Fastest hybrid 90 coupler I have ever built

[gadiLhv]

So I started experimenting with the PCB modeler. While I had zero idea how the syntax works, it took me less than half an hour to build the hybrid 90 structure with waveguide ports. Pretty darn impressive, if you ask me!

Couple of notes and questions:

1. Where did you get the idea for PCB modelling syntax? I haven't experimented with multi-layer yet, but this is really useful.
2. Is there support for ground flooding? E.g. GCPWs and such.
3. I would also add tapering with an offset. As much as turn is cool, sometimes you want an X,Y offset, rather than calculating the degree to turn, such that you will get a smooth transition.

The code below isn't optimized or working, but it will, very soon, it seems.

import emerge as em


MSL_50_W = 0.4
MSL_30_W = 0.7

H90_Feed_L = 2.5
H90_QWL_L = 4
H90_W = 4.5
H90_Taper_L = 0.25

Sub_Thick = 20*0.0254

Port_H_Fact = 3.25
Port_W_Fact = 7.5

m = em.Simulation('Hybrid_90')



pcb = em.geo.PCB(Sub_Thick,                 # Thickness in arbitrary unit
                 1e-3,                      # meter to unit multiplier (mm)
                 em.GCS,                    # Coordinate System 
                 material=em.lib.DIEL_FR4,  # Material
                 layers=2)                  # Number of layers

# pcb.new(0,0,MSL_50_W,(1,0)).store('p1').straight(H90_Feed_L).split((1,0),MSL_70_W).straight(H90_QWL_L).store('QWL1').merge()
pcb.new(-H90_QWL_L*0.5,H90_W*0.5,MSL_50_W,(1,0)).store('p1').straight(H90_Feed_L).\
    split((1,0)).taper(H90_Taper_L,MSL_30_W).\
    straight(H90_QWL_L).\
    split((1,0)).taper(H90_Taper_L,MSL_50_W).\
    split((1,0)).straight(H90_Feed_L).store('p2')
    
pcb.new(-H90_QWL_L*0.5,-H90_W*0.5,MSL_50_W,(1,0)).store('p4').straight(H90_Feed_L).\
    split((1,0)).taper(H90_Taper_L,MSL_30_W).\
    straight(H90_QWL_L).\
    split((1,0)).taper(H90_Taper_L,MSL_50_W).\
    split((1,0)).straight(H90_Feed_L).store('p3')
    
pcb.new(-H90_QWL_L*0.5 + H90_Feed_L,H90_W*0.5,MSL_50_W,(0,-1)).straight(H90_W)
pcb.new(-H90_QWL_L*0.5 + H90_Feed_L + H90_QWL_L,H90_W*0.5,MSL_50_W,(0,-1)).straight(H90_W)

traces = pcb.compile_paths(merge=True)

wgp1 = pcb.modal_port(pcb.load('p1'), Sub_Thick*Port_H_Fact, Port_W_Fact)
wgp2 = pcb.modal_port(pcb.load('p2'), Sub_Thick*Port_H_Fact, Port_W_Fact)
wgp3 = pcb.modal_port(pcb.load('p3'), Sub_Thick*Port_H_Fact, Port_W_Fact)
wgp4 = pcb.modal_port(pcb.load('p4'), Sub_Thick*Port_H_Fact, Port_W_Fact)

pcb.determine_bounds(10,10,10,10)

diel = pcb.generate_pcb()
air = pcb.generate_air(10)

m.commit_geometry()
m.mw.set_frequency(1e9)

m.generate_mesh()

m.view()

[FennisRobert]

Its nice right :)?
Its hard to say exactly where i got it from. In general its an existing design pattern called the builder pattern. Its also called method chaining.
I used it extensively when doing advent of code.

In principle ground flooding shouldn't be too hard i think. Make a reactangle, dilate the traces and done. I'll implement the dilation from OCC soon so that you can do it manually. Then at some point ill just make the "floor plane" function.

Tiny remark: modal ports have to be on the boundary of your PCB. You probably want lumped ports

[gadiLhv]

Why do the modal ports (waveguide ports) have to be on the boundary? Because of the lack of ABC? Implementing a waveguide port ABC in FEM is pretty straight forward. If you want, I can try and help with that...

[FennisRobert]

I'm not sure if we are talking past each other. The Rectangular waveguide was the first boundary condition I implemented together with PEC. The lumped port, rectangular waveguide, absorbing boundary condition, they are all Robin Boundary Conditions. But boundary conditions that are on the outside of your domain require different parameters and setup. Modal ports, just like waveguide ports, are ports that are setup to be on the boundary. Lumped ports are computed to impose an internal impedance boundary plus a uniform excitation. You can technically put a modal port inside the domain (EMerge doesn't stop you) but it wont give you the correct physical results.

As a note: This is exactly the same in all FEM software. Comsol, HFSS they will all complain if you try to put wave ports on the inside of your domain (if they even let you).

[FennisRobert]

Also, isnt the dx dy offset variable in .straight what you are looking for maybe?

[FennisRobert]

Z-fighting can get a bit intense but... Open CASCADE is tha bomb! This can be created with:

traces = pcb.compile_paths(merge=True)
traces_out = em.geo.expand_surface(traces, 2*mm)
top_conductor = pcb.plane(pcb.z(2))
top_conductor = em.geo.subtract(top_conductor, traces_out)

Next release :)

[gadiLhv]

Should I want to implement this taper pattern

Rather than this one

What is the best way to do so?

[FennisRobert]

Right now, the best way to do it is with .straight: Say that you go from w1 to w2 then you can do:

pcb.new().....straight(length, w2, dx=taper_length, dy=(w2-w1)/2)

[gadiLhv]

Alright, alive and well.

I found that if I define the ports as such, it works:

p1 = m.mw.bc.LumpedPort(lp1,1)
p2 = m.mw.bc.LumpedPort(lp2,2)
p3 = m.mw.bc.LumpedPort(lp3,3)
p4 = m.mw.bc.LumpedPort(lp4,4)

But if I define them like this,

p1 = m.mw.bc.LumpedPort(
    lp1, 1,
    width=MSL_50_W, height=Sub_Thick,
    direction=em.ZAX,
    active=True, Z0=50
)

p2 = m.mw.bc.LumpedPort(
    lp2, 2,
    width=MSL_50_W, height=Sub_Thick,
    direction=em.ZAX,
    active=True, Z0=50
)

p3 = m.mw.bc.LumpedPort(
    lp3, 3,
    width=MSL_50_W, height=Sub_Thick,
    direction=em.ZAX,
    active=True, Z0=50
)

p4 = m.mw.bc.LumpedPort(
    lp4, 4,
    width=MSL_50_W, height=Sub_Thick,
    direction=em.ZAX,
    active=True, Z0=50
)

It doesn't. Any ideas?

[FennisRobert]

yes, this is a mistake that I make myself quite often. I'm not 100% sure if its the issue but:

You probably have MSL_50_W defined in mm. But if you pass it to LumpedPort it gets interpreted as meters. So you have to multiply them by 0.001. That is what I assume is going on. Other than that, this should give the same results. Same goes for height

[gadiLhv]

Wow... Nice catch!

Cheers