International Journal of Electrical and Computer Engineering (IJECE) Vol.
No.
February 2017, pp.
ISSN: 2088-8708.
DOI: 10.
11591/ijece.
pp 176-183 Noise Characterization in InAlAs/InGaAs/InP pHEMTs for Low Noise Applications Djennati.
Ghaffour Electronic and Electrical Engineering Department.
University of Abu Bekr Belkaid Tlemcen.
Algeria
Research Unit of Materials and Renewable Energies Article Info
ABSTRACT
Article history:
In this paper, a noise revision of an InAlAs/InGaAs/InP psoeudomorphic high electron mobility transistor .
HEMT) in presented.
The noise performances of the device were predicted over a range of frequencies from 1GHz to 100GHz.
The minimum noise figure (NFmi.
, the noise resistance (R.
and optimum source impedance (Zop.
were extracted using two A physical model that includes diffusion noise and G-R noise models and an analytical model based on an improved PRC noise model that considers the feedback capacitance Cgd.
The two approaches presented matched results allowing a good prediction of the noise behaviour.
The pHEMT was used to design a single stage S-band low noise amplifier (LNA).
The LNA demonstrated a gain of 12.
6dB with a return loss coefficient of 6dB at the input and greater than -7dB in the output and an overall noise figure less than 1dB.
Received Jul 10, 2016 Revised Sep 20, 2016 Accepted Oct 13, 2016 Keyword:
InAlAs/InGaAs/InP LNA Noise simulation pHEMT PRC model Copyright A 2017 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Djennati.
Electronic and Electrical Engineering Department.
University of Abu Bekr Belkaid Tlemcen, 22 Abi Ayed Abdelkrim Road.
Tlemcen, 13000.
Algeria Email: zakarya.
djennati@mail.
univ-tlemcen.
INTRODUCTION
The InAlAs/InGaAs/InP psoeudomorphic high electron mobility transistors .
HEMT) have exhibited excellent performances in terms of noise figure compared to MOSFET devices at RF and microwave fields .
In addition, they are mostly used in the design of low noise amplifiers .
and proven to be more advantageous that the AlGaAs/InGaAs pHEMTs .
Advanced circuit design that uses pHEMT, needs precise and accurate noise models, especially for their application in low noise MMIC.
Many noise models have been reported in the literature: the wellknown FukuiAos model .
, the equivalent noise temperature model by Pospieszalski .
and the flicker noise model for HEMTs .
Many approaches on HEMTAos noise parameters characterization ware investigated .
In 1974.
Pucel proposed a charge control based approach for a noise model in FET devices, the model is known as the PRC noise model .
Later in 1988, the model was used for HEMT devices in .
and later improved by Zhi .
where they included, in the PRC model, the temperature coefficients in order to calculate precise noise parameters for HEMTs.
In this paper, we try to predict the noise performances at high frequency of an InAlAs/InGaAs/InP pHEMT by extracting the basic noise parameters using two approaches: a physical modelling and a mathematical modelling.
This work is organized as follow: first, we discuss the different sources of noise in the pHEMT devices.
Then, we make use of ATLAS from SILVACO TCAD suit, to simulate the noise behavior of a fabricated InAlAs/InGaAs/InP 250 nm gate length pHEMT.
Next, we describe the analytical model based on the 2-port small-signal equivalent circuit of the device mentioned before and we calculate the Journal homepage: http://iaesjournal.
com/online/index.
php/IJECE IJECE
ISSN: 2088-8708
P,R,C noise coefficients in order to extract the noise parameters.
Finally, the modeled InAlAs/InGaAs/InP pHEMT is used to design a single stage low noise amplifier (LNA) for S-band frequencies.
pHEMTs Noise Sources in pHEMTs Every semiconductor device exhibits the basic noise mechanisms in a way that depends on its features and physical properties.
There are five dominant noise mechanism in pHEMTs: .
Thermal noise, generated by the parasitic resistances of gate, drain and source.
Diffusion noise, located in the channel when the pHEMT is operating in saturation mode.
G-R noise, due to electron-hole pair generated in the space-charge or by impact ionization.
Shot noise, due to the reverse biased gate current.
The flicker noise, neglected in this work since the studied pHEMT is dedicated to LNA applications.
But still, there are at least two situations where the flicker noise affects microwave circuits: the up-conversion in mixers and the phase noise of HEMT oscillators .
Physical Modelling of Noise in pHEMT The epitaxial structure of the studied InAlAs/ InGaAs/InP pHEMT in this paper is shown in Table 1.
it is comparable to the fabricated pHEMT from .
, the gate length is 250 nm and a total gate width of 100 m, its layout consist of 2x50 m fingers.
The efficacy of the physical model of the structure in Table1 has been tested in a previous work .
through a DC characterization where we obtained, by inverse modelling.
DC performances matched to data measurements (IDSS=56mA at VDS=1.
5v and VGS=0v.
gm,max=125 mS/m.
Table 1.
The Epitaxiale Structure of the a 250 nm InGaAs/InAlAs/InP pHEMT Material In0.
53Ga0.
In0.
52Al0.
In0.
52Al0.
In0.
52Al0.
In0.
7Ga0.
In0.
52Al0.
In0.
52Al0.
In0.
52Al0.
Layer Cap Barrier Delta doping Spacer 1 Channel Spacer 2 Delta doping Buffer InP (S.
) Substrate Thickness 5 nm 15 nm 2 nm 10 nm 14 nm 10 nm 2 nm 450 nm The physical model in ATLAS-Silvaco is represented as a two-port noiseless device with random voltage sources attached to its ports.
These external voltage source are small and random, they produce the same voltage at the deviceAos terminals as the intrinsic noise sources (V 1 at the input and V2 for the output, .
ee Figure .
Figure 1.
Block Diagram of the Modeled Noise Sources in pHEMT ATLAS models the noise in the pHEMT, by calculating the statistical behavior of V 1 and V2.
Two types of intrinsic noise were considered and defined with their parameters in the simulation: Diffusion noise and G-R noise couple with impact ionization noise.
A direct AC analysis was performed before the noise simulation.
the current gain obtained is presented in Figure 2 where a good matching with data measurements is achieved.
We believe that the physical model of the pHEMT is reliable and could be used in order to get an insight and estimation on the noise performances.
The noise simulation was performed on the pHEMT physical model under a bias point of 15%IDSS to obtain the best noise performances.
The spectral densities of the noise calculated at the input (V.
and the output (V.
of the device are presented in Figure 3.
Noise Characterization in InAlAs/InGaAs/InP pHEMTs for Low Noise Applications (Z.
Djennat.
A ISSN: 2088-8708 Figure 2.
Current Gain .
vs Frequency.
Simulated Against Measured at VGS=0v and VDS=1.
Figure 3.
Voltage Spectral Density of: .
Diffusion noise.
G-R noise.
Impact Ionization noise.
Total noise Analytical Modeling In The PRC analytical noise model in this work is based on the 2-port linear equivalent circuit shown in Figure 4.
At the input, a gate current noise source ig2 is in parallel with the output impedance composed of the channel charge resistance Ri in series with the gate-source capacitance Cgs, while in the output, a channel noise current source id2 is in parallel with the output resistance Rds.
Figure 4.
Intrinsic noiseless pHEMT equivalent circuit coupled with its noise sources IJECE Vol.
No.
February 2017 : 176 Ae 183
IJECE
ISSN: 2088-8708
The elements of the small-signal equivalent circuit were calculated analytically from measured Sparameters charts, their values are presented in Table 2.
Table 2.
Small-Signal Equivalent Circuit Parameters at T=300k with 15%IDSS and VDS=1v Rgd() Rg () Ri() Rs () Rds() Rd () Cgd.
F) Lg .
H) Cds.
F) Ls .
H) Cgs.
F) Ld .
H) gm.
S) Cpg.
F) E.
Se.
Cpd.
F) The PRC terms in this model are dimensionless noise coefficients related to the magnitude of the current noise sources .
g2 and id.
, and to the correlation between them.
P is referred to as the channel noise R is the gate noise coefficient and last.
C is the correlation coefficient between the gate noise current and the drain noise current.
It should be noted that the correlation coefficient C is a pure complex parameter due to the existence of the coupling capacitance between the gate and the drain C gd.
The PRC coefficients are obtained by .
Oo( The PRC coefficients were calculated using equations in .
by considering the values of the smallsignal equivalent circuit parameters in Table.
1 with the values of ig2 and id2 that were extracted from the simulation curves in Figure 3 .
g2 is related to V1 and id2 related to V.
The results are shown in Figure 5.
Figure 5.
Values of P ,R and C Coefficients Versus Frequency of the 250nm InGaAs/InAlAs/InP pHEMT The curves in Figure 5 are drawn in tendency lines due to the uncertainty in the calculation of the gate noise at a very low drain current value.
Figure 5.
shows a frequency independent drain noise coefficient P due to the hot electron effect in the channel.
On the other hand, the gate noise coefficient R is frequency dependent for the relatively low frequencies region (<7GH.
: a result of an increasing in the gate leakage current.
Concerning the correlation coefficient C shown in Figure 5.
, this parameter strongly depends on the gateAos length Lg.
We believe that the frequency dependence of the coefficients R and C is mainly linked to the shot noise caused by the reversed gate current.
In addition to the PRC coefficients, three other parameters were described by Pucel:
Noise Characterization in InAlAs/InGaAs/InP pHEMTs for Low Noise Applications (Z.
Djennat.
A ISSN: 2088-8708 The coefficients in .
allow us to calculate the main noise parameters:
| |
Oo Oo The model above does not include the feedback capacitance Cgd, the charges held by this capacitance are responsible for the voltage drop in drain region.
These charges also give a rise to gate leakage current component and therefore contribute to the noise behaviour of the device.
When considering the effect of Cgd, the expressions of analytical model became as expressed in .
The feedback capacitance Cgd added to the model in .
constitutes a major influence in the noise figure: the coupling capacitance make the drain noise current frequency dependent resulting to a rise in noise figure for frequencies beyond ft .
) (
RESULTS AND DISCUSSIONS
For a precise prediction of the deviceAos noise performances, a precise noise model is needed.
In this section, we extract and compare the main important noise parameters: the minimum noise figure NFmin, the equivalent noise resistance Rn and the optimum source impedance Zopt.
The curves of the mathematical model illustrate a good matching within the predicted range by TCAD simulation.
While both model and TCAD curves are matched in NFMIN and Zopt, an average matching was achieved in Rn, nevertheless, the decreasing slop was correctly predicted.
ATLAS-SILVACO uses the impedance field method, to determine the internal noise parameters in Figure 6, by characterizing the entire deviceAos behaviour based on the fluctuation of the two correlated equivalent noise sources.
Figure 6.
shows a frequency dependence of NFMIN at 15%IDSS.
An increase from 0.
7 dB to 1.
2 dB is observed for frequencies less than 10 GHz, these values are well known for devices with a gate width of 100 m.
In Figure 6.
, a considerable decrease of the noise resistance mainly due to the combined effect of a higher transconductance and a reduced gate noise However, the sheet carrier density values, the diffusion coefficients and the GenerationRecombination parameters needed to be adjusted in order to achieve a good matching of the pHEMTAos noise The physical modelling offers a better insight on the physical mechanisms controlling the noise.
Meanwhile, the mathematical model depends on the small-signal equivalent circuit elements, which is advantageous for pHEMTAos optimization and improvements for its implementation in microwave circuits such as LNAs .
ow noise amplifier.
In addition, in the low-noise application, the pHEMT is generally embedded in the circuit and the values of the parasitic capacitances in the equivalent circuit will differ from the discrete elements used in this characterization, but the intrinsic elements still remain the same.
Also, the access resistances especially the input access resistance Rg, will obviously generate thermal noise contributing to the overall noise figure.
So from the low noise circuit designing perspective, it will be more useful to provide the whole model including the effect of access resistances.
However we believe that the presented analytical model is straightforward solution to predict the noise behavior of pHEMTs during its design process.
Following the noise modeling, the parameters of the small-signal equivalent circuits of the 250nm 2x50m pHEMT were to design a single stage LNA for S-band applications.
The purpose was to demonstrate the potential of InAlAs/InGaAs/InP pHEMT in low noise applications.
The corresponding single stage LNA schematic along with the values of the passive components and the biasing conditions are illustrated in Figure 7.
It is designed to operate in a frequency range of 2.
4 GHz to IJECE Vol.
No.
February 2017 : 176 Ae 183
IJECE
ISSN: 2088-8708
2 GHz in 50 system.
The common source configuration of the LNA is has a noise figure close to the transistorAos minimum noise figure .
and it is less prone to oscillations than other topologies.
A high gate resistance was used to at the input terminal of the transistor to prevent large current flowing to the input and hence an optimum gate voltage injection is achieved.
The series capacitors Cin and Cout isolate the DC bias sources for the RF input and RF output respectively.
The drain capacitor CD improves the input return loss and provides an output matching for the LNA.
The resistor RD1 is used to prevent however, it can reduce the overall gain to the amplifier and needs to be minimized.
LD and RD2 provide an adequate output match network and ensure the stability of the amplifier.
The LNA has been implemented in the FET lineal model from ADS Agilent.
The performances obtained are shown in Figure 8.
The amplifier shown in Figure 8.
, exhibited a gain around 12.
6dB at the frequency of 3GHz, with a return loss coefficient at the input -2.
4dB and grater that -7dB at the output as illustrated in Figure 8.
The minimum noise figure and the noise factor in Figure 8.
, are less than 1dB for all the operating frequencies as predicted by the noise characterization in Figure 6.
In must be noted that the passive components of the LNA are considered ideal.
The results obtained can be used for accurate design with real process design kit.
Figure 6.
Plots of NFmin vs.
Rn vs.
and Zopt chart .
of the 250nm InAlAs/InGaAs/InP pHEMT Figure 7.
Schematic Circuit of Single Stage S-band Low Noise Amplifier Noise Characterization in InAlAs/InGaAs/InP pHEMTs for Low Noise Applications (Z.
Djennat.
A ISSN: 2088-8708 .
Figure 8.
Performances of Single Stage LNA: .
Noise figure.
Return Loss Coefficients CONCLUSION In this work, a noise characterization was performed on InAlAs/InGaAs/InP pHEMT using two approaches: PucelAos PRC model based on analytical equations and a physical modelling by simulation on Atlas-Silvaco.
We observed that both mathematical and physical model can predict the noise performances of InAlAs/InGaAs/InP pHEMTs.
The improved analytical model and the physical model ware adjusted to obtain matching results between the main noiseAos parameters (NFmin,Rn and Zop.
in a wide frequency range .
-100GH.
A single stage S-band LNA was simulated from the small signal equivalent circuit.
The LNA demonstrated a gain of 12.
6dB and a noise figure less that 1dB as predicted.
Future work will be devoted to model the thermal noise by including temperature variation in order to have a full insight on the noise performances of InAlAs/InGaAs/InP pHEMT.
ACKNOWLEDGEMENTS
The authors wish to thank Professor M.
Missous from the school of Electrical and Electronic Engineering at ManchesterAos University, for providing the pHEMT structure and data measurements used for this work.
REFERENCES