HFSS loop simulation to extract inductance

Hi,

I've been practically living in this forum for some time, reviewed the HFSS book and examples I've found online. I feel that I am very close to achieving my simulation goal, but I'm stuck and would like some feedback. I apologize for the long post in advance.

I have a 1.5 turn silver loop I created by using HFSS's helix command. I did 1.5 turns so that the two ends of the loop are on opposite sides from each other (the faces of the loop ends are parallel, lie in the same plane, and are separated effectively by the diameter of my loop). I want to simulate this structure so that I can find its inductance by either:

1) L=Im(Z11)/w or
2) finding Gamma = (Zo-Zl)/(Zo+Zl) and comparing it to S11. Zo would be 50 and Zl is j*w*L

In either case, I have an L value that I found using a multi turn loop inductor calculator online. I want to see if HFSS will give me approximately the same L value at a particular frequency of interest)

Heres what I've tried for simulation setups so far:

1) I created an box with air property that is at least quarter wavelength from the loop in all directions. Then i created a long cylinder (PEC) from each of the loops ends until they reached one of the airbox's walls. I guess that this would be my "probe" for coax stimulation. I created the inner coax conductor (PEC) with the same radius as my probe on the other side of wall (outside the airbox). I then created the concentric coax dielectric layer (vacuum) , sizing its radius relative to the inner conductor to achieve 50ohm coax impedance. Since the coax is outside the airbox, I did not create an outer conductor since the coax is exposed to the background. I created the waveport on the face of the dielectric (not the face that touches the airbox wall), and assigned a terminal line from outer radius to inner radius of coax (I'm using driven terminal solution method). Of course, the length of the coax cable changes my results so I tried deembedding. I drew a deembeding vector from the coax face that has the waveport to the coax face that is touching the airbox. Lastly, I gave all six faces of the airbox the radiation BC, except that HFSS crapped out due to the two airbox walls that have the coax touching it. I took the radiation BC off these two walls and just made it an infinite PEC ground plane (not sure if this was the right thing to do). The inductance of the loop that I extracted from the resulting S and Z parameters were wrong.


2) Instead of having the coax on the outside of the airbox, I tried putting it inside. I shortened the length of my "probe" so that I could bring the coax inside the airbox. Since the coax is now inside, I created the coax's outer conductor (PEC) that I did not have before. Also, I created a PEC cap touching the waveport, so that HFSS knows to send the wave through the coax and into the loop. Again I had the deembeding vector point from one coax end (with the waveport) to the other end (touching the "probe"). I gave all six of my airbox walls radiation BC. Again, my simulations failed to give me the loop inductance I was expecting.


3) I tried then using a lumped port. With this setup, the loop was a little different. I just swept a circle around an axis to form a closed loop. But instead of actually closing the loop, i swept it like 358 degrees, so there was a gap between the two loop ends. (I hoped that HFSS would still consider this structure as a 1 turn loop). I created a 2D PEC sheet that touched the faces of the loop ends. I applied a lumped port excitation to the sheet, drawing the vector from one loop face to the other. I'm not very confident with lumped ports, so I did not know what to do next (ie. was I supposed to draw some kind of ground plane?) Needless to say, this simulation failed to give me the loop inductance I was expecting.


First of all, if anyone can comment on my three simulation setup attempts, I would appreciate that.

Secondly, I have some explicit questions that I hope can be addressed:
1) If I'm not interested in the radiation field patterns of the loop, but I just want to find the inductance, do I really need the airbox? I think I should still enclose the loop in a box of air so keep it from touching HFSS's background, but is the airbox's size and its radiation BC important in extracting inductance of loop?
2) The inner conductor of coax and the probe are both PEC. The probe connects to my silver loop. The silver/PEC boundary is not a problem, is it?
3) Should the deembeding vector be the length of my coax, or coax+probe? Also, I am deembeding from the waveport through the coax to the loop. Is this the right direction?
4) The HFSS describes a spiral inductor and uses a ground ring so to act as a return loop. Is a ground ring necessary for my loop structure? If so, how would it connect to my loop?


Thanks for your feedback.

1. You need the airbox, otherwise the pattern of fields around the inductor will be incorrect, and so will the results.

2/3. I don't like your whole probe/coax method. The probe and coax have inductance (and capacitance) of their own which gets mixed into the HFSS results (S+Z parameters). So of course the results are wrong. BTW, there is no need to draw a coax like you did - the whole idea of a "waveport" is that it acts as if an infinite transmission line, whose cross-section is the port, attached to the model where the port is located.

4. There must be a path for current to go in a complete loop, otherwise you are simulating a capacitor not an inductor. Thus the HFSS example uses a ground ring. The side-effect is that the ground ring adds inductance of its own. If your inductor has many turns, this extra inductance may be negligible. With 1.5 turns, I think not.

5. Your third simulation sounds like it could almost work. It would be a 1-turn inductor. Since it begins and ends in the same place, no ground ring is needed. Make a 2D sheet (don't define it as PEC), define it as a lumped port, and draw an integration line from one loop end to the other. I think this should give correct results.

6. To simulate an inductor with 1.5 (or any n.5) turns, avoiding the error of the ground ring inductance, I used the following method. Let's say the inductor is oriented along the x-axis (trace begins at [0,0,0], ends at [2,0,0], inductor coils centered on [1,0,0]). I made the model big in the Y and Z directions (air box). On each X face I put a waveport (centered at [0,0,0],[2,0,0]). I defined the two X faces overall as master and slave boundaries. Thus the overall model was like an infinite chain of inductors (hopefully far apart enough from each other not to mutually couple much). Because of the infinite chain, there was no need for a ground ring. The model could be resized in the X direction to include more or less of the trace in the induction calculation. I believe my results with this model were pretty accurate.