Surface Shader lighting independent of geometry

[sAmmY17]

Hi there,

I´m trying to render a sphere by some simple raytracing inside a cube mesh using a surface shader and shade the sphere correctly. This is the shader I build for this:

Shader "Custom/SurfaceSDF"
{
    SubShader
    {
        Tags {"RenderType" = "Opaque" "Queue" = "Geometry"}
        LOD 200

        CGPROGRAM

        #pragma surface surf NoLighting
        #pragma target 3.0

        struct Input
        {
            float3 worldPos;
        };

        UNITY_INSTANCING_BUFFER_START(Props)
        UNITY_INSTANCING_BUFFER_END(Props)

        fixed4 LightingNoLighting(SurfaceOutput s, fixed3 lightDir, fixed atten)
        {
            return fixed4(s.Albedo, s.Alpha);
        }

        float intersect(in float3 ro, in float3 rd, in float3 ce, in float ra)
        {
            float3 oc = ro - ce;
            float b = dot(oc, rd);
            float c = dot(oc, oc) - ra * ra;
            float h = b * b - c;
            if (h < 0.0) return -1.0; // no intersection
            h = sqrt(h);
            return -b - h;
        }

        void surf (Input IN, inout SurfaceOutput o)
        {
            float3 rd = normalize(IN.worldPos - _WorldSpaceCameraPos);
            float d = intersect(_WorldSpaceCameraPos, rd, float3(0, 0, 0), 1);
           
            clip(d);

            float3 hit = _WorldSpaceCameraPos + d * rd;
            float3 norm = normalize(mul(unity_WorldToObject, hit - float3(0, 0, 0)));
            o.Normal = norm;
            o.Albedo = o.Normal*0.5 + 0.5;
            o.Alpha = 1;
        }
        ENDCG
    }
    FallBack "Diffuse"
}

In general, this works and gives a correctly rendered sphere with correct normals when I use the NoLighting light model as you can see in the first image. But when I use the Lambert or Standard lighting model, the edges of the Cube give strange shading effects, as you can see in the second image.

So my question is, if the underlying mesh somehow has an effect on the lighting and not only the computed normals? If yes how could I resolve this issue so that the sphere is shaded correctly?

Thanks for your help in advance!

[bgolus]

Surface Shaders always expect the o.Normal value to be a tangent space normal vector. In other words a normal direction relative to the mesh’s texture UV orientation. If you output world or object space normals, it doesn’t know that, and will still treat them as tangent space normals, and transform whatever value you give to it from tangent space to world space. For this shader to work you need to either edit the generated code to comment out the one line that does that transform, or convert the SDF normal into tangent space.

I have an example of doing that in my triplanar normal mapping Surface Shader example here:

bgolus/Normal-Mapping-for-a-Triplanar-Shader/blob/master/TriplanarSurfaceShader.shader

See the WorldToTangentNormalVector function. It does require the Input struct have both float3 worldNormal; and INTERNAL_DATA to work.

Be aware that this particular implementation is an approximation. If you’re using something like the default Unity cube it’ll be pretty close to the same as ground truth. If you’re still seeing weird artifacts, you may need to use the correct inverse matrix version here:

Flat lighting without separate smoothing groups?

So I spent some time pondering this and realize what the problem is. The Transform node using World to Tangent is wrong, and has always been wrong. More over, so are all other examples of transforming from world space to tangent space. So, if you have a matrix that transforms from one space to another, and you need to go the opposite way, you need the inverse matrix. The usual view of a tangent to world matrix is that it’ll be orthogonal, that is to say the normal, tangent, and bitangent are a…

[sAmmY17]

Thanks so much, that helped a lot!