Light Cache

[Andrew1]

Hey Guys,

 

I have a question, is there some kind of rule or calculation regarding Light cache subdivision vs sample size setting.

I have noticed that rendering image for example 2000 sub and 0.05 sample size, gave me better results than 2000 sub and 0.005 sample size.

The first one gave me clean render the second one showed white spots of unwanted illumination in a few corners.

Vray seems to produce very nice renders but I have noticed that there is no one that knows it through out, I mean there are always some problems or issues and all you can hear is do this or that if that does not work then it must be it, then there is something wrong with geometry, then geometry is fine, then there is something else and it goes on and on.

I have studied this render engine for past 1.5 years, I must say that I do like the renders but sometimes there are problems that just drive me crazy.

There are times that for example irr + LC in some cases can never get rid of certain artifacts, does not matter how high you set the preset.

I am recently using qmc + lc and this seems to be the most reliable solution,

at least for me.

 

any comments would be appreciated.

 

Cheers,

 

Andrew

[Christopher Nichols]

Well think about it. You are now using much smaller sample areas. With the large one, you get a smoother result because it blends together nicely, with the smaller one, it can get more detail, but unless you seriously increase your subsivision, you will not get a smooth result.

 

Keep in mind that LC is not ment for detail. It is for mainly used for the secondary bounces, which does not care about the detail as much. You get all your high detail in your primary bounce. So for that use, the irradiance map or the brute force QMC, either of which are designed for detail. LC power is that it can get a huge amount of bounces (basically infinite), in a very small amount of time. For more information on LC, you can always check the free help page online:

http://www.spot3d.com/vray/help/150R1/examples_lightcache.htm

[Jeff Mottle]

Check out the Visualization Insider Article " Critical Settings" here: http://www.cgarchitect.com/upclose/VI/

[FlytE]

So chris on that basis, would you say that light cache is better kept for indoors scenes? In the past I have had quite alot of problems with lc and exteriors, im wondering if its something to do with that....

[Andrew1]

Chris and Jeff thanks for your kind help,

I was just thinking that there maybe a ratio aspect in the settings subdivision/sampling.

I have red the "visualization Insider" as well as the vray help file,

I am still having a bit of a problem of understanding it fully,

I just don't learn from reading I do read it but I can not retain it.

 

Chris said, "You are now using much smaller sample areas. With the large one, you get a smoother result because it blends together nicely, with the smaller one, it can get more detail, but unless you seriously increase your subsivision, you will not get a smooth result."

 

So with a larger sample my detail will still be targeted just more likely blended ?, I have had an issue where I could not get a good detail with my molding along ceiling line, and by increasing subdivision in LC this helped a lot,

Does the subdivision determine the detail and samples determine the quality of detail or the other way around.

 

I have been using qmc as primary, and lc as secondary with great results and time efficiency, just sometimes I get stuck with odd situations, and also I noticed that vray glitches and does not comply with setting changes.

 

Chris said "Keep in mind that LC is not ment for detail. It is for mainly used for the secondary bounces, which does not care about the detail as much. You get all your high detail in your primary bounce. So for that use, the irradiance map or the brute force QMC, either of which are designed for detail. "

 

I am aware that you are the master blaster of vray and I must listen to you !! jk, I get all my high detail from my qmc as primary but if I don't have much subdivision in my lc that won't impact my qmc calculation regarding the detail? as I wrote above, I have noticed that tweeking the subdivision in LC as secondary bounce, it picked up detail from my scene almost entirely.

I thought that both of them are respnsible in their own way for detail level rendered.

 

I am slowely getting a hang of it, regarding time management and the detail quality, but for sure there is still tons to learn.,

 

Thanks for the input guys,

Now back to F9 and see:D

 

Cheers !

[Christopher Nichols]

I use LC all the time. It is not for detail it is for spreading the light. The ONLY place you really need detail is with your primary bounce. That is why things like IR map, and FinalGather exist. In order to get the rest of the general illumination, a good method in the past was photon mapping, until LC came along, was way more effecient, and was way easier to use. Of course Photon mapping is still very much a valid thing when it comes to things like caustics.

 

Anyway... back to your questions.

 

I still use LC for exteriors, but it's contribution to the scene is so mimimal I often turn it off. It does not take up much time to calc, but can take away some precious ram.

 

If oyu are doing stills, or an animation with moving objects there is no need to turn your LC above lets say 1000 sub. Also, screen space is fine, and I often use the default. If I start to see crappy spots, I may raise the sample size.

 

Now... when doing it for baking, that is a different story. I need to raise the Subdivs because I need to spread those samples accross the whole scene. It is also a balance of sample sizes which is now in world space. If it is a huge space like a stadium those samples sinze may need to be like 10 feet (3+ meters). I need to raise those so that I get smooth results. Once the sample size is set, I slowly raise the subdivs, monitoring the ram to make sure it is not out of control. I have gone as high as 4000, but generally try to keep it on the low end if possible.

 

hmmm... this may be a good tutorial for my new DVD...

[Tim Saunders]

hmmm... this may be a good tutorial for my new DVD...

Good Idea.

I was going to suggest that Andrew watches your videos, and quite frankly if you have been studying vray "for the past 1.5 years", I am supprised you have not seen the videos yet. Especially since you don't learn as well by reading. Chris SHOWS you how to do the stuff. Not that everything is suddenly going to "click" after seeing the videos, but they will help. There is no set answer for everything in vray, or any renderer for that matter, because EVERY scene is different.

[Andrew1]

Thanks Guys,

 

I am gonna have to invest into those videos,

It's just like Tim says that every situation is different, and a different approach maybe necessary,

Well thanks again, this site is very informative , I realy enjoy this forum:D ,

 

Cheers !

 

Andrew