detail enhance on low irmap

[glyph]

I've been revisiting an old flythrough I did to get a better version for my online portfolio.
It's an office interior with several room areas, a fair amount of leather chairs, patterned glass, frosted glass and shiny metal ceiling light fittings with light mat tubes (those lights are particularly slow to render and still have a small bit of noise).

Anyway I've found that IRMap at 'low' settings, with detail enhancement on, lightcache at World scale at 3000 subdivs, glossies in lc, and then extra Ambient Occlusion from the global settings, is giving me a nice clean result at ok speeds.
(render is c 500 x 400)

I just want to check if this is a good method in theory or if the settings are going to be fighting each other?
Is detail enhancment with Low IRmap and added AO going to be similar and faster than medium IR Map and no detail enhance?
I've done loads of tests but lost track of it a bit!

[super gnu]

if youre talking about a precalculated imap, then it will be faster to use only a decent imap ( although a finely detailed one can use lots of ram rendering)

-this is simply because you only need to do a lighting calc every 25th ish frame. even if it takes a few hours to calculate, its still faster than having to do something every frame (i.e detail enhancement)

in some cases a low imap and det.enh can produce a good result, but ive found on interiors particularly, a low set of min and max values tend to miss fine details entirely and you get odd artifacts in the imap around shadow gaps etc. youd think simply increasing the detail enhancement radius would fix this, but i dont think it decides very intelligently where to apply it. it will do detail enhancement inside the shadow gap, but on either side of the shadow gap, since its a convex surface, your stuck with your low detail imap.

this is what i have found anyway.

for example ive had a scene with a clock on a wall. with pure brute force i get beautiful shadows around it from light bouncing off a nearby wall. these stretched about a metre from the clock, which was quite small.

this was very hard to recreate with the imap, so i tried detail enhancement. even with the radius set larger than the entire scene, i was unable to get the same nice shadows. it seems to only do the effect where ao would be applied, which isnt always where you would benefit from it.

i prefer a super sharp ( low interp. samples, and very high hsph subdivs) imap, with say -4, 0 for the min and max. this gives you nice sharp details and the speed advantage of a cached lighting solution.

[glyph]

Yes I'm using precalc lc and imap - I did have d.e. on for the precalc.. is d.e. only applied at final render?

Thing is, I was doing tests at 'medium' imap settings with d.e. off and these results are much better -
but that might be because I was previously doing lc at :
a) screen mode
b) world mode but with too small a radius so had to lower the subdivs a lot to stop crashes..
- both ended up with undefined detail
- now I have plenty of subdivs using a reasonable radius of 100mm

but doing a single frame test doesn't replicate what happens when you calc a flythrough LC !
In the past I often used custom settings on the imap but the presets seem ok for this resolution.

[super gnu]

i dont understand exactly why vray needs the imap to be calculated specially for d.e., but the actual lighting calculation part of the d.e. is done at rendertime, per frame. you can see this easily by comparing rendertimes.

this is the thing about vray though.. maybe in your particular case d.e. works particularly fast, so it becomes the right option. in my experience though, it slows down rendering a lot and never gives a result that id like ( the sharpness of brute force with much lower rendertimes)

id not think the lightcache has much to do with it. as long as its relatively smooth and clean ( you can check by doing a render with lc/lc) you should be ok, its only used for the secondary bounces, (as a basis for the finely detailed primary bounce.)

youre more likely to find its your interp. samples for the imap. the defaults do a good job of compromising between detail and hiding blotchiness, but if you want sharp lighting from imaps, you have to reduce the interp. samples, and then increase the hsp. subdivs to compensate fro the increased visibility of the blotches.