By means of the small-signal simulation mode of MINIMOS-NT, various high-frequency
data for a one-finger InGaP/GaAs HBT with an emitter area of m
m were extracted. This high-power device has been used for
power amplifier circuits for mobile communication. Figure 6.1 shows
the simulated device structure and the pad parasitics (capacitances and
inductances) of the measurement environment used for the S-parameter
measurement in the two-port pad parasitic equivalent circuit.
![]() |
The pad capacitances of the equivalent circuit are
150
fF,
75
fF, and
24
fF. The
parasitic inductance values are
1
pH,
75
pH, and finally
50
pH. The resistive parasitics are
neglected, since a rather small device and therefore only low currents are
considered.
Figure 6.2 shows a comparison between measured and simulated
collector currents
and an almost perfect match of the curves in
the small-signal area of the figure. A further increase of the input power
causes harmonics in the device, which cannot be obtained by the linear
small-signal mode (see Section 2.6).
The combined Smith/polar charts with a radius of one in Figure 6.3 show a comparison of simulated and measured
S-parameters at
3
V, with current densities
kA/cm
,
kA/cm
,
and
kA/cm
, respectively, for the frequency range between 50
MHz and
10
GHz. For the same device the high-frequency figures of merits current
gain
and the squared absolute value of the current ratio parameter
were extracted. The cut-off frequency
and maximum
oscillation frequency
are found at the intersection of these curves
with the 0
dB line. The lower right side of Figure 6.3 shows a
comparison of the simulated and measured
and the absolute value of
. The measurement data ends at 10
GHz, whereas the simulation
could be continued to
GHz showing another important advantage of
simulators to measurement equipments. In addition, a mixed-mode circuit was set up
to compare large signal measurement data in the small-signal range.
The AC-simulation takes about 200s CPU-time on a
2.4
GHz Intel Pentium IV with
1
GB memory running under Suse Linux 8.2 for a S-parameters computation with 20 frequency steps. A
number of 20 steps is more than sufficient to produce the graphs. For
comparison, the conventional small-signal equivalent-circuit approach
takes about 590
s CPU-time at the same machine for 200 time
steps at only one given frequency. As stated in the introduction, many
time steps have to be performed to ensure appropriate accuracy in the
time-domain to obtain sufficient accuracy for one frequency. To avoid
this number of time steps for all frequencies required, only one
frequency is used to extract an equivalent circuit valid in a specific
frequency range. The time for such a post-processing of the transient
simulation results to obtain the S-parameters at all frequencies is
not included. Thus, the more accurate approach can speed up the
frequency-domain simulation by about 98% (taking one frequency into
account).
![]() ![]() |
![]() ![]() ![]() ![]() |