Você está na página 1de 26

Daylighting Computation Methods

From Dot Chart to Digital Simulation Benjamin Geebelen, ir.-arch. K.U.Leuven, dept. of Architecture

Introduction
Whatever our surroundings, light is all around us. At all times, every point of every surface emits, reflects, absorbs and possibly transmits rays of light in an infinite number of directions. It is not hard to imagine that, in most real-world situations, light transfer is too complex a matter to be described analytically. This paper gives an overview of the tools that were devised to find reliable approximations for the distribution of light in a scene. It will focus on the simulation of skylight.

1 Historical background1
Throughout the ages lighting design, like most aspects of construction, was a question of craftsmanship and unwritten rules, of precedent and experience. It was the Industrial Revolution that brought the most rapid change in the applications, requirements and solutions for daylighting. Sparked by phenomena of radical sociological change and technological innovation, a whole range of new building types was invented, such as art galleries, railway stations, assembly halls, libraries, elementary schools and exhibition halls, building on advances in the glazing and framing technologies. With gas and oil lighting still being too expensive, polluting and dangerous, new functional briefs and increased urban densities entailed new problems and requirements for daylight availability. The need for daylighting design tools thus originated in an era in which photometry was in its infancy and luminance had not even been defined. Early daylighting regulations concentrated on the amount of direct daylight from the sky and consisted of simple geometric rules, such as the sky-line rule2 or minimal glazing-area-to-floor-area ratios. By the 1920s the first photometers had been invented and photometry had evolved sufficiently to allow more precise methods. These were invented mainly as a means to adjudicate disputes concerning the obstruction of light by a proposed building. Two of the earliest computation tools are the Waldram diagram, devised by P.J. and J.M. Waldram, and Pleijels pepper-pot diagrams. Throughout the 20th century, until the advent of personal computers, more graphical methods were conceived, such as the BRS protractor. Slowly the attention turned towards the internally reflected component. In times when interiors were often clad in dark natural finishes and covered in the grime of contemporary artificial light sources, it had been neglected. However, after the widespread introduction of electrical lighting, and a confrontation with the massive solar gains and heat losses of the large glass faades so popular in the 1970s, the
1 2

An interesting historical overview can be found in [2], on which this section was based.

The sky-line rule states that a room will be well lit if there is a unobstructed view of the sky from the point of interest.

significance of reflected light for achieving well-lit rooms with sensible window sizes became evident. It was recognized as a key feature to help reduce the variation of illuminance levels across a room. Mathematical prediction techniques were formulated, striking a balance between the complexity of inter-reflections and the simplicity of available design data. Some of these formulae are still useful design tools today. While the use of these graphical or hand-calculation methods is not fundamentally challenging, treating large numbers of reference points is likely to become a tedious and lengthy activity. It is not surprising that some of the first software tools consisted of digital translations of known hand-calculation methods. From the early 1980s on, researchers in the field of computer graphics began to investigate the possibilities of global illumination, the realistic simulation of light transfer within a scene. First initiatives were aimed at visual appeal without much care for physical accuracy, but they laid the groundwork for powerful lighting simulation tools. During the 1990s research initiatives multiplied with different orientations, ranging from user-friendly design tools to integrated energy-performance assessment tools.

2 Scale models
Architects have been using scale models for centuries to assess different aspects of their designs. Many still use scale models today to study the volumetric composition of their designs or to communicate with clients or consultants. Unlike thermal, acoustic, structural or hydrodynamic models, models for lighting studies are not subject to scaling effects. Since the wavelengths of light are so short with respect to the size of buildings and scale models, its behavior is largely unaffected. The light distribution in models with carefully duplicated geometry and material properties will qualitatively and quantitatively match the distribution in the actual room. This makes the scale model a very useful and intuitive lighting design tool, which every architect is familiar with. Even on a small budget the simplest model can give an immediate impression of the light distribution in a room or the dynamics under changing sun positions. Scale models can also be used for measuring quantitative data. However, for this purpose the models need to be built with more care. All joints have to be covered with masking tape and finishes have to match the real building materials as closely as possible. In order to obtain relevant data, the lighting conditions need to be controlled or at least monitored. In an outdoor set-up the simultaneous recording of the sky and the indoor conditions is far from straightforward, and the considerable impact of the luminance distribution of the sky may complicate an analysis of the results. Under artificial skies, i.e. hemispherical skies or mirror boxes, the luminance distribution can be kept constant, which facilitates the measuring procedure. There are a few limitations to the use of scale models for lighting studies: It is very difficult to include artificial lighting. Even if the intensity of artificial light sources can be simulated, the luminance distribution of the luminaires cannot. Some finishes, such as fabrics or brickwork, may be difficult to scale. This may cause errors.

Many artificial skies are only able to simulate overcast sky conditions. Exceptions are hemispherical skies with individually controlled lamps and the so-called onelamp artificial skies [15]. Different studies have indicated that scale models are not the most accurate of simulation tools. Reasons for this include poorly represented surface finishes, light leaks in the model, inexact luminance distributions of the artificial skies, imprecise placement of measurement instruments Photographs taken under artificial skies cannot be used to judge color in the scene. Hardly any architectural firms own an artificial sky or heliodon. Quantitative studies can therefore entail high costs. A testing facility and an operator need to be hired and the model needs to be transported. Due to the restricted time frame, and because the production of the models requires a great amount of care, so as not to introduce any light leaks, it may be difficult to make quick alterations to the model and compare different design alternatives.

3 Graphical, tabular and hand-calculation methods


A whole range of simplified methods has been developed, varying in ease-of-use, accuracy and applicability. They can be categorized in different ways: According to treatment of the direct and reflected components: do they produce one or both, or do they produce the total daylight factor in a single step? According to applicability: can they handle vertical, horizontal or sloping windows? Can they handle saw-tooth roof lights? According to daylight conditions: which kinds of sky luminance distributions can they handle? Typical for simplified methods is that they cannot handle complex sky luminance distributions. They are only applicable to azimuthally invariant sky models, mostly uniform or overcast. According to output: do they produce mean, minimum or maximum values, or can they handle arbitrary reference points? According to form: do they consist of tables, equations, protractors, nomograms, dot charts or diagrams? According to underlying light transfer model: are they based on the Flux Transfer method, the Lumen Method, or do they have another foundation? The computation of the direct and the reflected components are so different that many simplified methods merely produce one of both. It is then possible to choose one method for each component and use them in conjunction, e.g. a protractor for the direct and an equation for the reflected component. However, it is always wise to choose methods that make similar allowances for deterioration of decorations, dirt on the windowpanes, framing and window bars, and transmission losses of the glazing type. A detailed discussion of all available methods would go beyond the scope of this section. Only the most common ones are briefly discussed. A more general overview can be found in [1]. 3.1 The direct component

Because the direct component of the daylight factor depends on relatively few parameters the shape of the windows, their transmission characteristics and their

position relative to the reference point its computation has often been captured in graphical or tabular form. One of the oldest methods, dating back to 1923, is the Waldram diagram. This kind of diagram consists of a grid of squares, each of which represents an equal portion of daylight factor, on which one can draw the projections of windows and obstructions as seen from the reference point. In order to know the direct component of the daylight factor, one simply needs to count the squares within the outline of the projection. The diagram was designed in such a way that vertical edges remain vertical in the projection. Horizontal edges, however, need to follow the shape of the so-called droop lines in order to take the cosine law of illumination and the nonuniform luminance of the sky vault into account. The one shown in Fig. 1 is based on the luminance distribution of an overcast sky and allows for glazing losses. This method offers fairly good accuracy.
Angles of azimuth 90 80 70 60 50 40 30 20 10 0 10 20 30 40 50 60 70 80 90

80

70

60

50 40 40 30 20 30 20 10
Droop lines of horizontal edges at right angles to plane of window Droop lines of horizontal edges parallel to plane of window

Unobstructed view of sky

Angles of altitude

90 70

60

50

External obstruction

Fig. 1 Waldram diagram for CIE Overcast Sky and vertically glazed apertures, including corrections for glazing losses. As an example a large window and an obstructing tower are indicated. Each square indicated in fine lines corresponds with a daylight factor of 0.1%.

Pleijel followed a similar approach for the design of his pepper-pot diagrams. Here the direct component of the daylight factor can be obtained by counting the dots that fall within the contours of the projection. The great advantage of this kind of diagram is that the density of the dots accounts for the non-linearity of the illumination, so that projections can be made without deformations. The drawback, however, is that counting the dots can become a very tedious task.

Fig. 2 Pepper-pot diagram or dot chart for the sky component of the daylight factor on horizontal planes (from [16]). This diagram applies to vertical windows and the CIE Standard Overcast Sky.

The BRS Daylight Protractors are probably the most widely used graphical tools [3]. They come in pairs: one primary protractor or daylight scale, and one auxiliary protractor or correction scale. Protractors are available for different sky types and various slopes of glazing. The main advantage of protractors is that they can be used straight onto plans and sections of the proposed room. They are very easy to use. Fig. 4 and Fig. 4 show an example for vertical glazing. The primary protractor is placed onto a section drawing. It provides the sky component or the equivalent sky component of an external obstruction as the difference between the readings of the top and bottom edge. The secondary protractor is placed onto a plan drawing and delivers the correction factors for windows of finite length. For the externally reflected component an additional correction factor of 0.2 is usually used. Protractors are less practical for irregular compositions. However, it is usually possible to assume an average simple outline.

Fig. 3

BRS protractors for the CIE Standard Overcast Sky and vertical windows.

Fig. 4

The use of Building Research Station protractors.

During the very early stages of design, when scale drawings are not yet available, one can use the BRS Simplified Daylight Tables. For very simple geometric compositions these tables provide the sky component.

Sky component of the daylight factor [%] 1.30

2.50 2.40 2.40 2.40 2.30 2.30 2.20 2.20 2.10 2.00 2.00 1.90 1.90 1.80 1.70 1.60 1.50 1.40 1.30 1.10 0.99 0.83 0.68 0.53 0.39 0.25 0.14 0.06 0.02 0.2

3.70 3.70 3.60 3.60 3.50 3.40 3.40 3.30 3.20 3.10 3.00 2.90 2.80 2.70 2.60 2.40 2.30 2.10 1.90 1.70 1.50 1.20 0.97 0.74 0.52 0.34 0.18 0.09 0.02 0.3

4.90 4.80 4.70 4.60 4.50 4.50 4.40 4.30 4.10 4.00 3.90 3.80 3.60 3.50 3.30 3.20 2.90 2.70 2.50 2.20 1.90 1.60 1.30 0.98 0.70 0.45 0.26 0.11 0.03 0.4

5.90 5.90 5.80 5.70 5.50 5.40 5.30 5.20 5.00 4.80 4.70 4.60 4.40 4.20 4.00 3.80 3.60 3.30 3.00 2.60 2.20 1.90 1.50 1.20 0.82 0.54 0.30 0.12 0.03 0.5

6.90 6.80 6.70 6.60 6.40 6.30 6.20 6.00 5.80 5.60 5.40 5.30 5.10 4.90 4.60 4.40 4.10 3.80 3.40 3.00 2.60 2.20 1.70 1.30 0.97 0.62 0.34 0.14 0.04 0.6

7.70 7.60 7.40 7.30 7.10 7.00 6.80 6.60 6.40 6.20 6.00 5.80 5.60 5.40 5.10 4.80 4.50 4.20 3.80 3.30 2.80 2.40 1.90 1.50 1.00 0.70 0.38 0.16 0.04 0.7

8.40 8.30 8.20 8.00 7.80 7.60 7.50 7.30 7.00 6.70 6.50 6.30 6.10 5.80 5.60 5.20 4.90 4.50 4.00 3.60 3.10 2.60 2.10 1.60 1.10 0.75 0.42 0.20 0.05 0.8

9.00 8.80 8.70 8.50 8.20 8.10 7.90 7.70 7.40 7.10 6.90 6.70 6.50 6.20 5.90 5.60 5.20 4.80 4.30 3.80 3.30 2.70 2.20 1.70 1.20 0.82 0.44 0.21 0.05 0.9

9.60 9.40 9.20 9.00 8.70 8.60 8.40 8.10 7.90 7.50 7.30 7.10 6.80 6.50 6.20 5.90 5.50 5.00 4.60 4.00 3.40 2.90 2.30 1.80 1.30 0.89 0.47 0.21 0.05 1

10.70 11.60 12.20 12.60 13.00 13.70 14.20 14.60 14.90 15.00 10.50 11.10 11.70 12.30 12.70 13.30 13.70 14.00 14.10 14.20 10.30 10.90 11.40 12.00 12.40 12.90 13.30 13.50 13.60 13.70 10.10 10.60 11.10 11.80 12.20 12.60 12.90 13.20 13.20 13.30 9.80 9.60 9.30 9.10 8.70 8.30 8.10 7.80 7.50 7.20 6.80 6.40 5.90 5.40 4.90 4.30 3.70 3.10 2.50 1.90 1.40 0.92 0.49 0.22 0.06 1.2 10.20 10.70 11.30 11.70 12.00 12.40 12.50 12.60 12.70 10.00 10.50 11.10 11.40 11.70 12.00 12.20 12.30 12.30 9.80 9.50 9.10 8.70 8.50 8.20 7.80 7.50 7.10 6.70 6.20 5.70 5.10 4.50 3.80 3.20 2.50 1.90 1.40 0.95 0.50 0.22 0.06 1.4 10.20 10.80 11.10 11.40 11.70 11.80 11.90 11.90 10.00 10.40 10.70 11.00 11.20 11.30 11.40 11.50 9.60 9.10 8.80 8.50 8.20 7.80 7.40 7.00 6.40 5.90 5.30 4.60 3.90 3.30 2.60 2.00 1.40 0.95 0.50 0.22 0.06 1.6 10.00 10.20 10.50 10.70 10.80 10.90 10.90 9.50 9.20 8.80 8.50 8.10 7.60 7.20 6.60 6.00 5.40 4.70 4.00 3.30 2.60 2.00 1.40 0.96 0.51 0.22 0.06 1.8 9.70 9.40 9.00 8.60 8.20 7.80 7.30 6.70 6.10 5.40 4.70 4.00 3.30 2.60 2.00 1.50 0.96 0.51 0.23 0.07 2 9.90 9.60 9.20 8.80 8.40 7.90 7.40 6.80 6.20 5.50 4.80 4.00 3.30 2.60 2.10 1.50 0.96 0.52 0.23 0.07 2.5 10.00 10.10 10.20 10.30 9.70 9.30 8.90 8.50 8.00 7.50 6.90 6.20 5.60 4.80 4.10 3.40 2.70 2.10 1.50 0.97 0.52 0.23 0.07 3 9.80 9.40 9.00 8.60 8.00 7.50 6.90 6.30 5.60 4.90 4.10 3.40 2.70 2.10 1.50 0.97 0.52 0.23 0.07 4 9.90 9.50 9.10 8.60 8.10 7.60 6.90 6.30 5.70 5.00 4.20 3.40 2.80 2.10 1.50 0.98 0.53 0.24 0.08 6

5 4 3.5 Ratio height of window above working plane : distance from window [-] 3 2.8 2.6 2.4 2.2 2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

1.20 1.20 1.20 1.20 1.10 1.10 1.10 1.10 1.00 1.00 0.97 0.94 0.90 0.86 0.82 0.77 0.71 0.65 0.57 0.50 0.42 0.33 0.24 0.16 0.10 0.06 0.03 0.01 0.1

9.90 9.50 9.10 8.60 8.10 7.60 7.00 6.30 5.70 5.00 4.20 3.40 2.80 2.10 1.50 0.98 0.53 0.24 0.08

Ratio width of window to one side of normal : distance from window [-]

Table 1

BRS Simplified Daylight Table for vertical glazed rectangular windows.

3.2

The internally reflected component (IRC)

It is far more difficult to obtain a good estimate of the reflected component, since it depends on scene geometry and material properties of all surface finishes in the scene. Because of the endless possibilities it is hard to parameterize both and translate them into graphical form. Most simplified methods for the internally reflected component therefore consist of equations and nomograms. A number of sources prescribe different formulae for the internally reflected component, with different parameters and allowances. However, most of them can be traced back to only a few basic ways of abstracting the scene. The Split-Flux Method regards the scene as consisting of only two surfaces: the floor with the part of the vertical walls below the center of the window and the ceiling and the part of the vertical walls above the center of the window. It is then assumed that the light from the sky is distributed over the lower part, and the externally reflected light over the upper part. For clear glazing the average direct illuminance of the lower part can then be expressed as: 8

Ed ,l = g .

E g ,sky . Ag Al

(1)

where Ed ,l = the average direct illuminance of the lower part,

g Eg,sky Ag Al

= = = =

the specular transmittance of the glazing, the illuminance of the window due to the sky, the area of the window and the area of the lower part of the scene.

The average direct illuminance of the upper part can be expressed in a similar way: Ed ,u = g . E g , ground . Ag Au the average direct illuminance of the upper part, the specular transmittance of the glazing, the illuminance of the window due to the ground, the area of the window and the area of the upper part of the scene.
(2)

where Ed ,u = g = Eg,ground = Ag = Au =

Both Eg,sky and Eg,ground are estimated based on the horizontal illuminance under an unobstructed sky. The illuminance due to the sky for a vertical surface can fairly easily be computed as follows:
E g ,sky = Eh .Chv .Cobstr
(3)

where Eg,sky = the illuminance of the window due to the sky, Eh = the horizontal illuminance in the open field, Chv = a correction factor to transform from horizontal to vertical illuminance (0.5 for a uniform sky, roughly 0.4 for a CIE Overcast Sky) and Cobstr = a correction factor for exterior obstructions. For an overcast sky BRE lists values for the product of both correction factors depending on the angle of obstruction measured from the center of the window [4]. These can be found in Table 2. Alternatively Cobstr can be approximated as
Cobstr = 1

obstr 90

(4)

where obstr = the angle of obstruction [].

Table 2

Correction factors for the illuminance of the windows due to the sky according to BRE.

For the illuminance of the window due to the ground, the ground is regarded as a perfectly diffuse surface with constant luminance
Lground = ground . Eh
(5)

where Lground = the luminance of the ground, ground = the reflectance of the ground, usually taken as a minimum of 0.1 and Eh = the horizontal illuminance under an unobstructed sky. The illuminance of the window can then be approximated according to

E g , ground =

E .Lground = ground . h . 2 2

If we regard both parts of the scene as two parallel infinite planes, we can compute the average illuminance and luminance of the upper part of the scene as follows: Eu = and Lu = u . where Eu 1 u .Ed ,u + u . l .Ed ,l = . 1 l . u
(8)

Ed ,u + l .Ed ,l 1 l . u

Eu = the average illuminance of the upper part of the scene,

Ed ,u = the average direct illuminance of the upper part of the scene, Ed ,l = the average direct illuminance of the lower part of the scene,
Lu = the average luminance of the upper part of the scene, u = the average reflectance of the upper part of the scene and l = the average reflectance of the lower part of the scene.
The average internally reflected component of the daylight factor can then be computed as:

DF

IRC

g . Ag u . ground u .l .Chv .Cobstr . + 1 u . l 2 Au Al

10

Angle of obstruction [] No obstruction 10 20 30 40 50 60 70 80

Ch v.Cobstrn [%] 39 35 31 25 20 14 10 7 5

(6)

(7)

(9)

where DF
IRC

g Ag Au Al u

= = = = = =

the internally reflected component of the daylight factor, the specular transmittance of the glazing, the area of the window, the area of the upper part of the scene, the area of the lower part of the scene, the average reflectance of the upper part of the scene,

l = the average reflectance of the lower part of the scene, ground = the reflectance of the ground, Chv = a correction factor to transform from horizontal to vertical illuminance and Cobstr = a correction factor for exterior obstructions.
This only applies to clear glazing. For diffuse glazing the average direct illuminance values should be computed differently: Ed ,l = g . and Ed ,u = g . Eg ,sky + Eg , ground Ag . . 2 Au
(11)

E g ,sky + E g , ground Ag . 2 Al

(10)

The average internally reflected component of the daylight factor can then be computed as follows:

DF

IRC

g . Ag u 1 l ground . . + . A A 2 + Chv .Cobstr . 1 u . l 2 u l

(12)

A slightly simpler variation of the Split-Flux principle is the Integrated-Sphere Approximation, in which the entire scene is regarded as a closed sphere with constant luminance. For this situation the average luminance can be estimated according to
L= Ld 1
(13)

where
L = the average luminance of the inner surfaces of the scene, Ld = the average luminance of the inner surfaces of the scene due to direct illumination and = the average reflectance of the inner surfaces of the scene.

Using equations (1) and (2) we can estimate Ld as follows:

11

Ed ,u E . Au + l . d ,l . Al Ld = A E .A E .A 1 = . u . g . g , ground g . Au + l . g . g ,sky g . Al A. Au Al .A = g g .(u .E g , ground + l .E g ,sky ) A.

u .

(14)

where

u l Au Al A Ed ,u

= = = = = =

the average reflectance of the upper part of the scene, the average reflectance of the lower part of the scene, the total surface area of the upper part of the scene, the total surface area of the lower part of the scene, the total area of all inner surfaces of the scene, the average direct illuminance of the upper part of the scene,

Ed ,l = the average direct illuminance of the lower part of the scene, g = the specular transmittance of the glazing, Ag = the area of the window, Eg,ground = the illuminance of the window due to the ground and Eg,sky = the illuminance of the window due to the sky.
If we substitute Eg,ground and Eg,sky using equations (3) and (6), we can compute the average internally reflected component of the daylight factor as follows:

DF

IRC

g . Ag . u . ground + l .Chv .Cobstr A.( ) 1 2

(15)

where DF
IRC

ground Chv
Cobstr

g Ag A u l

= the average internally reflected component of the daylight factor, = the specular transmittance of the glazing, = the area of the window, = the total area of all inner surfaces of the scene, = the average reflectance of the upper part of the scene, = the average reflectance of the lower part of the scene, = the reflectance of the ground, = a correction factor to transform from horizontal to vertical illuminance and = a correction factor for exterior obstructions.

This is the method prescribed by BRE [4]. A value of 0.1 is assumed for ground and Table 2 lists values for the correction factors. A common and easy-to-use alternative for equations are nomograms. These can be obtained for different cases. The example shown in Fig. 5 allows the fast computation of the average internally reflected component for side-lit rooms. Find the point on scale A indicating the appropriate window-area-to-total-surface-area ratio. Find the point of scale B indicating the average reflection factor of the interior surfaces. Connect both points with a line. The point in which this line intersects with scale C

12

indicates the average internally reflected component of the daylight factor disregarding external obstructions. If there are obstructions, find the point on scale D that indicates the angle of obstruction and connect it with the point you just found on scale C. The intersection point of this line with scale E indicates the average internally reflected component with obstructions. The great advantage of nomograms is their ease-of-use. However, for complex geometries they may offer insufficient accuracy. Moreover, each nomogram is based on an assumption of the distribution of reflectance values. One cannot differentiate between the upper and lower part of the scene, which can cause discrepancies with the values that are computed using an equation.

13

Fig. 5

Nomogram for the average internally reflected component of the daylight factor (from [10]).

3.3

Combining methods

It is now possible to combine a method for the direct component and one for the internally reflected component to compute the total daylight factor:

14

DF = DF DC + DF IRC

= DF SC + DF ERC + DF IRC
where DF DFDC DFIRC DFSC DFERC = = = = = the daylight factor in a reference point, the direct component of the daylight factor, the internally reflected component of the daylight factor, the sky component of the daylight factor and the externally reflected component of the daylight factor.

(16)

The direct component is generally regarded as consisting of the sky component, due to the unobstructed portion of the sky, and the externally reflected component, due to obstructions. The latter is usually computed in the same way as the former and then corrected with a reflectance. For a uniform sky BRE prescribes a reflectance of 0.1, for a CIE Standard Overcast Sky 0.2. When combining methods, especially from different sources, special attention should go to: the transmission of the glazing: is it included in both methods? Is it included implicitly or explicitly? Do both methods employ the same value? additional allowances: does either method include a correction for dirt on the glazing? the output of the methods: do they provide point values, or average, maximum or minimum values? 3.4 Single-step methods

A few methods approximate the daylight factor without differentiating between a direct and a reflected component. Most of them use regression to express the daylight factor as a linear combination of the illuminance of the window due to the sky and the illuminance due to the reflection off the ground. By measurements in scale models appropriate coefficients have been established for a number of room geometries, material properties, glazing transmittances, etc. The best-known method of this type is the Lumen Method, where the daylight factor can be found as:

DF = Ev . Ag . g .K u
or
DF = (E g ,sky .K u ,sky + E g , ground .K u , ground ) Ag . g .

(17)

(18)

Ev = the vertical illuminance under an unobstructed sky, Eg,sky = the illuminance of the glazing due to the sky, Eg,ground = the illuminance of the glazing due to the ground, Ag = the glazing area, g = the transmittance of the glazing and Ku, Ku,sky, Ku,ground = coefficients of utilization. Coefficients of utilization have been published by different sources for a variety of geometries, glazing systems, sky types, etc.

where

15

3.5

Estimating annual daylight availability

Once the daylight factor in a reference point or on the working plane is known, the annual daylight availability can be estimated. Based on the function of a space, the required internal illuminance level can be determined. If we divide this by the daylight factor, we obtain the required external illuminance level. Based on meteorological data we can find the percentage of the working year during which this external illuminance level is attained (Fig. 6).
Daylight availability
Percentage of 100% working year during which 95% indicated illuminance is 90% attained 85% 80% 75% 70% 65% 60% 55% 50% 0 2000 6AM-6PM 7AM-3PM 4000 7AM-5PM 8AM-4PM 6000 8AM-6PM 9AM-5PM 8000 10000 12000 14000 Minimal illuminance in open field [lux]

Fig. 6

Daylight availability in the Belgium based on meteorological data from Brussels.

4 Digital simulation
Since the advent of personal computers numerous attempts have been made to develop software tools that predict the lighting in a proposed room. This section discusses the theories behind different approaches. Of each approach examples are given. However, since most software evolves rather rapidly, there is a distinct risk that the information presented here will quickly be outdated. Discussions of individual programs are therefore deliberately kept concise. 4.1 Simplified algorithms

A number of programs simply offer a digital translation of the simplified hand calculation methods discussed above. The main benefit of this approach is an increased ease-of-use. The input is usually simple and efficient, and results are delivered instantaneously. The user no longer needs to compute azimuth or altitude angles, sky components, average reflectance values, total surface areas or internally reflected components. The drawback, however, is that there is very little improvement in the accuracy of the results. This type of software is excellently suited for the initial stages of architectural design, when important decisions have to be made, based on little information. They allow the designer to easily and quickly compare different design alternatives. That the accuracy of the output is so low can be justified by the fact that there are too many unknowns to allow a precise result. At an early stage the designer is more interested in qualitative rather than quantitative results. 16

A good example of this kind of software is Leso-DIAL, which was developed at EPFL [14]. It computes the sky component analytically, which implies a slight improvement of accuracy as compared to manual methods, and estimates the average internally reflected component using the BRE Split-Flux Formula. The emphasis in the development of this program was on applicability in early design stages. It cannot handle complex geometry, but input is extremely straightforward and intuitive. Reflectance values are entered qualitatively, with values ranging from very dark to very light, and besides numerical output, it offers a diagnostics module, which suggests alterations to the design that would improve daylight availability. Another, much earlier, example is DAYLIT [1]. The goals were similar, but besides daylighting computations, the program includes electric lighting and thermal calculations. For its daylighting predictions it uses the Lumen Method as prescribed by the IES.

Fig. 7

Examples of the Leso-DIAL user interface.

4.2

Light transfer simulation3

The ambition to simulate the light transfer between the surfaces of scene first emerged in the field of computer graphics. In the pursuit of more realism of digitally synthesized images, researchers sought ways of simulating the interaction of light and objects, of mimicking light being reflected, transmitted and refracted, of computing shadows and highlights. The resulting algorithms, generally referred to as global illumination models, all define approximating solutions for what is known as the rendering equation [11]. In its outgoing form this equation expresses the amount of light leaving a surface at a certain point in a certain direction as the sum of the surfaces own emittance and the light reflected and transmitted by the surface:

Algorithms are discussed only briefly in this section. More information can be found in [6] and [7].

17

& & & & & & L(p, o ) = Le (p, o )+ f (p, o )L(r , )cos( )d

(19)

where & & L(p, o ) = the luminance of the surface at point p in direction o , & & Le (p, o ) = the luminance of the surface at point p in direction o due to its own emittance, & & f (p, o ) = the function that describes how light arriving from direction & & is reflected or transmitted to direction o at point p, & r = the point from which the light arriving from direction originated, & = the angle between direction and the surface normal and = the total sphere around point p. This is clearly a recursive formula: in order to compute the luminance at a point p we need to compute the luminance values at all points r surrounding it. It is already a simplification of reality, since it does not account for an interaction with the medium or spectral effects. In order to fully simulate the light distribution in a scene we need to solve this equation for all points in that scene. Except for a limited number of ideal cases, this is an impossible task: the equation needs to be solved for an infinite number of points; around each point an infinite number of directions needs to be considered; the function f is hard to determine for most real-life materials. The following section discusses the main techniques to find an approximate solution for the rendering equation. Historically ray tracing was the first technique to be developed [19]. It overcomes the problems in solving the rendering equation by limiting the number of investigated directions and thus tracing individual light rays through the scene. There are two variants to this scheme. In forward ray tracing or ray casting light rays are followed in the direction of light propagation, i.e. from the light source towards the scene. More common, however, is backward ray tracing, which tracks the light back from the viewer to the light source. This recursive algorithm is the literal translation of the rendering equation. The original goal was to produce realistic images of a geometric scene. Such an image can be interpreted as a projection of the scene onto a rectangular screen between the scene and the viewer. This screen consists of an orthogonal grid of picture elements or pixels, each of which portrays a particular part of the scene with a uniformly colored square. The color and intensity of a pixel is generally determined by the point in the scene that is visible in the pixels center or by a grid of points, all visible within the pixels boundaries. These points are found by tracing eye rays from the viewpoint, through the pixel, towards the scene until they hit one of the scenes surfaces. An estimate for the rendering equation is then found for each of these points. In theory all directions around a point need to be considered. However, ray tracing limits this number to only the most important ones, i.e. the directions of the light sources in the scene, the direction of reflection for reflective surfaces, and the direction of transmission for transparent surfaces. For each of these directions an additional ray is traced. Rays that sample a light source will check whether the light 18

source is visible to the investigated point. Reflection and transmission rays will look for the nearest intersection with objects in the scene. For each of these intersections the process is repeated, thus resulting in a recursive tree of rays (Fig. 8 and Fig. 9).
R3

Source 2

S32

Object 3

S31

R1 S22 Object 2

Object 4

S12 T1 E1 Object 1 S21 R2

Pixel Eye Image

S11

Source 1

Fig. 8

Backward ray tracing.

Eye E1 Pixel 1 E2 Pixel 2

Source 1 Source 2

S11 Object 1 S12 R1 T1

Source 1 Source 2

S21 Object 2 S22 T2 R2

Source 1 Source 2

S31 Object 3 S32 T3 R3

Fig. 9 The recursive ray tree. Eye rays are indicated with E, light-source rays with S, reflection rays with R, and transmission rays with T. The Xs indicate light-source rays that are blocked by other objects or transmission rays that are not investigated because the material is not translucent.

This technique performs best with scenes that contain ideally specular materials and point light sources (Fig. 10). Diffuse material behavior can be reproduced by tracing additional random rays. Large light sources can be handled by super sampling, i.e. testing visibility at multiple points across their surfaces. However, since the ray tree grows exponentially with the number of rays per intersection, these measures have a considerable impact on computation time. In its classic form, ray tracing is a view-dependent algorithm: only those points of the scene are investigated that influence the colors of the pixels in the final image (Fig. 19

11). Moving the viewpoint therefore entails a complete new simulation. Moreover, the result does not really represent the light distribution in the scene, but merely a limited number of light transfers between an equally limited number of points in the scene.

Fig. 10 A typical image rendered with classical ray tracing: smoothly curved surfaces, sharp reflections, sharp shadows from a point light source, etc.

Fig. 11

In ray tracing only those points are investigated that are important for the rendered image.

The main counterpart of ray tracing is generally called radiosity and was first introduced during the mid 1980s [8][13]. This technique adopts a finite-element approach to overcome the difficulties in solving the rendering equation. By subdividing the scene into a limited number of patches and nodes, and by specifying that all light exchange needs to happen between those nodes, the number of possible light fluxes will also be limited (Fig. 12). The rendering equation is now computed for n nodes instead of an infinite number of points, and for each node only the n-1 directions of the other nodes are investigated instead of an infinite number of

20

directions. The result of this simplification is a system of n equations expressing the luminance in each node as a linear combination of the luminance values in the other nodes. If we can solve the system, the luminance of any point in the scene can be approximated by interpolation between its surrounding nodes. The n n coefficients of the system of equations are called form factors. Each form factor describes the light transfer between a pair of nodes: it is proportional to the fraction of light transported from one node to the other.

Fig. 12

The number of light fluxes in a radiosity approach is limited.

Radiosity introduces two major challenges: computing the form factors and solving the system of equations. Not only is it a difficult task to compute a single form factor, an average scene can easily contain several thousands of nodes, resulting in millions of form factors. In addition, it would take exceptional computing times to solve a system of several thousands of equations. Generally these problems are overcome by a combination of measures: form factors are approximated in an efficient way, e.g. by means of ray casting; the number of form factors is reduced by grouping nodes in a hierarchical way; form factors are only computed on demand; the system is solved iteratively, e.g. by using Jacobi, Gauss-Seidel or Southwell iteration.

21

Fig. 13 A typical image rendered with classical radiosity: diffuse materials, soft shadows, large light sources (image by D. Marini, Universit degli Studi di Milano).

In many ways radiosity is the opposite of ray tracing. It performs best with scenes that consist of ideally diffuse opaque surfaces and large diffuse light sources; in its classical form it is view-independent, and it is an efficient way of estimating the light distribution in a scene. Both techniques have their pros and cons, which often seem complementary (Table 3).
Ray tracing View-dependent Handles specular behavior best Handles any geometry Can handle transparency Does not compute the overall light distribution in the scene Has difficulties with indirect lighting Radiosity View-independent Handles diffuse behavior best Performs best with facetted shapes Performs best with opaque surfaces Does compute the overall light distribution in the scene Indirect lighting is treated correctly

Table 3

Comparison between classical ray tracing and classical radiosity.

Not surprisingly many of the best lighting simulation programs use hybrid algorithms, combining both a radiosity and a ray-tracing step. For any digital simulation to be appropriate for daylighting, it needs to be able to aptly simulate the sky, which is a vast light source of non-uniform luminance. Different algorithms will employ different sky models: the sky as a large hemisphere with a superimposed luminance distribution: this approach may perform well if sky luminance can be expressed mathematically. A ray tracer will then sample the hemisphere at a great number of random points, each with the appropriate luminance. For more accuracy the hemisphere can be virtual, as if of infinite size. Luminance is now determined based on the sample ray s altitude and azimuth; 22

the sky as a collection of light sources: this approach may be chosen when sky luminance is known at a collection of azimuth / altitude pairs. Ray tracers may prefer point light sources, whereas radiosity programs may think of the light sources as disks of constant luminance. Again, for more accuracy these light sources should be treated as if at an infinite distance from the scene.

An additional difficulty arises when we want to assess annual daylight availability. The number of individual daylighting conditions for a single assessment can range from several thousands to several hundreds of thousands, depending on the chosen time step. For Brussels, a time step of one hour will result in about 4 700 instances. A time step of one minute4 raises that number to 280 000! Considering that a single simulation of reasonable accuracy can easily require a few minutes of computation time, it is obvious that it is unrealistic to run a full simulation for each individual daylighting condition. This would require two weeks for a time step of one hour, and two years for a time step of one minute. Different approaches have been examined: the daylight-factor method: a single luminance distribution is assumed for all time steps, usually the CIE Standard Overcast Sky. The scene needs to be simulated only once, after which the result can be scaled for each time step using the openfield diffuse horizontal illuminance. This is comparable to the approach suggested for simplified methods and delivers the quickest result. The great disadvantage is that all directionality of the sky s luminance distribution is lost. Since the CIE Standard Overcast Sky represents a distribution in which the highest luminance values are concentrated in the area of the zenith, which does not apply to partly cloudy or clear skies, this method will underestimate the daylight availability for side-lit room and overestimate the daylight availability for top-lit rooms; interpolation between extremes [20]: the scene is simulated once for an overcastsky luminance distribution, and once for a clear-sky distribution. For each time step the result is then obtained by interpolating between these two. The weight factor can be based on the cloud ratio or the effective sunshine probability. This approach is also very efficient, but still neglects the azimuthal dependence of luminance distributions. Moreover, an interpolation between two extreme conditions is not necessarily a good representation of an intermediate condition. interpolation between extremes with monthly sun positions [5]: to include the azimuthal dependence, the previous approach can be extended to include a circumsolar region. Typically a clear sky with sun is simulated for each hour of the 15th of every month, resulting in 150 additional simulations. This still leaves the problem of intermediate skies unsolved. classified weather data: Herkel and Pasquay have tried to group time steps into a set of some 450 categories of similar sun position, direct and diffuse illuminance [9]. This solution performs reasonably well, but results in a stepped cumulative daylight distribution. daylight coefficients: instead of simulating the entire sky, the sky vault is thought of as consisting of a set of discrete elements of constant luminance. The contribution of each element to the indoor light distribution is simulated to produce the daylight coefficients. For each time step the indoor light distribution

It has been argued that such short time steps are necessary to accurately model the behavior of lighting control systems [17].

23

can then be obtained as a linear combination of the daylight coefficients, using the sky elements luminance values as coefficients. Typically a set of 145 sky elements is used. This approach strikes a good balance between accuracy and computation time. 4.3 Examples of simulation software

By far the most well-known and popular package is Radiance [18]. Development began during the 1980s as a study of how rendering techniques such as ray tracing could be applied to lighting simulation. Previously the aim had merely been to produce good-looking images without much care for the physical correctness. Over the years Radiance has evolved into a set of some 50 different programs, constituting one of the most powerful and most accurate simulation suites currently available. It has often been used as the backbone of other simulation programs. The core algorithm behind Radiance is ray tracing. However, to account for the contribution of diffuse indirect light, a mesh of nodes is introduced into the scene in which irradiance is cached. These are used for indirect light source sampling. The program has seemingly endless possibilities. The user can define complex shapes and material behavior, add luminance patterns, define sky luminance distributions, etc. In addition, its accuracy has been extensively validated and documented [12]. In its original form, which can be obtained free of charge, the program has no graphical user interface, which makes it rather daunting for beginners. However, a recent AutoCAD interface, called Desktop Radiance, makes it far more user friendly. Many experts will prefer to produce initial scene descriptions with Desktop Radiance and then manually adapt the resulting files and use the command-line version for meticulous manipulation. Like Radiance, SuperLite was also developed at LBL. This program uses radiosity for the reflected component in combination with Monte-Carlo techniques for the direct component. Its modeling and visualization capabilities are rather limited, but it can be useful for early design stages. Genelux uses a variant of forward ray tracing, which its developers call photons generation . Particularly interesting is that it is a web-based tool. Users can upload their models onto a server and order the simulations of their choice. Results can be downloaded after completion. One of the programs that were derived from Radiance is ADELINE. It was first released in 1994 and combines Radiance and SuperLite to produce illuminance levels, daylight factors, comfort levels and photo-realistic images. In addition, it delivers lighting data that can be used for thermal simulation. It has a graphical user interface and a built-in geometrical modeler, but ease-of-use could be improved considerably. For predictions on an annual basis it interpolates between three luminance distributions for every hour of the 15th of every month, i.e. an overcast sky, a clear sky without sun and a clear sky with sun. Lumen Micro, which for some years was the American industry standard for electrical lighting, was enhanced with a daylighting module by the late 1980s. It uses radiosity to produce numerical results fairly quickly but takes slightly more time to produce rendered images. Its modeling capabilities may prove too limited for advanced simulations.

24

LightScape also uses radiosity, but offers an additional ray-tracing step to add specular effects. It can produce illuminance and luminance values for any point or surface in the scene, as well as highly realistic images. The software has recently been purchased by AutoDesk and has been incorporated in Autodesk VIZ 4. In recent years more attention has gone to the accurate and efficient estimation of annual daylight availability. Interesting in this respect is Passport-Light, developed in the framework of the EC project Daylight Europe . It uses backward ray tracing to compute daylight coefficients and can be used as a pre-processing step for timestep prediction. A similar effort was made by the developers of DAYSIM. This adapted version of Radiance produces daylight coefficients in a parallel manner.

References
[1] Ander, G.D., Milne, M. and Schiler, M., Fenestration Design Tool: A Microcomputer Program for Designers , in: Proceedings of the 2nd International Daylighting Conference, Long Beach, CA, 187-193 (1986). Baker, N., Fanchiotti, A. and Steemers, K. (eds.), Daylighting in Architecture A European Reference Book, James & James, London (1993). BRE, Digest 309 Estimating daylight in buildings: Part 1, Building Research Establishment, Garston (1986). BRE, Digest 310 Estimating daylight in buildings: Part 2, Building Research Establishment, Garston (1986). Erhorn, H., de Boer, J. and Dirksmller, M., ADELINE An Integrated Approach to Lighting Simulation , in: Proceedings of Daylighting 98, Ottawa, Natural Resources Canada, 21-28 (1998). Foley, J., van Dam, A., Feiner, S. and Hughes, J., Computer Graphics: Principles and Practice, Addison-Wesley, Reading (1990). Glassner, A.S., Principles of Digital Image Synthesis, Morgan Kaufmann, San Francisco (1995). Goral, C.M., Torrance, K.E., Greenberg, D.P. and Battaile, B., Modeling the Interaction of Light between Diffuse Surfaces , in: Computer Graphics 18(3), 213-222 (1984). Herkel, S. and Pasquay, T., Dynamic link of light and thermal simulation: on the way to integrated planning tools , in: Proceedings of the 5th International IBPSA Conference, Prague, IBPSA, 307-312 (1997).

[2] [3] [4] [5]

[6] [7] [8]

[9]

[10] Hopkinson, R.G., Architectural Physics Lighting, Her Majesty s Stationery Office, London (1963). [11] Kajiya, J.T., The Rendering Equation , in: Computer Graphics 20(4), 143-150 (1986). [12] Mardaljevic, J., Validation of a lighting simulation program under real sky conditions , in: Lighting Research & Technology 27(4), 181-188 (1995). [13] Nishita, T. and Nakamae, E., Continuous-Tone Representation of ThreeDimensional Objects Taking Account of Shadows and Interreflection , in: Computer Graphics 19(3), 23-30 (1985).

25

[14] Paule, B., Bodart, M., Citherlet, S. and Scartezzini, J.-L., Leso-DIAL Daylighting Design Software , in: Proceedings of Daylighting 98, Ottawa, Natural Resources Canada, 29-36 (1998). [15] Schouwenaars, S. and Wouters, P., One-lamp artificial sky and solar simulator for daylight measurements on scale models , in: Proceedings of International Building Physics Conference, Eindhoven, FAGO, TU/e, 283-290 (2000). [16] van Santen, C. and Hansen, A.J., Licht in de architectuur, J.H.De Bussy, Amsterdam (1985). [17] Walkenhorst, O., Luther, J., Reinhart, C. and Timmer, J., Dynamic annual daylight simulations based on one-hour and one-minute means of irradiance data , in: Solar Energy 72(5), 385-395 (2002). [18] Ward Larson, G. and Shakespeare, R., Rendering with Radiance The Art and Science of Lighting Visualization, Morgan Kaufmann, San Francisco (1998). [19] Whitted, T., An Improved Illumination Model for Shaded Display , in: Communications ACM 23(6), 343-349 (1980). [20] Winkelmann, F. and Selkowitz, S., Daylighting simulation in DOE-2: theory, validation and applications , in: Proceedings of the Building Energy Conference, Seattle, WA, 326-336 (1985).

26

Você também pode gostar