1 |
.\" RCSid $Id: genBSDF.1,v 1.21 2021/12/12 20:17:02 greg Exp $ |
2 |
.TH GENBSDF 1 9/3/2010 RADIANCE |
3 |
.SH NAME |
4 |
genBSDF - generate BSDF description from Radiance or MGF input |
5 |
.SH SYNOPSIS |
6 |
.B genBSDF |
7 |
[ |
8 |
.B "\-c Nsamp" |
9 |
][ |
10 |
.B "\-n Nproc" |
11 |
][ |
12 |
.B "\-r 'rcontrib opts...'" |
13 |
][ |
14 |
.B "\-W" |
15 |
][ |
16 |
.B "\-s 'x=string;y=string'" |
17 |
][ |
18 |
.B "\-t{3|4} Nlog2" |
19 |
][ |
20 |
.B "{+|-}C" |
21 |
][ |
22 |
.B "{+|-}a" |
23 |
][ |
24 |
.B "{+|-}forward" |
25 |
][ |
26 |
.B "{+|-}backward" |
27 |
][ |
28 |
.B "{+|-}mgf" |
29 |
][ |
30 |
.B "{+|-}geom unit" |
31 |
][ |
32 |
.B "\-dim Xmin Xmax Ymin Ymax Zmin Zmax" |
33 |
] |
34 |
[ |
35 |
.B "geom .." |
36 |
] |
37 |
.br |
38 |
or |
39 |
.br |
40 |
.B genBSDF |
41 |
.B "\-recover tempdir" |
42 |
.SH DESCRIPTION |
43 |
.I GenBSDF |
44 |
computes a bidirectional scattering distribution function from |
45 |
a Radiance or MGF scene description given on the input. |
46 |
The program assumes the input is in Radiance format unless the |
47 |
.I \+mgf |
48 |
option is specified. |
49 |
The output conforms to the LBNL Window 6 XML standard for BSDF data, |
50 |
and will include an MGF representation of the input geometry if the |
51 |
.I \+geom |
52 |
option is given, followed by one of "meter," "foot," "inch," |
53 |
"centimeter," or "millimeter," depending on the scene units. |
54 |
The default is to include the provided geometry, |
55 |
which is assumed to be in meters. |
56 |
Geometry output can be supressed with the |
57 |
.I \-geom |
58 |
option, which must also be followed by one of the above length units. |
59 |
.PP |
60 |
Normally, |
61 |
.I genBSDF |
62 |
computes components needed by a backwards ray-tracing process, |
63 |
.I \+backward. |
64 |
If both forward and backward (front and back) distributions are needed, the |
65 |
.I \+forward |
66 |
option may be given. |
67 |
To turn off backward components, use the |
68 |
.I \-backward |
69 |
option. |
70 |
Computing both components takes about twice as long as one component, but |
71 |
is recommended when rays will be impinging from either side. |
72 |
.PP |
73 |
The |
74 |
.I \+C |
75 |
option specifies that the output XML should include color information, |
76 |
which is interpreted by the rendering programs. |
77 |
The default option |
78 |
.I \-C |
79 |
reduces all BSDF data to grayscale. |
80 |
.PP |
81 |
The |
82 |
.I \-a |
83 |
option turns off reciprocity averaging for tensor tree output. |
84 |
Normally on (+a), this ensures that each tensor-tree hemisphere follows symmetry |
85 |
implied by Helmholtz reciprocity, and is designed to reduce ray sampling noise. |
86 |
However, for some systems, reciprocity averaging can actually make the output worse. |
87 |
.PP |
88 |
The geometry must fit a rectangular profile, whose width is along the X-axis, |
89 |
height is in the Y-axis, and depth is in the Z-axis. |
90 |
The positive Z-axis points into the room, and the input geometry should |
91 |
not extend into the room. |
92 |
(I.e., it should not contain any positive Z values, since the putative |
93 |
emitting surface is assumed to lie at Z=0.)\0 |
94 |
The entire window system should be modeled, including sills and |
95 |
edge geometry anticipated in the final installation, otherwise |
96 |
accuracy will be impaired. |
97 |
Similarly, materials in the description should be carefully measured. |
98 |
.PP |
99 |
Normally, the input geometry will be positioned according to its actual |
100 |
bounding box, but this may be overridden with the |
101 |
.I \-dim |
102 |
option. |
103 |
Use this in cases where the fenestration system is designed to fit a |
104 |
smaller (or larger) opening or is offset somehow. |
105 |
.PP |
106 |
The variance in the results may be reduced by increasing the number of |
107 |
samples per incident direction using the |
108 |
.I \-c |
109 |
option. |
110 |
This value defaults to 2000 samples distributed over the incoming plane |
111 |
for each of the 145 Klems hemisphere directions. |
112 |
.PP |
113 |
On multi-core machines, processing time may be reduced by the |
114 |
.I \-n |
115 |
option, which specifies the number of simultaneous |
116 |
processes to run in |
117 |
.I rcontrib(1). |
118 |
The |
119 |
.I \-r |
120 |
option may be used to specify a set of quoted arguments to be |
121 |
included on the |
122 |
.I rcontrib |
123 |
command line. |
124 |
.PP |
125 |
The |
126 |
.I \-W |
127 |
option is passed to |
128 |
.I wrapBSDF(1) |
129 |
to prepare the XML file for WINDOW6. |
130 |
Any |
131 |
.I \-s |
132 |
parameters are passed to the |
133 |
.I \-f |
134 |
option of |
135 |
.I wrapBSDF, |
136 |
controlling XML fields such as |
137 |
the Manufacturer (e.g., -s m=MF) and device Name (e.g, -s n=NM). |
138 |
.PP |
139 |
The |
140 |
.I \-t4 |
141 |
mode computes a non-uniform BSDF represented as a rank 4 tensor tree, |
142 |
suitable for use in the Radiance rendering tools. |
143 |
The parameter given to this option is the log to the base 2 of the |
144 |
sampling resolution in each dimension, and must be an integer. |
145 |
The |
146 |
.I \-c |
147 |
setting should be adjusted so that an appropriate number of samples |
148 |
lands in each region. |
149 |
A |
150 |
.I \-t4 |
151 |
parameter of 5 corresponds to 32x32 or 1024 output regions, so a |
152 |
.I \-c |
153 |
setting of 10240 would provide 10 samples per region on average. |
154 |
Increasing the resolution to 6 corresponds to 64x64 or 4096 |
155 |
regions, so the |
156 |
.I \-c |
157 |
setting would need to be increased by a factor of 4 to provide |
158 |
the same accuracy in each region. |
159 |
.PP |
160 |
The |
161 |
.I \-t3 |
162 |
mode is similar to |
163 |
.I \-t4 |
164 |
but computes a rank 3 tensor tree rather than rank 4. |
165 |
This provides a much faster computation, but only works |
166 |
in special circumstances. |
167 |
Specifically, do NOT use this option if the system is not in fact isotropic. |
168 |
I.e., only use |
169 |
.I \-t3 |
170 |
when you are certain that the system has a high degree of radial symmetry. |
171 |
Again, the parameter to this option sets the maximum resolution as |
172 |
a power of 2 in each dimension, but in this case there is one less |
173 |
dimension being sampled. |
174 |
.PP |
175 |
The |
176 |
.I \-recover |
177 |
option is available for continuing calculations that were killed by |
178 |
the system or the user. |
179 |
Unfortunately, genBSDF puts its temporary files in a directory |
180 |
that is often cleaned up after reboot, so this may not always work. |
181 |
.SH EXAMPLE |
182 |
To create a BSDF description including geometry from a set of venetian blinds: |
183 |
.IP "" .2i |
184 |
genblinds blind_white blind1 .07 3 1.5 30 40 | xform -rz -90 -rx 90 > blind1.rad |
185 |
.br |
186 |
genBSDF -r @rtc.opt blind_white.mat glazing.rad blind1.rad > blind1.xml |
187 |
.PP |
188 |
To create a non-uniform, anisotropic BSDF distribution with a maximum |
189 |
resolution of 128x128 from the same description: |
190 |
.IP "" .2i |
191 |
genBSDF -r @rtc.opt -t4 7 -c 160000 blind_white.mat glazing.rad blind1.rad > blind12.xml |
192 |
.SH NOTES |
193 |
The variable resolution (tensor tree) BSDF representation is not supported |
194 |
by all software and applicatons, and should be used with caution. |
195 |
It provides practical, high-resolution data for use in the |
196 |
Radiance rendering programs, but does not work in the matrix formulation |
197 |
of the daylight coefficient method for example. |
198 |
Also, third party tools generally expect or require a fixed number of sample |
199 |
directions using the Klems directions or similar. |
200 |
.SH AUTHOR |
201 |
Greg Ward |
202 |
.SH "SEE ALSO" |
203 |
dctimestep(1), gendaymtx(1), genklemsamp(1), genskyvec(1), mkillum(1), |
204 |
pkgBSDF(1), rcontrib(1), rfluxmtx(1), rmtxcomb(1), rmtxop(1), rtrace(1) wrapBSDF(1) |