Using HFSS successfully at 77GHz Kate Moore Communications Sensors Information Systems Contract R&D | Consultancy Specialist Products
ROKE Presentation Title Using HFSS successfully at 77GHz • Data sheets give you the dielectric constant and the loss tangent Or do they? • The materials properties are rarely measured at high frequencies so the user must measure it themselves. 2
ROKE Presentation Title Material Properties Design Ø outer Ø inner a resonant structure… Track width Gap1 Gap2 Manufacture 3
ROKE Presentation Title Material Properties Measure 4
ROKE Presentation Title Material Properties Compare with HFSS predictions: vary ε r and tan δ until the measured results overlay the predictions Resonant Rings: HFSS comparison 2 0 -5 Ring 4 -10 -15 -20 S21 (dB) -25 -30 HFSS: -35 Er=3.11 tand=0.009 -40 straight sides -45 -50 75 76 77 78 79 80 81 82 83 84 85 Frequency (GHz) 5
ROKE Presentation Title Using HFSS successfully at 77GHz Track shape and profile 6
ROKE Presentation Title Track shaping and cross section • The tracks are not perfect shapes • At these frequencies it makes a difference 7
ROKE Presentation Title Track shaping and cross section • HFSS allows you to change the profile of the tracks easily 8
ROKE Presentation Title Track shaping and cross section • This shaping makes a difference 9
ROKE Presentation Title Using HFSS successfully at 77GHz Bond Wires 10
ROKE Presentation Title Bond Wires HFSS models of bondwires Real bondwires 11
ROKE Presentation Title Bond Wires • HFSS predictions for an identical circuit with different bondwires Return Loss plot loss for different bondwire Ansoft LLC LO Match1 ANSOFT 0.00 Curve Info Bondw ire 3 Bondw ire 2 Bondw ire 1 -5.00 -10.00 Y1 -15.00 -20.00 -25.00 75.00 76.00 77.00 78.00 79.00 80.00 Freq [GHz] 12
ROKE Presentation Title Using HFSS successfully at 77GHz Ports, complex and otherwise 13
ROKE Presentation Title Matching to ports Typical port measurement of a 77GHz device Matching to the port needs to include • Bondwire • Port impedance • Matching network (stub) 14
ROKE Presentation Title Complex ports – a bit quirky Smith Plot 1 90 100 80 110 0.50 70 • Positive S parameters! 120 60 130 50 0.30 140 40 150 30 160 20 • Positive S-parameters are possible if using 0.10 1.00 170 10 2.00 -0.50 -0.30 0.00 0.00 1.00 -1.20 -3.00 -2.50 -2.00 -1.80 -1.50 complex ports. The fields in the model are all 180 0 0.00 -0.10 -2.00 -170 -10 correct but the return loss looks scary. It is due -0.30 -1.00 -160 -20 to the maths involved in normalising the ports. -0.50 -150 -30 -140 -40 -130 -50 What are you matching to? -120 -60 -110 -70 -100 -80 -90 Port impedance from datasheet = 29 -75j Tune the stub to give a good match. What is the impedance of the circuit with the stub? 29-75j Which will not give a good match in real life. You must manually set the port to 29 + 75j
ROKE Presentation Title Matching to ports To get around the positive s parameters: Divide the port into two parts: 1. A resistive port corresponding to the real part of the port impedance. i.e. 29 2.A reactive section in series corresponding to the port reactance i.e. 0.0273pF
ROKE Presentation Title Lumped ports vs waveports Lumped ports easy and small Can fit nicely on the end of a bond wire.
ROKE Presentation Title Lumped ports 90 100 80 1.00 110 70 120 60 Simple port at the end 130 50 of a track. The 0.50 2.00 dielectric is 0.1mm 140 40 thick 150 30 A simple port should be fine. Shouldn’t it? 160 0.20 5.00 20 170 10 0.00 0.20 0.50 1.00 2.00 5.00 180 0 0.00 -170 -10 -0.20 -5.00 -160 -20 -150 -30 -140 -40 -0.50 -2.00 -130 -50 -120 -60 -110 -1.00 -70 -100 -80 -90
ROKE Presentation Title Lumped ports 90 100 80 1.00 110 70 120 60 The port may need to be 130 50 thinner. Comparing the 0.50 2.00 input impedance shows a 140 40 significant difference. 150 30 Which is correct? 160 0.20 5.00 20 170 10 0.00 0.20 0.50 1.00 2.00 5.00 180 0 0.00 -170 -10 -0.20 -5.00 -160 -20 -150 -30 -140 -40 -0.50 -2.00 -130 -50 Lumped ports are not invisible -120 -60 they have dimensions and -110 -1.00 -70 therefore parasitic L and C -100 -80 -90
ROKE Presentation Title Lumped ports 90 100 80 1.00 110 70 120 60 130 50 L can be reduced by 0.50 2.00 moving the ground plane 140 40 up towards the track, shortening the port. 150 30 160 0.20 5.00 20 170 10 0.00 0.20 0.50 1.00 2.00 5.00 180 0 0.00 -170 -10 -0.20 -5.00 -160 -20 -150 -30 -140 -40 -0.50 -2.00 -130 -50 -120 -60 Or use a wave port -110 -1.00 -70 -100 -80 -90
ROKE Presentation Title Optimetrics Tolerance analysis 21
ROKE Presentation Title Optimetrics • PCB manufacturing accuracy is critical. Optimetrics can be used effectively with parametric sweeps show the effects of over and under etching. Return Loss variation with etching ANSOFT 0.00 Curve Info Ideal w ith C Under 10 Under 10 C Under 20 Under 20 C -5.00 Return Loss (dB) -10.00 -15.00 -20.00 -25.00 75.0 76.0 77.0 78.0 79.0 80.0 81.0 82.0 83.0 84.0 85.0 Freq [GHz]
ROKE Presentation Title Conclusions • HFSS can be used at every step when designing at 77GHz. • Initial designs • Validation of measured material properties • Further iteration with optimised material values • Analysis of manufacturing tolerances Return Loss variation with etching ANSOFT 0.00 Curve Info Ideal w ith C Under 10 Under 10 C Under 20 Under 20 C -5.00 Return Loss (dB) -10.00 • However, you must be careful because effects which can -15.00 happily be ignored at lower frequencies are critical at mm -20.00 -25.00 75.0 76.0 77.0 78.0 79.0 80.0 81.0 82.0 83.0 84.0 85.0 waves Freq [GHz]
Using HFSS successfully at 77GHz Kate Moore Communications Sensors Information Systems Contract R&D | Consultancy Specialist Products
Recommend
More recommend
