NETWORK ANALYSIS FOR DIGITAL ENGINEERS - Advanced Signal Integrity for High-Speed Digital Design - page 408

10

0

− 10

− 20

− 30

Passivity
violation

− 40

− 50

− 60

70

−

− 80

0 5 10 15 20 25 30 35 40 45 50

Frequency, GHz

Figure 9-42 Bad S -parameter measurement of a transmission line, showing passivity
violations due to incorrect VNA calibration.

Passivity
violations

10

− 10

20

−

− 30

Calibration
looks marginal

− 40

− 50

− 60

− 70

− 80

0 5 10 15 20 25 30 35 40 45 50

Frequency, GHz

Figure 9-43 Bad S -parameter measurement of a transmission line, showing passivity
violations at high frequencies where the VNA calibration breaks down.

(0 dB is equal to an insertion loss of 1 on a linear scale). In this case, the
passivity violation was a result and an incorrect calibration of the vector network
analyzer. Another example of a severe passivity violation is shown in Figure 9-43,
where the calibration breaks down after about 17 GHz. Even below 17 GHz this
measurement looks questionable, due to the significant amount of nonperiodic
ripples on the waveform. Periodic ripples would probably be a result of reflections
or crosstalk, but nonperiodic noise on S 21 indicates an inaccurate measurement.
Figure 9-44 shows an example of a good measurement of a 5-in. microstrip
transmission line on a FR4 substrate. Note that the insertion loss does not rise
above 0 dB and that it does not contain any nonperiodic noise. It should be noted
that the S -parameters could still be nonpassive even if S 21 does not rise above
0 dB. Passivity can only be verified using the techniques outlined in Section 9.3.1;
however, simple observation of S 21 is often sufficient to catch gross errors.
The periodic behavior of S 11 can also be used to get an intuitive feeling
about the validity of the S -parameters by ensuring reasonable phase delays. If
the structure under test is a transmission line, the return loss ( S 11 ) should be
periodic and the distance between peaks (or dips) is related to the time delay
with equation (9-4). The propagation delay can be double-checked at various