode (LED),” 2014
output voltages are not consistent compared to the commercial [Online]. Available at: http://www.fiber-
meter. It shows that there are errors occur in the system. This optics.info/articles/light-emitting_diode_led.
may occurred because of the different diameter size of each [8] International Diabetes Federation. (2013). IDF Diabetes Atlas
person that makes different penetration trough the different (6thed.) [Online] Available: http://www.idf.org/diabetesatlas/
person. But then, increment pattern still can be seen from