two glass issues

Use this forum for general user support and related questions.
Forum rules
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
kintuX
Posts: 707
Joined: Wed Jan 10, 2018 2:37 am

Re: two glass issues

Post by kintuX » Tue Sep 07, 2021 2:23 pm

Yes, here's result w/ biDir.
bug-biDir.jpg

Check attached .blend and compare. Or send yours.
Attachments
doubleGlass.7z
(293.7 KiB) Downloaded 11 times

daros
Posts: 163
Joined: Thu Dec 12, 2019 3:25 pm

Re: two glass issues

Post by daros » Wed Sep 08, 2021 5:51 am

Unfortunately i can only share the luxcore files.
Anyway i imported your file and created a new scene and now it works... very strange. I do not understand. Same scene, same material....
But what about the simple plane scene i posted where the same plane suddently renders almost black?

kintuX
Posts: 707
Joined: Wed Jan 10, 2018 2:37 am

Re: two glass issues

Post by kintuX » Wed Sep 08, 2021 2:51 pm

daros wrote:
Wed Sep 08, 2021 5:51 am
But what about the simple plane scene i posted where the same plane suddently renders almost black?
Your files render same as you've shown, but differently tonemaped & crash when I want to save outputs.
Also go into an endless loop, and when paused it still continues... denoising...
luxcoreUI_Paused_endlessLoop.jpg

PS
:?: What app do you use to generate LuxCore scene file? Could it be corrupt?
To test-check, copy geo into new scene and render...

daros
Posts: 163
Joined: Thu Dec 12, 2019 3:25 pm

Re: two glass issues

Post by daros » Thu Sep 09, 2021 9:52 am

thank for your feedback kintuX.

Anyway this two scenes should be really identical, only the camera angle varies. Why one shows up black and the other transparent? How is that possible? May Dade can help me with this?


Luxcore 2.6 alpha 0
good.png
bad.png
Attachments
buono.zip
(2.58 KiB) Downloaded 4 times
cattivo.zip
(2.6 KiB) Downloaded 5 times

User avatar
Dade
Developer
Posts: 5522
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: two glass issues

Post by Dade » Thu Sep 09, 2021 10:40 am

It is a numerical precision problem (see https://forums.luxcorerender.org/viewto ... 635#p31627). You are rendering something 300m away from the origin with a thickness of few mm.

This is the distance between two consecutive hits in a path (first number is the path depth, second is the distance):

Code: Select all

1=38.8555
2=0.0337687
3=0.0219092
4=0.0219092
5=0.0219092
6=0.0219092
7=0.0219092
8=0.0219092
9=0.0219092
10=0.0219092
11=0.0219092
12=0.0219092
13=0.0219092
14=0.0219092
15=0.0219092
16=0.0219092
17=0.0219092
18=0.0219092
19=0.0219092
20=0.0219092
21=0.0219092
22=0.0219092
23=0.0219092
24=0.0219092
25=0.0219092
26=0.0219092
27=0.0219092
28=0.0219092
29=0.0219092
30=0.0219092
31=0.0219092
32=0.0219092
33=0.0219092
34=0.0219092
35=0.0219092
36=0.0219092
37=0.0219092
38=0.0219092
39=0.0219092
40=0.0219092
41=0.0219092
42=0.0219092
43=0.0219092
44=0.0219092
45=0.0219092
46=0.0219092
47=0.0219092
48=0.0219092
49=0.0219092
50=0.0219092
51=0.0219092
52=0.0219092
53=0.0219092
54=0.0219092
55=0.0219092
56=0.0219092
57=0.0219092
58=0.0219092
59=0.0219092
60=0.0219092
61=0.0219092
62=0.0219092
63=0.0219092
64=0.0219092
65=0.0219092
66=0.0219092
67=0.0219092
68=0.0219092
69=0.0219092
70=0.0219092
71=0.0219092
72=0.0219092
73=0.0219092
74=0.0219092
75=0.0219092
76=0.0219092
77=0.0219092
78=0.0219092
79=0.0219092
80=0.0219092
81=0.0219092
82=0.0219092
83=0.0219092
84=0.0219092
85=0.0219092
86=0.0219092
87=0.0219092
88=0.0219092
89=0.0219092
90=0.0219092
91=0.0219092
92=0.0219092
93=0.0219092
94=0.0219092
95=0.0219092
96=0.0219092
97=0.0219092
98=0.0219092
99=0.0219092
100=0.0219092
101=0.0219092
102=0.0219092
103=0.0219092
104=0.0219092
105=0.0219092
106=0.0219092
107=0.0219092
108=0.0219092
109=0.0219092
110=0.0219092
111=0.0219092
112=0.0219092
113=0.0219092
114=0.0219092
115=0.0219092
116=0.0219092
117=0.0219092
118=0.0219092
119=0.0219092
120=0.0219092
121=0.0219092
122=0.0219092
123=0.0219092
124=0.0219092
125=0.0219092
126=0.0219092
127=0.0219092
128=0.0219092
129=0.0219092
130=0.0219092
131=0.0219092
132=0.0219092
133=0.0219092
134=0.0219092
135=0.0219092
136=0.0219092
137=0.0219092
138=0.0219092
139=0.0219092
140=0.0219092
141=0.0219092
142=0.0219092
143=0.0219092
144=0.0219092
145=0.0219092
146=0.0219092
147=0.0219092
148=0.0219092
149=0.0219092
150=0.0219092
151=0.0219092
152=0.0219092
153=0.0219092
154=0.0219092
155=0.0219092
156=0.0219092
157=0.0219092
158=0.0219092
159=0.0219092
160=0.0219092
161=0.0219092
162=0.0219092
163=0.0219092
164=0.0219092
165=0.0219092
166=0.0219092
167=0.0219092
168=0.0219092
169=0.0219092
170=0.0219092
171=0.0219092
172=0.0219092
173=0.0219092
174=0.0219092
175=0.0219092
176=0.0219092
177=0.0219092
178=0.0219092
179=0.0219092
180=0.0219092
181=0.0219092
182=0.0219092
183=0.0219092
184=0.0219092
185=0.0219092
186=0.0219092
187=0.0219092
188=0.0219092
189=0.0219092
190=0.0219092
191=0.0219092
192=0.0219092
193=0.0219092
194=0.0219092
195=0.0219092
196=0.0219092
197=0.0219092
198=0.0219092
199=0.0219092
200=0.0219092
201=0.0219092
202=0.0219092
203=0.0219092
204=0.0219092
205=0.0219092
206=0.0219092
207=0.0219092
208=0.0219092
209=0.0219092
210=0.0219092
211=0.0219092
212=0.0219092
213=0.0219092
214=0.0219092
215=0.0219092
216=0.0219092
217=0.0219092
218=0.0219092
219=0.0219092
220=0.0219092
221=0.0219092
222=0.0219092
223=0.0219092
224=0.0219092
225=0.0219092
226=0.0219092
227=0.0219092
228=0.0219092
229=0.0219092
230=0.0219092
231=0.0219092
232=0.0219092
233=0.0219092
234=0.0219092
235=0.0219092
236=0.0219092
237=0.0219092
238=0.0219092
239=0.0219092
240=0.0219092
241=0.0219092
242=0.0219092
243=0.0219092
244=0.0219092
245=0.0219092
246=0.0219092
247=0.0219092
248=0.0219092
249=0.0219092
250=0.0219092
251=0.0219092
252=0.0219092
253=0.0219092
254=0.0219092
255=0.0219092
256=0.0219092
As you see, the first hit is after 38m, the second is after 3cm and than the path is stuck there, self-intersecting the same point. The max. path depth set to 256 is also absurd but that is a different matter.

If you set the epsilon to a more appropriate value for your huge scene size, it will render fine:

Code: Select all

scene.epsilon.min = 0.001
scene.epsilon.max = 0.001
Support LuxCoreRender project with salts and bounties

daros
Posts: 163
Joined: Thu Dec 12, 2019 3:25 pm

Re: two glass issues

Post by daros » Thu Sep 09, 2021 4:14 pm

I understand the point, thanks a lot dade.
However it would be cool to address somehow this kind of problem because thereare many projects several km big. Despite that, a glass on a table should be rendered well even if it has only 3mm thickness.
Do you think this is addressable?

Post Reply