Ray Tracer half transparent mirror

norbert_walter937 · September 2017

Hi, I'm using the Ray Tracer in Onshape for optical simulation. In my situation I need a half transparent mirror with transmission and reflection. In the Ray Tracer settings I can only set the reflection or the transmission. I need booth. It is possible to set both values? Or is there an other solution?
Thanks for all answares.

philip_thomas · September 2017

Of course, you could add two instances of the mirror and define one as transmissive and one as reflective.
Once you have done that, fire a ray - when i tested it, my document found the reflective one first.
Next (parametrically) delete the reflective mirror and then fire the same ray again.

I hope this is what you were looking for . . .

Image: https://us.v-cdn.net/5022071/uploads/editor/5t/clxacb3pdrpl.png

https://cad.onshape.com/documents/f5df8e6ed86348e905aafc66/w/026594bfb959f8fefb55f348/e/fbda7538c47ed7e594efcd37

ilya_baran · September 2017

This is not currently possible with that feature -- its source code would have to be modified to support this.

norbert_walter937 · September 2017

Thanks for your answare. The Ray Tracer is very interesting for me. Specially in combination with mecanical designs. In this case I can check the system functions over the completely design. I d'ont need a full function Ray Tracer. Easy functionality is enough. For a better integration is following necaessary:

transmission and reflection at same time
easy light source generator with multible beam array
roughness as property for objekts (for different reflection directions)
wave length as property (refraction calculation)
easy lens generator (nice to have)

Are this topics possible in a future version?

I think this features are necessary and usefully also for education.

ilya_baran · September 2017

The ray tracer is a sample custom feature developed by @maximilian_schommer while he was a summer intern with us -- it is not a part of Onshape's core functionality and so we do not have resources we can allocate to its development. If you are comfortable with code, you can add the functionality you describe yourself or you may be able to interest someone in the community (or may be an Onshaper on their own time) to build it.

philip_thomas · September 2017

Of course, you could add two instances of the mirror and define one as transmissive and one as reflective.
Once you have done that, fire a ray - when i tested it, my document found the reflective one first.
Next (parametrically) delete the reflective mirror and then fire the same ray again.

I hope this is what you were looking for . . .

https://cad.onshape.com/documents/f5df8e6ed86348e905aafc66/w/026594bfb959f8fefb55f348/e/fbda7538c47ed7e594efcd37

norbert_walter937 · September 2017

Thanks for yours answares.
@ilya_baran : I understand the situation. At the moment I have no experiences with code developping in Onshape. I'll try it soon.
@philip_thomas : This is exactly what I need. The solution is a usefully nice trick. Thanks :-)

maximilian_schommer · September 2017

Hi @norbert_walter937

The ray tracer originally did give both the reflected and refracted ray, and it would be pretty simple to make a check box so that functionality is added to it again.

I'll see if I can get to it soon! Thanks for the suggestion!

Maximilian Schommer

philip_thomas · September 2017

Perhaps to avoid a lot of unnecessary rays it would be more helpful to add the 'both' as a characteristic of the body and not of the ray!

maximilian_schommer · September 2017

Thanks @philip_thomas and @norbert_walter937 for the suggestions!

I was able to implement the semi-silvered mirror functionality, which appears in the setup as being able to check a box for reflection and refraction for each part selected. The other suggestions about roughness, wavelength, etc. are great ideas, and I'll hopefully get around to them soon! Release 4 is where the new functionality lives.

Maximilian Schommer

Image: https://us.v-cdn.net/5022071/uploads/editor/yx/d4k9j6m1o84z.png

Image: https://us.v-cdn.net/5022071/uploads/editor/o5/a2ae8u5az3e0.png

philip_thomas · September 2017

It occurred to me afterwards that asking for this meant you had to deal with recursion as every bounce produced two more rays to follow!
Nicely done!

ivan_skachko · December 2017

Ray Tracer is indeed a very useful and efficient custom feature, however I encountered a bug in it. As the attached image shows, RT works well tracing a beam in the FRONT plane (the same plane where I sketched lens' profile to be revolved), but fails for the beam with identical parameters in the TOP plane. The problem seems to stem from feature script code ( var rayEndPoint = getRayIntersection(context, id, rayToTrace); ) not being able to compute the intersection of the ray with the back surface of the lens for some reason. Looking at the script for this feature I am generally curious why it takes so much code to find an intersection between a line and a surface of a body.
I understand that Ray Tracer will not be developed further by Onshape, but I wanted to bring the issue to the attention of the users.

Image: https://us.v-cdn.net/5022071/uploads/editor/lj/vask49wff6x2.png

philip_thomas · December 2017

^^^ @maximilian_schommer

philip_thomas · December 2017

@ivan_skachko - Max (the author of this custom feature) may chose to shed light on the other part of your statement, but my understanding based on discussions before Max started working on this, was that the following is true. There is no call in Parasolid to determine the intersection of a line and a body/face. For that reason, Max's code generates an artificial entity (i think its a surface), to be able to get that. I hope that puts your mind at rest

konstantin_shiriazdanov · December 2017

philip_thomas said:

There is no call in Parasolid to determine the intersection of a line and a body/face.

but what about evDistance, it can use a Line for one of the sides of measurement as stated in documentation. so you can detect if it intersects the face and evaluate normal in intersection point. or i'm missing something?

ilya_baran · December 2017

@philip_thomas Parasolid definitely has such calls. They were just not exposed to FS as is. The problem with evDistance is that it returns some intersection, not just the closest one, which is what is needed here.

Jake_Rosenfeld · December 2017

@konstantin_shiriazdanov

There are issues with the evDistance approach. If you use it to intersect a Line and a cylindrical face which it passes through twice, it will be very hard to detect the situation, and hard to gather the other intersections using evDistance alone. Also looping though every face of the selected bodies and running an evDistance could get very expensive.

Parasolid does support arbitrary intersections between curves and faces, but this is not currently exposed to FeatureScript, which is where the extrude 'up_to_next' approach came from. It's an easy built-in way to find the next intersection from a location to a set of bodies.

philip_thomas · December 2017

I love it when all the smart people turn up to the same party

konstantin_shiriazdanov · December 2017

Jake_Rosenfeld said:

Also looping though every face of the selected bodies and running an evDistance could get very expensive.

my Fold FS for unfolding arbitary extruded surface evaluates evDistance every 0.5 mm along the bounding edges of the face. that what should be called "very expensive"

the ambiguity of evDistace in the case of several possible intersections seems to be the main problem

philip_thomas · December 2017

@konstantin_shiriazdanov - wait, did you write a flattener????
I am very excited (i want one

) - is it public?

Many thanks - Philip

konstantin_shiriazdanov · December 2017

hi @philip_thomas
they are all in the same document, because they share a little library of functions for curve transforms, the one you want is Fold FS with fold/unfold options. It works with rather common extracted surfaces but makes some geometry distrosions (about 0.5 % of length).
https://cad.onshape.com/documents/0bb13c1b6ed6d4a6dd75cf99/w/b4493d47a45c27ce485c84b9/e/1299ae499a697e3b28bca1e4

maximilian_schommer · December 2017

ivan_skachko said:

Ray Tracer is indeed a very useful and efficient custom feature, however I encountered a bug in it. As the attached image shows, RT works well tracing a beam in the FRONT plane (the same plane where I sketched lens' profile to be revolved), but fails for the beam with identical parameters in the TOP plane. The problem seems to stem from feature script code ( var rayEndPoint = getRayIntersection(context, id, rayToTrace); ) not being able to compute the intersection of the ray with the back surface of the lens for some reason. Looking at the script for this feature I am generally curious why it takes so much code to find an intersection between a line and a surface of a body.
I understand that Ray Tracer will not be developed further by Onshape, but I wanted to bring the issue to the attention of the users.

@ivan_skachko

Thanks for the bug find! I'll take a look at it when I have time! The reason that it takes so much code to find the intersection of a ray and a surface is because there is no function in onshape that lets you calculate that uniquely. I'm using a workaround, where to "project" the ray, I'm extruding a very thin surface, and finding where the surface intersects another body. It's a bit strange, but it's the only way I know of to trace rays reliably. There have been strange bugs I've had to deal with such as rays intersecting surfaces at locations like the center point of a revolve, where a normal isn't properly defined, and I'll take a look at this bug and see what the issue is. Thanks again for pointing it out!

Max

philip_thomas · December 2017

@konstantin_shiriazdanov - THANK YOU - i am thrilled!

Image: https://us.v-cdn.net/5022071/uploads/editor/dw/3w7korukcazb.png

konstantin_shiriazdanov · December 2017

@philip_thomas
i should specifically mention - this fs was made to make correct flats from faces of surfaces that might be obtained by extraction operation. and the face should be situated to the one side of the "profile edge". in your case i see that you apply it to the lofted face. if the resulting region mathches original face by area and length of bounding edges I would be surprised a lot

philip_thomas · December 2017

@konstantin_shiriazdanov - hmm

So the difference in surface area is 2.8% in this case.

My apologies for not understanding what you mean by - "the face should be situated to the one side of the "profile edge"".
Would you please explain?
Is there anything else i can do to improve accuracy?
I would like 1% or better.

Many thanks - Philip

konstantin_shiriazdanov · December 2017

philip_thomas said:
"the face should be situated to the one side of the "profile edge"

I meant that when you select face or edge to transform and then generatrix profile edge the generatrix profile shouldn't cross any edge you need to flatten (but it may be one of the bounding edges) because internally it doesn't differ the direction since evDistance return unsigned value.

>Is there anything else i can do to improve accuracy?
sorry, can't think about something suitable right now, it uses numerical algorhytms of evDistance and of finding length parameter and seem like most numeric mistake is in x coordinate which is evaluated by my numeric function. if i find how to improove it i'll tell you.

philip_thomas · December 2017

@konstantin_shiriazdanov - Please accept my apologies - your English is 100 times better than my Russian (Ukrainian?), but i have no idea what you just said!

I think you said that when flattened, the profile edge should not cross the surface being flattened.
My apologies again if this is way off.

As it stands right now, this custom feature is a huge help and I am very grateful!

Thank you for all your support in the forums

konstantin_shiriazdanov · December 2017

@philip_thomas
sorry, that i'm training my english on you)

philip_thomas said:
I think you said that when flattened, the profile edge should not cross the surface being flattened.

that is it! an example of what happens when selected profile edge cross flattening entities

Image: https://us.v-cdn.net/5022071/uploads/editor/06/ucimskpj1n82.png

philip_thomas · December 2017

@konstantin_shiriazdanov - got it - thank you!

MBartlett21 · December 2019

maximilian_schommer said:
[...]
Thanks for the bug find! I'll take a look at it when I have time! The reason that it takes so much code to find the intersection of a ray and a surface is because there is no function in onshape that lets you calculate that uniquely. I'm using a workaround, where to "project" the ray, I'm extruding a very thin surface, and finding where the surface intersects another body. It's a bit strange, but it's the only way I know of to trace rays reliably. There have been strange bugs I've had to deal with such as rays intersecting surfaces at locations like the center point of a revolve, where a normal isn't properly defined, and I'll take a look at this bug and see what the issue is. Thanks again for pointing it out!

Max

@maximilian_schommer

Here is an edited version of your FeatureScript that uses evRaycast rather than extrude up-to-next.

FeatureScript 1174;
import(path : "onshape/std/geometry.fs", version : "1174.0");


annotation { "Feature Type Name" : "Ray Tracer" }
export const rayTracer = defineFeature(function(context is Context, id is Id, definition is map)
    precondition
    {
        annotation { "Name" : "Origin Point", "Filter" : EntityType.VERTEX, "MaxNumberOfPicks" : 1 }
        definition.origin is Query;

        annotation { "Name" : "Direction Point", "Filter" : EntityType.VERTEX }
        definition.directionPoint is Query;

        annotation { "Name" : "Maximum Depth", "UIHint" : UIHint.REMEMBER_PREVIOUS_VALUE }
        isInteger(definition.maxDepth, POSITIVE_COUNT_BOUNDS);

    }
    {
        const bodies = qAttributeFilter(qBodyType(qEverything(EntityType.BODY), BodyType.SOLID), {});
        const directionPoints = evaluateQuery(context, definition.directionPoint);
        const originQ = definition.origin;
        const patternTransformOrigin = getRemainderPatternTransform(context, {
                "references" : originQ
        });
        const patternTransformDirections = getRemainderPatternTransform(context, {
                "references" : qUnion(directionPoints)
        });
        const originPoint = patternTransformOrigin * evVertexPoint(context, {
                "vertex" : originQ
        });
        for (var i = 0; i < size(directionPoints); i += 1)
        {
            var directionPoint is Vector = patternTransformDirections * evVertexPoint(context, {
                    "vertex" : directionPoints[i]
                });
            var direction is Vector = normalize(directionPoint - originPoint);

            try silent
            {
                definition.medRef = getVariable(context, "mediumRefractionIndex");
            }
            catch
            {
                throw regenError("Set Up has not been performed");
            }

            var rayToTrace = line(originPoint, direction);
            var traceRayDef = {
                "rayToTrace" : rayToTrace,
                "recursionDepth" : 1,
                "medRef" : definition.medRef,
                "maxDepth" : definition.maxDepth,
                "hitsDone" : 0,
                "bodies" : bodies
            };
            const subId = id + unstableIdComponent(i);
            setExternalDisambiguation(context, subId, directionPoints[i]);
            traceSingleRay(context, subId + "curve", traceRayDef);
            try
            {
                opExtractWires(context, subId + "extract", {
                            "edges" : qCreatedBy(subId + "curve", EntityType.EDGE)
                        });
            }
            catch
            {
                opPattern(context, subId + "pattern", {
                            "entities" : qCreatedBy(subId + "curve", EntityType.BODY),
                            "instanceNames" : ["1"],
                            "transforms" : [identityTransform()]
                        });
            }
        }
        opDeleteBodies(context, id + "deleteBodies1", {
                    "entities" : qCreatedBy(id + ANY_ID + "curve", EntityType.BODY)
                });
    });



export function traceSingleRay(context is Context, id is Id, definition is map)
{

    var rayToTrace = definition.rayToTrace;

    var rayIntersectData = getRayIntersection(context, id, rayToTrace, definition.bodies);
    if (rayIntersectData == false)
    {
        debug(context, rayToTrace);
        return false;
    }
    const rayEndPoint = rayIntersectData.intersection;
    var intersectFace = rayIntersectData.entity;

    var n1 = definition.medRef;
    var n2 = getAttributes(context, {
                        "entities" : qOwnerBody(intersectFace)
                    })[0].indexOfRefraction;
    if(!(n2 is number))
        return false;

    var isReflective is boolean = getAttributes(context, {
                "entities" : qOwnerBody(intersectFace)
            })[0].isReflective;
    var isRefractive is boolean = getAttributes(context, {
                "entities" : qOwnerBody(intersectFace)
            })[0].isRefractive;

    var parameterVector = rayIntersectData.parameter;

    var surfaceNormal = zeroVector(3);
    try silent
    {
        surfaceNormal = evFaceTangentPlane(context, {
                        "face" : intersectFace,
                        "parameter" : parameterVector
                    }).normal;
    }
    catch
    {
        surfaceNormal = evFaceTangentPlane(context, {
                        "face" : intersectFace,
                        "parameter" : jitter(parameterVector)
                    }).normal;
    }


    //Makes sure that the normal is facing the right direction for our reflection/refraction calculations
    if (dot(surfaceNormal, rayToTrace.direction) > 0)
    {
        n1 = n2;
        n2 = definition.medRef;
        surfaceNormal *= -1;
    }

    definition.hitDone += 
    if (definition.hitsDone < definition.maxDepth && isReflective)
    {
        var reflectedRay is Line = reflectRay(line(rayEndPoint, rayToTrace.direction), surfaceNormal);
        reflectedRay.origin = reflectedRay.origin + reflectedRay.direction * .000001 * inch;
        definition.rayToTrace = reflectedRay;

        traceSingleRay(context, id + "recurseReflect", definition);
    }
    if (definition.hitsDone < definition.maxDepth && isRefractive)
    {
        var refractedRay = refractRay(line(rayEndPoint, rayToTrace.direction), surfaceNormal, n1, n2);
        definition.rayToTrace = refractedRay;
        if (refractedRay != false)
        {
            traceSingleRay(context, id + "recurseRefract_R", definition);
        }
        else
        {
            if (!isReflective)
            {
                var reflectedRay is Line = reflectRay(context, line(rayEndPoint, rayToTrace.direction), surfaceNormal);
                reflectedRay.origin = reflectedRay.origin + reflectedRay.direction * .000001 * inch;
                definition.rayToTrace = reflectedRay;
                traceSingleRay(context, id + "recurseReflect_R", definition);
            }

        }
    }
}

export function jitter(paramVec is Vector)
{
    var jitterAmount = .0001; // This is so that at the very center of revolved lens, you don't throw an error.
    paramVec[0] = clamp(paramVec[0], jitterAmount, 1 - jitterAmount);
    paramVec[1] = clamp(paramVec[1], jitterAmount, 1 - jitterAmount);
    return paramVec;
}

export function getRayIntersection(context is Context, id is Id, rayToTrace is Line, bodies is Query)
{
    var rayResult = evRaycast(context, {
            "entities" : qSubtraction(qOwnedByBody(bodies, EntityType.FACE), qWithinRadius(qEverything(EntityType.FACE), rayToTrace.origin, TOLERANCE.zeroLength * meter)),
            "ray" : rayToTrace
        });
    if (rayResult == [])
    {
        opFitSpline(context, id + "endFitSpline", {
                    "points" : [
                            rayToTrace.origin,
                            rayToTrace.origin + rayToTrace.direction * inch
                        ]
                });
        return false;
    }
    rayResult = rayResult[0];


    opFitSpline(context, id + "fitSpline1", {
                "points" : [
                        rayToTrace.origin,
                        rayResult.intersection
                    ]
            });
    return rayResult;
}

export function reflectRay(rayToReflect is Line, normal)
{
    var reflectedRay = -2 * (dot(rayToReflect.direction, normal)) * normal + rayToReflect.direction;
    return line(rayToReflect.origin, reflectedRay);
}

//n1 is the material that the light is coming from, n2 is the material the light is going into.
export function refractRay(rayToRefract is Line, normal, n1, n2)
{
    var thetaI = angleBetween(-1 * rayToRefract.direction, normal);
    var radical = 1 - (n1 / n2) ^ 2 * (1 - cos(thetaI) ^ 2);
    if (radical < 0)
    {
        return false;
    }

    var v is Vector = n1 / n2 * normalize(rayToRefract.direction) + (n1 / n2 * cos(thetaI) - sqrt(radical)) * normalize(normal);

    return line(rayToRefract.origin, v);
}

PS:
It can also be patterned using feature pattern

maximilian_schommer · October 2020

Thanks @MBartlett21 , I just incorporated the adaption! The new release works on my test cases, but if anyone could let me know if they have issues with the new code that would be really helpful!

https://cad.onshape.com/documents/258a19506c555f400614c944/w/74e657ae2800105e2874589a/e/d7167c2f04d1012df61f2aeb

Max

Ray Tracer half transparent mirror

Best Answer

Answers

Howdy, Stranger!

Quick Links

Categories