PseudoTrap missing a bump/cut

[ianna]

It looks like a bug: PseudoTrap with a negative radius (a subtraction) is represented as a Trapezoid now:

[MarkusFrankATcernch]

Could you please post the constructor arguments for this shape?
Thanks!

[ianna]

PseudoTrap(dx1, dx2, dy1, dy2, dz, r, atMinusZ)

 <PseudoTrap name="YE1_b" dx1="0.293734*m" dx2="1.86356*m" dy1="0.3000*m" dy2="0.3000*m" dz="2.92934*m" radius="-1.1350*m" atMinusZ="true"/>

[MarkusFrankATcernch]

Then I think we have a problem:
On October 29 I made a commit on your request fixing:

36d4b01#diff-7219d47bc4ab7516e0ca6c4f35f2602f

There I implemented the code you gave me:
https://cmssdt.cern.ch/lxr/source/Fireworks/Geometry/src/TGeoMgrFromDdd.cc#0390
This was correct at the time of commit and you closed the issue.

Now the new source how it should be implemented:
https://cmssdt.cern.ch/lxr/source/SimG4Core/Geometry/src/DDG4SolidConverter.cc#0362
This one is different from the above.
The first solution is wrong and only this one is valid.
It happens to be that this was the implementation, which was present PRIOR to October 29th.

I conclude you have a conflict in the CMS code having 2 different translation mechanisms
for the PseudoTrap shape which are incompatible.

You first have to solve this, otherwise we are always in trouble only depending from which side you come.....

[MarkusFrankATcernch]

I convinced myself that the old solution (before October 29th) was correct.
This will also solve this very problem. I am convinced that https://cmssdt.cern.ch/lxr/source/Fireworks/Geometry/src/TGeoMgrFromDdd.cc#0390 is buggy. You will need very good arguments to go back to this one.

[ianna]

Yep, the PseudoTrap shape with a positive radius (a union) is also wrong:

[MarkusFrankATcernch]

Which version are you using ?
For me it seems to work:

[ianna]

@MarkusFrankATcernch - sorry to disappoint you, but the bump/cut is along a wrong axis, and also a wrong precision which leaves some debris:

[MarkusFrankATcernch]

Ok. Please send me the XML parameters of the 4 cases together with the corresponding pictures.
Otherwise it is only a try-and-error session.
Thanks!

[ianna]

I guess, there is no surprise that we have the same problem with translation to Root - the bite is lower then expected (see red outline) and is leaving a flat surface which is detected by an overlap checker. I need to check what happens in the G4 builder. As you pointed out, that code is different.

  <SolidSection label="testPseudoTrapSolids.xml">
    <PseudoTrap name="ptrap1" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="10*cm" atMinusZ="false"/>
    <PseudoTrap name="ptrap2" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="10*cm" atMinusZ="true"/>
    <PseudoTrap name="ptrap3" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="-10*cm" atMinusZ="false"/>
    <PseudoTrap name="ptrap4" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="-10*cm" atMinusZ="true"/>
  </SolidSection>

[MarkusFrankATcernch]

From left to right:

red:  <PseudoTrap name="ptrap1" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="3*cm" atMinusZ="false"/>
green:<PseudoTrap name="ptrap2" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="10*cm" atMinusZ="true"/>

blue: <PseudoTrap name="ptrap3" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="-3*cm" atMinusZ="false"/>
yellow: <PseudoTrap name="ptrap4" dx1="10*cm" dx2="3*cm" dy1="30*cm" dy2="10*cm" dz="30*cm" radius="-10*cm" atMinusZ="true"/>

I reduced for red, blue the radius so that the effect is better visible on the narrow ends.
To my understanding this is correct.

The effect you observe is real and has an easy explanation:
If the radius is such that you cut out a half-cylinder with radius=dx1 (10cm) or dx2 < dx1 (cube-ish trap), then what you cut out of a trap is always wider than the trap. This is pure math: the derivation of the trap side is (2*dz)/(dx1-dx2) whereas for the cylinder's circle it is infinite i.e. the tube at the edges is always outside the trap unless the trap is cube-ish. See the right figure, where the radius is increased so that the trap derivation is smaller than the one of the tube: then the edges hit the surface....

[MarkusFrankATcernch]

Hi Yana,
Could you please comment on the above to keep the loop active?
Thanks!

[ianna]

The idea is that a circle line goes through the trapezoid vertices.

[MarkusFrankATcernch]

The circle goes through the trap vertices.
In the case that r==x1 and x2<x1 the circle comes from the outside and consequently cuts out this piece.
If I set x1=x2 then the circle comes from the inside due to the above argumentation:

[ianna]

if x1!=x2 the centre of a circle should be lowered, so that the edges would still touch the plane

[ianna]

Close for now - I'll come back to it if the geometry validations reports it as an issue.