FiltronicSolid State Applications Notes Discrete FET / PHEMT Devices Revision A August 1996 Filtronic Solid State Discrete FET / PHEMT Applications Notes 2 INTRODUCTION These applications notes describe Filtronic Solid State (FSS) discrete FET / PHEMT devices and typical applications, effective August 1, 1996. Although every effort has been made to ensure the accuracy of this information (contained in these Applications Notes and the associated Datasheets), FSS assumes no responsibility for errors, omissions, or future specifications changes. Every reasonable effort will be made to provide updated or corrected applications information or specification changes to all concerned users. Filtronic Solid State discrete FET / PHEMT devices are generally sold by description only. RF specifications, in particular, are based on samples of devices from individual semiconductor wafers or fabrication lots (groups of 6-10 wafers). All semiconductor die are 100% DC tested, based on specific test conditions and methods as described herein. Test methods are based on MIL-STD-750, with modifications where necessary. Unless otherwise indicated, all specifications listed as minimum or as maximum are guaranteed at the temperatures indicated (nominally at 22ΕC) and under the conditions listed. “Typical” specifications are based on average values, intended to reflect the majority of actual devices manufactured, but are not guaranteed. Filtronic Solid State assumes no responsibility for performance or reliability of its discrete devices when operated outside the recommended electrical or environmental limits, or if operated in such a way as to exceed the Absolute Maximum Ratings as indicated on specific Datasheets. Applications information is intended as general guidance only, and FSS is not responsible for user-provided circuitry. Filtronic Solid State Discrete FET / PHEMT Applications Notes 3 TABLE OF CONTENTS Section Page 1.Introduction.........................................................................................................................2 2.Quick Reference Guide......................................................................................................4 3.Cross-Reference Guide .....................................................................................................5 4.General Device Description................................................................................................6 5.PHEMT Device Physics .....................................................................................................8 6.PHEMT Characteristics......................................................................................................11 7.Small-Signal S-Parameters and Lumped Element Models................................................20 8.Biasing Circuits and Stabilization Techniques....................................................................23 9.Large-Signal Models and Optimum Power Match Data.....................................................26 10.Example Circuits ..............................................................................................................29 11.Recommended Assembly Techniques.............................................................................39 12.Parametric Screening and Quality Assurance..................................................................40 13.Discrete Device Uniformity...............................................................................................46 14.Device Reliability ..............................................................................................................48 15.Appendix A.......................................................................................................................49 Filtronic Solid State Discrete FET / PHEMT Applications Notes 4 2. Quick Reference Guide Discrete GaAs MESFET and PHEMT Product Family MODEL STATUS FREQ. RANGE (GHz) GAIN (dB) POWER (dBm) NOISE FIGURE (dB) V DS (V) I DSS (mA) PAE (%) CHIP DIM. (µm) GATE W/L (µm) TYPE LF 3807 (1) 1 - 12 7 24 --- 7 200 40 430x460 700/.5 MES LF 3814 STD 1 - 12 6 27 --- 7 400 40 430x460 1400/.5 MES LF 3830 STD 1 - 8 7 31.5 --- 9 700 36 340x930 3000/.5 MES LF 3850 STD 1 - 8 7 33.5 --- 9 1100 38 380x900 5000/.5 MES LF 6828 STD 1 - 18 7 19 --- 5.5 80 26 310x440 280/.5 MES LF 6836 STD 1 - 18 6.5 21 --- 5.5 120 27 330x350 360/.5 MES LF 6872 STD 1 - 18 6 24 --- 5.5 230 26 370x500 720/.5 MES LP 7512 STD 1 - 40 8.5 2 --- 1.0 2 35 --- 330x460 200/.25 SHP LP 7612 STD 1 - 40 9.5 3 21 1.1 4 5 55 50 330x460 200/.25 DHP LP 6836 PRT 1 - 30 9.5 3 24 --- 8 100 45 330x350 360/.25 DHP LP 6872 STD 1 - 26 9.5 3 27 --- 8 220 45 370x500 720/.25 DHP LP 1500 ED 5 1 - 20 9.0 3 30.5 --- 8 465 48 340x800 1500/.25 DHP LP 1800 ED 5 1 - 20 9.0 3 31.0 --- 8 560 48 340x820 1800/.25 DHP LP 3000 ED 5 1 - 20 8.5 3 33.5 --- 8 930 45 340x850 3000/.25 DHP LP 5000 ED 5 1 - 20 8.5 3 36.0 --- 8 1550 45 340x860 5000/.25 DHP LEGEND: STD STANDARD PRODUCT OBS OBSOLETE; MODEL DISCONTINUED. PRT PROTOTYPE; APPLICATION DEVELOPMENT SAMPLES AVAILABLE ED ENGINEERING DEVELOPMENT; PROTOTYPE DEVICES AVAILABLE AS INDICATED MES Standard GaAs MESFET SHP SINGLE HETEROJUNCTION PHEMT DHP DOUBLE HETEROJUNCTION PHEMT NOTES: 1). LF 3807 to be discontinued 10/96. One-half of the LF 3814 can be used as a replacement. 2). Associated Gain at Optimum Noise Figure, measured at 18 GHz. 3). Power Gain at 1dB Gain Compression, device tuned for optimum power at 18 GHz. 4). Low-noise bias, measured at 18 GHz. 5). Prototype devices available 10/96. FREQ. RANGE is the recommended range based on the device’s available gain. Filtronic Solid State Discrete FET / PHEMT Applications Notes 5 3. Cross-Reference Guide This guide is based on general performance parameters; optimum circuit performance may require re-design, since the FSS device S-parameters may differ significantly from other manufacturers’ devices as listed below. Contact the Foundry for additional information. NEC NE24200, 32400 LP 7512 NE24200, 67300 LP 7612, 7512 HI-REL NE710000 NE720000 NE760000 NE761000 LP 7612 NE800100 NE800200 NE800400 LF 3814 LF 3830 LF 3850 NE900000 NE900100 NE900200 NE900400 LP 7612 LP 6836 LP 6872 LP 1800 MWT (Microwave Technology) MWT-1, A1 MWT-2 MWT-3, A3 MWT-4 MWT-6 MWT-7, A7, S7 MWT-8, A8 MWT-9, A9 LP 7612 LP 6836 LP 7612 LP 7612 LP 6872, LF 3814 LP 7612 LP 1500, LP 6872 LP 6872 MWT-10 MWT-11, A11 MWT-12 MWT-13 MWT-14 MWT-15 MWT-16 LP 7512 LP 1500 LP 6872, LF 3814 LP 6872, LF 3814 LF 3850 LP 6836, LP 6872 LP 6872 FUJITSU FHX04X - FHX06X FHR02X, FHR10X FSC10X, FSC11X FSX02X, FSX03X FSX51X, FSX52X LP 7512 FSC11X FSC51X LF 6828 FSX52X LP 6836 FLX252XV LP 3000 FLC081XP FLC151XP FLC301XP LP 6872, LF 3830 LP 1800, LF 3850 LP 5000 FLK012XP FLK022XP/XV FLK052XP/XV FLK102XP/XV FLK202XV LP 7612 LP 6836 LP 6872 LP 1500 LP 3000 FLR024XP/XV FLR054XV FLR104XV LP 6836 LP 6872 LP 1500 FLR016XP/XV FLR026XP/XV FLR056XV LP 7612 LP 6836 LP 6872 Filtronic Solid State Discrete FET / PHEMT Applications Notes 6 4. GENERAL DEVICE DESCRIPTION Filtronic Solid State offers discrete devices that can be classified into three categories: GaAs MESFETs, Single Heterojunction Pseudomorphic High Electron Mobility Transistors (SH-PHEMTs), and Double Heterojunction Pseudomorphic High Electron Mobility Transistors (DH-PHEMTs). The latter two device types feature AlGaAs/InGaAs heterojunctions, with GaAs buffer and cap layers. Notional cross-sections are shown below: AuGe / Ni / Au Alloyed Contacts N+ GaAs Active Layer N- Buffer Layer Drain Contact Source Contact Ti / Pt / Au Recessed Gate N++ Contact Layer AlGaAs / GaAs Superlattice Undoped GaAs Buffer InGaAs Channel Layer N+ AlGaAs Layer (5 x 10 17 cm -3 ) N++ GaAs Contact Layer Drain Contact Source Contact Ti / Pt / Au Mushroom Gate GaAs MESFET STRUCTURE PHEMT STRUCTURE Undoped AlGaAs Spacer Heterojunction X Y Z Both major device types, the MESFET and the PHEMT, share several common features, beginning with the gate structure. This is a standard Pt / Ti / Au layered structure, which forms a Schottky contact to the semiconductor; general properties of this type of contact can be found in numerous references, such as in Sze 1 . The nominal gate length for FSS’s MESFET is 0.5 µm, and 0.25 µm for the PHEMT devices. Gate definition is accomplished by Electron-Beam direct-write techniques, using a Cambridge E-Beam system; all other patterning is done with standard contact photolithography techniques. In both device types, a gate recess channel is etched into the semiconductor prior to gate metallization, to reduce parasitic gate-source resistance, as well as optimizing the device’s reverse breakdown voltage. Note that the gate and gate recess channel are also offset towards the source contact. The MESFET gate structure is trapezoidal in cross-section, being somewhat wider at the bottom than at the top. The PHEMT gates are mushroom structures, thereby reducing gate metal resistance that would result from the narrower 0.25 µm base. Drain and source contacts are alloyed AuGe/Ni/Au contacts, with overplated Au for bond pads. The Ge in the contact metal is driven into semiconductor during the alloy process, where it acts as an amphoteric dopant, thereby reducing the contact barrier height and forming the ohmic contact itself. The AuGe/Ni/Au has been established as a very stable and robust contact structure. All discrete devices are passivated with silicon nitride (Si 3 N 4 ), which doubles as scratch protection. Nominal die thickness is 100 µm, or 75 µm for power devices (≥ 1W). The larger power devices generally include plated via-holes through the die to minimize parasitic source inductance. The vias are defined by Reactive Ion Etching (RIE) techniques, a method also employed on FSS’s MMIC products. All semiconductor structures are grown by Molecular Beam Epitaxy (MBE), which provides compositional control to a precision of a few atomic monolayers, and very stable and repeatable doping level control. FSS has utilized MBE technology for a wide variety of device structures for more than 15 years. Basic material parameters are routinely characterized, such as: doping level, carrier mobility (both at room temperature and at 77°K), sheet charge density (for PHEMT devices), and wafer uniformity. 1. S. M. Sze, Physics of Semiconductor Devices, Wiley, New York, 1981, pp. 245-297. Filtronic Solid State Discrete FET / PHEMT Applications Notes 8 5. PHEMT DEVICE PHYSICS The basic High Electron Mobility Transistor (HEMT) structure is described in many references, and a review of this device’s salient features is included here for completeness. Sze 1 provides a basic introduction to heterojunction energy band structures, and a very complete review of HEMT technology can be found in Transactions on Electron Devices. 2 Consideration of the basic lumped element intrinsic MESFET model (see Sec. 7 and Liechti 3 ) and its connection to the underlying device structure shows that there are several factors that ultimately limit high- frequency response. The unilateral gain is given by: G f f U ≈ max 2 2 where G U = unilateral gain, f = frequency, and f max = the maximum frequency of oscillation. The gain decreases at 6 dB/octave until f = f max , where the unilateral gain becomes unity. This f max is approximately given by: f f r f T io T DG max ≈ ′ + 2 τ where r’ io = input-to-output resistance ratio, τ DG = 2πR G C DG , the drain-to-gate RC time constant, and f T = the unity current gain, defined as: f g C T m GS ≡ 1 2π where g m = intrinsic device transconductance, and C GS = gate-source capacitance. The frequency f T is defined as that frequency at which the current through C GS (¹i C ¹ = 2πf C GS V C ) is equal to the that produced by the current generator (¹i G ¹= g m V C , where V C is the voltage developed across the capacitor C GS ). At frequencies above f T , the current through C GS is greater than that provided by the current generator, and thus f T represents a fundamental high-frequency limit. Combining expressions for the transconductance and the gate- source capacitance: C Z L h g v Z h GS S G m sat S ≅ ≅ ε ε where ε S = semiconductor dielectric constant, Z = gate width, L G = gate length, v sat = saturated carrier velocity, and h = depletion layer depth. The result becomes: f v L T sat G ≈ 2π Fundamentally the high-frequency performance is optimized by reduction of the device’s gate length, along with maximizing the saturated carrier velocity. In addition, the extrinsic resistances R G and R S must be minimized, as well as feedback elements such as C DG . While gate length is limited by the photolithography technology, the carrier velocity can be enhanced by use of alternative materials, i.e., GaAs instead of Si, or by use of a two-dimensional electron gas (2DEG) transistor. This latter device incorporates a heterojunction, which is a junction of two different semiconductors. One particular heterojunction of interest in this case is a Al 0.3 Ga 0.7 As/GaAs N+/P- combination, which results in a energy band diagram as show: Gate Metal N+ Doped AlGaAs Undoped P- GaAs Undoped AlGaAs Spacer Fermi Level Heterojunction 2-D Electron Gas Conduction Band X Y Z HEMT ENERGY BAND DIAGRAM Filtronic Solid State Discrete FET / PHEMT Applications Notes 9 Since the Al 0.3 Ga 0.7 As and GaAs have different bandgap energies, their conduction band levels will not line up, resulting in a discontinuity as shown. A thin electron inversion layer is formed on the GaAs side of the heterojunction, populated by carriers that have crossed over from the doped Al 0.3 Ga 0.7 As side; a very thin “spacer” layer further separates the electron layer from effect of ionized dopant impurities. This electron inversion layer has been termed a “two-dimensional electron gas” since there may be, under certain conditions, quantization of the allowed energy levels. Fairly accurate prediction of the inversion layer sheet charge density results from the appropriate quantum mechanical treatment. Also note that, by design, the Al 0.3 Ga 0.7 As layer thickness and doping level are chosen to ensure that, at zero applied gate bias, this layer is fully depleted of mobile carriers. Therefore the only mobile charge carriers reside in the 2D electron gas layer. The structure is designed to achieve a 2D electron gas sheet charge density that is as high as possible, typically 1.0-2.5 x 10 12 cm -2 , since this determines the I DSS per unit gate width for the device. Application of a bias voltage to the drain terminal, with the source grounded, causes current flow in the X-axis direction. Control of the 2D electron layer is by application of the gate voltage, since the sheet charge density is directly related to the electric field applied across the heterojunction. This heterostructure, used in place of standard MESFET structure, came to be known by various names, e.g., MODFET, TEGFET, and HEMT, for High Electron Mobility Transistor. An approximate expression for the transconductance is given by: g Z v d d m S sat ≅ + ε ∆ where d + ∆d = total Al 0.3 Ga 0.7 As layer thickness (including spacer). While this has the same functional form as with the MESFET, higher g m is achieved at much lower values of drain voltage, since the greater mobility in the undoped GaAs results in lower “knee” voltages (i.e., the onset of velocity saturation). In addition, the layer thickness is substantially less than a MESFET’s depletion layer thickness. The goal with HEMT design is to optimize the heterostructure in order to maximize low-field carrier mobility, saturation velocity, extrinsic transconductance, and to minimize the input capacitance C GS . The Pseudomorphic HEMT structure utilizes a different combination of semiconductors, as shown in the figure below: Gate Metal N+ Doped AlGaAs Undoped P- GaAs Undoped AlGaAs Spacer 2-D Electron Gas Undoped InGaAs PSEUDOMORPHIC HEMT ENERGY BAND DIAGRAM Filtronic Solid State Discrete FET / PHEMT Applications Notes 10 This structure features a AlGaAs/InGaAs heterojunction, rather than the AlGaAs/GaAs combination used for the HEMT device. The relative compositional ratio of gallium and indium is varied to establish the desired conduction band discontinuity, but there are material-related limits that must be considered. One important factor is that AlGaAs and InGaAs are not lattice-matched, i.e., the interatomic lattice spacing is different. Thus a grown heterojunction using this combination will contain a certain amount of strain, which ultimately limits the thickness of the InGaAs layer. Because the two semiconductors are not lattice- matched, the structure is referred to as a “pseudomorphic” HEMT device. The use of InGaAs provides for even higher mobility and saturated carrier velocity, along with improved 2-D electron gas confinement; in addition, by suitable modification of the basic PHEMT structure, the second heterojunction can be utilized for an addition 2-D electron gas layer parallel to the first. Typically this is done by using another AlGaAs layer below the InGaAs channel layer. FSS’s PHEMT technology features both types of PHEMT structures, as mentioned earlier; the DH-PHEMT structure gives much larger I DSS per unit gate width, and therefore much improved power density. The various device structures and their materials-related performance parameters are summarized in the Materials-Related Performance Parameters Table below. Note first the improvement in low-field mobility resulting from the use of a 2-D electron gas in the HEMT and PHEMT devices; there are also dramatic increases in extrinsic transconductance and high-frequency performance as well. The last two entries represent typical FSS performance from the SH- and DH-PHEMT structures. Also note the substantial improvement in effective carrier velocity resulting from the use of InGaAs vs. GaAs, which directly translates into higher cut-off frequency f T . References: 1). S. M. Sze, op. cit., pp. 122-129. 2). Morkoc, H. and Abe, M., ed., “Special Issue on Heterojunction Field-Effect Transistors”, IEEE Trans. Elec. Devices, Vol. ED-33, No. 5, May 1986. 3). Liechti, C. A., “Microwave Field-Effect Transistors - 1976”, IEEE Trans. Microwave Theory and Tech., Vol. MTT-24, No. 6, June 1976, pp. 279-300. SUMMARY OF MATERIALS-RELATED PERFORMANCE PARAMETERS DEVICE TYPE LG (µm) HALL MOBILITY (cm 2 /V) vsat (cm/s) IDSS (mA/mm) gm (mS/mm) fT (GHz) fmax (GHz) GaAs MESFET 0.50 3000 6.9 x 10 6 170 250 22 45 HEMT 0.65 6000 N/A 120 280 N/A N/A HEMT 0.25 6000 6.3 x 10 6 220 320 40 80 PHEMT 1.00 7000 11.0 x 10 6 290 310 15 N/A SH-PHEMT 0.25 7000 7.8 x 10 6 220 450 55 105 DH-PHEMT 0.25 5000 7.8 x 10 6 325 450 50 100 Filtronic Solid State Discrete FET / PHEMT Applications Notes 11 6. PHEMT CHARACTERISTICS Current-Voltage (I-V) Characteristics Based on the material presented in Section 5, it is clear that the fundamental charge transport and current control mechanisms in the PHEMT device differs significantly from the classical GaAs MESFET, and while there are many similarities between the devices extrinsically, there are also several important differences. The first of these lies in the basic I-V curves, and in the derived transconductance (G M ) and Drain-Source current (I DS ) vs. Gate voltage characteristics. Typical examples of these are given below for the LP 7612 device, which is a 0.25 x 200 µm discrete DHPHEMT. Note the low “knee” voltage, at about V DS = 0.5V, representing the transition from the linear to the saturated region. One of the most striking aspects is the large value of Drain- Source current attained with positive gate bias; for the DHPHEMT device, this maximum Drain-Source current, denoted I DSSM , can be as much as 75 to 100% above I DSS . In contrast, the MESFET might show 20-25% more current under the same forward gate bias. With both devices, the amount of forward gate bias that can be applied is limited to +0.7 to +0.8V, since larger values will drive the Gate into forward conduction. This PHEMT phenomenon has a number of implications for circuit designers. The transconductance vs. V GS curve shows a pronounced peak at about 80% of V P , the pinchoff voltage, where the current is effectively reduced to zero. The maximum extrinsic G m value is approximately 385 mS/mm, corresponding to an intrinsic transconductance of about 500 mS/mm. For optimum small-signal gain, an operating point of 70-80% of I DSS is recommended. Because the device’s I-V curves show symmetry about the V GS = 0V curve (i.e., I DSS ), the standard guidelines for Class A, AB, and C operation are different than for the MESFET. It is generally accepted that for true Class A operation, i.e., no distortion of the amplified signal, a quiescent operating point of I DS = 50% I DSS is the correct choice. This is based on consideration of the MESFET’s I-V curves and the derived G M vs. gate voltage characteristic. Notional examples of these are shown on the following page. TYPICAL PHEMT I-V CHARACTERISTICS I DS AND G M VS. GATE VOLTAGE LP7612 LP7612 DRAIN VOLTAGE (V) GATE VOLTAGE (V) D R A I N C U R R E N T ( m A ) D R A I N C U R R E N T ( m A ) Gate Voltage Start .8V Stop -1.2 Step -.2 V G M ( m S ) 150 120 90 60 30 0 150 120 90 60 30 0 150 120 90 60 30 0 0 1.2 2.4 3.6 4.8 6 .8 .4 0 -.4 -.8 -1.2 Filtronic Solid State Discrete FET / PHEMT Applications Notes 12 Class A operation requires, in the ideal case, a distortionless amplification of the input signal, and for a sinusoidal waveform, this also requires that no harmonic signals are generated. The standard for the measure of maximum “linear” power is the output power at 1 dB of gain compression (i.e., 1 dB below the small-signal gain), denoted P 1dB . In the case of the notional MESFET I-V curves given above, once the input signal drives the dynamic operating point into the gate-drain breakdown region (located in the lower right-hand corner), signal distortion can be expected. The load-line shown above is established by the output matching circuit, which must also resonate out the device’s output admittance. For maximum output power, the load-line must be chosen to maximize the peak-to-peak voltage and current amplitudes, since: P V I RF P P P P ≅ − − 1 8 In addition, biasing the device at less than 50% of I DSS can cause clipping of the output current waveform as the FET is driven beyond pinchoff (i.e., V GS > V P , the pinch- off voltage). Examination of the transconductance vs. gate voltage characteristic also shows that the region of constant G M is limited. For optimum linearity, MESFETs are biased at 50% of I DSS , and for larger small-signal gain, bias values of 60-75% are used. Higher power- added efficiency can be achieved at operating points less than 50% of I DSS , but with degraded linearity. Because the PHEMT I-V curves show that I DSSM is 150-175% of I DSS , it can be anticipated that for Class A operation, a quiescent bias point (QP) of 55-75% of I DSS is the correct choice. This bias point allows for a large current swing while avoiding current cut-off and its associated generation of harmonics. The selection of the quiescent Drain voltage then becomes a compromise between allowing for the largest voltage swing, while avoiding the Gate-Drain breakdown region. IDSSM IDSS 50% QP VGS = +0.5V VGS = 0V VGS = -VP V DS I DS INPUT SIGNAL APPLIED TO GATE OUTPUT CURRENT DELIVERED TO LOAD V GS 0.0 G M NOTIONAL MESFET I-V CURVES MESFET TRANSCONDUCTANCE CHARACTERISTIC GATE-TO-DRAIN BREAKDOWN REGION LOAD-LINE Filtronic Solid State Discrete FET / PHEMT Applications Notes 13 Power Linearity and Intermodulation Distortion (IMD) Because the PHEMT transconductance vs. gate voltage characteristic shows a peak at 70-80% of I DSS , as opposed to a relatively broad “plateau” of nearly constant G M , it is reasonable to suspect that the inherent linearity of the PHEMT is less than that of the MESFET. As is shown below, however, with a proper choice of the quiescent operating point, the PHEMT linearity and Intermodulation Distortion (IMD) performance is exceptionally good. Regardless of what measure is used to assess the fundamental linearity, the PHEMT meets or exceeds the “classical” power transfer model. It is useful to review the classical theory of nonlinearity in memoryless systems, 1 before presenting actual device data. A truly distortion-free two-port system will accept any input signal and produce a output signal that is a scaled version of the input, with only a linear phase shift. This means that the system transfer function H(jω) is given by: ( ) ( ) H j K j t o ω ω · − exp where K and t o are constants. For practical two-port systems with amplification, some non-linearity is of course inherent, but if the deviation from ideal linearity is mild, the two-port can be represented by a power series expansion of the input voltage, as follows: e k e k e k e with e A t o i i i i · + + · 1 2 2 3 3 1 cosω where e i = input signal, e o = output signal, A = input signal amplitude, ω 1 = signal frequency, and k 1,2,3 = expansion series coefficients. Expanding the power series and combining like terms, the output signal consists of a DC component, and components at the fundamental, 2nd, and 3rd harmonic frequency. The fundamental component has an amplitude which is, expressed in log units: G k k A dBm · + ¸ ¸ _ , 20 3 4 1 3 2 log where G = power gain. This is compared to the linear power gain: G k dBm o · 20 1 log where G o = linear or small-signal power gain. Note that if k 3 > 0, gain expansion occurs, while if k 3 < 0, then the two-port exhibits gain compression. If one now applies a second fundamental frequency or tone, the simultaneous two-tone input produces output signal components at: DC, ω 1 , ω 2 , 2ω 1 , 2ω 2 , and 3ω 1 , 3ω 2 ; there will also now be intermodulation products, i.e., second-order products at ω 1 ∀ ω 2 , and third-order products at: 2ω 1 ∀ ω 2 and 2ω 2 ∀ ω 1 . The third-order products are of particular concern, since they fall within the passband of a typical single- octave system, whereas the other terms do not. In a system with impedance R, the output power of the fundamental in the “linear” region of the power transfer characteristic will be, in dBm units: P R dBm o · ¸ ¸ _ , ¹ ' ¹ ¹ ¹ ¹ ; ¹ ¹ ¹ 10 2 10 1 2 3 log k A Once the two-port begins deviating from linearity, the term k 1 A is replaced by: k A k A 1 9 4 3 3 + . For all practical purposes, the two-port can be considered essentially linear if operated more than 10 dB below the 1 dB compression point. The power of the 3rd-order intermodulation products is given by: P k A R dBm 2 3 4 3 3 2 3 1 2 10 2 10 ω ω − · ¸ ¸ _ , ¹ ' ¹ ¹ ¹ ¹ ; ¹ ¹ ¹ log Note that the slope of the 3rd-order products increases at a 3:1 ratio compared to the linear power expression. Thus if the linear output power is extrapolated, it will intersect the 3rd-order product extrapolation, as a function of input power. This intersection point, the 3rd- order intermodulation “intercept point,” is useful for characterizing non-linear systems. This behavior is depicted in the figure below, along with an idealized two- tone IMD spectrum. 1. Ha, Tri T., Solid-State Microwave Amplifier Design, Wiley, New York, 1981, pp. 203-209. Filtronic Solid State Discrete FET / PHEMT Applications Notes 14 Using the relationships given above and solving analytically for the intercept point yields the expression: P k k R dBm IP · ¸ ¸ _ , ¹ ' ¹ ¹ ¹ ¹ ; ¹ ¹ ¹ 10 2 3 10 1 3 3 3 log Thus the intercept point in this example is independent of input power, and from this one also derives the well- known expression: P P dBm IP dB · + 1 10 63 . Another often-used expression relates the 3rd-order IMD product power levels to the fundamental power level, as follows: P P P dBm IP 2 1 2 1 3 2 ω ω ω − ≅ − 3 1 1 1 Input Power (dBm) Output Power (dBm) P IP 3rd-order products fundamental FUNDAMENTAL AND 3RD-ORDER IMD PRODUCT POWER TRANSFER CURVES ω ω 1 ω 2 2ω 2 - ω 1 2ω 1 - ω 2 ω 1 ω 2 ω INPUT SPECTRUM OUTPUT SPECTRUM Filtronic Solid State Discrete FET / PHEMT Applications Notes 15 Typical two-tone IMD performance data is presented below for the LP 7612 and LP 6872 discrete DHPHEMT devices, illustrating the effect of the quiescent OP. This effect is especially pronounced for the LP 7612, showing quite clearly that the operating current needs to be greater than 50% of I DSS in order to show zero gain expansion and optimum IMD performance. The power linearity was assessed by four methods: 1). Examination of the power transfer curve; qualitatively, the presence or absence of gain expansion. 2). Curve-fitting the P IN vs. P OUT data to the “classical” theory. 3). Intercept Point determination by graphical extrapolation. 4). Intercept Point determination by direct calculation, using: P P P dBm IP 2 1 2 1 3 2 ω ω ω − ≅ − The devices were characterized using the test equipment configuration shown below. Low-loss triple-sleeve coaxial tuners match the DUT to 50 ohms, which are manually tuned for optimum power and/or gain. The discrete device is eutectically die-attached onto a gold-plated copper groundbar/heatsink, which in turn is mounted between a pair of 50 ohm microstrip lines (on alumina). Bond wires connect the DUT’s Gate and Drain to the microstrip lines. Coaxial-to-microstrip launchers complete the measurement fixturing. The TWT amplifier is operated more than 25 dB below its 1dB compression point, and does not contribute intermodulation products. Input and output losses are corrected up to the coaxial-to- microstrip adapters. SIGNAL GENERATORS TWTA TUNER BIAS TEE POWER METER SENSOR -VG DUT TUNER +VD POWER METER SENSOR SPECTRUM ANALYZER TEST CONFIGURATION FOR TWO-TONE INTERMODULATION MEASUREMENTS Filtronic Solid State Discrete FET / PHEMT Applications Notes 16 LP 7612 5V 75% IDSS -30 -20 -10 0 10 20 30 40 3.3 4.3 4.8 6 6.9 8.1 8.5 9.7 10.7 11.7 12.9 14 14.5 15.7 16.6 17.6 18.6 19.6 20.6 21.6 22.6 23.6 INPUT POWER (dBm) POUT 1-TONE POUT EXT GP P3rd EXT P3rd ABS PINT CALC This data was taken on a LP 7612 DHPHEMT discrete device, as described above, biased at V DS = 5V, I DS = 75% I DSS . The actual data, corrected to a single- tone basis, is shown as the “POUT 1-TONE” and “P3rd ABS”. The first is the fundamental tone P OUT vs. P IN data, while the second is the actual (absolute) power level of the 3rd-order IMD products. Note that the actual 3rd-order data follows the expected 3:1 slope line, whose intersection with the linear extrapolation of the small- signal power gain shows a 3rd-order Intercept Point of +35 dBm. The single-tone output power at 1 dB gain compression, P 1dB = +21.5 dBm, with a Power-Added Efficiency PAE = 47%. Thus the device exceeds the expected: P 1dB + 10.6 dB = +32.1 dBm, and the PAE is consistent with Class A operation. The calculated Power Gain is shown as “GP” on the plot above, showing no gain expansion or other nonideal behavior prior to compression. Also shown as “PINT CALC” are the values calculated from: P P P IP 2 1 2 1 3 2 ω ω ω − ≅ − , showing constant IP3 values in the linear region, and reasonably good agreement with the graphical determination. Finally, the P OUT vs. P IN data fits well to the power transfer model, with k 1 = 3.89, and k 3 = -0.422. Filtronic Solid State Discrete FET / PHEMT Applications Notes 17 LP 7612 5V 50% IDSS -30 -20 -10 0 10 20 30 40 -0.3 0.8 1.8 2.9 3.6 4.5 5.4 6 7 8.1 9.1 9.7 10.7 11.8 12.9 13.7 14.6 15.2 16.2 17.2 18.2 19.2 INPUT POWER (dBm) POUT 1-TONE POUT EXT GP P3rd ABS P3rd EXT PINT CALC In the plot shown above, the operating current has been reduced to 50% of I DSS ; the single-tone P 1dB = +22.2 dBm, with a PAE of 59%, which indicates Class AB operation. Note that the extrapolation method for determining IP3 shows a value of +30.5 dBm, somewhat lower than the expected 22.2 + 10.6 = 32.8 dBm. Gain expansion is seen at input power levels between +5 to +9 dBm, and the power transfer model fits poorly to the actual data. The table below summarizes the bias effects on device linearity. The “75D” data shows the effect of de-tuning the output match to the device by 1 dB, to simulate a non-ideal circuit matching environment. There is no discernible effect on linearity. LP 7612 DHPHEMT IMD PERFORMANCE VS. OPERATING POINT V DS = 5.0V TEST FREQ. = 18 GHz I DS (%) P 1dB (dBm) PAE (%) Gain Expansion Calculated IP3 (dBm) Graphical IP3 (dBm) 30 21.7 59 0.3 dB 28-30 35.0 50 22.2 59 0.8 dB 28.5 30.5 75 21.5 47 None 35.0 35.0 75D 20.5 38 None 35.0 35.5 100 20.7 30 None 28.0 30.0 Filtronic Solid State Discrete FET / PHEMT Applications Notes 18 Similiar data for the LP 6872 device is shown below: LP 6872 8V 50%IDSS -20 -10 0 10 20 30 40 50 13.4 14.5 15.5 16.4 17.5 18 19.1 20.3 21.3 21.9 22.9 24 25 26 27 28 29 30 31 32 33 34 INPUT POWER (dBm) POUT 1-TONE POUT EXT GP P3rd ABS P3rd EXT PINT CALC LP 6872 DHPHEMT IMD PERFORMANCE VS. OPERATING POINT TEST FREQ. = 18 GHz V DS (V) I DS (%) P 1dB (dBm) PAE (%) Gain Expansion Calculated IP3 (dBm) Graphical IP3 (dBm) 8 30 27.9 42 0.1 dB 38-39 38 8 50 28.3 44 0.2 dB 39-41 43 7 75 28.2 49 0.1 dB 36-39 42 8 75 28.2 43 0.3 dB 35-38 40 8 85 28.3 40 None 41-42 42 The LP 6872 DHPHEMT, nominally P 1dB = 27 dBm at 18 GHz, exhibits good IMD performance over a wider range of bias conditions, but once again optimum linearity is achieved with a quiescent operating point set at or above 50% I DSS . As a further degree of freedom, the operating voltage can be adjusted over the 6-9V range without significant degradation of device linearity. This may be useful for optimizing the gain and power matching conditions. Filtronic Solid State Discrete FET / PHEMT Applications Notes 19 Operating Current “Pulling”: 0 10 20 30 40 50 60 -5 0 5 10 15 20 25 PIN (dBm) Ids 5v100p Ids 5v70p Ids 5v30p Ids 5v50p I DS vs. Input Power at V DS = 5V I DSS = 56 mA mA The data presented above shows the typical current “pulling” effect for the LP 7612 DHPHEMT discrete device, biased at V DS = 5V, at a test frequency of 15 GHz. The vertical axis shows the Drain supply operating current as a function of input power for 30, 50, 70, and 100% of I DSS . For most bias conditions, the operating current pulls up as the device is driven into saturation; for the PHEMT this is due to instantaneous current (i.e., RF current) excursions above I DSS , as discussed in Sec. 6. This data shows input power levels equating to 12 dB of gain compression; note the rapid decrease in current at extreme drive levels, P IN > 18 dBm, that indicates the onset of Gate conduction current. Continuous RF input power levels that result in large values of gain compression is not recommended for reliable operation. Note that for a quiescent operating point (OP) at 100% I DSS , the onset of Gate conduction current occurs at a slightly lower input power level, and that the operating current decreases significantly. Any appreciable amount of Gate current results in a shift of the OP due to an ohmic drop of about 75-100 mV due to the Gate metal resistance, which is about 7Ω for this device, in addition to the circuit’s grounding resistor, if any. Filtronic Solid State Discrete FET / PHEMT Applications Notes 20 7. SMALL-SIGNAL S-PARAMETERS AND LUMPED-ELEMENT MODELS A complete listing of small-signal scattering parameters for all production-released FSS discrete MESFET and PHEMT devices is available from an authorized Sales Representative or directly from the foundry. This DOS-formatted 3.5” diskette contains S- parameters files in the “Touchstone” format, which is denoted by the suffix “.s2p” on the file names. This diskette also contains several “Readme” text files that contain additional measurement-specific information, such as bond wire lengths, etc. The S-parameter files can be viewed using the MS-DOS™ Editor, or with MS Windows™ Notepad (the user will need to “Associate” the file with “Text File [Notepad.exe]”). The S-data file contains one or two comment lines, followed by a header line such as: “# S GHZ MA 50” This designates: S-parameters, frequency units in GHz, Magnitude-Angle format (or “RI” Real-Imaginary), and 50 ohms normalization. The header line is followed by the actual frequency and S-parameter data. The Touchstone format can be directly imported into standard CAD programs such as HPEEsof Libra™ or MDS™. Note that some files contain a “n” in the filename, such as “lp76_1n.s2p”, which indicates that bias-dependent Noise Characterization data is appended to the file, also in Touchstone format. For S-parameter characterization, the discrete device was eutectically die-attached onto gold-plated molybdenum carriers (for optimum heatsinking), along with two JMicrotechnology Probepoint™ substrates. The devices were then wire-bonded (using 0.0007” or 18 µm gold wire) from the appropriate bond pads to the substrates. These substrates accommodate the GGB Picoprobes™, which are in a ground-signal-ground configuration on a 150 µm pitch. The input and output reference planes of measurement are therefore up to the bond wires; all S-parameter files contain bond wire effects, as detailed in the “Readme” files. Actual measurements were taken by a Hewlett-Packard 8510C Network Analyzer. Lumped-element small-signal models were created by directly fitting to the measured S-parameter data, by optimization for minimum error between actual and modeled data. The resulting model was then subjected to verification testing by comparison with known parameters such as gate resistance and bond wire effects. The basic lumped-element model is as follows: I = v G e DS C M - j T D ω 1+ j f F L G R G C GS R I GATE C DG C DC R S L S SOURCE I DS R DS C DS R D L D DRAIN + _ VC LUMPED-ELEMENT MODEL Filtronic Solid State Discrete FET / PHEMT Applications Notes 21 FSS FET AND PHEMT LUMPED ELEMENT MODELS ELEMENT LF 6836 LF 6872 LF 3814 LF 3830 LF 3850 LP 7512 (low/up) LP 7612 (low/up) LP 6872 L G , nH 0.24 0.14 0.14 0.08 0.06 0.14 0.17/0.20 0.14 R G , Ω 10.7 5.3 2.0 1.18 0.20 5.3/6.4 7.0/4.4 1.0 C GS , pF 0.58 1.00 1.88 5.17 7.10 0.28/0.27 0.57/0.51 1.49 R I , Ω 0.80 0.16 2.79 0.0 0.57 1.50 0.01 1.3 C DC , pF 0.02 0.04 0.14 0.09 0.06 0.05/0.04 0.02 0.01 G M , mS 90 130 200 470 580 90/100 90/100 230 T D , pS 0.03 0.07 0.68 0.01 0.04 0.01 0.04/0.10 0.03 F, GHz 41 45 25 25 30 55/55 50/44 41.2 R DS , Ω 300 525 76.8 34.0 30.0 228/282 497/598 173 C DS , pF 0.05 0.12 0.12 0.49 0.68 0.03 0.06 0.16 C DG , pF 0.04 0.09 0.03 0.05 0.35 0.04 0.01 0.04 R D , Ω 1.2 0.76 0.09 0.04 0.10 0.10 0.16/1.5 0.20 L D , nH 0.19 0.11 0.12 0.09 0.07 0.10 0.16/0.06 0.13 R S , Ω 4.2 1.1 1.82 1.60 0.12 2.94/3.26 2.74/3.26 0.96 L S , nH 0.02 0.04 0.02 0.04 0.003 0.02/0.03 0.03 0.03 V DS , V 5.5 5 7 9 9 2 5 8 I DS , % 50 50 50 50 50 25 50 50 I DSS , mA 123 240 246 589 712 40 68 240 NOTES: 1). Model elements optimized to the following accuracies: S 11 : Magnitude: ≤ 2% Phase: ≤ 3° S 22 : Magnitude: ≤ 4% Phase: ≤ 5° S 21 : Magnitude: ≤ 10% Phase: ≤ 5° S 12 : Magnitude: ≤ 10% Phase: ≤ 5° Accuracy defined with respect to measured S-data. 2). low/up: “low” optimized for 2-12 GHz band. “hi” optimized for 10-26 GHz band. Filtronic Solid State Discrete FET / PHEMT Applications Notes 22 Typical modeled vs. measured S-parameter comparison is given below: B1 S11 dat a S[ 1. 1] B2 S11 model S[ 1. 1] 0.0 0.2 0.5 1.0 2.0 5.0 inf 0.0 0.2 0.5 1.0 2.0 5.0 -5.0 -2.0 -1.0 -0.5 -0.2 Fr equency 2. 0 t o 12. 0 Ghz LP 7612 Vds5V I ds=50% l dss=68mA Opt i mi zed f or 2 t o 12 GHz B1 S22 dat a S[ 2. 2] B2 S22 model S[ 2. 2] 0.0 0.2 0.5 1.0 2.0 5.0 inf 0.0 0.2 0.5 1.0 2.0 5.0 -5.0 -2.0 -1.0 -0.5 -0.2 Fr equency 2. 0 t o 12. 0 Ghz LP 7612 Vds5V I ds=50% l dss=68mA Opt i mi zed f or 2 t o 12 GHz B1 S12 dat a S[ 1. 2] B2 S12 model S[ 1. 2] Fr equency 2. 0 t o 12. 0 Ghz LP 7612 Vds5V I ds=50% l dss=68mA Opt i mi zed f or 2 t o 12 GHz 0.05 0.04 0.03 0.02 0.01 0.0 B1 S21 dat a S[ 2. 1] B2 S21 model S[ 2. 1] Fr equency 2. 0 t o 12. 0 Ghz LP 7612 Vds5V I ds=50% l dss=68mA Opt i mi zed f or 2 t o 12 GHz 6.0 5.0 4.0 3.0 2.0 1.0 0.0 150 120 90 60 30 0 180 -150 -90 -60 -30 -120 180 -150 -120 -90 -60 -30 0 30 60 90 120 150 Filtronic Solid State Discrete FET / PHEMT Applications Notes 23 8. BIASING CIRCUITS AND STABILIZATION TECHNIQUES Basic Self-Bias Configuration: A typical hybrid amplifier circuit utilizing discrete devices and implemented in microstrip is shown schematically below, along with a sketch of the actual circuit. In this particular example, the MESFET or PHEMT active device is “self-biased,” so that a single DC voltage supply is used; this is done by floating the DC ground of the device with discrete RF bypass capacitors. The device’s gate is DC grounded through the thin-film resistor as shown, and therefore the required negative gate-source voltage is established by the voltage drop across the source resistor. The drain bias is fed through a quarter-wavelength high-impedance line which is RF grounded at one end. For this illustration, conjugate gain matching is accomplished by use of “Series-L Shunt-C” matching networks on both the input and output circuits, implemented as microstrip lines and open-circuited stub lines. INPUT MATCHING NETWORK OUTPUT MATCHING NETWORK RF OUT RF IN SOURCE RESISTOR NETWORK WITH RF BYPASS CAPACITOR GATE GROUNDING RESISTOR WITH RF CHOKE +VS DC DE-COUPLING CAPACITOR (x2) 8/4 HIGH-IMPEDANCE MICROSTRIP LINE FOR RF CHOKE (SHORTED STUB) THIN-FILM RESISTOR LADDER FOR BIAS LEVEL SETTING “SERIES-L SHUNT-C” MATCHING NETWORK IMPLEMENTED AS MS LINE + OPEN-CIRCUITED LINE RF CHOKE + THIN-FILM RESISTOR FOR DC GATE GROUNDING Filtronic Solid State Discrete FET / PHEMT Applications Notes 24 Gate Grounding Resistor: In many instances, the gate DC grounding resistor and its associated RF choke can serve a dual role; in addition to providing the necessary DC ground for the gate of the active device, this network can provide stabilization at frequencies below the circuit’s passband. Typically (and especially for low-noise designs) the RF choke is implemented as a quarter-wavelength high- impedance microstrip line. Caution should be used for lengths other than 90° effective length in the passband, since the noise performance will be degraded. For low- noise devices (such as the LP 7512 and LP 7612) a resistor value of 5-15 ohms is satisfactory. For some applications it is desirable to eliminate the grounding resistor, and in these cases care must be taken to ensure that the circuit is stable below the passband, especially below 2 GHz, where the inherent gain of the PHEMT devices becomes very large. Alternately, the grounding resistor may be placed at the “top” of the RF choke, and in this configuration smaller or large resistor values may be used. This is sometimes desired for power devices, to avoid re-biasing if the device begins to draw gate current. Source Resistor Ladders: For self-biased circuits, the DC operating current is established by the voltage drop across the source resistor(s), so it is important to provide suitable combinations of resistors. In this instance, the PHEMT devices simplify the design process, since the Pinch-Off Voltage (VP) is nominally 0.8V, and is somewhat invariant with I DSS and device size (gate width). Suggested resistor-ladder values are given in the table below: Nominal IDSS (mA) RS Sequence (Ω) 30 30 / 15 / 7.5 / 3.25 / 1.63 60 20 / 15 / 7.5 / 3.25 / 1.63 120 5 / 2.5 / 1.25 / 0.63 / 0.31 220 2 / 1 / 0.5 / 0.25 / 0.125 450 1 / 0.5 / 0.25 / 0.125 / 0.06 650 0.8 / 0.4 / 0.2 / 0.1 / 0.05 These values allow for adjustment about a nominal 50% I DSS , which will occur at approximately 50% V P . For FSS’s MESFET devices, the Gate-Source voltage required for a 50% I DSS operating point can be provided for a specific wafer lot if needed. Stabilization: The general issue of stability is now considered, using a LP 7612 PHEMT device as an example. The data presented in the table following was taken at V DS = 5V, I DS = 50% I DSS , with no source bypass capacitors; bond wire inductances are included, with gate bond wires typically 0.004-0.006” long (2 per device), drain wires 0.006-0.008”( 2 per device), and source wires 0.006- 0.008” (4 per device). Note that the Stability Factor K is less than unity at all frequencies, indicating that device as is potentially unstable, depending upon the load and source impedances presented to the device. The Gain column presents either the Maximum Available Gain, G MAX , when k > 1.0, or Maximum Stable Gain, G MS , in the case where K < 1.0. For k < 0, the device is unstable in a 50Ω system. Examination of the Stability Circles 1 indicates that the largest regions of potential instability are in the load impedance plane, therefore careful attention must be paid when designing output matching networks. The Stability Factor k is defined as: k S S D S S where D S S S S ≡ − − + ≡ − 1 2 11 2 22 2 2 12 21 11 22 12 21 : where Sij = two-port power scattering parameters. The power gains G MAX and G MS are defined as: ( ) G S S k k G S S MAX MS ≡ t − ≡ 21 12 2 21 12 1 Stability circle definitions can be found in Vendelin 1 . 1. G. D. Vendelin, Design of Amplifiers and Oscillators by the S-Parameter Method, Wiley, New York, 1982. Filtronic Solid State Discrete FET / PHEMT Applications Notes 25 LP 7612 S-Parameters, Stability and Power Gain FREQ. S11 S21 S12 S22 GMAX GMS k S21 (dB) (GHz) MAG ANG MAG ANG MAG ANG MAG ANG (dB) 2 0.95 -37° 5.95 152° 0.02 73° 0.83 -9° 25.8 0.18 15.5 6 0.78 -96 4.34 109 0.03 51 0.77 -23 21.1 0.52 12.7 10 0.68 -137 3.13 79 0.04 45 0.74 -35 19.0 0.80 9.9 14 0.65 -166 2.41 56 0.05 48 0.74 -47 17.2 0.89 7.6 18 0.64 171 1.93 36 0.06 54 0.74 -61 15.3 0.84 5.7 22 0.67 152 1.63 18 0.08 57 0.73 -74 13.4 0.61 4.3 26 0.70 131 1.37 0 0.10 56 0.75 -86 11.3 0.35 2.7 30 0.72 115 1.24 -16 0.15 53 0.81 -101 9.2 0.01 1.8 As can be seen from this data, the LP 7612 provides excellent gain to 25 GHz and beyond, but does require stabilization at lower frequencies. For example, if the device is to be used in a 6 - 18 GHz gain stage, there is potential for instability below 5 GHz. One method to improve this situation is to apply the RF choke / grounding resistor network to the input of the device; using a 5 mil wide by 65 mil long (0.13 by 1.65 mm.) microstrip line (on 10 mil [0.25 mm] alumina), with a 20Ω grounding resistor gives the following: FREQ. k GMAX GMS 2 6.14 14.2 6 0.84 20.8 10 0.76 18.8 14 0.85 17.2 18 0.87 15.4 22 0.71 13.5 26 0.57 11.4 30 0.47 9.4 Now the stability is improved at frequencies below the passband, with little degradation of G MS in the passband of interest. For additional stabilization above the passband, the network can be modified to include a shunt capacitor to ground (implemented as an open- circuited stub), as shown below: Both distributed elements are quarter- wavelength (at the band center frequency f C ) high- impedance lines. Above the passband, the open- circuited line acts as a shunt capacitance to ground, thereby improving the stability. This line’s width may be varied to improve the roll-off characteristics above the passband. For low-frequency applications, classical resistive shunt feedback may be used to simultaneously stabilize and match the device. For example, a LP 7612, with two 50 pF source bypass capacitors, can be stabilized with a series R-L network (R=450Ω, L=1 nH) for operation in the 2-4 GHz band. This results in an insertion gain S 21 = 11.5 dB, and input/output return loss of better than -8 dB. Note that the phase of S 21 for this device is such that this technique is useful only for frequencies below about 8 GHz. λ/4 at fC 20Ω Filtronic Solid State Discrete FET / PHEMT Applications Notes 26 9. LARGE-SIGNAL MODELS AND OPTIMUM POWER MATCH DATA Introduction Two large-signal models were utilized to characterize the non-linear properties of FSS’s discrete MESFET and PHEMT devices, the Root model (developed by Hewlett-Packard) and the classical Curtice model. The Root model extraction consists of a file that is directly read by Libra™ (Release 6.0) and MDS™. The Curtice model can also be used with these CAD programs, but was originally derived for use with SPICE. Curtice model elements are presented in tabular form in this section, but the Root model extractions are available only on the FSS Devices Data Diskette. For convenience, Optimum Power Match data is presented below for the devices that were Root-modeled. The Root model represents a departure from standard non-linear models in that there are no closed- form analytical equations that model the I-V curves and non-linear capacitance effects as in the Curtice model. Therefore the Root model does not assume any particular functional relationships for the I-V curves, and is inherently more accurate. One disadvantage of the model is that it does not provide any additional insight into the device physics, so it cannot be compared to theory directly; it is a “file based” model. Root Model Background The Root model replaces several of the elements in the linear model (Sec. 7) with “contour maps” of charge vs. terminal voltage. Specifically, C GS , C GD and C DS are replaced with fitted contour functions that map the charge in the device as a 2-dimensional function of terminal voltages; the device’s parasitic elements, such as R G , L G , etc., are modeled with lumped elements. For example, the charge under the gate, Q G , is fitted to such a contour by measuring the device’s S- parameters at many points along the I-V curves; from this data the voltage-dependent capacitances C G and C GD are mapped as shown in the figure below. Essentially the Root model derives an admittance matrix which, along with the parasitic elements, can predict the non-linear behavior of a device without assuming a specific functional behavior. The result is a fairly accurate high-frequency large-signal model. There are some limitations, however, for example, the requirement to measure S-data over the device’s I-V curves is impractical with large power PHEMT devices, since they are inherently unstable in certain regions. Temperature effects and modeling of Class B and C operation are also not addressed by the model. Despite these limitations, the Root model represents one of the most accurate large-signal models available today that is supported in commercial CAD software. ( ) ( ) C V ,V Y V ,V G GS DS 11 meas GS DS · Im ,ω ω ( ) ( ) C V ,V Y V ,V GD GS DS 12 meas GS DS · −Im ,ω ω V DS V GS Q G dQ G = -C GD dV DS dQ G = C G dV GS Root Model Q G Charge Contour Filtronic Solid State Discrete FET / PHEMT Applications Notes 27 SUMMARY OF FSS DISCRETE DEVICE LARGE-SIGNAL MODEL EXTRACTIONS: DEVICE MODEL LP 7612 Root LP 6872 (half-cell) Root LF 6836 Root LF 3814 Root LF 6872 Root LF 3830 Curtice LF 3850 Curtice POWER PERFORMANCE AND OPTIMUM POWER MATCH AS PREDICTED BY THE ROOT MODEL: DEVICE FREQUENCY (GHz) OPTIMUM POWER MATCH P 1dB (dBm) POWER GAIN (dB) MAG PHASE LP 7612 5V, 50% IDSS = 63 mA 12 18 24 30 0.492 0.460 0.440 0.502 75.4° 102.0 132.1 161.1 18.4 18.6 18.6 18.3 14.8 11.0 8.9 7.9 LP 6872 8V, 50% IDSS = 240 mA 6 12 18 24 0.455 0.433 0.506 0.607 105.0 141.0 165.1 -176.0 26.3 27.2 27.4 27.5 19.7 12.7 9.8 8.4 LF 6836 5.5V, 50% IDSS = 142 mA 12 18 0.404 0.345 87.2 121.4 19.7 20.3 9.7 7.1 LF 3814 7V, 50% IDSS = 270 mA 4 8 12 0.526 0.448 0.500 92.4 122.2 143.2 23.3 25.3 25.6 16.7 9.7 7.2 LF 6872 8V, 50% IDSS = 257 mA 6 12 0.486 0.432 76.7 112.5 23.0 24.7 12.4 7.6 Filtronic Solid State Discrete FET / PHEMT Applications Notes 28 NOTES: 1). Root Model files are found on the FSS Devices Diskette; these files have the extension “.raw” The file should be directly read into the simulator; for example, in Libra the Root file should be place in the “Data” subdirectory in a specific Project Directory. These files are in ASCII text format. Do not scale the models. Predicted maximum power (at 1 dB compression) is generally within 1-1.5 dBm of actual values; simulated power performance should be taken as a relative performance indication, not as an absolute. 2). The LP 6872 Root Model extraction was done for a half device, due to inherent instability in a 50Ω system. The user should place two instances of the “HPFET” item, and connect the Gate, Drain, and Source connections together to form the complete, composite 6872 device. CURTICE MODEL EXTRACTIONS: PARAMETER AND UNITS LF 3830 LF 3850 BETA, 1/V 0.0644 0.3668 GAMMA, 1/V 0.5 0.5 VOUTO, V 5.0 9.0 VTO, V -3.73 -1.735 A0, A 0.01 0.01 A1, A/V 0.001 0.001 A2, A/V 2 -0.001 -0.001 A3, A/V 3 -0.0001 -0.0001 TAU, sec. 0.0 0.0 R1, Ω 0.0 0.0 R2, Ω 0.0 0.0 VBO, V 100 100 VBI, V 0.6292 0.6119 RF, Ω 0.0 0.0 IS, A 1.019 x 10 -11 1.019 x 10 -11 N 1.228 1.186 RDS, Ω 44.84 33.88 CRF, Fd 1 x 10 -6 1 x 10 -6 RD, Ω 0.2 0.474 RG, Ω 0.3 0.01 RS, Ω 0.5 0.24 RIN, Ω 4.401 -2.028 CGSO, Fd 5.89 x 10 -12 1.07 x 10 -12 CGDO, Fd 0.0 0.0 FC 0.5 0.5 CDS, Fd 2.94 x 10 -13 1.67 x 10 -12 CGS, Fd 0.0 0.0 CGD, Fd 1.83 x 10 -13 3.91 x 10 -13 LG, H (gate inductance) 1.07 x 10 -10 6.30 x 10 -11 LD, H (drain inductance) 6.60 x 10 -11 6.20 x 10 -11 LS, H (source inductance) 4.00 x 10 -12 5.00 x 10 -12 LBG, H (gate pad inductance) 3.50 x 10 -11 3.50 x 10 -11 LBD, H (drain pad inductance) 2.50 x 10 -11 2.50 x 10 -11 LBS, H (source pad inductance) 2.50 x 10 -11 2.50 x 10 -11 Filtronic Solid State Discrete FET / PHEMT Applications Notes 29 10. EXAMPLE CIRCUITS Gain Stage for the LP 7612 DHPHEMT: A gain stage was developed for the LP 7612 discrete device, implemented in microstrip on a 10 mil alumina substrate; the device is assumed to be die- attached onto a groundbar, with source bypass capacitors (2 at 8 pF each). This will allow self-biasing, although the topology shown below does not include biasing chokes. Optimization of the gain response assumes that this stage will be utilized in a balanced configuration, with 90° quadrature hybrids (e.g., Lange couplers). ELEMENT TYPE DIMENSIONS (mils) INPUT CIRCUIT: (Element nomenclature after conventions used in Libra™, Touchstone™) 1 MLIN W=10 L=40 (50Ω INPUT LINE) 2,3 MLEF 2 STUBS (OPEN-CIRCUITED MLIN) W=11.9 L=17.5 4 MCROS W1=10 W2=6 W3=W4=11.9 5 MLIN W=6 L=25.5 6,7 MLEF 2 STUBS W=13.8 L=11.6 8 MCROS W1=6 W2=21.1 W3=W4=13.8 9 MLIN W=21.1 L=15.9 10,11 MLEF 2 STUBS W=13.5 L=20.3 12 MCROS W1=21.1 W2=10.2 W3=W4=13.5 13 MLIN W=10.2 L=7 14 MTEE W1=10.2 W2=10.2 W3=2 15 MLIN W=2 L=67 (GATE GROUNDING CHOKE) 16 TFR RS=50 W=2 L=1 (GATE GROUNDING RESISTOR) 17 MLIN W=10.2 L=3 OUTPUT CIRCUIT: 1 MLIN W-5.3 L=28.6 2,3 MLEF 2 STUBS W=9 L=9.4 4 MCROS W1=5.3 W2=11.6 W3=W4=9 5 MLIN W=11.6 L=27 6 MLEF 1 STUB W=6.4 L=22.6 7 MTEE W1=11.6 W2=10 W3=6.4 8 MLIN W=10 L=40 (50Ω OUTPUT LINE) INPUT CIRCUIT OUTPUT CIRCUIT LP 7612 NOT SHOWN: RF CHOKES FOR DRAIN BIASING, DC BLOCKING CAPACITORS TUNING PADS MAY BE ADDED AS DESIRED. 50Ω INPUT/OUTPUT MLINS MAY BE SHORTENED AS NEEDED. Filtronic Solid State Discrete FET / PHEMT Applications Notes 30 Simulated linear (small-signal) performance of this gain stage is shown below; the SSG was 9.75 dB over the 4 to 20 GHz band, with t0.25 dB of gain ripple. Minimum stability (K-factor) was 0.7 at 8 GHz In a balanced configuration with Lange couplers (W=1.2 mils, S=0.85 mils, and L=97 mils), the SSG was 9.5 t0.5 dB, with a minimum k-factor of 2.0. The improvement in stability in a balanced configuration is typical, since the potentially unstable regions in the source and load planes are transformed to areas outside the normal Smith chart (Γ ≥ 1.0). The Root model was utilized to simulate the power performance, and the gain stage as described above showed 5 dB of power slope across the band. SINGLE-ENDED GAIN STAGE PERFORMANCE 12.0 11.0 10.0 9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.5 1.4 1.3 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 2.0 24.0 Prototype LP 7612 Gain Stage Linear Analysis 7612gain_tb S21 7612gain S[2,1] db 7612gain_tb K1 7612gain K Frequency 2.0 GHz/DIV Filtronic Solid State Discrete FET / PHEMT Applications Notes 31 The output circuit can be re-optimized for better power performance, resulting in the modified output section shown below. The optimum power match data presented in the preceding section was used for this re- optimization, and in a self-biased balanced configuration, a pair of 7612 devices provided 19.0 to 20.8 dBm of linear power (simulated) over the 6-18 GHz band as shown. Note that the devices were biased at about 70% of I DSS , and that the Root model correctly predicted no gain expansion prior to compression. ELEMENT TYPE DIMENSIONS (mils) MODIFIED OUTPUT CIRCUIT: 1 MLIN W=2.1 L=39.7 2 MLEF STUB W=11.5 L=25.5 4 MTEE W1=2.1 W2=10 W3=11.5 5 MLIN W=10 L=40 (50Ω OUTPUT LINE) POWER PERFORMANCE WITH MODIFIED OUTPUT STAGE MODIFIED OUTPUT CIRCUIT MAY BE MEANDERED TO FIT SPACE RESTRICTIONS AND ALLOW FOR TUNING OF ELEMENT LENGTH 13.0 12.0 11.0 10.0 9.0 8.0 7.0 25.0 20.0 15.0 10.0 5.0 0.0 -5.0 -10.0 20.0 7612balpow_tb PF1 7612balgain PF dBm 7612balpow_tb GAIN 7612balgain OUT_EQN dB Frequency, GHz 6.0 12.0 18.0 Linear Power (P1dB): 6 GHz 12 GHz 18 GHz Power 5.0 dBm/DIV Prototype LP 7612 Gain Stage w/Power Match Balanced Configuration Vds=5.5V Ids=45mA (per device) Self-Bias RS=8 ohms Filtronic Solid State Discrete FET / PHEMT Applications Notes 32 6-18 GHz Power Stage for the LP 6872 DHPHEMT: ELEMENT TYPE DIMENSIONS (mils) INPUT CIRCUIT: 1 MLIN W=10 L=40 (50Ω INPUT LINE) 2,3 MLEF 2 STUBS W=6.9 L=13.7 4 MCROS W1=10 W2=3.2 W3=W4=6.9 5 MLIN W=3.2 L=12 6 MLEF 1 STUB W=17.1 L=4.4 7 MTEE W1=3.2 W2=6.9 W3 =17.1 8 MLIN W=6.9 L=34.6 9,10 MLEF 2 STUBS W=24 L=20.6 11 MCROS W1=6.9 W2=37.5 W3=W4=24 12 MLIN W=37.5 L=16 13 MTEE W1=37.5 W2=37.5 W3=3 14 TFR RS=50 W=3 L=2.5 (GATE GROUNDING RESISTOR) 15 MLIN W=3 L=48 (GATE GROUNDING CHOKE) 17 MLIN W=37.5 L=3 OUTPUT CIRCUIT: 1 MLIN W=4.3 L=20.4 2,3 MLEF 2 STUBS W=30.4 L=16.7 4 MCROS W1=4.3 W2=4.6 W3=W4=30.4 5 MLIN W=4.6 L=33.8 6 MLEF 1 STUB W=14.2 L=26.7 7 MTEE W1=4.6 W2=10 W3=14.2 8 MLIN W=10 L=40 (50Ω OUTPUT LINE) INPUT CIRCUIT OUTPUT CIRCUIT LP 6872 NOT SHOWN: RF CHOKES FOR DRAIN BIASING, DC BLOCKING CAPACITORS TUNING PADS MAY BE ADDED AS DESIRED. 50Ω INPUT/OUTPUT MLINS MAY BE SHORTENED AS NEEDED. Filtronic Solid State Discrete FET / PHEMT Applications Notes 33 The Root model was utilized to optimize the power performance of this circuit, and with two LP 6872 PHEMTs, in a self-biased balanced configuration, produced over 30 dBm of linear power, with about 9 dB of power gain. The model did exhibit some irregularities under conditions of large input power levels which created convergence problems during the simulation. In the plot below, the circuit is not fully compressed by 1 dB across the band, but does exhibit acceptable output power over the 6-18 GHz band. Simulated SSG = 9.5 dB, with t0.4 dB of gain ripple. POWER PERFORMANCE IN A BALANCED CONFIGURATION: Output Power Power Gain 15.0 14.0 13.0 12.0 11.0 10.0 9.0 8.0 7.0 6.0 5.0 31.0 30.0 29.0 28.0 27.0 26.0 25.0 24.0 23.0 22.0 21.0 0.0 20.0 Frequency 5.0 GHz/DIV Prototype LP6872 Power Stage Balanced Self-Biased at 22.5dBm Input Power Vd=9.0V Ids=140mA Idss=210mA RS=2.5 ohms 6872balpower_tb PF1 6872balpower PF dBm 6872balpower_tb GAIN 6872balpower OUT_EQN dB Filtronic Solid State Discrete FET / PHEMT Applications Notes 34 The plot below shows predicted compression characteristics in this stage, with a slight amount of gain expansion (about 0.15 dB), consistent with the quiescent bias point of 67% (see Sec. 6). Simulated P 1dB = 30.7 dBm at 16 GHz, with a saturated output power of about 31.5 dBm. COMPRESSION CHARACTERISTICS AT 16 GHz Small-Signal Gain 32.0 27.0 22.0 17.0 12.0 Power 2.0 dBm/DIV Prototype LP6872 Power Stage Vd=9.0V Ids=140mA Idss=210mA RS=2.5 ohms 6872balpower_tb PF1 6872balpower PF dBm 6872balpower_tb GAIN 6872balpower OUT_EQN dB 11.0 10.0 9.0 8.0 7.0 0.0 26.0 Frequency, GHz 16.0 P1dB = 30.7 Psat = 31.5 Filtronic Solid State Discrete FET / PHEMT Applications Notes 35 2 GHz 1 Watt Power Stage for the LP 6872: A 2 GHz, 1W power stage utilizing two LP 6872 DHPHEMT devices is shown below, implemented in microstrip on Duroid (ε R = 2.2, h = 10 mils [0.25 mm]), and using a lossy input matching network to achieve adequate stability and flat SSG. 1 The output circuit is designed for optimum power match, while the input circuit is derived from a second-order all-pass network, with the PHEMT’s input capacitance C GS (as modified by the Miller effect) incorporated as one of the network’s elements. The simulated SSG is 17.5-18.0 dB at 21 GHz, with an input return loss of -12 dB, and a worst- case stability factor of 1.3 (at 2.5 GHz). The implementation of this circuit utilizes a combination of lumped element (chip) components and microstrip elements. The circuit is predicted to provide +30.2 dBm of linear power at 2 GHz, with a Power-Added Efficiency of 38%. 1. Arell, T., and Hongsmatip, T., “2-6 GHz Commercial Power Amplifier,” Applied Microwave, Winter 1993, pp. 51-56. ELEMENT TYPE DIMENSIONS (mils) OR VALUE (units) INPUT CIRCUIT: 1 MLIN W=30 L=100 (50Ω INPUT LINE) L1 IND L=1.6 nH L2 IND L=1.1 nH C1 CAP C=1.3 pF R1 RES R=18 ohms OUTPUT CIRCUIT: 1 MLIN W=27 L=35 2 MTEE W1=27 W2=30 W3=34 3 MLEF 1 STUB W=34 L=57 4 MLIN W=30 L=100 (50Ω OUTPUT LINE) L1 C1 L2 R1 -V G 100 pF (x4) LP 6872 (x2) LAYOUT MAY NEED TO BE MODIFIED TO ACCOMODATE ACTUAL CHIP COMPONENT SIZES NOT SHOWN: DRAIN BIASING CIRCUIT Filtronic Solid State Discrete FET / PHEMT Applications Notes 36 PREDICTED POWER PERFORMANCE 16.5 16.0 15.5 15.0 14.5 14.0 13.5 13.0 12.5 12.0 11.5 11.0 10.5 10.0 0.0 20.0 6872cbandpow_tb PF1 6872cbandpow PF dBm 6872cbandpow_tb GAIN 6872cbandpow OUT_EQN dB Frequency, GHz 1.5 2.0 2.5 3.0 Linear Power (dBm): 1.5 GHz 29.7 2.0 GHz 30.2 2.5 GHz 30.5 3.0 GHz 30.5 Power 5.0 dBm/DIV Prototype 2 GHz Power Amplifier Two LP 6872 PHEMTs Vds=9V Ids=290mA Dual Bias Vgs=-0.3V 32.0 30.0 28.0 26.0 24.0 22.0 20.0 18.0 16.0 14.0 12.0 10.0 8.0 6.0 Filtronic Solid State Discrete FET / PHEMT Applications Notes 37 6-18 GHz Low Noise Stage for the LP 7512 SHPHEMT: ELEMENT TYPE DIMENSIONS (mils) INPUT CIRCUIT: 1 MLIN W=10 L=40 (50Ω INPUT LINE) 2,3 MLEF 2 STUBS W=10 L=13 4 MCROS W1=10 W2=9.4 W3=W4=10 5 MLIN W=9.4 L=33 6,7 MLEF 2 STUBS W=10.8 L=14.8 8 MCROS W1=9.4 W2=20.4 W3=W4=10.8 9 MLIN W=20.4 L=40.3 10,11 MLEF 2 STUBS W=18.4 L=16.1 12 MCROS W1=20.4 W2=7.5 W3=W4=18.4 13 MLIN W=7.5 L=9.8 14 MTEE W1=W2=7.5 W3=2 15 MLIN W=2 L=67 16 TFR RS=50 W=2 L=1 17 MLIN W=7.5 L=3 OUTPUT CIRCUIT: 1 MLIN W=12 L=31.3 2,3 MLEF 2 STUBS W=27.7 L=20 4 MCROS W1=12 W2=10.3 W3=W4=27.7 5 MLIN W=10.3 L=17.3 6 MLEF 1 STUB W=20 L=16 7 MTEE W1=10.3 W2=10 W3=20 8 MLIN W=10 L=40 (50Ω OUTPUT LINE) INPUT CIRCUIT OUTPUT CIRCUIT LP 7512 NOT SHOWN: RF CHOKES FOR DRAIN BIASING, DC BLOCKING CAPACITORS TUNING PADS MAY BE ADDED AS DESIRED. 50Ω INPUT/OUTPUT MLINS MAY BE SHORTENED AS NEEDED. Filtronic Solid State Discrete FET / PHEMT Applications Notes 38 The simulated noise and gain performance of this stage is shown below; nominal performance was 1.9 dB noise figure from 12-18 GHz, with SSG = 8.5 t 0.6 dB over the 6-18 GHz band. Significantly lower noise figure can be achieved with this device for narrower bandwidths; the typical LP 7512 performance is 1.0 dB NF at 18 GHz. SINGLE-ENDED NOISE PERFORMANCE Gain 10.0 9.0 8.0 7.0 6.0 5.0 Power 2.0 GHz/DIV Prototype Single-Ended LP 7512 Low Noise Stage Vds=2V Ids=15mA 7512noise_tb S21 7512noise S[2.1] dB 7512noise_tb NF1 7512noise NF dB 6.0 5.0 4.0 3.0 2.0 1.0 2.0 20.0 Noise Figure Filtronic Solid State Discrete FET / PHEMT Applications Notes 39 11. RECOMMENDED ASSEMBLY TECHNIQUES Handling of Discrete Devices: Handling of FSS Discrete Devices must at all times be done with Electrostatic Discharge (ESD) precautions, as outlined below. The devices may be manually handled with clean sharp tweezers by grasping the device on its edges. Some devices include airbridge structures that are easily damaged with careless handling. Automatic handling equipment may be used, but soft, conductive pick-up collets are preferred. Collets must be grounded to prevent static charge accumulation. The optimum collet size is such that the device is contacted around its periphery and away from the active structures in the center. Rectangular cross-sections are preferred. Die Attach: The recommended die attach is the “eutectic” die attach, using a 80/20 Gold/Tin eutectic solder, which has a melt temperature of about 280°C. Eutectic die attach should be done under forming gas (90% N 2 , 10% H 2 ) for best results. If forming gas is not available, then clean dry N 2 gas should be used. Eutectic die attach in ambient air is not recommended. Pre-sized eutectic preforms are available in a variety of sizes; use a preform whose dimensions are close to the device being attached. The device can be placed using clean sharp tweezers, once the preform has melted on the heated stage (which is typically 285-290°C); the assembly should be removed from the heated stage as soon as the die is positioned. For large devices (greater than 0.5 x 0.5 mm or 20 x 20 mils), the assembly should be cooled at a rate of about 10°C per minute, to reduce thermal stresses. Maximum time at the eutectic temperature is 1 min. Other solder materials may be used; contact the Foundry for compatibility. Another acceptable die attach method is the use of conductive epoxy (gold or silver filled), which is assembled at room temperature and then cured per the manufacturer’s directions. The epoxy die attach should not be used for power devices, since the thermal resistivity of the device will be degraded. Die Placement: In general devices should be placed as close as possible to the matching circuits, regardless of the type of substrate material used. For example, with microstrip circuits implemented on alumina, the die should be placed within 0.001” (0.025 mm) of the input matching circuit to minimize the gate bond wire lengths. If source bypass chip capacitors are used, they should be placed as close as possible to the device to minimize source bond wire length. The optimum heatsink material is gold-plated copper, molybdenum, or Cu composites, which allow for efficient heat dissipation. Less desirable is kovar, and direct die-attach onto duroid or alumina should only be used for low noise applications. All heatsink material should be gold (100 µin min.) over nickel plated. Poor heatsinking will result in high operating channel temperatures and degraded device reliability. Lead Bond: The recommended lead bond technique is thermocompression wedge bonding with 0.001” (25 µm) diameter gold wire. This should be done on a heated stage at 230-240°C, with a heated bonding tool at 150- 160°C. Dry N 2 or forming gas is the preferred ambient during bonding. The bond tool force should be 35-38 gm. Ultrasonic bonding is not recommended. Contact the Foundry for additional information. Storage: FSS discrete devices should be stored in clean dry Nitrogen gas at room temperature (20-25°C). Alternately, storage at room temperature in clean dry ambient air is acceptable; in any case, the parts should be stored with proper ESD precautions. ESD Precautions: Standard ESD precautions for Class 1A devices (0-500V) should be observed in storing, handling, and assembling FSS discrete devices. Users should follow the measures outlined in MIL-STD-1686, “Electrostatic Discharge Control Program,” and MIL-HDBK-263, “Electrostatic Discharge Control Handbook.” Filtronic Solid State Discrete FET / PHEMT Applications Notes 40 12. PARAMETRIC SCREENING AND QUALITY ASSURANCE Filtronic’s Quality Assurance procedures for its discrete FET and PHEMT devices are patterned after the JANC level requirements of MIL-STD-19500J. FSS offers its discrete devices in three basic grades, as follows: • Commercial/Military Grade (C/MG) (Bulk purchase) • Sorted Military Grade (SMG) • High-Reliability Grade (HiRel) These grades differ with respect to the level of parametric screening and visual inspection. All grades are warranted to meet or exceed the performance specifications as indicated for each device type on the appropriate Datasheet. Test methods and visual inspection criteria are per MIL-STD-750, as detailed below. The discrete device fabrication, test, and Quality Conformance Inspection (QCI) sequence is as follows: GaAs SUBSTRATE INSPECTION AND PREPARATION MOLECULAR BEAM EPITAXIAL GROWTH EPITAXIAL CHARACTERIZATION Defect Density Sheet Carrier Density Hall Mobility LOT FORMATION 6-10 wafers/lot PROCESSING / FABRICATION FINAL ELECTRICAL AND VISUAL INSPECTION 100% DC TEST GROUP A TESTS Subgroups 1 and 2 10 devices / wafer ACCEPT/REJECT 10/1 GROUP B TESTS Bond Pull / Die Shear 5 devices / wafer ACCEPT/REJECT 5/0 LOT ACCEPTANCE INSPECTION Per Sequence Below ACCEPT/REJECT 10/1 PREPARATION FOR SHIPMENT COMMERCIAL/MILITARY GRADE Sample Visual Inspection Film Frame Packing SORTED MILITARY GRADE 100% Visual Inspection Packing (Waffle or Gel Pack) HIGH-RELIABILITY GRADE Wafer Lot submittal for JANS screening SCRIBE AND BREAK SAMPLE VISUAL INSPECTION FILTRONIC SOLID STATE DISCRETE DEVICE FABRICATION AND QUALITY ASSURANCE SEQUENCE STABILIZATION BAKE Filtronic Solid State Discrete FET / PHEMT Applications Notes 41 TABLE 1 GROUP A TESTS AND METHODS SUBGROUP TEST OR INSPECTION SYMBOL MIL-STD-750 METHOD 1 DC Characteristics: Saturated Drain-Source Current Transconductance Peak Transconductance (opt.) Pinch-Off Voltage Gate-Source and Gate-Drain Leakage Current Gate-Source and Gate-Drain Breakdown Voltage Forward Gate Voltage (opt. for JANS level) I DSS G M G M PEAK V P I GS , I GD BV GS , BV GD V GF 3413 3455 (60 Hz) 3455 (50%V P ) 3403 3401, 3411 3401, 3411 2 RF Characteristics: Optimum Noise Figure Associated Gain at NF MIN Max. Output Power at 1 dB Gain Compression Power Gain at P 1dB Power-Added Efficiency at P 1dB NF MIN G A P 1dB G P PAE (η) 3246.1 3255 NOTE: Noise Figure is not measured on power devices, and power performance (P 1dB , G P , PAE) is not characterized on low noise devices. LOT ACCEPTANCE INSPECTION SMALL LOT ASSEMBLY 10 devices / wafer DIE ATTACH / LEAD BOND INTERNAL VISUAL INSPECTION HERMETIC LID SEAL STABILIZATION BAKE GROUP A TESTS Subgroup 1 DC POWER BURN-IN 96 Hours, TCH = 175°C GROUP A TESTS Subgroup 1 ENDPOINT AND DELTA CALCULATIONS Per Datasheet and Delta Limits Table LOT ACCEPTANCE INSPECTION SEQUENCE: Filtronic Solid State Discrete FET / PHEMT Applications Notes 42 TABLE 2 GROUP B TESTS AND METHODS TEST OR INSPECTION MIL-STD-750 METHOD CONDITION SMALL LOT ACCEPT/REJECT n/c Bond Strength (Bond Pull) Die Shear 2037 2017 All leads 5/0 5/0 TABLE 3 LOT ACCEPTANCE INSPECTION TESTS AND METHODS TEST OR INSPECTION MIL-STD-750 METHOD CONDITION SMALL LOT ACCEPT/REJECT n/c Internal Visual Inspection Stabilization Bake DC Power Burn-In Endpoint Calculations Delta Calculations 2072 1031.4 1039 FSS Worksmanship Standards 24 hr. at 200°C in N 2 ambient 96 hr. at 175°C channel temperature Per Datasheet min/max values Per Table 4 Limits --- --- --- 10/1 10/1 TABLE 4 DELTA LIMITS DELTA PARAMETER SYMBOL DELTA LIMITS FOR DC POWER BURN-IN Saturated Drain-Source Current Transconductance Peak Transconductance (opt.) Pinch-Off Voltage Gate-Source and Gate-Drain Leakage Current Gate-Source and Gate-Drain Breakdown Voltage I DSS G M G M PEAK V P I GS , I GD BV GS , BV GD t20% t20% --- t20% t150% or 2.0ΦA t25% Filtronic Solid State Discrete FET / PHEMT Applications Notes 43 In summary, 25 devices are taken at random from each wafer submitted for QCI screening, with 5 devices used for Bond Pull/Die Shear testing, 10 for Group A testing, and 10 for Lot Acceptance Inspection (LAI). The devices used in Group A testing are die attached and lead bonded into 50 ohm microstrip evaluation circuits, and triple-sleeve low-loss coaxial tuners are used to provide optimum impedance matching for power, gain, and noise figure measurements. The LAI devices are assembled and sealed into 100 mil ceramic leaded packages for the power burn-in. Strict adherence to JANC requirements calls for a sample of 22 per inspection lot, with an accept/reject limit of 22/0 (no failures allowed). This corresponds to a Lot Tolerance Percent Defective (LTPD) of λ=10; the FSS screening uses a sample size of 20 total, with 2 failures allowed (20/2). The equivalent LTPD level is approximately λ=25. The less stringent LTPD is justified by the omission of High-Temperature Reverse Bias (HTRB) screening prior to the power burn-in, which would serve to remove “infant mortality” devices. The use of microstrip test circuits for 10 of the sample devices is done to minimize RF performance degradation due to package parasitics. For the Sorted Military Grade, Group A and LAI testing is done on each wafer. For Commercial/Military Grade, this screening is done on a “run” basis; a fabrication run is defined as a group of 6-10 wafers that are fabricated and processed simultaneously as a group. Therefore as a minimum, each wafer run of 6-10 wafers will have at least one wafer submitted for Group A and LAI testing; devices purchased under the C/MG are not necessarily from a wafer that has been screened, but the purchaser can request devices from the screened wafer in a given lot for a nominal surcharge. Devices purchased to the High-Reliability Grade are subjected to a more rigorous screening procedure per MIL-19500 JANS level. The various grades and the associated test and inspection methods are summarized below: DISCRETE DEVICE GRADES AND SCREENING PROCEDURES GRADE GROUP A TESTING LAI TESTING SHIPMENT METHOD 100% DC DATA ON DISKETTE VISUAL INSPECTION METHOD FILM & RING WAFFLE OR GEL PACKED C/MG COMMERCIAL PER EACH RUN PER EACH RUN ONLY NOT AVAILABLE NOT AVAILABLE SAMPLE BASIS C/MG MILITARY UPGRADE EACH WAFER EACH WAFER ONLY NOT AVAILABLE AVAILABLE SAMPLE BASIS VISUAL YIELD PROVIDED SMG EACH WAFER EACH WAFER N/A YES PROVIDED 100% BASIS HI-REL JANS JANS N/A YES PROVIDED 100% BASIS NOTES: 1). Minimum order quantity for C/MG option is 1/4 wafer; approximate quantities available from an authorized FSS Sales Representative or from the Foundry. 2). Hi-Rel screening per customer Source Control Drawing. Filtronic Solid State Discrete FET / PHEMT Applications Notes 44 WAFER FABRICATION INITIAL VISUAL INSPECTION MIL-750 METHOD 2070, 2072 SAMPLE BASIS: 35 DIE PER WAFER PCM TESTING STABILIZATION BAKE MIL-750 METHOD 1032 200C FOR 24 HRS 100% DC TESTING GROUP A, SUBGROUP 1 SCRIBE & BREAK SMALL LOT BUILD 35 MIN. PACKAGED DEVICES DIE ATTACH, LEAD BOND AND SEAL INTERNAL VISUAL INSPECTION MIL-750 METHOD 2071 GROUP B SUBGROUP 1 SEM INSPECTION MIL-750 METHOD 2077 BOND PULL TEST MIL-750 METHOD 2037 3 DEVICES MIN., ALL LEADS DIE SHEAR TEST MIL-750 METHOD 2017 3 DEVICES MIN. LID SEAL STABILIZATION BAKE 175C FOR 24 HRS DC PARAMETRIC SCREEN GROUP A, SUBGROUP 1 SMALL LOT INSPECTION QCI / LOT QUALIFICATION DC TESTED DIE FOR VISUAL INSPECTION AND SHIPMENT PREP FILTRONIC SOLID STATE MIL-19500J JANS SCREENING SEQUENCE Filtronic Solid State Discrete FET / PHEMT Applications Notes 45 PRE-BURN-IN DC PARAMETRIC TESTING PRE-CONDITIONING DC POWER BURN-IN MIL-750 METHOD 1039 T CH = 150C FOR 160 HRS POST-BURN-IN DC PARAMETRIC TESTING PDA DETERMINATION GROUP B SUBGROUP 5 ACCELERATED LIFE TEST MIL-750 METHOD 1027 5 DEV. / 0 REJECTS PRE-BURN-IN PARAMETRIC TESTING DC POWER BURN-IN T CH = 225C FOR 240 HRS POST-BURN-IN PARAMETRIC TESTING DELTA CALCULATIONS GROUP B SUBGROUP 2 ACCELERATED LIFE TEST MIL-750 METHOD 1027 5 DEV. / 0 REJECTS PRE-BURN-IN PARAMETRIC TESTING DC POWER BURN-IN T CH = 225C FOR 240 HRS POST-BURN-IN PARAMETRIC TESTING DELTA CALCULATIONS GROUP B SUBGROUP 3 OPERATING LIFE TEST MIL-750 METHOD 1026 8 DEV. / 0 REJECTS PRE-BURN-IN PARAMETRIC TESTING DC POWER BURN-IN T CH = 150C FOR 1000 HRS POST-BURN-IN PARAMETRIC TESTING DELTA CALCULATIONS GROUP A SUBGROUP 2 RF TESTS 10 DEV. / 0 REJECTS (OPTIONALLY DONE ON CARRIER-MOUNTED DEVICES) 100% VISUAL INSPECTION QA CERTIFICATION PREPARATION FOR SHIPMENT SHIPMENT CSI Filtronic Solid State Discrete FET / PHEMT Applications Notes 46 13. DISCRETE DEVICE UNIFORMITY The uniformity of FSS’s discrete PHEMT and MESFET devices depends fundamentally on the materials and processing technology utilized in the manufacturing process. In addition to the Parametric Screening and QCI testing as outlined in Sec. 12, typical device uniformity data is presented below, using the LP 7612 DHPHEMT as the example. Uniformity is presented at three levels: wafer uniformity, same-lot wafer-to-wafer uniformity, and lot-to-lot uniformity. The DC parametrics given are as measured during the 100% screening, while the RF data presented was taken from a RF-probeable version of the LP 7612. This modified 7612 device was directly measured on-wafer using Cascade Microtech™ GSG coplanar probes. While the S-data of this modified 7612 is not identical to the production version, the data does represent the process and material variation across a wafer. TYPICAL 100% DC PARAMETRIC TEST DATA LP 7612 LOT: 3059 WAFER: #4 TEST DATE: 4/30/96 TOTAL DIE TESTED: 21566 TOTAL PASSED DEVICES: 14032 PARAMETER UNITS MEAN STD.DEV. σ LIMITS UP/LOW C PK Drain-Source Current, I DSS mA 70.7 6.6 40/85 0.72 Pinch-Off Voltage, V P V -0.84 0.19 -0.25/-1.50 1.06 Transconductance at I DSS , G M mS 67.6 21.5 50/--- 0.27 Transconductance at 50% V P mS 89.9 8.5 60/--- 1.17 Gate-Drain Breakdown, BV GD V -11.5 1.02 -8/--- 1.15 Gate-Source Breakdown, BV GS V -10.9 1.70 -6/--- 0.98 Leakage Current, I GSO µA 0.53 0.56 ---/10 0.30 TYPICAL RF UNIFORMITY: WAFER BASIS LP 7612 LOT: 3051 WAFER: #3 10 SITES ACROSS WAFER V DS = 5V I DS = 50%I DSS PARAMETER FREQ (GHz) MAGNITUDE PHASE MEAN σ MEAN σ S 11 5 15 25 0.922 0.645 0.616 0.005 0.002 0.010 -60° -126 -149 3° 3 3 S 21 5 15 25 4.73 2.27 1.49 0.15 0.04 0.07 135 80 57 2 3 3 S 12 5 15 25 0.048 0.068 0.036 0.005 0.009 0.007 59 6 3 2 2 2 S 22 5 15 25 0.755 0.564 0.572 0.018 0.030 0.068 -26 -55 -68 2 4 4 Maximum Available Gain (dB) 5 15 25 19.9 11.7 7.4 0.4 0.4 0.1 --- --- Filtronic Solid State Discrete FET / PHEMT Applications Notes 47 TYPICAL RF UNIFORMITY: LOT BASIS LP 7612 LOT: 3051 WAFERS #3, 4, 5, 6 ONE SITE AT RANDOM PER WAFER V DS = 5V I DS = 50%I DSS PARAMETER FREQ (GHz). 3 4 5 6 MAG. ANG. MAG. ANG. MAG. ANG. MAG. ANG. S 11 5 15 25 0.929 0.641 0.599 -55° -120 -144 0.913 0.646 0.633 -65° -133 -157 0.913 0.639 0.617 -62° -129 -152 0.914 0.639 0.626 -65° -133 -157 S 21 5 15 25 4.88 2.49 1.63 138 84 62 4.82 2.25 1.45 132 77 52 4.97 2.38 1.55 134 81 59 4.90 2.28 1.48 131 76 51 S 12 5 15 25 0.054 0.080 0.054 60 9 1 0.044 0.057 0.024 58 3 10 0.064 0.088 0.062 56 7 0 0.046 0.061 0.027 57 2 5 S 22 5 15 25 0.719 0.512 0.503 -29 -62 -74 0.767 0.589 0.610 -24 -51 -64 0.731 0.508 0.509 -31 -61 -73 0.770 0.576 0.594 -26 -53 -67 Maximum Available Gain (dB) 5 15 25 19.6* 12.5 7.6 --- 20.4* 11.5 7.6 --- 18.9* 12.3 7.5 --- 20.3* 11.4 7.5 --- NOTE: Maximum Available Gain with * indicates Maximum Stable Gain, since Stability Factor k < 1.0. Lot-to-Lot Uniformity: All lots are screened as outlined in Sec. 12, and must meet all parametric test specifications. In this way, minimum device performance is assured as detailed in the appropriate Datasheet. Due to varying user requirements, lots (groups of 6-10 wafers) are occasionally targeted to specific I DSS ranges. This targeting accounts for much of the lot-to-lot variation, but comparison of different lots with similiar I DSS values will exhibit good uniformity, on the order of same-lot wafer- to-wafer uniformity. Users are encouraged to obtain samples from a specific lot in order to verify performance in a specific circuit or application; the quantity of available die from a particular wafer or wafer lot is available from the Foundry. Filtronic Solid State Discrete FET / PHEMT Applications Notes 48 14. DEVICE RELIABILITY A summary of reliability investigations that have been conducted on FSS’s MESFETs, PHEMTs, and MMICs is included in Appendix A. For MESFETs, the measured MTTF is 1.0 x 10 8 hrs. at a channel temperature of 120°C, with an Activation Energy of 1.52 eV, and An Instantaneous Failure Rate of less than 10 FITs. For PHEMT devices, in addition to the work summarized in the Appendix, recent work has shown that the MTTF for DHPEMTs exceeds 1.2 x 10 7 hrs. at T CH = 120°C, based on an accelerated burn-in of samples at 225°C. These samples have accumulated 1,045 hrs. with no failures. Filtronic Solid State Discrete FET / PHEMT Applications Notes 49 APPENDIX A Filtronic Solid State Device and Process Reliability 50 Device and Process Reliability InGaAs / AlGaAs PHEMT Devices Revision B January 1998 Filtronic Solid State Device and Process Reliability 51 Scope This report presents a current assessment of the reliability of InGaAs/AlGaAs Pseudomorphic High Electron Mobility Transistor (PHEMT) discrete and MMICs devices, as presently manufactured at Filtronic Solid State (FSS). GaAs MESFET reliability results from earlier research conducted at FSS is included for completeness. The report concludes with a summary of work in progress. Background The reliability of semiconductor devices can be modeled by the "log-normal" distribution, which is a probability distribution function described by a shape parameter and a median lifetime, as follows: { } p t t n t t m ( ) exp ( / ) ≡ − ¹ ' ¹ ¹ ¹ ¹ ; ¹ ¹ ¹ 1 2 2 2 2 σ π σ where σ = shape parameter, t = time, and t m = median lifetime. The interpretation of this function in the context of device reliability is that integrating p(t) from zero to some time t f gives the probability that a given device has failed. Another way to interpret this function is that the area under the curve p(t) from zero to the median lifetime t m is equal to 0.5, indicating that there is a 50% chance that a device will have failed. The function is normalized such that the total area under the curve is equal to one. This failure probability distribution assumes one dominant failure mechanism, with no early or “freak” failures. The median lifetime of a population of devices is closely related to the Mean-Time-to-Failure (MTTF), and generally the two quantities are interchangeable. The MTTF is one of the primary parameters that must be determined for a given device or process, and this is generally done by exposing a sample of devices to accelerated life testing at elevated temperatures. The MTTF is determined by the definition of a device "failure," which in turn may be affected by a number of degradation or failure mechanisms. One generally attempts to identify a single dominant failure mechanism, and MTTF at some elevated temperature is assumed to follow the Arrhenius relationship: AF E k T T A o A ≡ − ¸ ¸ _ , ¹ ' ¹ ¹ ; ¹ exp 1 1 where AF = Arrhenius Acceleration Factor, E A = Activation Energy, k = 8.62 x 10 -5 eV/ K (Boltzmann constant), T O = standard operating temperature, and T A = acceleration temperature. The Activation Energy (usually expressed in electron volts, eV) characterizes the degradation process, and clearly should be as large as possible for reliable device operation at elevated temperatures. For example, assuming E A = 1 eV, the acceleration factor resulting from device operation at 250°C, compared to operation at 120°C is AF = 1,537, meaning that a MTTF of 10 6 hours will be reduced to 651 hours at the elevated temperature. Another commonly used assessment in reliability analysis is the Instantaneous Failure Rate (IFR), which for a log- normal distribution is given by: 1 ( ) { } ( ) IFR n t t t efrc n t t m m ≡ − ¹ ' ¹ ¹ ; ¹ ¹ ' ¹ ¹ ; ¹ 2 1 2 1 2 2 2 exp / / σ σ π σ where efrc denotes the complementary error function. The IFR is usually multiplied by 10 9 , and the units of "FITs" are commonly used, 1 FIT meaning one failure in 10 9 hours of operation. In expanded form: 1 H. Fukui, et. al., “Reliability of Power GaAs Field-Effect Transistors,” IEEE Trans. on Elec. Devices, Vol. ED-29, No. 3, March 1982. Filtronic Solid State Device and Process Reliability 52 ( ) { } ( ) IFR n t t t e dt where z n t t m t z m · ¹ ' ¹ ¹ ; ¹ − ¹ ' ¹ ¹ ¹ ¹ ; ¹ ¹ ¹ ≡ − ∫ 2 1 2 1 2 1 2 2 0 2 exp / : / σ σ π π σ π To completely characterize the distribution, one must determine the shape factor and the median lifetime (or MTTF) at a given temperature, and the Acceleration Factor is used to predict the MTTF at some other temperature. The shape factor is taken to be invariant with temperature; it expresses the variation in time of device failures as the sample ages past the median lifetime. Ideally there are no device failures until the median lifetime is reached, at which time most of the devices fail at essentially the same time. For component and system lifetimes t s such that: t s << t m , the average IFR ≈ 1/MTTF. In the case of MESFET and PHEMT devices, the integrity of the Schottky contact formed by the Gate electrode is of critical importance to reliable device operation. Changes in the metal-semiconductor barrier height will affect all other channel parameters; therefore, the temperature that is substituted into the Arrhenius formula is the average channel temperature, determined by the following: T P P P CH IN DC J OUT · + − Θ where: T CH = Channel temperature, P IN = RF input power, P DC = DC power dissipated in device, Θ J = channel-to-heatsink thermal resistivity, and P OUT = RF power delivered to a load. The channel region for these devices is defined as the region extending from the Drain contact to the Source contact, along the total periphery of the Gate, if the device is a multi-finger or interdigitated topology. In the case of FSS devices, the thermal resistivity is measured by direct-contact nematic liquid- crystal thermography. This method allows near-optical resolution, and features less than one micron can be imaged. MESFET Device Structure and Reliability Assessment The MESFET structure used at FSS is a classical recessed, offset gate structure, as shown below: The semiconductor structure shown is grown by Molecular Beam Epitaxy, with the gate structure defined either photolithographically or by direct-write electron-beam techniques. The device is passivated with a 2000A layer of silicon nitride (Si 3 N 4 ). The Gate metallization system in a refractory Ti/Pt/Au structure, with additional overplated gold, and the ohmic contacts are alloyed AuGe/Ni/Au with overplated gold; both the Gate and ohmic contacts are industry-standard technologies. AuGe / Ni / Au Alloyed Contacts N+ GaAs Active Layer N- Buffer Layer Drain Contact Source Contact Ti / Pt / Au Recessed Gate N++ Contact Layer GaAs MESFET STRUCTURE Filtronic Solid State Device and Process Reliability 53 Samples of 0.5 x 285 (Gate length x width) µm discrete MESFETs have been evaluated using both high temperature storage (non-operating), and elevated temperature DC operating lifetest over the last several years. The primary failure mode was a decrease in Saturated Drain-Source current (I DSS ) and Pinch-off voltage (V P ); a device failure was defined to be a change of more than 20% from the initial value. No significant change in transconductance (g M ) was observed, at least when measured at zero Gate-Source bias (V GS ). Gate-Source breakdown voltage (BV GS ) decreased initially, but then stabilized throughout the study. For the operating lifetest, a failure was defined as a change of more than t20% in DC bias current I DS . The results from both types of tests were similar, indicating that the dominant failure mechanism was independent of electric field strength or current density. The failure data from both types of tests was combined to determine an activation energy of 1.52 eV, with a projected median lifetime (or MTTF) of 1.0 x 10 8 hours at a channel temperature of 120°C. The data showed a dispersion (shape factor) of σ = 0.49, with a projected Instantaneous Failure Rate (IFR) of less than 10 FITs at the 120°C channel temperature. The probable degradation mechanism was diffusion of Gate metal into the active layer, commonly known as "gate sinking." Elemental Au, in particular, acts as an acceptor impurity in GaAs, with a ionization energy of 0.09 eV (measured above the valence band). The diffusion of metal into the channel can reduce the effective channel thickness, consistent with a reduction of I DSS and V P . The stability of the transconductance at V GS = 0V indicated that neither the Gate-Source parasitic resistance nor the contact resistance had degraded. This finding demonstrated excellent stability of the ohmic contacts at high temperatures. Transconductance stability at bias conditions other than V GS = 0V was not characterized, nor were any RF parameters measured. Initial changes in breakdown voltage were related to changes in surface defect density (trapping sites); this is an example of an “annealing” effect, and does not affect long-term reliability. MESFET MMIC Reliability A 10-piece sample of 0.5 µm gate length MESFET-based MMICs was exposed to a step-stress test at baseplate temperatures of 125, 175, and 225°C for 3600 hours at each temperature. No failures were noted during the 125 or 175°C trials, and 2 out of 10 devices failed after 2500 hours during the 225°C trial, which corresponds to a channel temperature of approximately 245°C. Thus the MTTF at 245°C is at least 2500 hours, and given an acceleration factor of AF = 43,692, the projected MTTF at a channel temperature of 120°C would be at least 1.1 x 10 8 hours, which is comparable to the result for discrete MESFETs, assuming a comparable E A . FSS MMICs also incorporate TaN thin-film deposited resistors and Si 3 N 4 metal-insulator-metal (MIM) capacitors, which have been exposed to high temperature storage at 250°C for 1000 hours. Observed changes in sheet resistance and capacitance per unit area were both less than 3%. For a reliability estimate, taking 10% change as the definition of a failure, the MTTF at 250°C should then be at least 3000 hours, and the projected MTTF at 120°C would then be 1.8 x 10 8 hours, although the activation energy for these two components was not verified. Filtronic Solid State Device and Process Reliability 54 PHEMT Device Structure The PHEMT structure used at FSS differs somewhat from the MESFET structure, as shown below: The PHEMT device features a slimier Gate structure as the MESFET (i.e., the refractory Ti/Pt/Au system), but in this device the Schottky contact is formed with AlGaAs (the Al mole fraction is <0.3), rather than doped GaAs. For reduced Gate resistance, the a “mushroom” structure is formed by a direct-write electron-beam lithography process; each Gate section measures 0.25µm at its base. The use of a wider bandgap, less heavily doped contact material provides for more Gate stability at high temperatures over long periods of time. By design, the AlGaAs layer is fully depleted of carriers at zero applied Gate bias, and Drain-Source current flows in the “2-dimensional electron gas” formed at the AlGaAs/InGaAs heterojunction. The Source and Drain ohmic contacts are formed with the identical AuGe/Ni/Au ohmic metallization system as used in the MESFET. All metallization systems are formed by an evaporation process, with overplated Au; the devices are passivated as with the MESFETs. The structure shown above is a Single Heterojunction PHEMT (SHPHEMT), which is a low-noise, low- power dissipation design. FSS power PHEMT devices utilize a Double Heterojunction PHEMT structure (DHPHEMT), which features two AlGaAs/InGaAs heterojunctions, one on each side of the InGaAs channel layer. The doping levels, mole fractions, and layer thicknesses are based on proprietary designs developed by FSS to provide superior current and power density performance. AlGaAs / GaAs Superlattice Undoped GaAs Buffer InGaAs Channel Layer N+ AlGaAs Layer (5 x 10 17 cm -3 ) N++ GaAs Contact Layer Drain Contact Source Contact Ti / Pt / Au Mushroom Gate PHEMT STRUCTURE Undoped AlGaAs Spacer Heterojunction Filtronic Solid State Device and Process Reliability 55 Discrete PHEMT Device Reliability Reliability Study Test Device: The device selected for the baseline PHEMT reliability study was a discrete DHPHEMT device, the LP7612, which is a 0.25 x 200 µm, “π-gate” (non-interdigitated) design. This device features the following nominal performance: RF Output Power at 1dB compression: V DS = 5V, I DS = 50% I DSS , f = 18 GHz +21 dBm RF Gain at 1 dB compression: V DS = 5V, I DS = 50% I DSS , f = 18 GHz 8.5 dB Minimum Noise Figure V DS = 2V, I DS = 33% I DSS , f = 18 GHz 1.25 dB Saturated Drain-Source current, I DSS : 65 mA Pinch-Off Voltage at I DS = 1 mA, VP: -0.75 V Gate-Drain Breakdown Voltage at I GD = 1 mA, BV GD : -9 V Gate-Source Leakage current at VGS = -5V, I GSO : 1 µA The test devices were assembled into hermetically-sealed, ceramic packages, the P-70 stripline package, thereby allowing insertion into test fixtures for characterization and high-temperature burn-in. Experimental Protocol: To establish the basic reliability parameters, a discrete DH-PHEMT wafer was randomly selected from FSS’s standard production material, and subjected to the usual lot screening procedures. The Lot Acceptance Inspection and Quality Conformance Inspection screening procedures are detailed in FSS’s Applications Notes, Discrete FET/PHEMT Devices, Rev. A, Nov. 1996. A random group of 100 LP7612 devices was taken from this wafer, and subjected to the following screening: 100% DC Parametric Testing (Go / No Go) Visual Inspection Die Attach into P-70 ceramic packages Wire Bond Lid Seal Stabilization Bake (200°C for 24 hrs, non-operating) Pre Burn-In DC Test DC Power Burn-In (T CH = 175°C for 168 hrs) Post Burn-In DC Test Post Burn-In RF Test (S 21 at 12 GHz) The primary goal of this effort was to determine the extrapolated MTTF at a 150°C (channel temperature), based on the assumption that the Arrhenius equation models the dominant failure mechanism. Randomly selected groups of the packaged LP7612 devices were subjected to high-temperature DC power burn-in, at channel temperatures of 290, 275, 268, and 260°C. Channel temperature was achieved by a combination of baseplate temperature and self-heating in the devices. The LP7612 thermal resistivity had been previously characterized by use of direct-contact liquid crystal thermography. The typical value is 300-310 °C/W (at V DS = 3.0V) at room temperature, and is assumed to rise to approximately 330°C/W at the temperatures used for this experiment. (The thermal conductivity of GaAs decreases as a function of temperature due to increased phonon scattering.) Filtronic Solid State Device and Process Reliability 56 Each temperature group was continuously monitored during the burn-in, and periodically tested at room temperature to determine if a failure had occurred. A failure was declared if a device exhibited any change in Saturated Drain-Source current (I DSS ), DC transconductance (g M ), or pinch-off voltage (V P ) greater than 15%, compared to the initial values. Also monitored were: Gate-Drain and Gate-Source breakdown voltages (BV GD , BV GS ), leakage current (I GSO ) and RF insertion gain at 12 GHz (S 21 ). Bias conditions used for the high temperature burn-in were: V DS = 3.0V, I DS = 30 t 5 mA. High-Temperature Burn-In Test System: This test station uses three independent, proportionally-controlled hotplates, each with 10 test positions. All 30 test positions are independently controlled by a DC bias system under the overall control of a computer. Custom software allows the setting of each test position to a desired drain-source voltage and operating current. This software allows for automatic polling of all test positions at user-selected intervals during the burn-in. Thus a complete record of each device’s DC characteristics (I DSS , g M , V P , and I GSO ) at temperature was kept, although only room temperature DC test data was used for failure determination. The test devices were mounted with biasing circuits that served to suppress low- frequency oscillations that can occur with high transconductance PHEMTs. Results: At the onset of the study, the failure criteria that were established to be as follows: I DSS : > t 15% g M : > t 15% These criteria were selected based on a typical application of this device, as a small-signal gain element. Failures were declared when either of these parameters had degraded or changed by more than the stated amount, compared to the initial (zero time), room temperature values. As discussed previously, all other DC parameters were monitored and measured during the course of the burn-in, although they were not used for failure declaration. During the course of the experiment, it was discovered that the transconductance showed essentially little or no degradation, so this parameter was not a factor. The following tables present the cumulative failures as a function of time for each experimental group: Group #1: T CH = 290°C V DS = 3.0V I DS = 40mA Baseplate Temp. = 240°C Number of Devices: 8 TIME (hrs.) FAILURES CUMULATIVE FAILURES (%) 63 2 25 126 1 38 165 1 50 174 1 63 262 3 100 Group #2: T CH = 275°C V DS = 3.0V I DS = 35mA Baseplate Temp. = 240°C Number of Devices: 18 TIME (hrs.) FAILURES CUMULATIVE FAILURES (%) 195 6 33 261 1 39 307 1 44 349 1 50 424 1 56 465 2 67 727 3 83 Filtronic Solid State Device and Process Reliability 57 Group #3: T CH = 268°C V DS = 3.0V I DS = 25mA Baseplate Temp. = 240°C Number of Devices: 8 TIME (hrs.) FAILURES CUMULATIVE FAILURES (%) 769 2 25 1286 1 38 1562 2 63 Group #4: T CH = 260°C V DS = 3.0V I DS = 25mA Baseplate Temp. = 235°C Number of Devices: 11 TIME (hrs.) FAILURES CUMULATIVE FAILURES (%) 359 1 9 656 1 18 664 1 27 1005 2 45 1013 1 55 1531 1 64 1800 1 73 Analysis: Initial estimates of MTTF and shape factor for each group were made by plotting cumulative failures on log- probability paper, and by graphically determining the best linear fit. This method gave the following results: GROUP MTTF (hrs) SHAPE FACTOR (σ) T CH = 290°C 150 0.4 T CH = 275°C 330 1.0 T CH = 268°C 1350 0.7 T CH = 260°C 1150 0.8 For a refinement of the MTTF estimates from this graphical method, each group’s results were fitted to the IFR relationship as given above, and this analysis is presented in the following series of graphs. The comparison between the theoretical Instantaneous Failure Rate behavior and the actual experimental results serves several purposes: first, the IFR relationship as presented assumes a single failure mechanism (i.e., no “early” failures), second, the graphically fitted MTTF can be compared to the exact log-normal expression, and lastly the shape factor can be validated. Filtronic Solid State Device and Process Reliability 58 GROUP #1 T CH = 290°C FITTED IFR IFR, MTTF=130 HR, SIGMA=0.76 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 25 50 75 100 125 150 175 200 225 TIME (HR) I F R ( H R ) GROUP #1 T CH = 290°C AVERAGE FAILURE RATE AFR 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 63 126 165 174 262 TIME (HR) A F R ( H R ) The peak failure rates are in rough agreement, 0.008 vs. 0.014, but the average rate peaks at 174 hrs. vs. the IFR peak at 115 hrs. No significant early failures are apparent. Filtronic Solid State Device and Process Reliability 59 GROUP #2 T CH = 275°C FITTED IFR IFR, MTTF=350 HR, SIGMA=0.99 0 0.0005 0.001 0.0015 0.002 0.0025 0.003 25 125 225 325 425 525 625 725 825 TIME (HR) I F R , ( H R ) GROUP #2 T CH = 275°C AVERAGE FAILURE RATE AFR 0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018 195 261 307 349 443 727 TIME (HR) A F R The IFR peak failure rate is significantly higher than the average rate, 0.0025 vs. 0.0014, but this group may have included some early failures. The first monitoring interval at 195 hrs. revealed 6 failed devices, and it is likely that an earlier monitoring time would have resolved the high average rate. The peak rates are at 225 vs. 349 hrs. Filtronic Solid State Device and Process Reliability 60 GROUP #3 T CH = 268°C FITTED IFR IFR, MTTF=1350 HRS, SIGMA=0.72 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 0.0009 10 250 500 750 1000 1250 1500 1750 2000 TIME (HR) I F R ( H R ) GROUP #3 T CH = 268°C AVERAGE FAILURE RATE AFR 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 0.0009 769 1286 1562 TIME (HR) A F R ( H R ) The peak average failure rate compares well with the peak IFR, 0.0008 vs. 0.0009, with peak rates at 1250 vs. 1500 hrs.; no average failure rate data was available after 1562 hrs. Filtronic Solid State Device and Process Reliability 61 GROUP #4 T CH = 260°C FITTED IFR IFR, MTTF=1250, SIGMA=0.80 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 0.0009 5 250 500 750 1000 1250 1500 1750 1800 TIME (HR) I F R ( H R ) GROUP #4 T CH = 260°C AVERAGE FAILURE RATE AFR, TCH=260C 0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 359 660 1009 1531 1800 TIME (HR) A F R The IFR and average failure rate plots are in good agreement, both in terms of peak rates, and peak rate time in hours. No early failures were apparent. Filtronic Solid State Device and Process Reliability 62 FILTRONIC SOLID STATE RELIABILITY STUDY LP7612 DISCRETE DHPHEMT ARRHENIUS PLOT 332 3.0E+08 150 1350 1150 1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+09 1.0E+10 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 1/T ( x 10 -3 K -1 ) MT TF (hr s) 2.14 eV 290C 275C 268C 260C 150C DC POWER BURN-IN BIAS: V DS =3V I DS =35mA DEVICE HOURS: >45,000 E A = 2.14 eV (BEST FIT) EXTRAPOLATED MTTF AT T CH = 150C FIT RATE = 3.3 Filtronic Solid State Device and Process Reliability 63 MTTF Extrapolation and E A Estimate: A least-squares linear fit to the results of the four experimental results yields the Arrhenius plot shown on the preceding page, giving an extrapolated MTTF at T CH = 150°C of 3.0 x 10 8 hrs, with and Activation Energy estimate of 2.14 eV. This MTTF corresponds to a 3.3 FITs average failure or hazard rate. The Characteristics of PHEMT Aging: The PHEMT device shows exceptional stability at high temperatures, especially with regard to transconductance and RF power gain (S 21 ). The majority of the experimental devices showed little or no degradation in g M , and typically less than 0.5 dB change in Insertion Gain (S 21 ), which was measured at 12 GHz. It may be inferred from the S 21 measurements that the g M vs. V GS characteristic (transfer curve) is stable during aging. If used in a self-biased amplifier circuit (DC-grounded Gate, DC-floated Source), the additional stabilization of the operating current from the bias resistors will result in excellent bias point stability. Typical aging effects are shown below on one of the devices from the 275°C group: T CH = 275C, S/N 9 0 10 20 30 40 50 60 70 80 90 100 0 131 195 261 349 465 TIME (hrs) GM (mS) IDSS (mA) VP (mV) BVGD (V) BVGS (V) IGSO (uA) This particular device exhibited +0.3 dB of change in S 21 over the course of the burn-in, implying stability of the g M at the nominal RF bias point of I DS = 50%I DSS . Most of the experimental devices in the four groups showed an increase in S 21 over the duration of the burn-in, typically 0.2 to 0.5 dB (despite significant degradation of the package leads). This g M stability can be related to the PHEMT’s transconductance expression: g v Z d d M sat s i ≅ + ε where v sat = saturated carrier velocity, Z = Gate width, ε S = semiconductor dielectric constant, d+d i = AlGaAs layer thickness. The AlGaAs layer thickness is fixed at the time of the epitaxial growth, and to first order, does not change with device aging. By contrast, the MESFET expression contains the depletion layer depth as follows: ( ) g v Z h where h V x V V q N M sat s s G bi D ≅ ≅ + + ε ε : ( ) 2 Filtronic Solid State Device and Process Reliability 64 where h = depletion depth, V(x) = potential along channel, V G = applied Gate voltage, V bi = Schottky contact built-in voltage, q = electron charge, and N D = active layer doping. The gate sinking mechanism can cause localized compensation of the MESFET’s active layer doping, thereby altering the N D term. The MESFET is sensitive to any changes in the depletion layer depth, whatever the cause, whereas the PHEMT fundamentally is less sensitive. Both devices, however, can exhibit changes in IDSS as channel parameters change, as seen in the following expressions: ( ) PHEMT I v Z V d d MESFET I q v Z a h N DSS sat s P i DSS sat bi D : : ≅ + ≅ − ε where: V P = Pinch-Off voltage, a = active layer (channel) depth, and h bi = built-in (zero bias) depletion depth. For the PHEMT, if the pinch-off voltage is dependent on the Schottky contact built-in voltage (and therefore to the barrier height), while the MESFET expression contains dependencies on both the contact properties as well as the active layer doping. (Note that the I DSS of the 275°C sample PHEMT tracks V P closely.) Hydrogen Sensitivity FSS LP7612 discretes have been studied as to their sensitivity to molecular hydrogen (H 2 ), a potential concern for hermetically sealed long-duration space flight applications. Amplifiers and linearizers built with LP7612s were exposed to various hydrogen concentrations (up to 4%), at elevated temperatures for times exceeding 2,000 hrs. These experiments demonstrated that, with suitable circuit topologies (e.g., self-biased), components built with FSS devices could perform acceptably over a mission duration of 15 years or more. S-Level Qualification of PHEMTs FSS LP7612 PHEMT devices have been repeatedly qualified to MIL-STD-19500 JANS grade, for use in space flight hardware. LP7612 discretes are presently “in-flight” on a number of telecommunications satellites. Key elements of this qualification procedure included a 240 hour DC operating high-temperature burn-in, and a 1000 hour operating life test. Failures were defined as devices exhibiting changes greater t15% (1000 hour life test) or t20% (240 hour burn-in) in I DSS , g M , or V P , and t0.5 dB change in the power gain (S 21 ) at 12 GHz. A separate limit of t100% or 1 µA (whichever was greater) was imposed for the reverse leakage current (I GSS ). Samples of the wafer lot under evaluation must complete both the 240 hour burn-in, at a channel temperature of 225°C, and a 1000 hour life test at a channel temperature of 150°C; the sampling method is based on LTPD (Lot Tolerance Percent Defective) levels of 30% and 50%. PHEMT MMIC Reliability MMICs based on the DHPHEMT active device technology are currently undergoing reliability testing at FSS, following the discrete PHEMT results. The LMA411, a two-stage, self-biased, low noise MMIC, has demonstrated exceptional stability at a 275°C channel temperature, with a MTTF exceeding 800 hrs. This channel temperature was achieved with a 240°C baseplate temperature in addition to self-heating. Operating current changes were less than 5%, and the small-signal gain degraded less than 0.3dB for all test devices. The self-biased circuit topology is resistant to changes in I DSS and V P , so these results are consistent with the discrete results. All of the test MMICs suffered severe physical deterioration due to the extremely high baseplate temperature; the testing was discontinued after it became apparent that the package and off-chip components could not withstand the test conditions. Filtronic Solid State Device and Process Reliability 65 Fabrication Process Stability All FSS discrete and MMIC devices are fabricated by essentially identical processes. The epitaxial specifications and evaluation criteria, fabrication processes and test methods, and DC/RF test specifications and test methods are fully documented. Statistical Process Control (SPC) methods are utilized to monitor the fabrication processes, with key parameters continuously tracked on trend charts. The Division is quality certified to ISO 9001 standards. All discrete and MMIC wafers and wafer lots are evaluated and screened by a tailored MIL-19500 JANC screening procedure, as detailed in FSS’s Applications Notes, Discrete FET/PHEMT Devices, Rev. A, November 1986. Current Reliability Characterization Effort: FSS is continuing its assessment of the reliability of its devices, and the following experiments are planned: • Accumulation of additional LP7612 high-temperature Arrhenius data • Reliability of large gate periphery discrete devices • Effects of RF drive (to 1dB compression levels) on reliability • Verification of degradation mechanisms by microscopic analysis (e.g., Auger and SIMS) • Effects of high RF power level exposure, pulsed and CW, to 5W levels Prepared By: Michael Jon Bailey Product Engineering Manager Reviewed By: Bill Ireton Manager of Semiconductor Operations