PDA

View Full Version : Affect Specular in Distant Light Causes Fireflies



medzo
07-15-2018, 03:24 PM
Hi

No matter what I do I found that Affect Specular in Distant Light always causes fireflies... sun can't produce specular highlights?!? it must be a bug.

Matej

Ma3rk
07-15-2018, 04:07 PM
The interaction of lighting, surfacing, camera & rendering settings are so significantly different in 2018 that you can't just jump in where you were & expect the same results.

A couple of vids I'd recommend.

There are several by Andrew Comb, but this one definitely, although the audio hum is hideous:

https://www.youtube.com/watch?v=72AAwSFx4nA

https://www.youtube.com/watch?v=qeLLGPwiYOs

And then there's Rebel Hills series:

https://www.youtube.com/watch?v=lY3K...MnKF7-4W84I5Tz

This series is extracted from one he's soon to release but allowed some early bird copies. Very worthwhile series to pick up when you can. Not only explains the differences & what aspects they affect, his workflow to using the Nodal surfacing & new render buffers really makes sense & probably not something you'd hit on by yourself.

That being said, much of the forum chatter on fireflys, noise etc. I suspect are mostly self inflicted from simply not bothering to take the time to learn how the new system is all inter-connected. So, they'll flail about.

medzo
07-15-2018, 04:13 PM
Found the solution. Turn off Normalize and crank up intensity to 100000 lx :)

jwiede
07-15-2018, 07:32 PM
That being said, much of the forum chatter on fireflys, noise etc. I suspect are mostly self inflicted from simply not bothering to take the time to learn how the new system is all inter-connected. So, they'll flail about.

That's right, keep blaming the customers -- after all, it's not like there's any historical basis to suggest doing so is a bad idea. :rolleyes:

JohnMarchant
07-16-2018, 07:54 AM
I think NT/LWG3D really missed a trick with 2018. There are some radical changes in surfacing and rendering in 2018 that a few presets would have gone a long way to helping us out. We can save render presets and surface presets so how hard can it be. Just half a dozen render presets for indoors and outdoors rendering would have saved allot of grief and time.

Yeah i know you cant have a catch it all preset and of course we still need to learn the basics but like the surface presets it would have helped out allot.

Marander
07-16-2018, 10:56 AM
I think NT/LWG3D really missed a trick with 2018. There are some radical changes in surfacing and rendering in 2018 that a few presets would have gone a long way to helping us out. We can save render presets and surface presets so how hard can it be. Just half a dozen render presets for indoors and outdoors rendering would have saved allot of grief and time.

Yeah i know you cant have a catch it all preset and of course we still need to learn the basics but like the surface presets it would have helped out allot.

I completely agree. This makes it specially not attractive, efficient or usable for new users or as a secondary application (together with the various UI flaws).

Tobian
07-18-2018, 06:33 AM
Can you actually show us, because I rarely get fireflies except for some indirect specular hits, if i really crank it up bright. This is an inherent issue with any PBR lighting system, because physically realistic light sources cause fireflies. Some renderers are better at dealing with that noise, and hopefully some of those features will make it into the next version.

John, given I have to tailor my render settings per scene, I am not sure what presets you think *COULD* exist to optimise LightWave. There could be some internal optimisations and workflow redesigns which could help, but that's software re-design, not 'presets'. The essential issue is by default the settings are set to a minimum for maximum speed, but that's going to result in crappy renders unless your scene is very simple, the problem is if you then crank all those settings up to high, then you get hideous slow render times on scenes which don't need settings that high. Different materials and environments require different amounts of sample per the type which is causing the issue. The tools are there, in the buffers, to go through your scene and optimise it, so there's the minimum number of samples per ray type... we're just so far past 'low medium high' in the world of indirect specular and diffuse it's not funny.

JohnMarchant
07-18-2018, 07:18 AM
Can you actually show us, because I rarely get fireflies except for some indirect specular hits, if i really crank it up bright. This is an inherent issue with any PBR lighting system, because physically realistic light sources cause fireflies. Some renderers are better at dealing with that noise, and hopefully some of those features will make it into the next version.

John, given I have to tailor my render settings per scene, I am not sure what presets you think *COULD* exist to optimise LightWave. There could be some internal optimisations and workflow redesigns which could help, but that's software re-design, not 'presets'. The essential issue is by default the settings are set to a minimum for maximum speed, but that's going to result in crappy renders unless your scene is very simple, the problem is if you then crank all those settings up to high, then you get hideous slow render times on scenes which don't need settings that high. Different materials and environments require different amounts of sample per the type which is causing the issue. The tools are there, in the buffers, to go through your scene and optimise it, so there's the minimum number of samples per ray type... we're just so far past 'low medium high' in the world of indirect specular and diffuse it's not funny.

I didnt say that, there is no catch all render setup that will please everyone and every scenario. Im not saying they will optimise LightWave for all scenarios. I have never, i think, found a preset that actually fitted exactly what i wanted without some adjustment. However i would not say that we should scrap the whole preset system because of that. Presets are an initial guide for people and especially for those who do not understand the software fully. As far as i know presets for rendering setups are new to 2018, so whats the point then if we wont use them or there is no point to them. There is a point to them albeit a little point.

What i said was meant at all the threads that popped up after 2018 release saying how crappy they looked in many instances and many were from people who were longtime LW users. This was of course like me because we did not understand the changes internally that had taken place. There were even threads with exactly the same settings as possible in 2015 and 2018 and the differences in some cases were stark, not because 2018 renderer was worse, it was because we did not understand it.

I think NT/LWG3D could have provided basic presets for some render setups or else whats the point of having the preset system exposed to render setups. I would have killed off a few of these threads and also if you were a new LW user it would have helped you out. As it was Andrew Comb came to the rescue with some excellant videos to explain what was going on. Even if they did provide them of course we will always adjust them to suit our needs. I would have thought maybe 6 to 8 render presets were probably enough to get started.

Tobian
07-18-2018, 08:23 AM
Right, and that's the problem. I'm not disagreeing with you, in that it would be nice, it's that I don't know what those presets would be, and I am Andrew Comb :D To be fair there is a big article on noise reduction and all of the stuff on environment lights are explained in the manual, it's just that no one actually read those things. The main issue is that the workflows weren't compatible, and any automatic conversion would have flaws. There was no way round that except for better learning materials (which yes SHOULD have happened, but someone dropped the ball there!) An example is the paradigm shift of specular. If you loaded an old asset, it imported and converted old specular to new specular, and disabled glossy reflections, so that it wouldn't look hideous; because the paradigms are different, there's no way to 'automatically' convert, you just have to learn. Unlike many versions before it, you couldn't just load and hit f9 and everything would work as before, no render preset would fix and old asset, and lighting set up in the old way...

JohnMarchant
07-18-2018, 09:03 AM
Hi Andrew,

Sorry did not know it was you there :):). Oh they would only be maybe 3 interior and 3 exterior, using IES Lights, Enviroment lights and such, of course you cant hit all the possibilities but some. The main point would have been to help out those that were new to LW but also those of us that are new to LW 2018. Your videos did a very good job of expaning these differences and i have always recommened them to others if they had asked. Many of the old LightWavers wanted to dive in straight away and see what all the fuss was about and to be honest i dont read the manual for every new itteration of Adobe After Effects or Photoshop, i learn by trial and error and then read the manual later in slow time or as a reference. How many people read the Apple terms and conditions every time its updated and at the last count it was i think 50 pages, not many i will wager but we should.

Agreed the learning materials were very spartan in early January which is why yours and Attiis were invaluable, also Craig Moinins YT videos were good as well. It would have shut down some of these threads here and in other places slagging of the 2018 renderer like it was crap when in reality it was us that needed to unlearn and relearn how LW works. I still see threads talking about fireflies and such and maybe 30 mins of so of your video was enough for me to get a clean render with no fireflies at all.

Its the users fault without a doubt, its not like it could not deliver what it was meant to, or that it was seriously stuffed internally, i just think NT/LWG3D could have helped a little bit more than they did. The Manual is superb and because its online it can be kept up to date with changes and such, also some good tutorials in there as well. I just wish it were possible to download a snapshot of it, of course its only as in date as the snapshot but it would be good for reading on a train or airplane journey if you dont have internet access.

Agreed old assets set up the old way are never going to work in the new renderer and look the same, the only thing in the old assets worth saving is the model, maybe the animation and UV's and texturing. The rest of the setup needs to start from scratch.

raymondtrace
07-18-2018, 09:33 AM
I just noticed that Ben added the sample scene to the noise reduction documentation in late May (thanks!). https://docs.lightwave3d.com/display/LW2018/Removing+Noise+workflow

The best way for people to learn the new rendering is through controlled exercises with sample scenes. Then they can apply that knowledge to their own scenes.

jwiede
07-18-2018, 01:36 PM
The best way for people to learn the new rendering is through controlled exercises with sample scenes. Then they can apply that knowledge to their own scenes.

If everyone learned adequately through unsupervised experimentation, that'd be a reasonable position. Alas, many do not, and as paying customers, still need to be educated.

jwiede
07-18-2018, 02:51 PM
Agreed the learning materials were very spartan in early January which is why yours and Attiis were invaluable, also Craig Moinins YT videos were good as well.

In terms of what Newtek provides with/for LW2018, though, the learning materials are still quite thin. Saying "it's a complete re-learn" and then not providing extensive beginner walk-throughs, tips and tricks, and significantly more troubleshooting guidance, isn't "dropping the ball", it's deflating the ball and throwing it away. If Newtek had significantly enhanced what's there in the seven months since release, this would be a non-issue, but that hasn't happened.

The whole point of the new docs system was to support significant additions of multimedia content, yet the additions to date have been minor. Even if that only meant providing workflows and walk-throughs that relied heavily on third-party content and resources, that still would have at least been an attempt. We were told a primary criteria of the new docs system is that it makes dynamic changes and substantial additions much easier to do. This is the exact kind of scenario where such significant additions are needed, and seven months is more than enough time to produce such content in the new system.

So why hasn't that occurred yet, and when will it occur?

raymondtrace
07-18-2018, 04:55 PM
If everyone learned adequately through unsupervised experimentation, that'd be a reasonable position. Alas, many do not, and as paying customers, still need to be educated.

That comment may be contrary to what was quoted. I'm not suggesting unsupervised experimentation. The recent addition of the sample scene file at https://docs.lightwave3d.com/display/LW2018/Removing+Noise+workflow is appreciated because it gives PBR newcomers a common point of reference. That sample scene is the educational tool that was missing in January. Customers can work along with the steps in that guide, on a common scene file, and see how each step in the flowchart affects the scene.

madno
07-19-2018, 12:01 AM
Even though I found the doc very helpful to understand the basics of the new sample concept I think it could be be more helpful if there would be added example / tut chapters regarding the most asked questions.

If you search the doc for "fireflies" you get one hit - and that one tells you, you can use point lights as firelies (meaning those firelies you want to have e.g. as glowwing sparks from a camp fire). Numerous people had / have questions about render fireflies but don't find much help in the docs - and I think I have not seen a helpful post from a LW3D employee as well.

Same with interpolated GI. A lot of questions about that with respect to indoor sceenes. As it seems there is a commen sense now, that LW 2018 is not good for GI interpolated indoor scenes. It would be great if there would be a chapter in the docs showing the opposite.

On the other hand we have, at least in my opinion, good information about noise. There is a nice chapter about how to analyse and deal with noise.

jwiede
07-20-2018, 01:19 PM
There is a nice chapter about how to analyse and deal with noise.

I disagree somewhat. The problem with that article is that it focuses initially on increasing camera and light samples to resolve noise, yet those settings are also most likely to lead to unacceptable render time increases (which isn't really mentioned in context). In fact, by their prominence, it suggests they should be the "first line of attack" in solving noise issues, when in reality they should really be among the last resorts (because of the severe negative impact raising them has on LW2018's render performance).

It's the general lack of "holistic" education on sampling -- which settings are less harmful than others w.r.t. render performance, etc. -- that represents one of the most significantly under-documented "genres" of the LW2018 docs. Type "sampling" into the docs' search field. The number one result is a tiny blurb in an appendix on "Sampling BG and GI" which is a one-sentence note about how backdrop* sampling settings interact. Number two is another appendix one-paragraph blurb about MIS sampling. The noise article we're discussing is number three. None of the subsequent articles offer any sort of holistic explanation about sampling in LW2018 either.

I've run across at least a dozen other similar cases where the docs currently make minimal-to-no effort to explain fundamental aspects on LW2018's systems' configuration and operation. Focusing almost exclusively on how those configurations and operations differ from prior versions is not the same thing. That kind of differential documentation represents secondary notational content, at best, not as the primary information provided by the docs on those systems. Whatever documentation testing occurred, it is also painfully obvious that new / inexperienced users were either not adequately represented in that testing, or their feedback was not given adequate weight/priority.

For example, I am not suggesting the docs are responsible for teaching the reader the generic basics of how renderers operate internally or similar level of explanation (though it's worth noting that most other renderer's docs actually do provide that information as well). I absolutely DO believe it is the docs' responsibility to educate readers on how the LW2018 renderer's configuration/parameters and operation MAP to those broader concepts, as well as provide a critically-important detailing of how those specific configuration/parameters and operations of LW2018's renderer interact. It is critical that the user understand how those elements affect systemic efficiency and performance, and how to optimize same for their own needs. Unfortunately, most of that information remains MIA.

This a widespead, very frequent, and (IMO) serious problem with the LW2018 docs. They intrinsically presume the reader is an experienced LW user who is familiar with how LW2015 and prior operated, as well as presuming the reader has a thorough a priori understanding of how LW2018's functions (as if, say, a beta-tester). Unfortunately, for customers who weren't beta-testing LW2018 for years and discussing with developers how and why it behaved the way it did (aka most customers), we do NOT have that knowledge, nor is it provided by the current documentation.

If there had been any significant effort in the seven months since release to provide the missing basic operational and behavioral information for LW2018, this wouldn't be an issue. Unfortunately, most of the content actually added in that period amounts to yet more articles and content reliant on readers having extensive a priori understanding of LW2018.

I'm not saying nor believe they need to fix those fundamental omissions overnight, but I do think seven months is more than enough time that customers should see significant improvements in that direction -- that hasn't happened, and I view that as a serious problem (as well as a serious discouragement to potential new and upgrade customers). Documentation of fundamentals needs to receive much greater prioritization, and (quite likely) participation in documentation efforts and/or staffing need to be increased as well.



*: BTW, using "BG" (presumably meaning "background") in the article title, but then using "backdrop" in the article itself, is confusing and unclear. These are the docs, they shouldn't be making a priori assumptions that readers automatically understand that "BG" or "background" is equivalent to "backdrop" in context, particularly in this context. The number two article on "Light MIS Samples" has similar expectations that readers have a priori knowledge (not available in the docs)