How to model SIW in HFSS
时间：03-31
整理：3721RD
点击：
Hi all,
I'm trying to model a length of SIW and extract the modal cutoff frequencies from the propagation constants but every time the software returns the rectangular cutoffs. I can't find my error, hoping somebody has had experience of this and can point me in the right direction.
I've tried a number of more complicated setups but the basic method is to create a box of substrate, e.g. duroid with epr=2.2. Placing a waveport on each end and then performing a discrete frequency sweep in the driven modal regime returns a cutoff spectrum that agrees with sqrt[(npi/a)^2 + (mpi/b)^2] i.e. conventional rectangular waveguide modes. I'm plotting Im[gamma] for each mode in port 1 then exporting the data and using mathematica to extrapolate the ideal cutoff eigenvalues by fitting Im[Sqrt[k_c^2 - (2pi f 10^9 sqrt(mu_r ep_r)/c)^2]] (I'm working with eigenvalues, HFSS outputs frequencies).
Placing a radiation boundary on the right and left walls and leaving the top and bottom alone - they are part of the 'outer' boundary and thus get treated as PEC - returns the parallel plate modes (mpi/b), including TEM, as expected.
To change this structure to an SIW one I have to make the walls explicit. Planes are attached to the longitudinal walls and assigned PEC boundaries then I either create a series of boxes I can subtract from the waveguide, leaving rectangular gaps in the side walls or draw it as two planes in union with a bunch of cylinders for the side walls - the cylinders cause bigger meshes and longer times and the equivalence formula is based upon strips anyway so it should yield similar results.
I'm using 'Che - analytical equivalence between SIW and rectangular waveguide (2008 )', which claims that an SIW guide with transverse centre-centre width a, via radius R and longitudinal via centre-centre spacing W is equivalent to a conventional rectangular guide with width a_eff, where a_eff is a function of a, R and W. As such, with a=10mm, R=0.25mm and W=1mm such that R/W=0.25 the cutoff spectrum should be largely the same for the SIW and conventional guides. With twice the spacing the R/W ratio falls to 0.125 and the equivalent width decreases to around 9.3mm. From 15Ghz in air the TE10 cutoff ought to rise to nearly 16 (in ep_r=2.2 it goes from 10 to 10.8 ). In eigenvalues it should go from 100pi to around 337 rad/cm.
What happens is that the Im[gamma] plot remains unchanged, as though it were only simulating the waveport itself... Looking at the field overlays on the waveport it goes through the modes in order TE10 TE20 TE01 TE11, whereas the TE01 and TE11 modes shouldn't propagate.
Plots of S21 are similarly unaffected, showing the onset of propagation at exactly the same point. I've tried attaching a block of PEC to the outside of the waveports, I've tried feeding the SIW with a short length of conventional waveguide. I embed the structure in varying lengths of substrate from several wavelengths down to just one or two and various widths from the size of the SIW to several times the width. I've used planar and 3D walls, scaled up the via radius and spacing, added PML boundaries, airboxes and explicitly set all the BCs. I've tried oversized waveports, a la microstrip, but as the field is enclosed by the SIW guide this doesn't feel right and introduces extra problems with the boundaries it touches. Field plots inside the substrate look like they should.
If anyone's successfully plotted the propagation constant of SIW guides with R/W != 0.25, or can spot the flaw in my models here, I'd be enormously grateful. I've attached the latest version of the file I'm playing with, It's a bit long (200 mm with 0.5 mm features) because I thought the ports were directly coupling, but it can bear to be cut down a little so it doesn't take forever to simulate. The geometry is all defined as project variables, z for the length.
Many thanks for any advice!
I'm trying to model a length of SIW and extract the modal cutoff frequencies from the propagation constants but every time the software returns the rectangular cutoffs. I can't find my error, hoping somebody has had experience of this and can point me in the right direction.
I've tried a number of more complicated setups but the basic method is to create a box of substrate, e.g. duroid with epr=2.2. Placing a waveport on each end and then performing a discrete frequency sweep in the driven modal regime returns a cutoff spectrum that agrees with sqrt[(npi/a)^2 + (mpi/b)^2] i.e. conventional rectangular waveguide modes. I'm plotting Im[gamma] for each mode in port 1 then exporting the data and using mathematica to extrapolate the ideal cutoff eigenvalues by fitting Im[Sqrt[k_c^2 - (2pi f 10^9 sqrt(mu_r ep_r)/c)^2]] (I'm working with eigenvalues, HFSS outputs frequencies).
Placing a radiation boundary on the right and left walls and leaving the top and bottom alone - they are part of the 'outer' boundary and thus get treated as PEC - returns the parallel plate modes (mpi/b), including TEM, as expected.
To change this structure to an SIW one I have to make the walls explicit. Planes are attached to the longitudinal walls and assigned PEC boundaries then I either create a series of boxes I can subtract from the waveguide, leaving rectangular gaps in the side walls or draw it as two planes in union with a bunch of cylinders for the side walls - the cylinders cause bigger meshes and longer times and the equivalence formula is based upon strips anyway so it should yield similar results.
I'm using 'Che - analytical equivalence between SIW and rectangular waveguide (2008 )', which claims that an SIW guide with transverse centre-centre width a, via radius R and longitudinal via centre-centre spacing W is equivalent to a conventional rectangular guide with width a_eff, where a_eff is a function of a, R and W. As such, with a=10mm, R=0.25mm and W=1mm such that R/W=0.25 the cutoff spectrum should be largely the same for the SIW and conventional guides. With twice the spacing the R/W ratio falls to 0.125 and the equivalent width decreases to around 9.3mm. From 15Ghz in air the TE10 cutoff ought to rise to nearly 16 (in ep_r=2.2 it goes from 10 to 10.8 ). In eigenvalues it should go from 100pi to around 337 rad/cm.
What happens is that the Im[gamma] plot remains unchanged, as though it were only simulating the waveport itself... Looking at the field overlays on the waveport it goes through the modes in order TE10 TE20 TE01 TE11, whereas the TE01 and TE11 modes shouldn't propagate.
Plots of S21 are similarly unaffected, showing the onset of propagation at exactly the same point. I've tried attaching a block of PEC to the outside of the waveports, I've tried feeding the SIW with a short length of conventional waveguide. I embed the structure in varying lengths of substrate from several wavelengths down to just one or two and various widths from the size of the SIW to several times the width. I've used planar and 3D walls, scaled up the via radius and spacing, added PML boundaries, airboxes and explicitly set all the BCs. I've tried oversized waveports, a la microstrip, but as the field is enclosed by the SIW guide this doesn't feel right and introduces extra problems with the boundaries it touches. Field plots inside the substrate look like they should.
If anyone's successfully plotted the propagation constant of SIW guides with R/W != 0.25, or can spot the flaw in my models here, I'd be enormously grateful. I've attached the latest version of the file I'm playing with, It's a bit long (200 mm with 0.5 mm features) because I thought the ports were directly coupling, but it can bear to be cut down a little so it doesn't take forever to simulate. The geometry is all defined as project variables, z for the length.
Many thanks for any advice!