Volume 7 Issue 2 pp. 246-257 • doi: 10.15627/jd.2020.21

Rizki A. Mangkuto,^{∗,a} Atthaillah^{b,c}

**Author affiliations**

^{a} Building Physics Research Group, Faculty of Industrial Technology, Institut Teknologi Bandung, Jl. Ganesha 10, Labtek VI, Bandung 40132, Indonesia
^{b} Engineering Physics Doctorate Program, Faculty of Industrial Technology Institut Teknologi Bandung, Jl. Ganesha 10, Labtek VI, Bandung 40132, Indonesia
^{c} Architecture Program, Faculty of Engineering, Universitas Malikussaleh, Jl. Cot Teungku Nie, Aceh Utara 24355, Indonesia

* Corresponding author.

armanto@tf.itb.ac.id (R. A. Mangkuto)

atthaillah@unimal.ac.id (Atthaillah)

**History:** Received 2 October 2020 | Revised 4 November 2020 | Accepted 16 November 2020 | Published online 30 November 2020

**Copyright:** © 2020 The Author(s). Published by solarlits.com. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

**Citation:** Rizki A. Mangkuto, Atthaillah, Phasor Method to Estimate Illuminances Due to Parallel Arrays of Light Sources, Journal of Daylighting 7 (2020) 246-257. https://dx.doi.org/10.15627/jd.2020.21

**Figures and tables**

Direct horizontal illuminance along a calculation row due to two parallel arrays of large numbers of identical light sources behaves like a periodic signal with a sinusoidal pattern, which contains useful information for design purpose. This study aims to describe, verify, and discuss the theoretical concept on the superposition of direct horizontal illuminance from both arrays in such configurations, and how to extract the information using the phasor method. Four different approaches are proposed to estimate the total direct horizontal illuminance *E _{T}(x)* and to verify the concept. Sensitivity analysis is also conducted to observe the influence of each input variable to the resulting

Phasor method, Horizontal illuminance, Sinusoidal pattern, Parallel array

*BZ British zonal*
*E _{P} Horizontal illuminance on a point [lx]*

Δ

Calculation of illuminance on the workplane or reference plane is one of the most important tasks to perform in lighting engineering and design. Starting from the prominent work of Lambert in the 18^{th} century [1], most of the procedures for illuminance calculation have been well-established in the last century [2,3] and thus have become standard knowledge for lighting engineers and designers.

The advance of computational lighting modelling and simulation in the past decade [4-6] also greatly reduces the need for manually performing routine illuminance calculation, and so contributes in providing better and more accurate design recommendations. Nevertheless, it is important for any lighting engineers and designers to understand the underlying principles used in the calculation processes, to minimise the risk of having unreliable results [7]. The knowledge of the underlying principles shall enable engineers or designers to write their own (computer) programs or algorithms whenever necessary; and would be a valuable material for education purpose [7-9].

Despite seeming to be well-understood, one of the topics that tends to be overlooked is the the horizontal illuminance (*E*_{P}) calculation due to parallel arrays of identical light sources mounted at an equidistant interval, forming a periodic spatial pattern. An example would be the configuration of pendant luminaires in a library (Fig. 1), which consists of two parallel rows (or arrays) of luminaires. Since the mounting height and the distance between adjacent luminaires at the same array are typically constant, the direct *E*_{P} values at a given calculation row parallel to the arrays also follows a periodic spatial pattern. For typical interior luminaires, the periodic pattern is expected to be a sinusoidal one, provided the number of luminaires in each array is large enough.

Figure 1

**Fig. 1.** Illustration of two parallel arrays of luminaires (highlighted in yellow) with a periodic pattern. (adapted from public domain image: Main Reading Room of the New York City Public Library on 5th Avenue, ca. 1910–1920).

The direct *E*_{P} along the calculation row due to a single array of luminaires therefore behaves like a periodic signal, stationary in time (assuming the sources are constantly turned on) but varying in one-dimensional space of interest. Combining both arrays of luminaires together shall yield another sinusoidal pattern, which is basically the superposition of two sinusoidal signals. The pattern contains several important information, such as the values and positions of maximum and minimum illuminances along the calculation row, which can be useful for design purpose.

The example can be also extended, for instance, in the lighting arrangement above the audience seats in theatres. To a certain extent, one can also think on arrays of windows or clerestories on both long sides of a corridor under the CIE overcast sky. All those scenarios shall reveal a sinusoidal pattern of *E*_{P} along a calculation row parallel to the sources. When the calculation row is located for instance on a diffuse floor, the resulting illuminance pattern may also be correlated with luminance difference and spatial contrast, which is considered an integral part of the architectural features of the space [10,11]. Spatial contrast is found to be an important factor in indoor (e.g. [11-15]) and outdoor spaces (e.g. [16-19]). In a typical indoor office setting, for instance, it is known that luminance diversity and non-uniform lighting distribution may be positively correlated with and the occupants’ impression of excitement and preference [20,21], although a space with too excessive luminous variability may be perceived as uncomfortable [22].

The concept of spatial brightness and perceived adequacy of illumination (PAI) in lighting design has also been promoted in the past decade [12,23,24]. The existence of spatial patterns is also typically indicated with luminance contrast, in the case of indoor spaces [11,14]; and mean, minimum, and maximum illuminances in the case of outdoor spaces (or any space with relatively low reflections from the environment), particularly streets or roads [17-19].

The information contained in the sinusoidal pattern is however seldom discussed in literatures and standards. This study therefore aims to describe and discuss the theoretical concept on the superposition of direct *E*_{P} from both sources’ arrays in such configurations, and to verify it with some worked examples. The article is structured as follows: Section 2 provides the theoretical concept of the relevant calculations for obtaining the direct *E*_{P }due to two parallel, periodic arrays of identical sources. Section 3 describes the methods applied in verifying the proposed concept, followed with sensitivity analysis on the input variables. Sections 4 and 5 provides the results and discussion, whereas Section 6 concludes the article.

Consider an array comprising a large number of identical light sources, i.e. those with identical luminous flux and luminous intensity distribution (LID), mounted at a constant height *z* above the reference plane. The LID of the sources is assumed to be cosine-like, i.e. closely follows that of BZ1 or up to BZ5 type in the British Zonal (BZ) classification system [25,28]. Adjacent sources are separated by a constant distance *λ*, as illustrated in Figs. 2(a), 3(a), and 4(a). Suppose there exists a row of calculation points, parallel with the sources array at a lateral distance *y*.

Figure 2

**Fig. 2.** Plan view of (**a**) a periodic array and (**b**) two parallel, periodic arrays of identical light sources with a constant distance *λ*. The black dots in the middle represent the row of calculation points for horizontal illuminance.

Figure 3

**Fig. 3.** Elevation view of (**a**) a periodic array and (**b**) two parallel, periodic arrays of identical light sources with a constant interval *λ*, as per Fig. 2.

Figure 4

**Fig. 4.** Perspective view of (**a**) a periodic array and (**b**) two parallel, periodic arrays of identical light sources with a constant interval *λ*, as per Fig. 2.

The direct *E*_{P} values on the calculation points are expected to vary according to a sinusoidal pattern, as illustrated in Fig. 5. One can thus observe the maximum, median, and minimum* E*_{P} values (*E*_{max}, *E*_{med},* E*_{min}) along the calculation row. Since the LID of the sources is assumed to follow BZ1 or up to BZ5 type, the source’s luminous intensity at any zenith angle *γ* reads:

\[
I_\gamma=\left\{\begin{matrix}I_0\cos^n{\gamma},\mathrm{\ \ }\ 0\le\gamma\le\pi\mathrm{/2}\ \\0,\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \pi\mathrm{/2}<\gamma\le\pi\\\end{matrix}\right.
\]

where *I*_{0} is the maximum luminous intensity at *γ* = 0; while *n* = 4, 3, 2, 1.5, and 1 respectively for BZ1, BZ2, BZ3, BZ4, and BZ5. Therefore, *E*_{max} will be achieved at position that is the nearest to an individual source (e.g. directly beneath it), while *E*_{min} will be achieved at position that is the farthest to any individual source (e.g. midway between two adjacent sources).

From Fig. 5, it is clear that the median (also the mean) of the *E*_{P} values is:

\[
E_{med}=\frac{E_{\mathrm{max}}+E_{min}}{2},
\]

whereas one can define *E*_{o} as the difference between *E*_{max} and *E*_{med}, which is equal to the difference between *E*_{med} and *E*_{min}, which also reads:

\[
E_o=\frac{E_{\mathrm{max}}-E_{min}}{2}
\]

Figure 5

**Fig. 5.** Sinusoidal pattern of direct horizontal illuminance due to a periodic array of identical sources.

The direct* E*_{P} values at any position* x* along the calculation row (*E*(*x*)) can therefore be described in the standard cosine form in Eq. (4) as follows:

\[
E(x)=E_o\cos{(}kx+\phi)+E_{\mathrm{med}}
\]

where *k *is the wavenumber, equal to 2*π*/*λ*; and *ϕ* is the phase difference relative to the origin as shown in Fig. 5. It is common to say that according to the illustration of Fig. 5, *ϕ* is negative (i.e. the signal is lagging behind). The distance *x*_{max} (measured from *x* = 0) at which *E*_{max} occurs is thus:

\[
x_{max}=-\frac{\phi}{2\pi}\lambda=-\frac{\phi}{k}
\]

Consider now another array comprising a large number of identical sources mounted at a constant height above the reference plane, also with a constant interval *λ*, but with a spatial phase difference *δ*. Suppose this array is located parallel with the first one, at a lateral distance *Y*, as shown in Figs. 2(b), 3(b), and 4(b). The mounting height of the sources in the second array is not necessarily the same with those in the first array (*z*), but it is practical and realistic to assume they are the same. If the contribution from each array is now measured individually, one can thus write:

\[
{E_1(x)=E_{\mathrm{o,}1}\cos{(}kx+\phi_1)+E_{\mathrm{med,1}}\mathrm{\ ,}\ \atop E_2(x)=E_{\mathrm{o,}2}\cos{(}kx+\phi_2)+E_{\mathrm{med,2}}\mathrm{\ ,}\ }
\]

where *E*_{1}(*x*) and* E*_{2}(*x*) are respectively the direct horizontal illuminances at any position* x* along the calculation row due to the first and the second array. Notice that both arrays have the same wavenumber *k*, since the interval between adjacent sources in both arrays is constant.

Suppose that the contribution from both arrays are to be measured altogether. The resulting direct horizontal illuminance *E _{T}*(

In this case, the cosine terms in *E*_{1}(*x*) and* E*_{2}(*x*) shall be described as the real (Re) parts of the complex notations *ξ*_{1}(*x*) and* ξ*_{2}(*x*), which read:

\[
\xi_1(x)=E_{\mathrm{o,}1}[\cos{(}kx+\phi_1)+j\sin{(}kx+\phi_1)]=Eo,1ej(kx+ϕ1)=Eo,1ejkxeϕ1 ,
\atop \xi_2(x)=E_{\mathrm{o,}2}[\cos{(}kx+\phi_2)+j\sin{(}kx+\phi_2)]=Eo,2ej(kx+ϕ2)=Eo,2ejkxeϕ2 ,
\]

where *j* = √–1.

Since *k* is constant for both signals, the sum of both complex notations, *ξ _{T}*(

\[
{\xi_T(x)=\xi_1(x)+\xi_2(x)=e^{jkx}(E_{\mathrm{o,}1}e^{j\phi_1}+E_{\mathrm{o,}2}e^{j\phi_2})\atop=e^{jkx}(E_{\mathrm{o,}T}e^{j\phi_T})\ =\ E_{\mathrm{o,}T}e^{j(kx+\phi_T)}\ .}
\]

The variables *E*_{o,T} and *ϕ _{T}* can be determined using the phasor diagram as illustrated in Fig. 6. One can thus write:

\[
E_{o,T}=\sqrt{(E_{o,1}\cos{\phi_1}+E_{o,2}\cos{\phi_2})^2+(E_{o,1}\sin{\phi_1}+E_{o,2}\sin{\phi_2})^2}
\]

\[
\phi_T=\arctan{\frac{E_{o,1}\sin{\phi_1}+E_{o,2}\sin{\phi_2}}{E_{o,1}\cos{\phi_1}+E_{o,2}\cos{\phi_2}}}
\]

\[
E_T(x)=Re{{}\xi_T(x)}+E_{\mathrm{med,1}}+E_{\mathrm{med,2}}=E_{\mathrm{o,}T}\cos{(}kx+\phi_T)+E_{\mathrm{med,1}}+E_{\mathrm{med,2}}.
\]

Figure 6

**Fig. 6.** Phasor diagram representing superposition of direct horizontal illuminance due to two parallel, periodic arrays of identical sources.

From Equation (11), the maximum, median, and minimum values of *E _{T}*(

\[
E_{max,{T}}=E_{\mathrm{o,}T}+E_{\mathrm{med,1}}+E_{\mathrm{med,2}},
\]

\[
E_{\mathrm{med},T}=E_{\mathrm{med,1}}+E_{\mathrm{med,2}},
\]

\[
E_{min,{T}}=-E_{\mathrm{o,}T}+E_{\mathrm{med,1}}+E_{\mathrm{med,2}},
\]

Having presented the concept, given a configuration of two parallel arrays of identical sources as previously described, one can estimate the* E _{T}*(

Figure 7

**Fig. 7.** Illustration of discrete *E*_{1}(*x*), *E*_{2}(*x*), and *E _{T}*(

**Approach 1**(phasor addition from regression model of discrete individual illuminance data):- Perform curve fitting using sinusoidal regression on the discrete
*E*_{1}(*x*) and*E*_{2}(*x*) data. For both datasets, define the wavenumber*k*= 2*π*/*λ*, where*λ*is constant for both arrays of sources. Ensure that the coefficient of determination (*R*^{2}) for both sinusoidal models are very close to 1. - Observe
*E*_{o,1},*ϕ*_{1},*E*_{med,1},*E*_{o,2},*ϕ*_{2}, and*E*_{med,2}values from the obtained models. - Add both models of
*E*_{1}(*x*) and*E*_{2}(*x*) using the phasor method as per Equations (9) until (11), to obtain*E*_{o,T},*ϕ*, and_{T}*E*_{med,T}. **Approach 2**(phasor addition from observation of discrete individual illuminance data):- Observe
*E*_{min},*E*_{max}, and*x*_{max}from the discrete*E*_{1}(*x*) and*E*_{2}(*x*) data. - Calculate
*E*_{o}and*E*_{med}based on*E*_{min}and*E*_{max}for each dataset using Equation (3). - Calculate
*k*= 2*π*/*λ*. - Calculate
*ϕ*= –*kx*_{max}for each dataset using Equation (5). - Add both models of
*E*_{1}(*x*) and*E*_{2}(*x*) using the phasor method as per Equations (9) until (11), to obtain*E*_{o,T},*ϕ*, and_{T}*E*_{med,T}.

Results from both approaches shall be verifiable by directly observing the addition of *E*_{1}(*x*) and *E*_{2}(*x*) for each discrete point, thus giving discrete *E _{T}*(

**Approach 3**(regression model from observation of discrete total illuminance data):- Perform curve fitting using sinusoidal regression on the discrete
*E*(_{T}*x*). Also define the wavenumber*k*= 2*π*/*λ*and ensure that*R*^{2}à - Observe
*E*_{o,T},*ϕ*, and_{T}*E*_{med,T}values from the obtained models. **Approach 4**(observation of discrete total illuminance data):- Observe
*E*_{min,T},*E*_{max,T}, and*x*_{max,T}from the discrete*E*(_{T}*x*) data. - Calculate
*E*_{o,T}and*E*_{med,T}based on*E*_{min,T}and*E*_{max,T}for each dataset using Equation (3). - Calculate
*k*= 2*π*/*λ*. - Calculate
*ϕ*= –_{T}*kx*_{max,T}for each dataset using Equation (5).

Based on the obtained coefficients in Approach 1 until 3, the *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} can be estimated. The corresponding values obtained in all approaches are expected to be close to each other, provided the sinusoidal pattern is achieved. It should be noted that only a hypothetical configuration is described here, which are relevant to explain the theoretical concept of phasor addition in direct horizontal illuminance calculation. The function of the space is irrelevant in this part, because the concept is meant to be general, and thus is not affected by the absolute values of the horizontal illuminance. Material reflectance is assumed zero for all surfaces, because only direct horizontal illuminance is considered throughout the study. Dimensions are provided conceptually (with symbols) in Figs. 2-4, again because this section only provides the general concept, which is not affected by the absolute values of space dimensions. To further illustrate and test the concept, several examples are given in Section 3.

To give an illustration of the proposed concept in Section 2, consider the configuration of two parallel arrays of identical sources in Fig. 8. The distance between arrays (*Y*), the distance between adjacent sources (*λ*), and the mounting height (*z*) are set constant as 3 m. Each array consists of 14 identical downlight sources, each of having LID according to either BZ1 or BZ5 type, with the maximum luminous intensity of 1800 cd (i.e. yielding *E* = 200 lx at a point 3 m directly beneath the source’s centre). The observed space in Fig. 8 can be thought as a large reading hall in libraries (as in Fig. 1), provided a nighttime situation and a relatively low or negligible mean room surface reflectance. The specified maximum luminous intensity and the corresponding direct horizontal illuminance can thus be attributed to the chosen function.

There are 60 calculation points with an interval of 0.1 m along the calculation row, located at *x* = –1.5 until 4.5 m, according to the plan view in Fig. 8. The calculation row is at lateral distance *y* = *y*_{P}; where *y*_{P} = 0, 0.5, 1.0, or 1.5 m. The spatial phase difference *δ *between both arrays is defined as 0, 0.5, 1.0, or 1.5 m, all shifted towards the right-hand side of the first array. Table 1 lists all variables considered (LID,* y*_{P}, and *δ*) in the test. All combinations are considered, except the ones with *y*_{P} = *δ* = 1.5 m, because *E*_{1}(*x*) and *E*_{2}(*x*) will cancel each other out with respect to *E*_{med}, so that the resulting *E _{T}*(

While the length of the calculation row is only 6 m, it is necessary to introduce as many as 14 sources in each array to ensure the sinusoidal pattern is conserved in the calculation row. If there are too few sources in each array, the resulting *E*_{2}(*x*) due to shifted sources in the second array will be asymmetrical, because the calculation points on the left-hand side will see less contribution from the sources compared to the right-hand side.

The direct *E*_{P} at the *i*-th calculation point (*E*_{P,i}) due to an individual source *S _{i} *can be simply determined using the formula of point light source as follows:

\[
E_{P,i}=\frac{I_\gamma}{r_i^2}\cos{\alpha}=\frac{I_\gamma}{r_i^2}\frac{z}{r_i}=\frac{I_\gamma}{r_i^3}\cdot\ z.
\]

where *α* is the angle between the line of sight and the normal of the calculation point; *r _{i}* is the distance between

\[
E_{P,i}=\frac{I_0}{r_i^3}\left(\cos{\gamma}\right)^n\cdot\ z=\frac{I_0}{r_i^3}\left(\frac{z}{r_i}\right)^n\cdot\ z=\frac{I_0z^{n+1}}{r_i^{n+3}}.
\]

The total direct *E*_{P} at the *i*-th calculation point (*E*(*x*)) due to *N* number of sources is thus simply the sum of *E*_{P,i} from each source.

\[
E(x)=\sum_{i=1}^{N}E_{P,i}.
\]

After performing routine calculations to obtain *E*(*x*) at the entire 60 points along the calculation row, the *E*_{1}(*x*) and *E*_{2}(*x*) values are determined. In turn, the four approaches described in the end of Section 2 are applied to estimate the sinusoidal model of *E _{T}*(

To observe the influence of each input variable to the resulting pattern of *E _{T}*(

In turn, the input (*n _{i}*,

\[
n\prime_i=\frac{n_i-\mu_n}{\sigma_n}\mathrm{\ \ ;\ \ \ }y_P\prime_i=\frac{y_{Pi}-\mu_{y_P}}{\sigma_{y_P}}\mathrm{\ \ ;\ \ }\delta\prime_i=\frac{\delta_i-\mu_\delta}{\sigma_\delta}\mathrm{\ \ ;\ \ }i=1,2,...,30
\]

\[
{\mathrm{\Delta E}\prime_i=\frac{\mathrm{\Delta}E_i-\mu_{\mathrm{\Delta E}}}{\sigma_{\mathrm{\Delta E}}}\mathrm{\ \ ;\ \ }\ x{\prime_{max,{T},}}_i=\frac{x_{max,{T},i}-\mu_{x_{max,{T}},i}}{\sigma_{x_{max,{T}},i}}\mathrm{\ ;}\ \atop i=1,2,...,30,}
\]

The regression analysis for is performed according to the following linear models in Equations (20) and (21):

\[
\mathrm{\Delta\ E}\prime=\beta_1\prime\ n\prime+\beta_2\prime\ y_P\prime+\beta_3\prime\delta\prime+\varepsilon\prime
\]

\[
x\prime_{max,{T}}=\beta_1\prime\ n\prime+\beta_2\prime\ y_P\prime+\beta_3\prime\delta\prime+\varepsilon\prime
\]

where *β *’ is the standardised regression coefficient (SRC), and *ε *’ is the intercept or residual error. The SRC values represent the influence of each input variable on the output, where SRC = 1 suggests a highly positive influence, SRC = –1 suggests a highly negative influence, and SRC = 0 suggests no influence at all.

The resulting values of *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} obtained using the four approaches are summarised in Table 2 (for BZ1 LID type) and Table 3 (for BZ5 LID type). Meanwhile, Tables 4 and 5 reports the absolute difference between the largest and smallest obtained values among the four approaches, for both LID types, and for each combination of *y*_{P} and *δ*. It is observed that the differences between values obtained from all approaches are very small, i.e. ≤ 0.5 lx for the illuminances and ≤ 0.04 m for *x*_{max,T}. Therefore, the proposed concept is verified and any of the four approaches can be applied for the considered scenarios.

Table 2

**Table 2.** Obtained values of *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} using the four approaches in the configuration with BZ1 LID type.

Table 3

**Table 3.** Obtained values of *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} using the four approaches in the configuration with the BZ5 LID type.

Table 4

**Table 4.** Absolute differences between the largest and smallest values of *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} obtained with the four approaches in the configuration with BZ1 LID type.

Table 5

**Table 5.** Absolute differences between the largest and smallest values of *E*_{min,T}, *E*_{max,T}, *E*_{med,T}, and *x*_{max,T} obtained with the four approaches in the configuration with BZ5 LID type.

The greatest difference between illuminances obtained using the four approaches is 0.5 lx, achieved for *E*_{min,T} in the configuration with *y*_{P} = 1 m,* δ *= 1.5 m, BZ1 type. Meanwhile, the greatest difference between *x*_{max,T} values from all approaches is 0.04 m, i.e. less than half of the interval between calculation points, achieved in the configuration with (*y*_{P}, *δ*) = (0.5 m, 0.5 m); (0.5 m, 1.0 m); and (1 m, 1 m); all with BZ5 type. The greatest differences between *E*_{max,T} values and between *E*_{med,T} values from all approaches are found to be only 0.2 lx, hence practically no difference at all.

To give further illustrations, Fig. 9 displays the plots of *E _{T}*(

Figure 9

**Fig. 9.** Total direct horizontal illuminance *E _{T}*(

Table 6 displays the SRC (*β *’) of each input variable in Equation (20), based on the sensitivity analysis results. The adjusted coefficient of determination (*R*^{2}_{adj}) of the linear model of Δ*E* ’ is 0.921, and thus is sufficiently high to ensure reliable results.

The largest SRC in the Δ*E* ’ sensitivity model is 0.899, obtained for *n’*. In other words, the choice of LID type of the source can significantly affect Δ*E*. When *n* = 1 (BZ5 type), Δ*E* is relatively small so that the fluctuation between *E*_{max,T} and *E*_{min,T} may be unnoticeable, because the sources’ LID is wide (120° beam angle) and thus create smaller contrast within points along the calculation row. Conversely, when *n* = 5 (BZ1 type, 66° beam angle), the fluctuation may be noticeable, since the narrow LID corresponds to smaller beam angle, yielding a greater contrast along the row. For any *n* values in between, the resulting performance can be estimated accordingly due to the linear behaviour of the variables.

Meanwhile, the lateral position of the calculation row (*y*_{P}) and spatial phase difference (*δ*) are relatively less influential on the fluctuation (SRC = –0.341 and –0.131 respectively). The impacts from both variables are negative, meaning that the larger the *y*_{P} and *δ* values, the smaller the spatial contrast. It should be noticed however that the variation of *y*_{P} and *δ* is periodic, so that the predicted effects only apply within the range between 0 and *λ*/2. Beyond that range, the effect would be symmetrical and not be continuously increasing (or decreasing).

Figure 10

**Fig. 10.** Average *x*_{max,T} in all configurations against *y*_{P}; error bars denote the standard deviations.

While the Δ*E* ’ sensitivity model is highly linear, the *x* ’_{max,T} model is not (*R*^{2}_{adj} = 0.321); so that the results cannot be held reliable. It is noticed however that the largest SRC in that model is of *y*_{P}’ (0.619). Thus, an additional analysis is performed by plotting the average *x*_{max,T} in all configurations against the *y*_{P} values. Figure 10 displays the resulting plot, suggesting a logarithmic relation (*R*^{2} = 0.998) between both variables. The standard deviation from the average values, however, can be considerably large, particularly for *y*_{P} = 1.5 m. The large deviations can be attributed to the significant difference of both LID types, which can be observed from pattern comparison as follows.

The sinusoidal patterns of *E _{T}*(

Figure 11

**Fig. 11.** Total direct horizontal illuminance *E _{T}*(

With respect to *x*_{max,T}, it is observed that for 0 ≤ *y*_{P} ≤ 1 m, the *x*_{max,T} values are close to zero, meaning that the maximum *E _{T}*(

As an alternative to visualise the results, several samples of the scenes with BZ1, *δ *= 0; BZ1, *δ *= 1.5 m; BZ5, *δ *= 0; and BZ5, *δ *= 1.5 m are modelled and simulated in *DIALux *4.12 [28,29], with light loss factor of 1.0. In addition, to visualise the illuminance distribution on arbitrary vertical planes, two parallel black walls are introduced at *y* = –0.5 m and 3.5 m (i.e. 0.5 m beyond each sources’ array). The resulting false colour images showing direct illuminance distribution on the surfaces in those scenes are displayed in Fig. 12. Note that the same illuminance scale, between 150 and 550 lx, are employed in those four scenes.

Figure 12

**Fig. 12.** False colour images of direct illuminance on surfaces in scenes with (**a**) BZ1, *δ *= 0, (**b**) BZ1, *δ *= 1.5 m, (**c**) BZ5, *δ *= 0, (**d**) BZ5, *δ *= 1.5 m, simulated in *DIALux *4.12.

It is observed that while the maximum luminous intensity *I*_{0} = 1800 cd is constant for both BZ types, the maximum *E*_{P} values in the BZ1 scenes are smaller than those in the BZ5 scenes, due to the wider LID in the latter. For the same reason, the *E*_{P} variations in the BZ5 scenes are also less visible than those in the BZ1 scenes. Regardless of the *δ *values, the highest spatial contrast in the BZ1 scenes are found between the points beneath the sources and those on the middle row. Nevertheless, increasing the *δ* value up to 1.5 m (= *λ*/2) may reduce this contrast, relative to *δ* = 0. In the BZ5 scenes, shifting the *δ* value is not expected to dramatically change the visual observation of the horizontal surface.

The following remarks can be made regarding the phasor method proposed in Section 2 as Approaches 1 and 2. Both approaches can be employed in the situation where the individual *E*_{P} from the source is known or can be determined. Approach 2 is relatively simpler than Approach 1, which requires a certain sinusoidal regression calculator; but if there are too few discrete *E*_{P} data, the estimation of *x*_{max} value in Approach 2 may be incorrect, because the actual *x*_{max} value may be located between (i.e. not exactly on) the observed calculation points.

Meanwhile, the direct observation of *E _{T}*(

In turn, one may still question the usefulness of such concept in lighting design. To answer that, it is first to be understood that an engineering design process requires a known relation between input (design parameters) and output variables (performance indicators). Once the values of the output variables have been defined, for instance based on standards, regulations, or user requirements, the designer shall be able to determine the required input variables values. In this case, the resulting *E _{T}*(

An obvious example would be the case to determine the position of calculation row having uniform pattern of *E _{T}*(

While the case of *y*_{P} = *Y*/2 is quite intuitive due to symmetry, the case of arbitrary *y*_{P} values (0 < *y*_{P} < *Y*/2) may be less obvious, so that the configuration of both arrays should be designed properly. In that case, one can start with the desired sinusoidal pattern of *E _{T}*(

Equivalently, one can construct the phasor diagram as in Fig. 6 and proceed with conducting the analysis to obtain* E _{o}*

Figure 13

**Fig. 13.** Phasor diagram to determine *E _{o}*

As mentioned earlier, several main assumptions are required in the entire concept. First, all sources should possess a cosine-like LID, so that the resulting *E*_{1}(*x*), *E*_{2}(*x*), and *E _{T}*(

The second main assumption is that both arrays have a constant spatial wavelength *λ*, which is necessary to enable the use of phasor diagram, where one can put the term *e ^{jkx}* in Equation (8) aside and recall it later once the superposition has been done. In cases where

Finally, to give another perspective on the applicability of the proposed concept, it is worthwhile to thoroughly read the proposal of Cuttle et al. [12,23,24] regarding the appearance-based, perceived adequacy of illumination (PAI), and the eventual reception of his proposal during the past decade. While they argue that horizontal illuminance is not necessarily related with PAI and spatial brightness [12], some (perhaps many) will keep saying that horizontal workplane illuminance is still very much what matters in the design practice [30]. In this current study, though still based on horizontal illuminance, the resulting patterns are never expected to be uniform (except for the special case of *y*_{P} = *Y*/2 and *δ* = *λ*/2), and thus can be related to some extent with the concept of target/ambience illuminance ratio (TAIR) [24,30]. For instance, though not exactly the same, the *E*_{min} can be compared in analogy with the ambient illuminance, whereas *E*_{max} with the target illuminance.

As a follow-up of this current study, further investigations are suggested on more scenarios involving sinusoidal patterns of illuminance in various realistic situations, for instance in large halls or residential roads. Furthermore, since this study only considers regularly-spaced arrays of sources, scenarios in future studies can also be extended to irregularly-spaced arrays. The proposed phasor method in this study is thus expected to become a tool to better understand the physical phenomena of light superposition from parallel arrays of sources, and eventually to refresh the design practice in such cases.

Direct horizontal illuminance along a calculation row (*E _{T}*(

Based on the sensitivity analysis, the choice of LID type significantly affects Δ*E*, which is the ratio between (*E*_{max,T} – *E*_{min,T}) and *E*_{med,T}. For BZ5 type, Δ*E* is relatively small so that the illuminance fluctuation may be unnoticeable, because the sources’ LID is wide (120° beam angle) and thus create smaller contrast within points along the calculation row. For BZ1 type, the fluctuation may however be noticeable, since the narrow LID corresponds to smaller beam angle, yielding greater spatial contrast.

Meanwhile, the impact of lateral position of the calculation row (*y*_{P}) and the spatial phase difference (*δ*) on Δ*E* patterns are inconsistent. In the tested scenarios, for 0 ≤ *y*_{P} ≤ 1 m, the maximum *E _{T}*(

This research is supported by the Institute of Research and Community Service of Institut Teknologi Bandung (LPPM ITB), under the P3MI 2020 research scheme.

R. A. M. designed the research framework, performed the computation and simulation, and wrote the manuscript. At. reviewed and co-wrote the manuscript.

The author declares that there is no conflict of interest.

- J.H. Lambert, Photometria, sive de Mensura et Gradibus Luminis, Colorum et Umbrae, Eberhard Klett, Augsburg, 1760.
- D.L. DiLaura, K.W. Houser, R.G. Mistrick, G.R. Steffy (eds.), The Lighting Handbook: Reference and Application, 10th Edition, Illuminating Engineering Society, New York, 2011.
- P.R Boyce, Editorial: Illuminating engineering rises from the grave, Lighting Research and Technology 51 (2019) 487–487. https://doi.org/10.1177/1477153519852246
- C.E. Ochoa, M.B.C. Aries, J.L.M. Hensen, State of the art in lighting simulation for building science: a literature review, Journal of Building Performance Simulation 5 (2012) 209–233. https://doi.org/10.1080/19401493.2011.558211
- Z. Huang, A. Sanderson, Light field modelling and interpolation using Kriging techniques, Lighting Research and Technology 46 (2013) 219–237. https://doi.org/10.1177/1477153512473996
- A.A. Baloch, P.H. Shaikh, F. Shaikh, Z.H. Leghari, N.H. Mirjat, M.A. Uqai, Simulation tools application for artificial lighting in buildings, Renewable and Sustainable Energy Reviews 82 (2018) 3007–3026. https://doi.org/10.1016/j.rser.2017.10.035
- R.H. Simons, A.R. Bean, Lighting Engineering: Applied Calculations, Architectural Press, Oxford, 2001.
- C.F. Reinhart, T. Dogan, D. Ibarra, H.W. Samuelson, Learning by playing – teaching energy simulation as a game. Journal of Building Performance Simulation 5 (2012) 359–368. https://doi.org/10.1080/19401493.2011.619668
- I. Beausoleil-Morrison, Learning the fundamentals of building performance simulation through an experiential teaching approach, Journal of Building Performance Simulation 12 (2019) 308–325. https://doi.org/10.1080/19401493.2018.1479773
- J.A. Lynes, Designing for contrast rendition, Lighting Research and Technology 14 (1982) 1–18. https://doi.org/10.1177/096032718201400101
- S. Rockcastle, M. Amundadottir, M. Andersen, Contrast measures for predicting perceptual effects of daylight in architectural renderings, Lighting Research and Technology 49 (2016) 882–903. https://doi.org/10.1177/1477153516644292
- J. Duff, K. Kelly, C. Cuttle, Perceived adequacy of illumination, spatial brightness, horizontal illuminance and mean room surface exitance in a small office, Lighting Research and Technology 49 (2017) 133–146. https://doi.org/10.1177/1477153515599189
- J.M. Monteoliva, A. Villalba, A. Pattini, Daylighting metrics: an approach to dynamic cubic illuminance, Journal of Daylighting 5 (2018) 34–42. https://doi.org/10.15627/jd.2018.6
- V.W.L. Lo, K.A. Steemers, Methods for assessing the effects of spatial luminance patterns on perceived qualities of concert lighting, Lighting Research and Technology 52 (2019) 106–130. https://doi.org/10.1177/1477153519841098
- F. Selim, S.M. Elkholy, A.F. Bendary, A new trend for indoor lighting design based on a hybrid methodology, Journal of Daylighting 7 (2020) 137–153.
- A. Lauria, S. Secchi, L. Vessella, Visual wayfinding for partially sighted pedestrians – The use of luminance contrast in outdoor pavings, Lighting Research and Technology 51 (2018) 937–955. https://doi.org/10.1177/1477153518792978
- Y. Jiang, S. Li, B. Guan, G. Zhao, D. Boruff, L. Garg, P. Patel, Field evaluation of selected light sources for roadway lighting, Journal of Traffic and Transportation Engineering (English Edition) 5 (2018) 372–385. https://doi.org/10.1016/j.jtte.2018.05.002
- D.M. Kretzer, High-mast lighting as an adequate way of lighting pedestrian paths in informal settlements?, Development Engineering 5 (2020) 100053. https://doi.org/10.1016/j.deveng.2020.100053
- S. Fotios, C. Robbins, Research Note: Describing average illuminance for P-class roads, Lighting Research and Technology, in press. https://doi.org/10.1177/1477153520911193
- D. Cetegen, J. Veitch, G. Newsham, View size and office illuminance effects on employee satisfaction, Proceedings of Balkan light, Ljubljana, Slovenia (2008).
- D. Tiller, J. Veitch, Perceived room brightness: pilot study on the effect of luminance distribution, Lighting Research and Technology 27 (1995) 93–101. https://doi.org/10.1177/14771535950270020401
- K. Wymelenberg, M. Inanici, A study of luminance distribution patterns and occupant preference in daylit offices, Proceedings of PLEA 2009 - the 26th conference on passive and low energy architecture, Quebec City (2009).
- C. Cuttle, Towards the third stage of lighting profession, Lighting Research and Technology 42 (2010) 73–93. https://doi.org/10.1177/1477153509104013
- C. Cuttle, A new direction for general lighting practice, Lighting Research and Technology 45 (2013) 22–39. https://doi.org/10.1177/1477153512469201
- W.R. Stevens, Building Physics: Lighting – Seeing in the Artificial Environment, Pergamon Press, Exeter, 1969.
- J.A. Lynes, Does the British Zonal System have a future?, Lighting Research and Technology 11 (1979) 150–153. https://doi.org/10.1177/14771535790110030201
- Stats.Blue, Sinusoidal Regression Calculator, Available online at: http://stats.blue/Stats_Suite/sinusoidal_regression_calculator.html. 2018. zzzzzz
- DIAL GmbH. DIALux. Available online at: https://www.dial.de/en/dialux/.
- R.A. Mangkuto, Validation of DIALux 4.12 and DIALux evo 4.1 against the analytical test cases of CIE 171:2006, LEUKOS 12 (2016) 139–150. https://doi.org/10.1080/15502724.2015.1061438
- C. Cuttle, Lighting by Design. 2nd Edition, Routledge, London, 2008.

Copyright © 2020 The Author(s). Published by solarlits.com.