INTRODUCTION
Microwave amplifiers constitute an area of prime interest for the high-frequency industry. Most of them use III-V semiconductors such as Gallium Arsenide (GaAs) and Indium Phosphide (InP) because of their higher performance in terms of gain and noise. At the component level, pseudomorphic Heterojonction Transistors (pHEMTs) are widely used in Monolithic Microwave Integrated Circuits (MMICs). The aim of this study is the modeling of such transistors subsequently used to design a low-noise amplifier.
MATERIALS AND METHODS
Small signal equivalent circuit: One of the most widely used small-signal
HEMT equivalent circuit is shown in Fig. 1 (Dambrine
and Cappy, 1988; Berroth and Bosh, 1990; Anholt
and Swirhun, 1991a). According to its structure and operation mode, this
circuit can be divided into two parts namely the intrinsic and the extrinsic
parts. The intrinsic part corresponds to the active part of the transistor i.e.,
the channel. The extrinsic part corresponds to the connecting zone (access lines
and electrode components).
Extraction of small signal parameters: Determination of the elements
of a linear model is based on an experimental characterization of the transistor.
The small-signal extraction method proposed in this study is based on S-parameters
measurements. It involves the use of 2 sets of measurements at different bias
conditions; pinched or cold and hot device measurements.
|
Fig. 1: |
Small signal equivalent circuit of a HEMT |
The measured S matri is converted to a impedance matrix (Z) whose elements
Zij have real and imaginary part: real (Zij) and imag
(Zij) (Freckey, 1994).
The extrinsic elements can be obtained from S-parameters measurements under
cold and pinched off biasing conditions: a zero drain source voltage Vds
and a gate source voltage much lower than the pinch-off voltage Vp
(i.e., Vds = 0 V and Vgs << Vp). In fact,
cold pinched off bias conditions can simplify the topology of the small signal
equivalent circuit as shown in Fig. 2. For more efficiency,
the π circuit model in Fig. 2 can be transformed to a
T circuit (Caddemi et al., 2006; Anholt
and Swirhun, 1991b). Thus, the Z-parameters Zpij can be expressed
as:
|
Fig. 2: |
Small signal equivalent circuit of HEMT at zero drain bias
and gate voltage pinch-off |
Where, subscript p is pinched off condition. In order to simplify the circuit
analysis, the capacitive π network is transformed into a T network consisting
of 3 capacitances Cg, Cs and Cd (Caddemi
et al., 2006; Anholt and Swirhun, 1991b;
Khalaf, 2000):
The parasitic resistances Rg, Rs and Rd can
be deduced from the real part of 1, 2 and 3 (Khalaf, 2000).
The parasitic inductances Lg, Ls and Ld can
be extracted from the slope of the curve of ω* imag (Zpij) vs.
ω2 (Caddemi et al., 2006; Khalaf,
2000; Chen et al., 2006; White
and Healty, 1993). When the conduction in the channel is removed (i.e.,
deeply in pinch off Vds = 0, Vgs << Vp),
it is possible to extract the parasitic (extrinsic) capacitances Cpg
and Cpd.
Frequencies of some GHz, the effects due to parasitic inductances and resistances
can be neglected and thus have little influence on the imaginary parts of the
admittance matrix, assuming Cgso = Cgdo and neglecting
Cds (Dambrine and Cappy, 1988; Freckey,
1994; Anholt and Swirhun, 1991b; White
and Healty, 1993; Chigaeva et al., 2000).
We can write these equations:
Once all the extrinsic elements are determined, we can directly extract the
intrinsic elements (Ri, Cgs, Cgd, Rds,
Cds, Gm and τ) from the intrinsic Y-parameters according
to the expressions proposed by Dambrine and Cappy (1988),
Berroth and Bosh (1990), Khalaf (2000),
Chigaeva et al. (2000), Shirakawa
et al. (1995) and Wurtz (1994):
Where:
Description of device technology: The fabricated high breakdown InGaAs-InAlAs-InP pHEMT (sample-1841) and InGaAs-AlGaAs-GaAs pHEMT (sample-1891) were chosen for this research.
These devices with 1 μm gate length and 200 μm (2x100 μm) gate
width are used. The epitaxial layer structure of the used device is shown in
Fig. 3. Looking at the structure from bottom to top, a lattice-matched
undoped InAlAs buffer layer of thickness 4500 Å is grown on top of an
InP semi insulating substrate. A highly strained, undoped InGaAs, channel is
grown well below the critical thickness of this composition (140 Å). The
spacer is a lattice matched, undoped InAlAs layer of thickness 100 Å used
to spatially separate the heavily doped delta-region from the active channel.
A supply layer is formed with thickness 150 Å to supply electrons into
the 2DEG with Delta-doping sheet density of 3.6x1012 cm-2.
Noise figure characterization: The main target of the fabricated devices (GaAs and InP HEMTs and pHEMTs) is the design and implementation of broadband low noise amplifiers for SKA applications. Thus, it is important to get an accurate analytic expression for calculating the minimum noise figure of these devices. Since, the noise figure of a FET is affected by both bias point and generator impedance, the minimum noise figure, NFmin defined here is an absolute minimum noise figure obtained by adjusting both bias and generator impedance.
Using the four equivalent element values Gm, Cgs/Fc Rs and Rg, determined by S-parameter measurement and small-signal parameter extraction Fukui empirically derived a simple expression for NFmin:
Kf is a fitting factor which depends on the material system used.
Applying this empirical form to the fabricated devices used in this study and
comparing with experimentally measured NFmin for the same devices,
the best extracted fit values for the fitting factor k is 2.8 for the GaAs based
samples and 3.4-3.6 for the InP based samples, f is the frequency.
|
Fig. 3: |
Epitaxial layer structure of used transistor (Inp device) |
RESULTS AND DISCUSSION
The direct method presented in this research was demonstrated through the extraction
of the small-signal equivalent circuit of two pHEMTs. Practically, the values
of the extrinsic components were optimized to best fit with measurements (Fig.
4). Table 1 shows the extrinsic and intrinsic element
values of the small-signal equivalent circuit of the 2 following pHEMTs: the
VMBE- 1841-A132-InP (biased at Vds = 1.5V and Vgs = -0.2V)
and the VMBE-1891-A322-GAs (biased at Vds = 0.5V and Vgs
= - 0.4V). The curves displayed in Fig. 4 showed a good agreement
between obtained and measured S-parameters (Table 2). It should
be noted that the stability factor k as well as the maximum available gain G
max (with k≥1) or the most stable gain MSG (if k<1) are critical
parameters for microwave circuits designers. By comparing the GaAs-pHEMT and
the InP-pHEMT it is noted that the InP-pHEMT exhibits a lower minimum noise
factor (NF min) (Fig. 5 and 6),
a higher gain G max and a higher cut-off frequency Fc (Table
3):
Table 1: |
Extracted parameter values for the two transistors |
|
Table 2: |
Errors between measured and modelled values of S-parameters |
|
Table 3: |
Transistor performance parameters |
|
|
Fig. 4: |
Comparison between measured (circles) and calculated (lines)
S parameters of pHEMT devices (a, b) VMBE-1841-A132-InP under the bias conditions:
Vds = 1.5V, Vgs = -0.2V (c, d) VMBE-1891-A322-GaAs
under the bias conditions: Vds = 0.5V, Vgs = -0.4V |
|
Fig. 5: |
Comparison of the measured minimum noise figure (room-temperature
minimum noise figure of a 1x200 μm device from each sample up to 2
GHz) (circles) with minimum noise figure calculations based on Fukuis
analysis (lines) |
|
Fig. 6: |
Maximum Gain Gmax, stability factor K for the:
(a) VMBE-1841-A132-InP under the bias conditions: Vds = 1.5V,
Vgs = -0.2V (b) VMBE-1891-A322-GaAs under the bias conditions:
Vds = 0.5V, Vgs = -0.4V |
• |
InP-pHEMT: S21 = 14 dB, Gmax = 22 dB,
NFmin = 0.14 dB and FC = 28 GHz |
• |
GaAs-pHEMT: S21 = 2 dB, Gmax = 13 dB,
NFmin = 0.198 dB and Fc = 8.6 GHz |
CONCLUSION
This research presents a direct extraction method of the elements of small signal equivalent circuit of 2 transistors PHEMT one on the GaAs substrate and the other on the InP substrate of which the length and the width of the gate, respectively 1 and 200 μm. This modelling is essential for any active or passive component and which precedes any design of a radio frequency circuit.
The technique suggested is of experimental type, rests to measures of the parameters of dispersion S, followed by a method of optimization with an aim of studying the linear behavior and the performances ultra high frequencies of the transistor. The results obtained show a good agreement between simulated and measured S parameters. The analysis of the performances ultra high frequencies such as the gain, the factor of noise NFmin and the frequency band justifies the choice of transistor pHEMT on the InP substrate for the application low noise.