mirrors don't seem to be reflecting as they should

[charlie baker]

while i have been an accurender user since release 2 i don't use it more than a few times a year, so i rarely get a good knowledge base up before i forget it all again. then i need to ask for help - so here goes.

this time i am now designing a vertical farm in an old office building- for this to work avoiding using huge amounts of artificial lighting we are proposing to use mirrors on the outside of the building to bounce light that would have been lost to the un-windowed parts of the elevation. but the light doesn't seem to be bouncing on the attached model - the reflections are but not the sunlight so the bit of the room that i know will be lit by mirror outside bouncing to another on the ceiling (i built a card model to prove it pic attached) still shows up as relatively dark - using the lighting analysis function of ar4 it was showing no discernible difference at all.

is this a limitation in me or the software?

some thoughts would be appreciated, the file is attached

thanks

 

ps more on this project is here if interested:

http://headstretcher.blogspot.com/2011/07/vertical-farming-in-manch...

[Roy Hirshkowitz]

First of all, the units are incorrect. I changed the units to meters to better match the scale of the model.  Always check this first, it was causing your textured columns to be rendered incorrectly.  Now, on to the real problem:

 

Bouncing light of a mirror-like surface to illuminate other surfaces is what's known in rendering as a reflective caustics path.  It is a very difficult problem for rendering software to handle.  It was not handled at all by previous version of AR.  In nXt it is possible to do, although it's not perfect (it doesn't work well for all models) and you do need to do a little work in terms of telling nXt which surfaces to consider.  It does work better with the Path Tracer.

 

I did the following:

 

1. I switched to the Path Tracer in Lighting->Advanced.
2. I exploded the block containing the outside solar reflectors.
3. I selected the reflectors and tagged them as Caustics using the Object Properties tab.  It is not necessary to select the ceiling reflector since it doesn't receive light directly.
4. I let the thing render for about ten minutes.

You can pretty clearly see the additional light in the first image (the second has no caustics.)  In particular, the reflected light on the floor from the ceiling shows up as a discrete rectangle (expected with a mirror-like surface.)  There is also a rectangular reflection on the ceiling.

 

 

 

[Roy Hirshkowitz]

BTW,  your metal material is only 50% reflective or so.  That means the double bounce will only be about 25% reflective.  When I change the material color to 253 253 253 (a reflectivity of about 98.4%), I get a very different result.  You'll need to calibrate this value carefully to the actual material if you want to get a good simulation.  To get the reflectivity from the color, divide by 255 and square the result.  To get a color from reflectivity, take the square root and multiply by 255.

 

[charlie baker]

you're a star thanks for the first reply - i can see the outline now but  in your other reply which showed up on my mail doesn't seem to show up here, you mentioned the reflectance but i can't seem to find the setting in the material editor which makes me feel a bit daft

 

also once i've got this working, will i be able to pull up a lighting analysis graphic with the false colours like the old accurender used to to do?

 

if not will i be able to pull of lighting values any other way - the agronomists need to know lux levels for plant growth.

 

thanks in advance

[Roy Hirshkowitz]

By reflectance I wast just referring to the color.  I made the mirror color 253,253,253 instead of 180.  Not sure why you see only one response-- the second picture is better. 

 

No illuminance data directly.  By running your cursor over the image you can get luminance (in cd/m^2).  You can calculate illuminance (lux) for any matte surface by taking that luminance number, multiplying by the reflectance (calculate that from the color as mentioned above)-- and divide by PI.

[charlie baker]

this is getting there now - image attached, haven't remodelled mirror steelwork yet - , the geometry is quite fiendish trying to work out optimal mirror shape.

just a couple of things, i've checked the 2 units variable in autocad and they're both showing up as metres - you mentioned in your first reply that you had to change them to get the materials to render properly - am i missing a setting somewhere?

 

on the mouse over illuminance - is that the 3 values at the bottom of the render window? if so;

q1 how come there are 3 of them?

q2 how come the value changes so radically if i change materials, i'd have thought luminance was the amount landing there so not connected to what it was landing on?q2a if it is dependent on the material, how do i calculate the reflectance from the colour?

q3 there are quite a lot of different values as i mouse over the single area - so do i need to make the floor a single texture free colour to even those out?

 

sorry for being so thick!

but thanks for your help - we're only a little practice and we have managed to do what a major corporate engineering company we were told to work with couldn't - they were using IES so i've had to do it for them using your, as ever, marvellous software

[Roy Hirshkowitz]

Sorry for the late response-- I was out of town for a few days.  The mouse-over shows luminance.  Luminance is the amount of light reflected back toward the viewer.  It is definitely dependent on both the received light and the reflective properties (material) of the surface in question.  Illuminance, on the other hand, is the amount of light received by a surface and is not affected by the material. Calculate illuminance from luminance using the procedure above.

The three values are for each RGB channel-- they'll be the same if your material is gray.

The different values can be due to material variations (bumps, etc..) or to noise in the Path Tracer.  The PT noise will eventually resolve, the variability in the material will not.

[charlie baker]

much as i thought, but thought i'd better check, i gave my agronomist the proportional difference between the direct and reflected for now, i'll leave PT running overnight to smooth out the variations, as i'd used a smooth material. but at risk of further exposing my lack of knowledge in this area, and i did try googling first to try and avoid having to ask yet more, how do you work out reflectance from the colours?

(i do quite the like the fact that urbed (staff numbers -9) and mcneel (staff numbers about 50 isn't it?) have managed to resolve an issue that IES and Ecotect and Buro Happold - (staff numbers massive) haven't.)

what we're all wondering is how do you do light shelves properly on other software if you can't reflect more than once. but also what is the bounces setting in accurender that you can set actually doing?

will this exercise make it worthwhile considering putting the lighting analysis from previous versions into net? that would complete the supremacy of net over the above.

thanks for your help so far

 

[Roy Hirshkowitz]

Caustics paths are difficult to do properly and can be very time consuming.  Modern software like Maxwell can handle these paths very well (although slowly)-- some older software such as Radiance could do some of them as well with some added user intervention.

Most lighting analysis software is really only set up for diffuse reflection. They can calculate this fairly quickly, and it is adequate for things like code-compliance simulations.  This will change in the future, no doubt.

*****

I can easily produce false color images of Luminance.  I can't do Illuminance easily without making some changes to the engine. (Illuminance information was generated automatically in previous versions when using radiosity (not for caustics, though)).  There isn't a lot of demand for lighting analysis tools-- I think yours is the second request I've had.

 

[charlie baker]

maybe it's an area of market expansion once people know about the gaps in the industry software, <he suggested hopefully>

in the meantime how to i calculate the reflectance?

thanks

[Roy Hirshkowitz]

Take the color and divide by 255.  Square the result.

 

r = (c/255)^2

 

Use gray colors to make your life easier.  Make sure the surface for which you are calculating the illuminance from the nXt luminance has a matte material (Highlight Intensity = 0.0).