Is adding supporting loops to control shading behaviour/normal map gradients a good practice?

Macebo

Hello everyone.

So i am at the low poly stage and in order to capture well this kind of detail



I usually go for a single smothing group, like here:



But this results in bad gradients as you can see.

From what i've read, when mipmapped, normal maps with too much gradients results in artifacts.
So to solve it i added a loop surrounding that part: 



It gets rid of the ugly shadows.

So my question is: is this a common practice in production, so i can keep doing it? Is there another way to overcome this issue? Splitting smothing groups is not an option, because from what i've gathered from my tests it doesn't capture well the detail i mention earlier.

Or am i being too paranoid about performance optimization in 2017?

Thanks.

musashidan

Gradients can cause mipping issues but with the modern averaged mikkt method gradients are expected. Adding supporting edges to control shading is a common method so really it's the usual case of running tests and seeing what works. I've yet to encounter any mipping errors from the averaged mikkt method and I've been using it exclusively for about 2 years.

what I would say is that the extra geo to capture the small cutout detail(that's causing the shading error) is unnecessary. The normal map will pick that up.

ZacD

I agree with musashidan, there's a lot of extra geometry. You could easily get by with 3 circle loops, let the normal map do it's job. and face weighting the normals will get rid of gradients if it is causing issues. 

Macebo

@musashidan & @ZacD  Thanks for your time. I believe that adding an extra loop is actually my case. I got rid of one extra loop. It looked great



But only at that  angle... From this one, because of the averaged normals in that area, it looks like this



I didn't know that we can use face weighted normals for a high to low workflow, i thought it was somethign apropriate only for mid poly stuff :O

Macebo

Lol just ignore my latest issue, it's a completely different issue than averaged normals. It's waviness due to lack of geometry