Measurement 63 (2015) 128–136

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Optoelectronic sensor device for monitoring the maceration

of red wine: Design issues and validation
F. Jiménez-Márquez 1, J. Vázquez ⇑,1, J.L. Sánchez-Rojas
Microsystems, Actuators & Sensors Group, Escuela Técnica Superior de Ingenieros Industriales, Universidad de Castilla-La Mancha, Avenida Camilo José Cela
s/n, 13071 Ciudad Real, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Enologists usually rely on spectrophotometers to perform the chromatic characterization
Received 6 August 2014 of red wines. This work reports an optoelectronic instrument based on absorbance mea-
Received in revised form 31 October 2014 surements aiming not only at the assessment of the chromatic characteristics of finished
Accepted 2 December 2014
red wines, but at the supervision of the gradual maceration of fermenting grape musts
Available online 13 December 2014
as well. Maceration is a chemical process that takes place during the early stages of the fer-
mentation that finally results in red wine. It is through maceration that wine acquires its
Keywords:
distinctive chromatic features. Key issues related to the selection of optoelectronic compo-
Color
Absorbance
nents, the system design and its final validation using diluted wine samples are thoroughly
LED discussed. Resolution figures lie in the range of one thousandth of an absorbance unit.
Photodetector Maceration has been successfully monitored with the proposed instrument using grape
Fermentation must samples extracted during the first three days from two different fermentations.
Wine maceration Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction of ethanol. One of the factors that determine the color of a

The alcoholic fermentation that turns grape must into
wine is a chemical process that remains active over a span
of several days and is periodically supervised by winemak-
ers. Although the time a wine fermentation is in progress
depends on a number of factors, like the amount and type
of yeast utilized, the sugar concentration of the grape must
and the temperature of the tanks, it can be claimed that
sugar content in the original grape must is virtually
depleted in many cases after twelve days counting from
the day when yeasts were inoculated [1]. During this time,
the grape must undergoes changes that alter both its
chemical composition and its appearance due to the con-
sumption of sugars by yeast and the subsequent formation
⇑ Corresponding author. Tel.: +34 926295300/3899.
E-mail addresses: k.jimenez.marquez@gmail.com (F. Jiménez-
Márquez), javier.vazquez@uclm.es (J. Vázquez), joseluis.saldavero@uclm.
es (J.L. Sánchez-Rojas).
1
Both authors have contributed equally to this work.
finished red wine is the maceration length; that is, the
amount of time that grape skins are kept in the fermenting
grape must once the fermentation has begun. Conse-
quently, the possibility of monitoring the color evolution
of grape musts, as a given fermentation is in progress,
would provide enologists with valuable information as to
when it is the right moment to remove the grape skins
from the fermentation tanks. Unlike other parameters of
relevance in winemaking, like the density, which is regu-
larly monitored by enologists during the whole fermenta-
tion span, the chromatic characteristics are usually
characterized once the fermentation is finished.
There is not a single method to characterize the color
for a given application. Some of the available techniques
rely either on RGB coordinates [2,3]; the CIELab color space
[4,5]; or on measuring optical densities at multiple discrete
wavelengths [6]. In the case of winemaking, the chromatic
characteristics of a given wine are usually determined by
enologists from absorbance measurements carried out at

http://dx.doi.org/10.1016/j.measurement.2014.12.009
0263-2241/Ó 2014 Elsevier Ltd. All rights reserved.
 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136 129

three wavelengths in the visible spectrum, using a spectro- Absorbance at a given wavelength Ak is calculated using
photometer. Short macerations produce rosé wines, the Lambert–Beer law, which, as demonstrated in [11],
whereas longer maceration times are required for red holds for LED-based photometers applied to a wide range
wines [1]. of sample thicknesses:
The current design has been previously tested with
actual grape must samples extracted periodically at differ- I0
Ak ¼ log ¼ alc ð1Þ
ent stages of a number of fermentations carried out at lab- I
oratory scale. Details relating to both the winemaking
procedure and the analysis of the measurements carried where I and I0 are the radiation intensities after passing
out on the extracted samples, aiming at monitoring the through the sample under test and through distilled water,
evolution of the chromatic characteristics of grape musts respectively (see Fig. 1). a is the absorptivity, l is the path
during the maceration phase, can be found in a previous length of absorption and c is the concentration of the
work [7]. The present work does not intend to focus on absorbing species [12].
the chemical aspects of wine fermentations; it is rather Let A420, A520 and A620 be the absorbances at 420, 520
devoted to cover in detail key issues related to the design, and 620 nm using a 1-cm sample thickness, the common
the selection of optoelectronic components and the subse- reference in colorimetry. The expressions (2)–(4) provide
quent validation of the developed sensor device. The sys- the color intensity, the hue and the brightness, respectively.
tem described in this work is part of a multi-purpose
optoelectronic instrument initially conceived to obtain CI ¼ A420 þ A520 þ A620 ð2Þ
not only the chromatic characteristics of a fermenting
grape must, but its refractive index at different stages of A420
the fermentation too. Measuring the refractive index H¼ ð3Þ
A520
instead of the density represents an alternative method
to follow the fermentation kinetics [7,8].
ðA420 þ A620 Þ 100
dAð%Þ ¼ A520   ð4Þ
2. Chromatic characterization 2 A520
Eqs. (2)–(4) represent an advance with respect to the
The chromatic characteristics of fermenting grape pioneering technique proposed by Sudraud to describe
musts are not gradually acquired during the whole fer- wine color, where only absorbances at 420 and 520 nm
mentation, but mainly at its early stages instead, when were required in the calculations. In spite of this advance,
the maceration occurs. Upon completion of the maceration the result of the analysis involving a third wavelength fails
process, the fermentation is still in progress and the fer- to reflect the overall visual perception of a wine’s color
menting grape must undergoes a slight but steady color and, although the application of the CIELab universal color
loss [7]. The spectrum of red wines has a maximum at appreciation system would represent an improvement
520 nm, due to anthocyanins and their flavylium combina- compared to the Glories method, the results are difficult
tions, and a minimum in the region of 420 nm. The current for winemakers to interpret [9].
approach to color analysis in winemaking requires absor-
bance measurements at 420 and 520 nm that take into
account the contributions of red and yellow to overall
color, with an additional measurement at 620 nm to
include the blue component in young red wines [9]. This
approach, known as Glories method, has previously been
applied to finished red wines by other authors [10]. The
chromatic parameters involved in the Glories method are
three: color intensity (CI), hue (H) and brightness (dA
(%)). The color intensity, of special relevance in enology,
is a measure of the amount of color and may vary signifi-
cantly among different grape types. Its evolution as a given
fermentation progresses has been reported previously [1].
The hue is an indicator of color development toward
orange, whereas the brightness is related to the shape of
the spectrum, since a distinct absorbance peak at 520 nm
is typical of bright red wines [9]. There is no target specifi-
cation in terms of either expected values or validity ranges
to apply the Glories method. This lack of strict specifica-
tions is a consequence of the fact that wine cellars follow
their own protocols to produce a certain wine variety. Such
protocols, it should be noted, are rather flexible, as they are
influenced by factors like the sugar content of the Fig. 1. Procedure for determining radiation intensities at 420, 520 and
harvested grapes, which may vary significantly from one 620 nm using distilled water and a test sample from which absorbances
season to the next. can be determined.
130 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136
3. Optoelectronic system design
Both the optoelectronic elements and the test cuvette
are lodged inside a customized frame designed with Solid-
Works [13] and made out of nylon, a material that provides
mechanical stability and avoids misalignments between
the light emitters’ optical axes and their paired detectors.
LEDs have been chosen as light sources despite their limi-
tations in terms of spectral bandwidth, focusing and light
power since the number of wavelengths commercially
available for LEDs is considerably larger than for other
sources like laser diodes.
Fig. 2 depicts the relative positions of the three LED
lamps, their matching detectors and the must sample. A
printed circuit board holding both the three LED drivers
and the conditioning electronics of the six photocurrents
present in the system is connected to the frame. The
instrument sits in a metal cage that remains closed while
measurements are in progress.
Voltage signals were sampled by a data acquisition
board (NI USB-6211 by National Instruments) featuring a
16-bit A/D converter and a sampling rate of 250 kS/s. The
acquired samples were processed and displayed by means
of a LabVIEW [14] virtual instrument running on a laptop
computer.
3.1. LED and photodetection electronics
Absorbance measurements were performed using one
LED and two photodiodes for every wavelength of interest.
One of the two photodiodes (which will be henceforth
referred to as signal photodiode) is aligned with the LED’s
optical axis and captures the attenuated light after passing
through the sample under test; whereas the second one
(the control photodiode) is positioned off the main optical
path and receives a residual fraction of the total emitted
radiation that has not been transmitted through the sam-
ple, as depicted in Fig. 3. The control photodiode, which
belongs to the LED driver, is intended to avoid a significant
dependence of the LED light intensity on both aging and,
mainly, temperature changes, as the luminous intensity
of an LED shows an exponential dependency on tempera-
ture [15]. As can be seen, the temperature compensation
circuit is placed in the feedback loop of the driving stage.
Fig. 2. Two views of the nylon frame and a test sample while the prototype is in operation.
Other LED drivers reported in the literature do not control
the light intensity but the current through the LED instead
[16,17]. Although their operation does not require an addi-
tional control photodiode, they are prone to drifts with
temperature.
Only one TL082 chip is needed for every wavelength, as
the TL082 integrates two Op Amps in the same package.
The third amplifier in Fig. 3 is not a TL082 since the signal
photodiode is integrated, along with an amplifier, in a
single package.
The LED driver operates linearly owing to the negative
feedback in a loop of mixed nature, both electrical and
optical. It comprises an Op Amp followed by a transistor-
based current amplifier that supplies the LED with proper
current levels. The diode Dp protects the emitter-base junc-
tion of the bipolar transistor against a potential breakdown
in case the negative feedback is momentarily interrupted.
The LED light was modulated in order to prevent stray
light sources from contributing to any of the three photo-
currents present in the signal photodetection stage, even
though the influence of such sources, like the ambient light
or the light emitted by the other two nearby LEDs, is neg-
ligible since the holding frame is deliberately opaque. A
previous frame, mechanized from transparent methacry-
late, failed to prevent unwanted light sources from
reaching the corresponding photodetectors. As a result,
modulation was imperative in that case.
The modulating voltage signal, output by the DAQ, is a
square waveform with a superimposed dc voltage adjusted
so that its low-level and high-level voltages are 0 V and
5 V, respectively. The selected modulation frequency was
230 Hz for the three LED drivers. (Using the same fre-
quency for the three drivers is convenient since the use
of hardware resources in the DAQ is optimized.) Resistors
Re were designed to set the rms value of the modulated
current through each LED at 20 mA. This pulsed current
drives the LEDs at reasonable light levels and is still well
below the LEDs’ maximum ratings.
A capacitor is connected in each of the transimpedance
amplifiers’ feedback loops for stability purposes, as it pre-
vents the system from oscillating. Design guidelines can be
found in [18,19].
The anode and the cathode of the control photodiode
are grounded so that the voltage across the device is zero.
 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136
Fig. 3. Electronic schematic: LED driver and photodetection stages for operation at one wavelength.
As can be seen in Fig. 3, its anode is connected to the com-
mon ground node (that is, the signal ground for both the
driving and the detection stages), whereas its cathode is
connected to a virtual ground. Reversing the connections
does not influence the magnitude of the generated photo-
current; however, such a configuration would have an
undesirable implication on the system design: if the
cathode were connected to the signal ground node, the
current-to-voltage converter would act as an inverting
stage, adding an unwanted phase shift. Since its voltage
output is fed back to the driving stage and compared with
the modulating signal, an additional inverting stage would
be necessary in order to cancel the introduced phase shift
prior to comparison with the modulating signal. Conse-
quently, connecting the control photodiode terminals as
in Fig. 3 contributes to simplify the design without com-
promising its performance.
3.2. Demodulation scheme
At an early stage of the development the voltage signal
of interest was recovered from the modulated waveform
using a two-stage analog circuit comprising a multiple
feedback filter (a type of second-order active bandpass fil-
ter) followed by a peak detector. Although this demodula-
tion technique performed satisfactorily, it was eventually
replaced with a synchronous detection scheme based on
a lock-in amplifier implemented in the LabVIEW virtual
instrument This new approach improved the former design
in terms of portability and performance: any spurious
source contained in the voltage signal sampled by the
DAQ is effectively filtered out without attenuating the sig-
nal to be recovered, provided its frequency lies above or
131
below the modulating frequency by only 2 Hz. (Matching
such a high selectivity by means of a conventional analog
filter would be a challenging task.) Resolution figures
achieved with this scheme are in the range of 1.0–2.4 thou-
sandths of absorbance units (AU) as can be seen further on
in Table 4.
4. Optoelectronic devices
4.1. LED lamps
LEDs were carefully chosen to meet the requirements of
the application. A given LED, intended to operate for the
type of absorbance measurements carried out in the cur-
rent work, should ideally observe the following require-
ments: (1) a peak wavelength that matches the
wavelength of interest (420, 520 and 620 nm); (2) a direc-
tional output beam characterized by a narrow viewing
angle 2h1/2, to prevent a significant fraction of radiated
light power from missing the photodetector active surface;
(3) a small spectral halfwidth Dk1/2, to minimize the influ-
ence of wavelengths different than those of interest; (4) a
large radiant power, to account for both the losses in the
test sample and the moderate photodetector responsivity
at certain wavelengths. Especially the latter can be an issue
in practice, as manufacturers do not usually provide radio-
metric magnitudes like the radiant power, but photometric
magnitudes instead, like the luminous intensity. This can
sometimes be misleading, since a high luminous intensity
source does not necessarily radiate an equally high radiant
intensity. Consequently, it is convenient to either measure
the radiant power of LED sources, or to figure it out analyt-
ically from the supplied data, prior to making a decision
132 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136

regarding the suitability of an LED lamp. This work supports the validity of the measurements. Experimental
addresses both alternatives. Table 1 compiles a number data for the three LED types lodged in the prototype were
of selected optical magnitudes of relevance for the three slightly corrected to take into account the minor influence
LED types included in the instrument and extracted from of the ambient light. The measurements, shown in Table 2,
their technical specifications, which are through-hole LED were obtained from a total number of 21 LEDs. Table 2 also
lamps encapsulated in T-13=4 (5-mm diameter) packages. includes the corresponding analytical estimates that were
As can be seen in Table 1, different manufacturers do not calculated using the data gathered in Table 1. Details on
supply the same set of technical specifications: Roithner the analytical procedure are outlined in Appendix A.
Lasertechnik provides the radiant intensity Ie instead of The radiant power values measured for the reddish LED
the luminous intensity Iv given by the other two manufac- are noticeably low (around 1 mW), which could be some-
turers. In order to find Iv from Ie for the mentioned LED, the thing unexpected taking into account that the correspond-
following equation was used: ing luminous intensity Iv is high. Two reasons account for
this: on the one hand, the luminous efficacy gv at 620 nm
Iv ¼ gv ðkÞIe ð5Þ is still rather high (260 lm/W); and on the other hand, this
particular LED type features an extremely narrow radiation
where gv(k) is the luminous efficacy, which is calculated as
beam (2h1/2 = 6°). As a result, the on-axis luminous inten-
the product of the photopic conversion constant K (stan-
sity Iv(0) is higher for this LED than for other LEDs featuring
dardized to be 683 lm/W [17]) and the photopic spectral
a wider radiation pattern but emitting the same optical
luminous efficiency function V(k). Discrete values for V(k)
power. The luminous intensity of LED lamps listed on data
at 5 nm intervals can be found in [20], and at 10 nm inter-
sheets is a single point, on-axis measurement at a specified
vals in [21–23].
test current [24]; this fact reveals that luminous intensity
figures supplied by manufacturers refer actually to Iv(0).
4.1.1. LED radiant power It is apparent from the comparison of the measured
The radiant power was measured in a lab at 25 °C with data with the analytical estimates in Table 2 that the latter
the photodiode sensor PM100D by Thorlabs, featuring a fall short of the target values found experimentally for the
detector operating in the range from 400 to 1100 nm (sili- LEDs included in the design. Owing to the number and nat-
con). In order to perform reliable measurements with the ure of the approximations involved in the analytical proce-
PM100D, it is necessary to enter the operating wavelength dure, the estimates can only be regarded as rough figures
of the light source to be measured. If the light is not mono- that are not expected to accurately match the measure-
chromatic, as it happens to be the case for LED lamps, the ments. Despite this limitation, the formula derived from
measurement accuracy is compromised to a certain extent. the analytical procedure yields power figures that can
Nevertheless, measurements carried out with this sensor prove useful to make an informed decision, based solely
on light sources characterized by a broad spectral response on technical data as supplied by manufacturers, as to
are still accurate, provided their spectral response shape is whether or not a given LED lamp is suitable for an applica-
symmetrical around the peak wavelength. This feature is tion prior to being set under test or even purchased. The
satisfied by the three LED types operating in the system analytical method can also be of practical interest if a pho-
and thus it is expected that measurements closely match todetector sensor is not available.
the actual optical power values. It should be noted that
the only optical power rating provided in the data sheets 4.2. Photodetection stage
(of 15 mW for the bluish LED) is in good agreement with
the mean value found experimentally (14.28 mW), which For a given LED optical output power, the suitability of
an LED is strongly dependent on its peak wavelength, since
Table 1
the response of the photodetectors utilized is not flat.
Selection of manufacturer data for the three LED lamp types included in the Table 3 illustrates this point by showing approximate val-
design. Variables are as follows: kp: peak wavelength; Dk1/2: spectral ues for the ratings of the two photodetectors used at the
halfwidth; h1/2: off-axis angle where the luminous intensity is one half the three wavelengths of interest, extracted from their corre-
on-axis intensity; Iv: luminous intensity; Ie: radiant intensity; Ue: radiant
sponding spectral responsivity curves. In order to simplify
power. Test conditions: 20 mA polarization current, 25 °C room
temperature. the design, the two types of selected photodetectors cover
the whole spectral range of interest. The OPT301 is an inte-
Manufacturer and reference
grated circuit containing a photodiode and a transimped-
Parameter Hewlett–Packard Kingbright Roithner ance amplifier on a single chip previously utilized in the
HLMP-EH08-UX000 L-7113VGC-H LED420-01
context of photometric detection [11], whereas the
Color Reddish-orange Green Bluish-violet
SFH213 is a silicon PIN photodiode mounted inside a 5-
kp,typ (nm) 621 520 420
mm diameter plastic package (that is, the same encapsula-
Dk1/2 (nm) 17 35 15
2h1/2 (°) 6 20 16
tion as the LEDs in the system), and especially suitable for
Iv,max. (mcd) 10,700 – – applications from 400 to 1100 nm.
Iv,typ. (mcd) – 14,000 164a As can be seen in Table 3, the photocurrent conversion
Iv,min. (mcd) 2750 8000 – ratings are fairly poor for shorter wavelengths. This aver-
Ie,typ (mW/sr) – – 60
age performance poses tight design constraints in the blu-
Ue,typ (mW) – – 15
ish-violet region of the spectrum regarding the minimum
a
Derived from the provided radiant intensity Ie using (5). LED optical power required to generate a sufficient current
 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136 133

Table 2
Measurements and analytical estimates of radiant power for the three LED types utilized in the design.

Number of LED lamps Radiant power measurements (±0.2%) Estimates

Ue,max (mW) Ue,min (mW) Ue,avg (mW) Ue (mW)
Reddish (8) 1.29 0.93 1.10 0.50 (max)
Green (9) 7.63 4.64 6.27 3.94 (typ)
Bluish (4) 14.99 13.82 14.28 5.23 (typ)

Table 3
Responsivity values for the two types of photodetectors utilized.

k (nm) Photodetector responsivity (A/W)

Burr Brown OPT301 OSRAM SFH213
420 0.17 0.11
520 0.31 0.27
620 0.42 0.42

in the photodetection stage that, in turn, will be converted

into a voltage signal while keeping both a reasonably wide
voltage range and an acceptable signal-to-noise ratio. Fur-
thermore, red wine absorption is distinctly stronger at 420
and 520 nm than at 620 nm, which prevents a significant
fraction of the emitted LED power from reaching the signal
photodetectors at these two shorter wavelengths.
A careful selection of two LED lamps from two different
manufacturers peaking at around 420 nm (HLMP-DB25-
B0000 by Avago Technologies and L53MBC by Kingbright)
failed to meet the design specifications owing to lack of
optical power and had to be eventually ruled out. The
LED lamp finally chosen to successfully operate at
420 nm features the same 5-mm encapsulation as the rest
of the LED lamps while emitting a remarkably high
radiation (15 mW), that comfortably meets the power
requirements at that wavelength.

5. System validation

A spectrophotometer Jasco V-670 was utilized to test

the performance of the system from eighteen wine sam-
ples diluted with different quantities of water and gath-
ered in three groups (six samples for each wavelength).
The wine was purchased from the winery ‘‘Galán de
Membrilla’’ (La Mancha, Spain). Dilutions were adjusted
to cover a range of absorbances evenly distributed, while
keeping the upper absorbance reading close to unity
regardless of the wavelength. The deviations found are
shown in Fig. 4. It is apparent from the measurements that
the prototype works best at 420 nm, whereas the largest
deviations occur at 620 nm, especially when absorbance
values approach unity. However, even though some of
the deviations found are somewhat large, the system still
performs reasonably well in terms of accuracy without
the need of extra corrections, provided its operation is
restricted to the absorbance ranges observed during a
typical fermentation process, as Fig. 4 illustrates: the
horizontal dashed bars added to the plots set the upper
and the lower absorbance limits extracted from actual
fermentation data obtained with the current prototype in
Fig. 4. Absorbance values of diluted wine samples measured both with a
spectrophotometer and with the customized prototype at (a) 420 nm; (b)
520 nm; (c) 620 nm. Prototype data were corrected using a calibration
factor for 520 and 620 nm. The horizontal bars set absorbance limits
obtained from fermentation data.
a previous work [7]. As can be seen, the deviations are
acceptable within the zones of practical interest, particu-
larly at 620 nm, where absorbance values are remarkably
low compared to the other two wavelengths (as expected
for any variety of red wine). Nevertheless, errors can be
readily minimized in the whole range of absorbances
under consideration by using the spectrophotometer mea-
surements as calibration data. In doing so, single
calibration factors for 520 nm and 620 nm appropriately
134 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136

introduced in the LabVIEW software are required for mea- complete stop. Results from two different fermentations
surements in the green and the reddish spectrum region. carried out on actual grape must samples in a laboratory
Results improved upon correction of the initial data as fermenter are discussed below, showing the evolution with
depicted in Fig. 4. time of color intensity, hue and brightness. The time span
chosen was three days from the beginning of the fermenta-
5.1. Error analysis and figures of merit tion, as the color intensity (the most relevant parameter in
practice for enologists) increased only during approxi-
In order to complete the validation of the system, three mately the first 33 h of the fermentation, to start a steady
relevant figures of merit were considered: resolution, decrease afterward (see Fig. 5).
repeatability and reproducibility. Resolution is calculated
as the voltage noise to sensitivity ratio. Since the voltage
levels delivered by the three demodulation stages show a
non-linear dependence with the absorbance, it follows that
the sensitivity depends on it, being higher for small absor-
bance values. In an attempt to take into account this non-
linear behavior, the absorbance range considered in this
work was split into two subranges (from 0 to 0.5 AU and
from 0.5 to 1 AU). In order to determine the voltage noise,
measurement errors were obtained after processing 20
sets of 1000 consecutive voltage readings each, being each
set sampled at different time intervals. The error value cor-
responding to every set was calculated as three times the
standard deviation from the 1000 readings. The third larg-
est error in this group of 20 error values was taken as the
noise figure necessary to calculate the resolution. In turn,
both repeatability and reproducibility were obtained as
three times the standard deviation of 20 mean values,
being each one of them the result of averaging 100 consec-
utive readings. The difference between the procedures to
determine repeatability and reproducibility figures is that,
in order to get the reproducibility, the test cuvette was
taken out from the frame and put back into place again
between readings (that is, 20 times altogether), whereas
this operation was not carried out in the case of the repeat-
ability. The corresponding figures of merit are summarized
in Table 4, which lie in the same range as those reported in
the technical data of the spectrophotometer utilized. Spe-
cifically, the photometric accuracy is reported to be
±0.003 AU in the range 0.5–1 AU for the Jasco V-670. As
can be seen, the system performs best in the range of small
absorbances, where the sensitivity is higher.

6. Maceration

A useful application of the proposed device consists of

providing winemakers with valuable information as to
when it is the right time to remove the grape skins from Fig. 5. Evolution of color intensity, hue and brightness measured on
the tanks while a given fermentation has not come to a actual grape must samples for three consecutive days.

Table 4
Relevant figures of merit considering two absorbance subranges: from 0 to 0.5 AU and from 0.5 to 1 AU.

Subrange LED color

(AU) Bluish Green Reddish
3
Resolution (10 AU) [0, 0.5] ±1.1 ±1.0 ±1.2
[0.5, 1] ±1.1 ±2.0 ±2.4
Repeatability (103 AU) [0, 0.5] ±0.6 ±0.6 ±0.5
[0.5, 1] ±0.9 ±0.9 ±0.7
Reproducibility (103 AU) [0, 0.5] ±1.0 ±3.3 ±3.4
[0.5, 1] ±1.5 ±8.8 ±6.3
 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136
In fermentation 1 the grape skins were removed after
reaching the maximum, whereas in fermentation 2 they
were kept inside the laboratory fermenter during the
whole fermentation. As can be seen, once the loss of color
intensity begins, it decays at a slightly faster pace in the
case of fermentation 1. The color loss in either case possi-
bly results from the contribution of several factors such as
condensation reactions, deposition of the coloring matter
or the absorption on the yeast walls. Longer maceration
periods are linked with high concentrations of
higher-molecular-weight tannins and it can result in the
extraction of undesirable flavorants [25].
As regards brightness, it shows a trend similar to that
found for the color intensity, peaking at around 33 h or a
few hours later, depending on the fermentation chosen.
The hue exhibits the reverse trend, showing a minimum
instead of a maximum by the time the color intensity
peaks. In any case, results prove that any of the three
parameters would be a suitable candidate to help wine-
makers detect an appropriate time to remove the grape
skins, either by checking a maximum or a minimum.
Although hue and brightness readings are in practice per-
formed (so far) on wines that are ready to be bottled rather
than on fermenting musts, it would be of interest to mon-
itor both parameters during the fermentation too, as the
corresponding readings can provide valuable information
to assess the role played by various pigments involved in
the color of wine.
7. Conclusions
A low-cost, portable and reliable optoelectronic system
has been designed and successfully tested with the pur-
pose of supervising the evolution of the chromatic charac-
teristics of grape musts while an alcoholic fermentation is
in progress. Measurements carried out with the custom-
ized prototype were compared with those obtained using
a spectrophotometer for both calibration and validation
purposes using samples of diluted wine, reaching a resolu-
tion in the range of one thousandth of an AU. The system
operates on discrete must samples so far, and it is foresee-
able that it will be gradually advanced into a fully auto-
mated version featuring real-time supervision capabilities
without major modifications of the current optoelectronic
design. To that end a customized hydraulic circuit, devised
to allow a continuous flow of fermenting grape must into
the optoelectronic instrument, will be designed and tested.
Although the system has been devised with the only pur-
pose of contributing to the automation of the winemaking
industry, the outlined aspects of the design can be of prac-
tical interest for applications other than winemaking that
also rely on absorbance measurements at wavelengths dif-
ferent than those reported here. Such applications have
found a widespread use in analytical chemistry and related
areas over the past years, using LED-based instruments
that, altogether, cover a broad spectral range, from ultravi-
olet to infrared wavelengths. To mention some recent
examples, they have been applied to the detection of pollu-
tants, pesticides and a variety of chemicals, either in water
or in air, [26–30], and to perform turbidity measurements
of water samples as well [31].
135
Acknowledgements
This work was supported by MINECO (Spain) under
grant number DPI2012-31203. The authors thank Pedro
Jiménez de los Galanes for the mechanization of the nylon
frame used in this work.
Appendix A. Analytical estimate of the LED radiant
power
The following analysis yields a simple formula from
which a rough analytical estimate of the radiant power
can be easily obtained using only two parameters usually
supplied by LED manufacturers: the on-axis luminous
intensity Iv(0), and the viewing half angle h1/2.
The luminous intensity Iv is the luminous flux per unit
solid angle along a particular direction:
dUv
Iv ¼ ðA:1Þ
dX
Iv is measured by manufacturers on the mechanical axis of
the LED lamp package, which is closely aligned with the
optical axis [24]. The actual radiation pattern can be
approximated by an exponential intensity source Iv(h), that
falls very rapidly with the off-axis angle h (being the axis
taken along the maximum luminous intensity), and is
given by [17,32]:
Iv ðhÞ ¼ Iv ð0Þ cosn h ðA:2Þ
At the viewing half angle, Iv(h1/2) = Iv(0)/2, and the
directivity parameter n can be estimated using the
following expression [17]:
lnð0:5Þ
n¼ ðA:3Þ
ln½cosðh1=2 Þ
Observe that, if n = 1, an exponential intensity source
becomes a Lambertian source, whose intensity distribution
is typical of planar LED structures [33]. In the present work,
where all LED lamps have a narrow viewing angle (and, as
a result, a high on-axis luminous intensity Iv(0)), n is much
greater than unity and has to be calculated for every LED.
The fraction of the luminous flux emitted by an exponen-
tial intensity source which is confined in a cone whose full
vertex angle is 2h is given by
Z Z 2p Z h
Uv ðhÞ ¼ Iv ðhÞdX ¼ Iv ð0Þ cosn h sin hdhd/
0 0
Iv ð0Þ2p
¼ ð1  cosnþ1 hÞ ðA:4Þ
nþ1
As the spatial radiation pattern of all LED lamps utilized
is quite narrow, most of the emitted energy is confined in a
cone with a small apex angle. However, it is preferable to
integrate over a whole hemisphere (by letting h = p/2) in
order to consider all the emitted radiation. Eq. (A.4) results
then:
I ð0Þ2p
Uv ðh ¼ p=2Þ ¼ v ðA:5Þ
nþ1
Using (A.5), the luminous flux Uv can be roughly esti-
mated using only the on-axis luminous intensity and the
viewing half angle from which n can be derived. Once Uv
is known, the radiant power emitted, Ue, can be readily
derived using:
136 F. Jiménez-Márquez et al. / Measurement 63 (2015) 128–136
Uv ¼ gv ðkÞUe ðA:6Þ
A major source of error involved in the outlined analysis
consists of using technical data related to discrete wave-
length values to derive the emitted power from the lumi-
nous flux. It is actually a continuous spectrum, rather
than a single wavelength, what contributes to the provided
luminous intensity ratings from which the luminous flux is
obtained. Besides, most manufacturers provide h1/2 to
characterize the angular distribution of LED lamps with a
single intensity peak. However, it has been suggested pre-
viously by other authors that manufacturers should pro-
vide in their technical specifications not only h1/2, but a
formula with coefficients corresponding to the shape of
the radiation pattern [34], what would effectively allow
the final users to have at their disposal a comprehensive
characterization of the LEDs.
References
[1] P. Ribéreau-Gayon, D. Dubourdieu, B. Donèche, A. Lonvaud,
Handbook of Enology Volume 1: The Microbiology of Wine and
Vinifications, John Wiley & Sons, Chichester, 2006.
[2] C. de la Torre, R. Muñiz, M.A. Pérez, A new, low-cost, on-line RGB
colorimeter for wine industry based on optical fibers, in: Proc. of the
IMEKO XIX World Congress – Fundamental and Applied Metrology,
2009, pp. 2554–58.
[3] J.S. Bajić, D.Z. Stupar, B.M. Dakić, M.B. Živanov, L.F. Nagy, An absolute
rotary position sensor based on cylindrical coordinate color space
transformation, Sens. Actuators, A 213 (2014) 27–34.
[4] J. Mutanen, J. Räty, E. Gornov, P. Lehtonen, K.E. Peiponen, T.
Jaaskelainen, Measurement of color, refractive index, and turbidity
of red wines, Am. J. Enol. Viticult. 58 (3) (2007) 387–392.
[5] I. Niskanen, J. Mutanen, P. Lehtonen, J. Räty, M. Yamaguchi, K.E.
Peiponen, T. Jaaskelainen, Multifunction spectrometer for optical
inspection of red wine, Sens. Instrum. Food Qual. Saf. 2 (2008) 58–65.
[6] P.K. Dasgupta, I.-Y. Eom, K.J. Morris, J. Li, Light emitting diode-based
detectors: absorbance, fluorescence and spectroelectrochemical
measurements in a planar flow-through cell, Anal. Chim. Acta 500
(2003) 337–364.
[7] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, High-
resolution low-cost optoelectronic instrument for supervising grape
must fermentation, Microsyst. Technol. 20 (2014) 769–782.
[8] F. Jiménez-Márquez, J. Vázquez, J. Úbeda, J.L. Sánchez-Rojas, Low-
cost and portable refractive optoelectronic device for measuring
wine fermentation kinetics, Sens. Actuators, B 178 (2013) 316–323.
[9] P. Ribéreau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu, Handbook
of Enology Volume 2: The Chemistry of Wine Stabilization and
Treatments, John Wiley & Sons, Chichester, 2006.
[10] I. Noiseux, W. Long, A. Cournoyer, M. Vernon, Simple fiber-optic-
based sensors for process monitoring: an application in wine quality
control monitoring, Soc. Appl. Spectrosc. 58 (8) (2004) 1010–1019.
[11] T.R. Dias, M.A.S. Brasil, M.A. Feres, B.F. Reis, A flow cell with a new
design to improve the utilization of the radiation emitted by LED and
employed as a radiation source for photometric detection, Sens.
Actuators, B 198 (2014) 448–454.
[12] J.D. Ingle Jr., S.R. Crouch, Spectrochemical Analysis, Prentice-Hall,
Upper Saddle River (New Jersey), 1988.
[13] <www.solidworks.com> (accessed 09.06.14).
[14] <www.ni.com/labview> (accessed 09.06.14).
[15] Temperature Compensation Circuit for Constant LED Intensity,
Application Brief I-012, Ref. 5963-7544EN, Avago Technologies,
2010.
[16] I. Al-Bahadly, R. Berndt, LED-based color sensing system, Sens.
Transducers J. 114 (3) (2010) 132–150.
[17] S.O. Kasap, Optoelectronics and Photonics. Principles and Practices,
second ed., Pearson Education, Harlow, 2013.
[18] J. Graeme, Photodiode Amplifiers. Op Amp Solutions, McGraw Hill,
New York, 1996.
[19] M. Johnson, Photodetection and Measurement – Maximizing
Performance in Optical Systems, McGraw Hill, New York, 2003.
[20] E.F. Schubert, Light-Emitting Diodes, second ed., Cambridge
University Press, New York, 2006.
[21] A. Ryer, Light Measurement Handbook, International Light,
Newburyport, 1998.
[22] N. Ohta, A.R. Robertson, Colorimetry. Fundamentals and
Applications, John Wiley and Sons, Chichester, 2005.
[23] D.R. Lide (Ed.), Sensitivity of the human eye to light of different
wavelengths, CRC Handbook of Chemistry and Physics, Internet
Version http://www.hbcpnetbase.com, CRC Press, Boca Raton, 2005,
pp. 10–217.
[24] A guide to human visual perception and the optical characteristics of
LED displays, Application Brief D-004, Ref. 5963-7073EN, Avago
Technologies, 2010.
[25] R.S. Jackson, Wine Science. Principles and Applications, Academic
Press, London, 2008.
[26] Rohit, L. Kanwar, K. Kesava Rao, Development of a low-cost portable
colorimeter for the estimation of fluoride in drinking water, Sens.
Actuators, B 149 (2010) 245–251.
[27] A.J. Palma, J.M. Ortigosa, A. Lapresta-Fernández, M.D. Fernández-
Ramos, M.A. Carvajal, L.F. Capitán-Vallvey, Portable light-emitting
diode-based photometer with one-shot optochemical sensors for
measurement in the field, Rev. Sci. Instrum. 79 (2008) 103105–
103108.
[28] M.E. Moragues, R. Montes-Robles, J.V. Ros-Lis, M. Alcañiz, J. Ibañez, T.
Pardo, R. Martínez-Máñez, An optoelectronic sensing device for CO
detection in air based on a binuclear rhodium complex, Sens.
Actuators, B 191 (2014) 257–263.
[29] D. Bueno, G. Alonso, R. Muñoz, J.L. Marty, Low-cost and portable
absorbance measuring system to carbamate and organophosphate
pesticides, Sens. Actuators, B 203 (2014) 81–88.
[30] M. Tedetti, P. Joffre, M. Goutx, Development of a field-portable
fluorometer based on phenanthrene- and tryptophan-like
compounds in natural waters, Sens. Actuators, B 182 (2013) 416–
423.
[31] H. Tai, D. Li, C. Wang, Q. Ding, C. Wang, S. Liu, Design and
characterization of a smart turbidity transducer for distributed
measurement system, Sens. Actuators, A 175 (2012) 1–8.
[32] P. Manninen, J. Hovila, P. Kärhä, E. Ikonen, Method for analysing
luminous intensity of light-emitting diodes, Meas. Sci. Technol. 18
(2007) 223–229.
[33] J.M. Senior, Optical Fiber Communications. Principles and Practice,
second ed., Prentice Hall, Hemel Hempstead, 1992.
[34] I. Moreno, C.-C. Sun, Modeling the radiation pattern of LEDs, Opt.
Express 16 (3) (2008) 1808–1819.