A Cryogenic Ultra-Low-Noise MMIC-based LNA with a discrete First Stage Transistor Suitable for Radio Astronomy Applications
11 A Cryogenic Ultra-Low-Noise MMIC-based LNAwith a discrete First Stage Transistor Suitable forRadio Astronomy Applications
Mark A. McCulloch, Simon J. Melhuish, and Lucio Piccirillo
Abstract —In this paper a new design of MMIC based LNA isoutlined. This design uses a discrete 100-nm InP HEMT placedin front of an existing InP MMIC LNA to lower the overallnoise temperature of the LNA. This new approach known asthe Transistor in front of MMIC (T+MMIC) LNA, possessesa gain in excess of 40 dB and an average noise temperatureof 9.4 K compared to 14.5 K for the equivalent MMIC-onlyLNA measured across a 27–33 GHz bandwidth at a physicaltemperature of 8 K. A simple ADS model offering further insightsinto the operation of the LNA is also presented and a potentialradio astronomy application is discussed.
Index Terms —Amplifiers, cryogenic, InP HEMT, low-noiseamplifier (LNA), microwave integrated circuit (MIC), monolithicmicrowave integrated circuit (MMIC), radio astronomy.
I. I
NTRODUCTION L OW noise amplifiers (LNAs) form the most importantpart of the sensitive coherent receivers that are used inradio astronomy and the advancement that has taken place inour understanding of the universe over the last few decades isin no small part down to the development of ever lower noiseLNAs. A good LNA should possess a reasonable amount ofgain, which is used to suppress the noise contribution of thefollowing components, and simultaneously it should contributeas little additional noise as possible to the overall system.Very low noise Ka-band LNAs were developed by theJodrell Bank Observatory (JBO) in the previous decade forthe European Space Agency’s
Planck
Low Frequency Instru-ment (LFI) [1]. These LNAs utilized the common microwaveintegrated circuit (MIC or hybrid) approach and were based onfour discrete Indium Phosphide (InP) 100-nm gate length highelectron mobility transistors (HEMTs), with the lowest-noiseamplifiers possessing an absolute minimum noise temperatureof 5 K and a band average (27–33 GHz) of 8.1 K.However, the last three decades have also seen the develop-ment of an alternative approach to low-noise amplification: themonolithic microwave integrated circuit (MMIC) LNA. Theseamplifiers integrate all of the LNA’s components (transistors,transmission lines, bias networks) onto a substrate with a largedielectric constant, for example Gallium Arsenide (cid:15) r ≈ . orIndium Phosphide (InP) (cid:15) r ≈ . . The current lowest-noiseKa-band MMICs [2] achieve an average noise temperature of15.5 K. The integration of these components gives the MMIC LNAseveral advantages over comparable MIC designs. Their in-tegrated nature reduces the number of components and bondwires required, making them easier and less time-consumingto integrate into a module (chassis). The use of substrateswith a large dielectric constant results in substantially smallerLNAs (the MMIC chips are typically < a r X i v : . [ a s t r o - ph . I M ] O c t arrays with a large number of pixels are planned and one of thepriorities will be to detect the very weak B-mode polarizationsignal. II. D ESIGN B ASIS
A. Cascaded Systems
The noise temperature of a cascaded system is given by thefamiliar Friss equation (1) [7], which shows that the overallnoise temperature of such a system is dominated by boththe noise temperature ( T ) and the gain ( G ) of the firstcomponent. T n = T + T G + T G G + . . . (1)Thus placing a discrete transistor in front of a MMICshould lead to two beneficial effects. Firstly, the MMIC’scontribution to the noise temperature of the overall amplifierwill be suppressed by the gain of this transistor. Secondly, theLNA’s noise temperature will be determined primarily by thenoise temperature of the transistor. Therefore if the transistorpossesses a noise temperature that is less than the noisetemperature of the MMIC, the resulting LNA will possess anoverall noise temperature only slightly higher than that of thetransistor. B. Linear Two-Ports
The advantages of a discrete matching network and dis-crete transistor can be illustrated by considering the work ofPospieszalski [8], who showed that the minimum noise ( T min ) of any linear two-port device is given by (2), where T is thestandard temperature (290 K), Z s is the source impedance, Z opt is the optimum source impedance, R s is the sourceresistance and R opt is the optimum source resistance. N isgiven by (3), where G opt is the optimal conductance and R n and G n are the noise resistance and conductance respectively. T n = T min + N T | Z s − Z opt | R s R opt (2) N = G n R opt = R n G opt (3)(2) shows that it should be possible to match all devicesto a given input impedance for which T n will equal T min .The transistor’s matching network can be designed to do this.However, whilst this is relatively “easy” to do with discretedevices, the integrated nature of MMICs means that it is notusually possible to match their 50 Ω inputs ideally to theinput impedance of the first transistor. Indeed it has beenshown that even at lower frequencies the on-chip matchingnetwork contributes several degrees more to the overall noisetemperature of the amplifier than an off-chip network [9].These issues become even more apparent when you considerthe use of LNAs at cryogenic temperatures. Typically an LNAdesigner may need to adjust matching network parametersiteratively to converge on a design optimized for cryogenicoperation. For MIC LNAs this is much easier to achievethan for MMICs, for which the design is fixed at the time of the wafer run, although variations on a design may beaccommodated on a wafer.Consequently we developed a design for an LNA known asthe Transistor in front of MMIC (T+MMIC) that exploits notonly an off-chip matching network, but also utilizes a discretetransistor for the first stage of amplification. This facilitates: • An accurate match to the first stage transistor, optimizingnoise performance. • Ease of modification to the design for cryogenic use. • Suppression of the noise of the less-than-ideally matchedMMIC. III. T HE D ESIGN
A. The Active Devices
The T+MMIC LNA is based around two active components:a high-quality transistor and a good MMIC LNA. The transis-tor (Fig. 1a) is a × µ m, 100-nm gate-length InP HEMT,that had originally been supplied to JBO by JPL Pasadena foruse in the Planck project. The transistor was fabricated underthe Cryogenic HEMT Optimization Program (CHOP) [10],and originated from wafer run 3. These particular transistors,known as Cryo-3, still offer state-of-the-art noise performanceand the use of such a transistor allows the T+MMIC LNA tobe compared with the
Planck
LFI Ka-band LNAs.The MMIC LNA (Fig. 1b) was developed as part of the Eu-ropean Commission’s FARADAY project [11]. Since this wasa radio astronomy project the MMIC LNA already possesseda good cryogenic noise temperature, typically around 20 K,and a gain in excess of 40 dB across its 26–36 GHz operatingband. The FARADAY MMICs were fabricated on InP byNorthrop Grumman Space Technologies (NGST). They utilizefour × µ m gate width, 100-nm gate length transistors. (a) (b)Fig. 1. The Active Devices: (a) a Cryo-3 transistor with approximatedimensions µm x µm , (b) a FARADAY MMIC LNA with dimensions µm x µm . B. Module Design
For ease of assembly and to allow the use of pre-existingdesigns and components the module that houses the T+MMICLNA is a merger of two existing JBO-designed LNA modules.The transistor section is based around the first stage of the
Planck
Planck
Cryo-3 input matching network, which avoided the need for a re-design of the module, though this did restrict thebandwidth of the amplifier to 27-33 GHz. The merger of twomodules also resulted in the transistor and the MMIC beingconnected by a rather long ( ∼ µ m PolyflonCuflon substrate [13], which has an electrical permittivity anda dielectric loss tangent of 2.05 and 0.00045 respectively. MMIC500.8 0.1 101000 0.10.8 1616 1050 VdVg . . . . G DS
TL1 TL2 TL3 TL40.2 0.2 Fig. 2. RF circuit schematic for the T+MMIC LNA. Resistors and capacitorshave units of Ω and pF respectively, bond wire lengths are in mm. Trans-mission lines (TL) 1, 3 & 4 all have an impedance of 50 Ω and lengths of1.05, 7.2 and 31 mm respectively. TL 2 which is used to match the the sourceimpedance to the transistor’s input impedance for minimum noise temperaturehas an impedance of 23 Ω and a length of 0.85 mm.Fig. 3. The assembled LNA. From left to right: probe, input matchingnetwork, gate bias, transistor, drain bias, 50 Ω microstrip line, MMIC, 50 Ω output transmission line. A schematic of the RF circuit can be seen in Fig. 2 C. Theroretical Performance
Since the
Planck amplifier used a Cryo-3 transistor for itsfirst 2 stages we can estimate the average noise temperature ofour LNA by using (1) and the average noise performance ofboth the
Planck amplifier and the MMIC. Considering the firsttwo stages (from (1) the third and forth stages are negligible)of the
Planck amplifier the noise temperature can be writtenas the following: T P lanck = x + xG (4)Where T P lanck is the average noise temperature of the
Planck amplifier (8.1 K), x and G are the noise temperatureand gain (8 dB) of the Cryo-3 respectively. For the T+MMICamplifier the following can also be written: T LNA = x + T MMIC G (5)Where T MMIC is the average noise temperature of the Fara-day MMIC ( ≈
20 K). Re-arranging (4) and (5) and eliminating x gives the expected noise temperature of the T+MMIC LNAas ≈
10K (6). T LNA = T P lanck G + T MMIC G ≈ K (6)IV. N
OISE T EST E QUIPMENT
A. RF System
The noise test set-up has been previously used to developLNAs for the
Planck
Satellite and the Merlin array of radiotelescopes. It consists of an HP (now Agilent Technologies)8350B sweep generator, with an 83550A plug-in module and a83554A mm wave source module to provide an LO to a downconverting mixer, and an 8970B noise figure meter on the IF.For room-temperature measurements an Agilent R347B noisesource is used to supply two different levels of input noiseto the LNA. Cryogenic measurements were made using the“hot” and “cold” load approach where a variable temperaturemicrowave-absorbing (and therefore emitting) waveguide loadis coupled to the input of the LNA. Since the load is placedwithin the cryostat and we are able to measure the temperatureof the load with great accuracy our results should have asimilar level of accuracy. S measurements of the load showthat the match to the LNA is reasonable ( S < -15 dB), thuswe are confident that T phys (cid:39) T RF . B. The Cryostat
To perform the cryogenic noise measurements the LNAwas placed in a cryostat that had previously been developedfor sub-2-K physical temperature noise temperature investi-gations [14]. The cryogenic cooling system is built around aPulse Tube Cooler (PTC), manufactured by Sumitomo HeavyIndustries (SHI), with two cooling stages; the first cools aradiation shield to ∼
50 K and pre-cools the 2nd stage, whilstthe 2nd stage cools (with no LNA) to ∼ temperature to be increased. An identical approach is usedto control the temperature of the hot / cold load. Both theLNA and the load are attached to the 3-K stage by means ofa thermal switch, which allows for a more rapid adjustmentin temperature than would be possible with only a resistorand a weak thermal link. Waveguide is used to carry the RFsignal out of the cryostat. The waveguide is composed of twosections; a section of thin-walled gold-plated stainless steelwaveguide (SS WG) and a section of brass waveguide (BRWG). These are used rather than copper as they offer reducedthermal conduction. The load is connected to the amplifiervia a waveguide thermal break and a length of stainless steelwaveguide; again this is to isolate thermally the LNA fromthe load.A cryogenic readout / control unit that we developedpreviously from a device used for the QUAD telescope isused to monitor key temperatures within the system and tocontrol the various heaters. A PID control loop sets load andLNA temperatures. We use silicon diode and ruthenium-oxidethermometers, calibrated against a rhodium-iron standard andGRTs to monitor the temperature. Fig. 4 shows the layout ofthe cryogenic system for noise temperature measurements. PTC SS e W G B R e W G F r i dge Fig. 4. The cryostat’s layout when configured for noise temperature mea-surements. An additional waveguide can be installed for S parameter mea-surements. The optional 1 K fridge is also shown. Reproduced from [14].
V. P
ERFORMANCE
The performance of the T+MMIC LNA was measured ata physical temperature of 8 K and the results are shown inFig. 5a. For comparison purposes the results for a Faradayonly MMIC LNA are also shown. For completeness, the roomtemperature performance is also shown in Fig. 5b. In all casesthe amplifiers were biased for minimum noise. To highlight theeffectiveness of this technique the
Planck
LFI average noisetemperature (8.1K) is also shown, though it should be notedthat this was measured at a physical temperature of 20 K. TheT+MMIC LNA was measured at 8-K owing to recent results[14] where an LNA was cooled to 2-K that showed that noisetemperature continues to decrease with decreasing physicaltemperature. We estimate that at cryogenic temperatures theuncertainty in our noise measurements is ± K whilst at roomtemperature we estimate that this increases to ± K. Theseestimates are based on repeated observations and are consistentwith other measurements that have used similar techniques[15].
27 28 29 30 31 32 33Frequency, [GHz] 010203040 N o i s e T e m pe r a t u r e , [ K ] G a i n , [ d B ] T+MMIC MMIC Planck 8K (a)
27 28 29 30 31 32 33Frequency, [GHz] 140160180200220240260 N o i s e T e m pe r a t u r e , [ K ] G a i n , [ d B ] T+MMIC MMIC 290K (b)Fig. 5. The T+MMIC LNA’s performance at (a) 8 K and (b) 290 K, comparedwith measurements from an LNA consisting of only an MMIC (same designand wafer).
VI. M
ODELING
For further insight into the operation of the T+MMIC LNAa model of the amplifier was produced using Agilent’s Ad-vanced Design System (ADS). The transistor stage (transistor,bias networks, input and output transmission lines) is fullysimulated in ADS, the outputs of which are then combinedwith the MMIC’s S parameters and noise behavior to give theLNA’s overall characteristics.
A. The Transistor Stage1) Cryo-3 Transistor:
The Cryo-3 transistor was simulatedusing a modified version (Fig. 6) of the standard 15-parameterequivalent circuit model (see [16]–[18] for a discussion onparameter extraction). This particular version was developedby M. Pospieszalski [19]. The equivalent circuit parameterswere measured as part of the
Planck project and are shown inTable I. Further details can also be found in [20]. For the roomtemperature simulations the passive extrinsic and intrinsicparameters are left unchanged from those of the cryogenicsimulation [21], whilst the Pospieszalski noise equivalenttemperatures [22] T d , T g and T a are adjusted appropriately.For room temperature modeling the transconductance g m isdecreased by ∼
20% over the cryogenic temperature value [23].The Voltage Controlled Current Source (VCCS) illustratedin Fig. 6 is the approach used in ADS to model a 3-porttransistor. The other parameters are simulated by using ideallumped components. The two transmission lines denoted TLare used to represent the inductance of the source pads. Thisinductance is modeled through the use of ideal microstrip lineswith impedance = 20 Ω , length = 63.5 µ m, (cid:15) r = 13.1.
2) Bias Networks:
The components in the transistor’s biasnetworks are also simulated using lumped components, whilstthe transmission lines are simulated using microstrip compo-nents.
Gate DrainSourceC gs C gd C ds R d R g R s C pg C pd g m L d L g R gs R ds τ C pd C pd C pd C pg Source
TL TL
Fig. 6. A transistor equivalent circuit, suitable for use in Agilent’s AdvancedDesign System. Developed by M. Pospieszalski [19].
3) Bond Wires:
Although ADS includes bond wire sim-ulation tools, the bond wire model only takes into accountthe wire’s inductance, not the capacitive effects that arisefrom the surrounding metallic structure and the bonding pointsthemselves. Since these effects can have a significant impacton the performance of the RF circuit, considerable work hasbeen done to investigate and describe the behavior of bondwires [24], [25] at mm-wave frequencies. The quasi-static
TABLE IC
RYO -3 EQUIVALENT CIRCUIT PARAMETERS g m
80 mS 67 mSExtrinsic Parameters Intrinsic ParametersRg 1 Ω Cgs 52 fFRd 5 Ω Cgd 24 fFRs 2.2 Ω Cds 10 fFCpg 4.6 fF Rds 135 Ω Cpg 4.6 fF Rgs 4 Ω Cpd 12 fF τ model outlined in [25] utilizes 4 transmission line elementsto help model the wire’s behavior. Ansys’ High FrequencyStructure Simulator (HFSS), an EM simulator, can also beused to simulate the behavior of bond wires. However, likethe quasi-static model the wire’s profile needs to be wellknown, which is not always possible, as was the case here.Therefore a simpler single transmission line approach, wherethe bond wires are simulated using high impedance ( ),ideal ( (cid:15) r = 1) transmission lines was used. This approach hadproved useful during the development of the Planck
LNAs[19]. Fig. 7 illustrates the difference between a 500 µ m (linearlength) bond wire that was simulated in ADS using the bondwire simulation component, an equivalent bond wire (Fig. 7a)simulated in HFSS and an ideal 150 Ω transmission line. Fig.7b shows that a single high impedance transmission line stillrepresents a good approximation to the HFSS simulation. B. Faraday MMIC S Parameters
To allow the MMIC part of the amplifier behavior to be sim-ulated in both the room temperature and cryogenic model, anequivalent MMIC from the Faraday Project was integrated intoa suitable test module. This LNA’s S Parameters were mea-sured using an Agilent Technologies PNA-X from which theMMIC’s S-parameters were extracted using the de-embeddingtechnique outlined in [26]. Since the de-embedding process(7) requires the S and T (8) parameters of the surroundingtest fixture to be known, models of the test module’s inputand output waveguide-to-microstrip transitions were producedin HFSS (Fig. 8) and the S-parameters simulated (Fig. 9). Theactual de-embedding was performed using the dedicated de-embedding components in ADS. (cid:2) T I (cid:3) − (cid:2) T I (cid:3) (cid:2) T MMIC (cid:3) (cid:2) T O (cid:3) (cid:2) T O (cid:3) − = (cid:2) T MMIC (cid:3) (7) (cid:20) T T T T (cid:21) = 1 S (cid:20) S S − S S S − S (cid:21) (8) (a) S11S21ADSTLHFSSFrequency, 250MHz - 50GHz (b)Fig. 7. (a) An HFSS model of a bond wire, ready for simulation. (b) Theresults of 3 different (ADS, HFSS and an ideal transmission line) approachesto the modeling of a 500 µm bond wire.Fig. 8. The HFSS model of the input broadband waveguide to microstriptransition. In the case of the cryogenic measurements the input and outputstainless steel and brass waveguides are also simulated in HFSS. C. Modeling Performance1) Gain:
The predicted room-temperature and 8-K S values were compared with the actual measured S valuesand can be seen in Fig. 10 and Fig. 11. The gain measured by
26 28 30 32 34 36 38 40Frequency, [GHz]−60−40−20020 R e t u r n Lo ss , [ d B ] −2.0−1.5−1.0−0.50.0 I n s e r t i on Lo ss , [ d B ] Input: S11 S21 Ouput: S11 S21
Fig. 9. The simulated results for the input and output broadband waveguideto microstrip transition and subsequent 50 Ω microstrip lines. the NFM is also shown. The slight ripple in the modeled S is due to the effects of the long input and output waveguidesnot being fully removed from the measurements of the MMIC.
2) Noise:
Using the Pospieszalski noise equivalent temper-atures and the equivalent circuit model, the noise performanceof the transistor stage can be modeled. Using this with (1) andthe noise performance of the MMIC, the noise performanceof the T+MMIC-based LNA can be predicted. The modeledand actual performance are compared in Figs. 10 (290 K) and11 (20 K).
26 28 30 32 34 36Frequency, [GHz]−40−200204060 G a i n , [ d B ] N o i s e T e m pe r a t u r e , [ K ] S21 T n Gain MMIC (T n ) Fig. 10. Modeled (dashed line) and measured (solid line) 290-K T+MMIC S and noise temperature ( T n ) results for 26–36 GHz. The T+MMIC’s gainrecorded by the NFM is also shown, along with the noise performance of theMMIC only amplifier. The models confirm that the amplifier is behaving asexpected from the Friss equation (1). Fig. 11 shows thatwithin the intended operating bandwidth (27-33 GHz) thenoise temperature of the T+MMIC LNA sits just above thenoise temperature of the transistor (first stage), with the gainsuppressing the noise contribution of the MMIC. Outside thisband, however, once the gain provided by the first stage
26 28 30 32 34 36Frequency, [GHz]0204060 G a i n , [ d B ] N o i s e T e m pe r a t u r e , [ K ] S21 T n Gain Cryo3 (T n )Cryo3 (Gain) Fig. 11. Modeled (dashed line) and measured (solid line) 8-K T+MMIC S and noise temperature ( T n ) results for 26–36 GHz. The modeled noisetemperature and the gain of the transistor are also shown, along with the gainof the T+MMIC LNA as recorded by the NFM. reduces, the noise of the MMIC becomes more significant andthe noise temperature of the T+MMIC LNA drifts towards thatof the MMIC, as can be seen in Fig. 10.VII. C ONCLUSION
MMIC LNAs are now the preferred choice for the LNAsrequired by radio astronomy, but their noise performance isstill inferior to that of MIC-based LNAs. One possible solutionis to use a discrete transistor in front of the MMIC. Thispaper has reported on the development of such an LNA, withan average noise temperature of 9.4 K. This is some 4–5 Klower than an equivalent MMIC LNA, representing a near improvement. Cryogenic cooling to 8 K has also resulted inan amplifier that almost matches the noise performance of thelowest-noise Ka-band LNAs so far developed, illustrating thatcooling below the typical 15–20 K that is currently used bymost radio observatories may prove beneficial. We have alsopresented a simple approach to modeling such an amplifier,showing that the MMIC can almost be regarded as a “blackbox” in terms of the amplifier’s development with only thetransistor’s equivalent circuit parameters and noise parametersneeding to be measured with a probe station. The modeled dataalso show that we have demonstrated effective suppressionof the (higher) MMIC noise by the lower-noise first-stagetransistor, within its operating band.VIII. D
ISCUSSION
Considering our earlier estimate of the expected noisetemperature of the T+MMIC LNA; at 20-K (Fig. 12) the noisetemperature is slightly higher (11.4 K) than that expected fromthe
Planck amplifier (10 K). This is likely due to the
Planck amplifiers using a slightly different version of the Cryo-3which had a thiner passivation layer. These transistors werefound to have a slightly better noise performance than the typeof Cryo-3 transistor that was integrated into this amplifier. During the development, the LNA was found to oscillateunder certain conditions. Examining the measured and mod-eled S parameters showed that the amplifier was conditionallystable at several fequencies within its design bandwidth. Thisis likely due to the large quantity of gain and could potentiallybe resolved by a lower gain MMIC.This approach would also be applicable to the experimentssuch as the Q/U-Imaging-Experiment (QUIET) which usedintegrated modules to observe the CMB, but encounteredcompression issues when using multiple MMICs in cascade toachieve sufficient gain [27]. This approach would offer slightlyreduced additional gain, thus preventing compression, whilstalso offering the potential for reducing the noise temperature.One obvious drawback of this technology is the need todevelop a new module for the integration of the MMIC andthe transistor. A preferred approach would be to mount thetransistor into its own module and connect via waveguide toan existing MMIC-based amplifier module.
27 28 29 30 31 32 33Frequency, [GHz]46810121416 N o i s e T e m pe r a t u r e , [ K ]
8K 20K Planck
Fig. 12. The noise temperature of the T+MMIC LNA at 20 K and 8 K,compared to the average noise temperature of the Planck amplifiers at 20 K. A CKNOWLEDGMENT
The authors would like to thank D. Shepard for machiningthe LNA module, E. Blackhurst for assembling the LNAand A. Galtress for his help in designing the module. Wewould also like to thank, Prof P. Wilkinson and Dr D. Georgefor supplying the Faraday MMIC LNA, Prof R Davis forsupplying the Cryo-3 transistor, Agilent for supplying ADSand our test equipment and Ansys for supplying HFSS. Thiswork was funded by the Science and Technology FacilitiesCouncil (STFC), Consolidated Grant ST/J001562/1.R
EFERENCES[1] R. Davis et al. , “Design, development and verification of the 30 and44 ghz front-end modules for the planck low frequency instrument,”
Journal of Instrumentation , vol. 4, no. 12, pp. T12 002 (36 pp.) –, 2009.[2] Y.-L. Tang et al. , “Full ka-band high performance inp mmic lna module,”in
Microwave Symposium Digest, 2006. IEEE MTT-S International ,2006, pp. 81–84.[3] J. A. Rubio-Martn et al. , “The quijote-cmb experiment: studying thepolarisation of the galactic and cosmological microwave emissions,” pp.84 442Y–84 442Y–11, 2012.[4] S. Bhaumik et al. , “Comparative study of hemts for lnas in squarekilometre array telescope,” in
Microwave Symposium Digest (MTT),2011 IEEE MTT-S International , 2011, pp. 1–4.[5] I. Buder, “Q/u imaging experiment (quiet): A ground-based probe ofcosmic microwave background polarization,” vol. 7741, San Diego, CA,United states, 2010, pp. The Society of Photo–Optical InstrumentationEngineers (SPIE) –.[6] M. McCulloch et al. , “A low noise ka-band amplifier for radio astron-omy,” in
Microwave Integrated Circuits Conference (EuMIC), 2012 7thEuropean , 2012, pp. 881–884. [7] H. Friis, “Noise figures of radio receivers,”
Proceedings of the IRE ,vol. 32, no. 7, pp. 419–422, 1944.[8] M. Pospieszalski, “On the measurement of noise parameters of mi-crowave two-ports (short paper),”
IEEE Trans. Microw. Theory Tech. ,vol. 34, no. 4, pp. 456 – 458, 1986.[9] B. Aja et al. , “Cryogenic low-noise mhemt-based mmic amplifiers for4 -12 ghz band,”
IEEE Microw. Wireless Compon. Lett. , vol. 21, no. 11,pp. 613 –615, nov. 2011.[10] J. Shell, “The cryogenic dc behavior of cryo3/az1 inp 0.1-by-80-micrometer-gate high electron mobility transistor devices,”
Jet Propul-sion Lab., La Ca˜nada Flintridge, CA, IPN PR , pp. 42–169, 2007.[11] D. Kettle et al. , “A lattice matched inp chip set for a ka band radiometer,”
IEEE Trans. Microw. Theory Tech. , pp. 4 pp. –, 2005.[12] Rubi˜no-Mart´ın et al. , “The QUIJOTE CMB Experiment,” in
Highlightsof Spanish Astrophysics V , 2010, p. 127.[13] “Polyflon cuflon data sheet,” Polyflon Company, Norwalk, USA.[14] S. Melhuish et al. , “Lna noise behaviour below 10k physical tempera-ture,” in
Microwave Integrated Circuits Conference (EuMIC), 2012 7thEuropean , 2012, pp. 393–396.[15] S. Munoz et al. , “Drain temperature dependence on ambient temperaturefor a cryogenic low noise c-band amplifier,” in
Microwave Conference,1997. 27th European , vol. 1, sept. 1997, pp. 114 –118.[16] M. Berroth and R. Bosch, “Broad-band determination of the fet small-signal equivalent circuit,”
Microwave Theory and Techniques, IEEETransactions on , vol. 38, no. 7, pp. 891–895, 1990.[17] G. Dambrine et al. , “A new method for determining the fet small-signal equivalent circuit,”
Microwave Theory and Techniques, IEEETransactions on , vol. 36, no. 7, pp. 1151–1159, 1988.[18] R. Tayrani et al. , “A new and reliable direct parasitic extraction methodfor mesfets and hemts,” in
Microwave Conference, 1993. 23rd European ,1993, pp. 451–453.[19] M. W. Pospieszalski, “Notes on modeling of five stage q-band amplifierusing cryo-3 µm hemts.”[20] M. Pospieszalski, “Cryogenic amplifiers for jansky very large array re-ceivers,” in Microwave Radar and Wireless Communications (MIKON),2012 19th International Conference on , vol. 2, 2012, pp. 748–751.[21] M. Pospieszalski et al. , “Millimeter-wave, cryogenically-coolable ampli-fiers using alinas/gainas/inp hemts,” in
Microwave Symposium Digest,1993., IEEE MTT-S International , 1993, pp. 515–518 vol.2.[22] M. Pospieszalski, “Modeling of noise parameters of mesfets and modfetsand their frequency and temperature dependence,”
Microwave Theoryand Techniques, IEEE Transactions on , vol. 37, no. 9, pp. 1340–1350,1989.[23] M. Pospieszalski and E. Wollack, “Characteristics of broadband inphfet millimeter-wave amplifiers and their applications in radioastronomyreceivers,” in
Millimeter Waves, 1997 Topical Symposium on , 1997, pp.143–146.[24] H.-Y. Lee, “Wideband characterization of a typical bonding wire formicrowave and millimeter-wave integrated circuits,”
Microwave Theoryand Techniques, IEEE Transactions on , vol. 43, no. 1, pp. 63–68, 1995.[25] F. Alimenti et al. , “Modeling and characterization of the bonding-wireinterconnection,”
Microwave Theory and Techniques, IEEE Transactionson , vol. 49, no. 1, pp. 142–150, 2001.[26]
De-embedding and Embedding S-Parameter Networks Using a VectorNetwork Analyzer , Application note 1364-1 ed., Agilent Technologies,http://literature.agilent.com/litweb/pdf/5980-2784EN.pdf.[27] R. Reeves, “Quiet coherent polarimeter modules,”
Journal of LowTemperature Physics , vol. 167, no. 5-6, pp. 929 – 35, 2012.
Mark McCulloch received his MPhys Physics withAstrophysics degree from the University of Manch-ester in 2008, and has recently submitted his Ph.Ddegree thesis at the University of Manchester. Forhis Ph.D he investigated potential enhancements toLNAs with the aim of lowering the noise tempera-ture of LNAs with special focus to future CosmicMicrowave Background observatories.
Simon Melhuish graduated in Physics from NewCollege, Oxford, and worked on telescope systemsfor Cosmic Microwave Background studies for hisPh.D at Jodrell Bank, University of Manchester. Hehas been responsible for elements of various radiotelescope projects, including the Very Small Array,QUaD and Clover.