High-Tc Superconductive WB Compressive Receivers_10.1.1.73

8/8/2019 High-Tc Superconductive WB Compressive Receivers_10.1.1.73

1/35

LYONS, ARSENAULT, ANDERSON, SOLLNER, MURPHY, SEAVER, BOISVERT, SLATTERY, AND RALSTON

High-TcSuperconductive Wideband Compressive Receivers

VOLUME 9, NUMBER 1, 1996 THE LINCOLN LABORATORY JOURNAL

33

High-TcSuperconductive

Wideband CompressiveReceiversW. Gregory Lyons, Duane R. Arsenault, Alfredo C. Anderson,T.C.L. Gerhard Sollner, Peter G. Murphy, Mark M. Seaver,Rene R. Boisvert, Richard L. Slattery, and Richard W. Ralston

sWideband compressive receivers are an attractive application of analog high-

transition temperature superconductive (HTS) microwave filters. Chirp filters

form the basis of compressive receivers, implementing a chirp-transformalgorithm in the analog domain for real-time spectral analysis. HTS tapped-

delay-line chirp filters are an enabling technology for instantaneous bandwidths

greater than 1 GHz, and have evolved sufficiently to support dispersive delays as

long as 40 nsec with multigigahertz bandwidths and time-bandwidth products

in excess of 100. Long dispersive delays have been obtained by using a bonded/

thinned-wafer technique to fabricate YBa2Cu3O7stripline devices on 5-mil-

thick, 2-in-diameter LaAlO3 substrates. These filters have produced better than

18-dB error sidelobes. In addition, a 3-GHz-bandwidth HTS compressive

cueing receiver was recently delivered to the Naval Research Laboratory to be

flown on the High-Temperature Superconductivity Space Experiment (HTSSE),

and demonstrations have been performed by combining HTS chirp filters withconventional compressive-receiver hardware. We propose a novel compressive

cryoreceiver architecture that combines HTS, cryoelectronic, and advanced

high-speed semiconductor technologies. The proposed receiver will rival the

sensitivity of a narrowband receiver while providing unprecedented

instantaneous wideband frequency coverage, and future developments will

extend the bandwidth capability. We make detailed comparisons to an all-digital

receiver and to channelized-filter receiver architectures. An HTS compressive

receiver is projected to be superior in overall size, weight, and power; its

applications include electronic warfare and dynamic molecular spectroscopy for

remote sensing.

T since1986 in the application of thin-film high-transition temperature (high-Tc) supercon-

ductors to passive analog microwave devices [13].High-quality, high-Tc superconductive (HTS) thinfilms with microwave surface resistances many ordersof magnitude below that of copper at 77 K can nowbe reliably deposited over a 3-in-diameter substrate

area. This advance has led to the implementation of alarge variety of HTS passive microwave device struc-tures. The planar nature of thin-film HTS structuresoffers a substantial size and weight advantage overlow-loss waveguide structures made from normalmetal, and the cryogenic operation of HTS devicesaffords the system engineer an opportunity to achievea very low-noise receiver front end. Planar HTS struc-


2/35



34 THE LINCOLN LABORATORY JOURNAL VOLUME 9, NUMBER 1, 1996

tures require two-dimensional lithographic tech-niques to define devices rather than tedious and inac-curate three-dimensional machining techniques that

waveguides and dielectrics typically require. Careful

design of HTS microwave devices in conjunction with low-loss dielectrics at cryogenic temperaturesmay also allow higher quality factors (Q) and lowerlosses to be obtained than with conventional

waveguide structures.Passive microwave devices were an early favorite in

the history of HTS thin-film development efforts be-cause of their simple single-layer structure. Research-ers soon learned, however, that a very low-loss passivemicrowave device depends on HTS film quality anddesign techniques to the same extent as an active mul-

tilayer Josephson-junction circuit. For example, weunderstand that a high-Q HTS microwave device

with good power-handling capability requiresroughly the same film quality as a low-noise two-

junction HTS circuit for magnetic sensing [4]. Good-quality HTS films can be readily obtained today anda variety of useful HTS microwave device structureshas been demonstrated. Researchers are now focusingon filter structures and HTS applications that willhave the greatest impact on their respective overallmicrowave system (e.g., radar system, communica-

tions satellite, or remote-sensing receiver).This enhancement in system-level performance

must be significant enough to justify the burdens ofcryogenic cooling, which include increased powerconsumption, increased size and weight, and poten-tially reduced reliability. The 4.2-K operating tem-perature of conventional superconducting microwavedevices severely limited their application because thecryogenic systems required to achieve 4.2 K are toocomplex and cumbersome. The advent of HTS de-vices eases this cryogenic burden. Operating tempera-

tures between 50 and 90 K allow the use of simpler,smaller, more reliable, and less power-hungry cryo-genic coolers, such as those used or planned for infra-red-imaging systems on remote-sensing satellites andmilitary platforms [5].

HTS chirp filters are an important example of apassive microwave device that has a significant sys-tem-level impact. They represent an enabling tech-nology because they support bandwidths beyond the

1-GHz limit of surface-acoustic wave (SAW) com-pressive receiver technology [6] and the 2-GHz limitof acousto-optic channelizer technology [7], and be-cause superconductivity is the only technology that

successfully supports multigigahertz bandwidths inan accurate chirp-filter structure [8, 9].

The chirp filter and analog chirp-transform algo-rithm form the basis of a spectral-analysis receiverknown as a compressive receiver. The term compres-sive receiver is derived from the receivers use of ananalog Fourier transform to perform a virtual chan-nelization of the wideband input and compress eachRF input tone into a narrow pulse. The detected out-put from a compressive receiver consists of thesenarrow, or compressed, pulses, each representing the

frequency bin of a transformed input signal, arrivingin sequential order in the time domain. The inputfrequency of the signals is determined by measuringthe time positions of these pulses. Since the detectedpulses can appear close together in time and becausethey are extremely narrow (the pulsewidth is inverselyproportional to the analysis bandwidth), high-speedlogic circuits are required to process the pulses. TheHTS wideband compressive receivers described inthis article and in Reference 10 represent a union be-tween a multigigahertz-bandwidth HTS-based ana-

log chirp transform and advanced high-speed semi-conductor circuits for pulse processing.

The virtual channelization function of a compres-sive receiver is power efficient, and provides the finefrequency resolution and sensitivity normally avail-able only from a narrowband receiver. These at-tributes are desirable in military electronic-warfareapplications and in dynamic molecular spectroscopyfor remote sensing. Both applications demand ex-tremely wide bandwidth coverage, constantly push-ing the state of the art in receivers. Military elec-

tronic-warfare applications push toward continuoustime coverage of tens of gigahertz of bandwidth withthe highest possible dynamic range and sensitivity.Remote-sensing applications, such as satellite-baseddynamic molecular spectroscopy of the atmosphere,often involve multigigahertz-wide molecular-excita-tion linewidths. The compressive receivers ability tosimultaneously measure the full extent of these broadlinewidths leads to greatly reduced integration times.


3/35



VOLUME 9, NUMBER 1, 1996 THE LINCOLN LABORATORY JOURNAL 35

Furthermore, the HTS wideband compressive re-ceiver has size, weight, and power advantages overconventional technology. By delivering improved per-formance over a wider bandwidth and in a more com-

pact design, the HTS wideband compressive receivercan meet the demands of existing applications.

Superconductive Chirp Filters

Chirp filters, which have been used extensively inpulse-compression radars, are the backbone of a com-pressive receiver. These filters are also known as dis-persive delay lines and linearly frequency-modulateddelay lines. Early chirp filters were made with folded-tape meanderlines or crimped coaxial cable. Folded-tape meanderlines produced a phase shift in each turn

of the meander, with the phase shift dependent on theturn-to-turn coupling. A meanderline was meticu-lously synthesized by manipulating the number ofsections, and for each section, manipulating the cen-ter frequency, the number of turns, and the turn-to-turn coupling. The crimped coaxial cable used back-

ward reflections created by impedance discontinuitiesat each crimp to provide the chirp filtering [8].

The acceptance of chirp filters as a system compo-nent did not occur until accurate and large time-bandwidth-product SAW chirp filters were devel-

oped. A wide variety of effects, however, limit thebandwidth of SAW chirp filters, including propaga-tion loss, transducer inefficiency, and difficulty infabricating the submicron dimensions required by

FIGURE 1. Generalized transversal-filter structure with time delays i and tap weights ai. Time-delayed samples of the

input signal are amplitude weighted by the appropriate aiand coherently summed to produce the filter output.

high-frequency transducers. Attempts have beenmade to build chirp devices with magnetostatic wave(MSW) media, but the tremendous dispersion inMSW materials makes this strategy impractical [7].

For other chirp devices, such as folded-tape meander-lines and crimped coaxial cable, insertion loss and thedevice accuracy limit system applications.

The concept of a superconductive chirp filter wasinitially proposed by J.T. Lynch, and subsequently re-duced to practice and patented by a Lincoln Labora-tory research team [11, 12]. This work grew out of aneffort by S.A. Reible to build superconductive analogconvolver structures in the gigahertz range, whichparalleled research at the time into SAW-based de-vices [13]. The two advantages of superconductors in

transmission-line structures are their extremely lowloss at microwave frequencies and their nondispersive(i.e., frequency independent) penetration depth.These advantages lead to long and compact electro-magnetic delay lines. Introducing filter taps into thesedelay lines can produce a superconductive chirp filterthat extends the bandwidth capability of chirp filtersbeyond the 1-GHz limitation of SAW devices.

A chirp filter is a form of a fixed tap-weight trans-versal filter, which is a general class of filters used toimplement matched filtering, correlation, convolu-

tion, and Fourier transformation. Figure 1 illustratesthe generalized transversal-filter structure. Filter tapsprovide samples of the input signal differentially de-layed in time by an amount i. These time-delayed

Delay

1

Delay

2Input

Output

a1

Delay

3

a2 a3

Delay

n

an

n


4/35




samples are amplitude weighted by a factor ai and co-herently summed to produce the filter output. Thenumber of information cycles of the waveform gath-ered coherently in the filter determines the signal pro-cessing gain, measured conveniently as the time-bandwidth product.

Figure 2 illustrates the operation of a proximity-

tapped superconductive chirp filter [14, 15]. Thetransmission-line structure is typically stripline, andupper and lower ground planes sandwich the signallines. This structure usually involves two substrates

with signal lines and lower ground plane on oppositesides of the bottom substrate, and the upper groundplane on the top side of the top substrate. For clarity,Figure 2(a) shows only the signal lines. A series ofbackward-wave couplers achieves the downchirp filterresponse (group delay increases as frequency de-creases) in direct analogy to a SAW chirp grating or

transducer array. Each coupler has a peak response atthe frequency for which the coupler is a quarter-wave-length long. Making the reciprocal of the length ofeach coupler a linear function of the length along theline causes the peak frequency response of the back-

ward-wave couplers to vary as a linear function of de-lay. Weighting of the taps is achieved by varying thecoupling strength between the two striplines formingeach backward-wave coupler. Line-to-line isolation

greater than about 55 dB must be maintained in theuncoupled sections of the filter.

Figure 2(b) illustrates the downchirp operation ofa superconductive chirp filter over a typical 3-GHzbandwidth with a dispersive delay of 40 nsec. The de-vice is symmetric and is operated by using either itsdownchirp or upchirp ports. An impulse function ap-

plied to the downchirp ports produces a downchirpsignal over the bandwidth of the chirp filter, asshown. The 6.0-GHz component of the impulsecouples to the output immediately while the 3.0-GHz component experiences the full filter delay.

Conversely, an upchirp signal with the proper fre-quency-delay characteristic applied to the downchirpports of the filter is compressed into a pulse of widthk/Bc and amplitude (TBc)

1/2 above the input ampli-tude, where Bc is the chirp-filter bandwidth, Tis thedispersive delay, and kis a constant near unity deter-

mined by the filter weighting function. This pulse isreferred to as a compressed pulse, and the action ofthe downchirp filter on the upchirp signal is calledmatched filtering. As an example, k= 1.33 for Ham-ming weighting, giving a 0.44-nsec mainlobe pulse-

width for Bc= 3.0 GHz. The compressed pulse hassidelobes whose ideal amplitude depends on the

weighting function, in addition to having a k/Bcmainlobe pulsewidth.

FIGURE 2. (a) Structure and operation of a proximity-tapped superconductive chirp filter. The upchirp ports have beenterminated into 50 . The electromagnetic delay lines are implemented in stripline and the taps are implemented by a

cascade of backward-wave couplers. (b) The downchirp impulse response is shown for a typical 3-GHz-bandwidth,

40-nsec-long chirp filter.

/4

Backward-wave

coupler

t

t

Input

Impulse

Downchirp

signal

Output

Downchirp

response6.0

3.0

0 40

Frequency(GHz

)

Amplitude

A

mplitude

Time (nsec)

Downchirp ports

(a) (b)

50

50


5/35




In direct analogy to a SAW grating or transducerarray, the effective number of quarter-wavelengthcouplers Neff active at a particular frequency fin asuperconductive chirp filter with dispersive delay T

and bandwidth Bc is

N fT

Beff

c

= .

Thus, unlike the typical transversal filter, such as acharge-coupled device (CCD), in which the time de-lay between taps is constant and energy is tappedfrom a signal (for all relevant frequencies) across thefull length of the device, the chirp filter effectivelytaps energy over only an Neff grouping of couplers at

a given frequency. This grating arrangement enablesthe chirp filter to produce a continuous downchirp orupchirp response, as shown in Figure 2(b).

Several techniques exist for designing supercon-ductive chirp filters. Initial work used the coupling-of-modes theory [16]. More recent designs based onS- and T-matrix circuit analysis have been made pos-sible by advances in computing power [17].

The material system of choice in early work on su-perconductive chirp filters was niobium on high-re-sistivity silicon. After initial demonstrations of the

chirp-filter concept and a demonstration of a chirp-transform algorithm [18], further research yielded de-vices with error-sidelobe levels of 32 dB, and band-

widths as large as 6 GHz [19]. An error-sidelobe levelis determined by the magnitude of nonidealities in achirp-filter response, which can be calculated by us-ing an analysis based on paired-echo theory [20]. In acompressed-pulse output, the error sidelobes riseabove the designed sidelobe levels obtained for a par-ticular filter weighting function. For example, withrespect to the mainlobe, the ideal peak sidelobe level

for Hamming weighting is 42.8 dB. An error-side-lobe level of 32 dB corresponds to a filter perfor-mance of 0.75-dB peak-to-peak amplitude accuracyand 5 peak-to-peak phase accuracy [21].

Reflectively tapped superconductive chirp filterswere also designed and fabricated on the basis of a well-defined impedance discontinuity at the tappoints [16]. However, this structure is susceptible tospurious reflections from defects and imperfections

and has no input-to-output isolation, requiring a cir-culator for operation.

HTS Chirp Filters

The advent of high-Tc superconductors has providedan opportunity to move the concept of superconduc-tive chirp filters into actual system applications. (Formore information on the impact of HTS materials onmicrowave applications, see the sidebar entitled Pas-sive Microwave Applications for High-TemperatureSuperconductors, on the next page.) Initial work fo-cused on materials and processing issues, with somedesign consideration peculiar to the high-dielectricconstant substrates. Historically, one of the first HTSdevices demonstrated was an 8-nsec, 3-GHz-band-

width YBa2Cu3O7 (YBCO) chirp filter [22]. Thisdevice was followed soon after by the demonstrationof a 12-nsec, 3-GHz-bandwidth YBCO chirp filter[23, 24]. Finally, a matched pair of 12-nsec, 3-GHz-bandwidth YBCO chirp filters, one flat weighted andthe other Hamming weighted, were used to generatea compressed pulse [25]. The matched filters exhib-ited 25-dB error sidelobe performance, consistent

with 2.2-dB peak-to-peak amplitude accuracy and14 peak-to-peak phase accuracy [21]. This perfor-mance is of comparable accuracy but at three times

the bandwidth of the widest bandwidth SAW devices.The successful matched-filter demonstration pavedthe way for the High-Temperature SuperconductivitySpace Experiment (HTSSE) compressive cueing re-ceiver described later.

Figure 3 shows the measured electrical characteris-tics of one of these 12-nsec Hamming-weighted fil-ters, with a comparison between the designed andmeasured frequency-domain response. Chirp filters

with the characteristics shown in Figure 3 were alsoused later in the HTSSE prototype receiver. As shown

in Figure 3(a), 5 dB of insertion loss is designed intothe filter, which limits the strength of the backward-

wave couplers enough to avoid distorting the inputsignal as it propagates through the tapped-delay-linechirp filter. Dissipation loss in the filter is too small tomeasure.

These first chirp filters were fabricated in a typicalstripline configuration with YBCO signal lines andtwo silver ground planes on LaAlO3 substrates. The


6/35




class of superconductors with sub-stantially higher transition tem-peratures (Tc) offers the opportu-nity for greatly simplifying thecryocooling apparatus in most su-perconductor applications, mak-ing many more applications botheconomically and technically

practical. The new supercond-uctors, known as high-tempera-ture, or high Tc, superconductors(HTS), are complex layered cop-per-oxide compounds that chal-lenge material scientists to masterdifficult crystal growth methods.

The workhorse material hasbeen YBa2Cu3O7 (YBCO). Ithas a respectably high transitiontemperature of 93 K, and can

more readily be grown in singlephase than other HTS materials,resulting in high-quality thinfilms. A significant milestone inthe development of thin films formicrowave applications was thereliable and reproducible achieve-ment of low surface resistance in

YBCO at 77 K by many laborato-ries around the world. Figure Aillustrates this achievement, circa

1992, with data compiled by H.Piel et al. [1]. Passive microwavedevices benefit tremendouslyfrom the orders of magnitudelower surface resistance affordedby HTS thin films compared tonormal metals.

Researchers have begun to

implement a wide variety of pla-nar thin-film HTS passive micro-

wave applications. In contrast,active HTS microwave devices,such as mixers, that make use ofthe Josephson effect are still onlyin the early stages of development.In the main text, we introduce theadvantages of planar HTS passivedevices and describe a chirp filterbased on long, tapped supercon-ducting delay lines.

Another good example of pla-

nar HTS structures is a narrow-band (high-Q) resonator-basedfilter that could previously beimplemented only as a cavity fil-ter. Figure B shows the size com-parison between a compact planarHTS filter and an electricallysimilar bulky cavity filter. This size

and weight difference becomesdramatic when numerous filters

are used in a microwave system forchannelizing a band of frequen-cies. Small size and weight are es-pecially important for systems onmobile platforms such as aircraftand satellites. Figure C illustratesthe effect of the low surface resis-tance of a superconductor on anarrowband microstrip filter [2].The insertion loss of a filter can beestimated as

LB

g

Qi

uii

n

0

1

434

=

(%), (A)

where L0 is the center-frequencyincrease in attenuation (in dB) be-cause of dissipation losses,B(%) isthe fractional bandwidth in per-

FIGURE A. Collection of measured surface resistance data at 77 K for thin

films of YBCO from nine laboratories. The data are plotted as a function

of frequency. The surface resistance of copper at 77 K and superconduct-

ing niobium at 7.7 K are shown for comparison.

P A S S I V E M I C R O W A V E A P P L I C A T I O N S F O R

H I G H T E M P E R A T U R E S U P E R C O N D U C T O R S

YBa2Cu3O7 (77 K)

0.1 1

105

106

104

103

102

101

10

5

55

51

5

Nb (7.7 K)

1.2.3.4.5.6.7.

8.9.

Conductus and HPU. HoustonKFA JulichNTTLincoln LaboratorySiemensUCLA

U. WuppertalRSRE (MgO)

Cu (77 K) 4

2

89

6

Frequency (GHz)

Su

rfaceresistance()

100

f2

3

7


7/35




cent, gi is the normalized seriesinductance and shunt capacitanceof the low-pass filter prototype,and Qui is the unloaded Qof theith resonator [3].

Equation A and the data ofFigure C indicate that the trans-mission-line resonators compris-ing the filter in Figure B have un-loaded Q values on the order of150 for Au at room temperature,250 for Ag at 77K, and 2000 for

YBCO at 77 K. Soon we expect todemonstrate planar HTS patch-based resonators and filters withunloaded Q values greater than

150,000. This unloaded Q ishigher than the unloaded Qval-ues of most normal metal cavities.However, this performance inresonator-based devices can comeat a price. The surface resistance ofa superconductor is nonlinear athigh fields (especially at tempera-

FIGURE B. Size comparison between a single (four-

pole) superconducting microstrip filter and a single(six-pole) dual-mode dielectric-loaded cavity filter.

The cavity filter is used in the input frequency multi-

plexer on a communications satellite. (Courtesy of

COMSAT Laboratories.)

FIGURE C.Measured transmission response at 77 K

of a four-pole-Chebyshev 1%-bandwidth YBCO mi-crostrip filter. Measured responses of the same filter

fabricated from silver (at 77 K) and gold (at 300 K) are

shown for comparison. The superconducting filter

exhibits a dramatic improvement in insertion loss

and filter shape factor.

tures approaching Tc), whichgenerates intermodulation distor-tion and associated spurious sig-nals. This intermodulation distor-tion must be characterized to

ensure satisfactory performanceof a device, particularly in trans-mitter applications [4].

Another example of an HTSpassive microwave application is amillimeter-wave phased-array an-tenna feed. The effect of a super-conducting feed network onphased-array antenna gain isshown for a 60-GHz array in Fig-ure D, from Reference 5. The gain

for the phased-array antenna iscalculated as

Gain dBi( ) log( / )

. log( )

=

+

20

8 69 10 4

0D

D

where dBi is gain in dB referred toisotropic radiation, D is the length

of one side of the array, 0 is thewavelength in free space, and isthe attenuation coefficient (innepers/m) of the microstrip feed-line conductor. Figure E, from

Reference 6, shows a prototypeHTS feed structure. A second ar-ticle in this issue describes the de-velopment of low-loss HTS-fer-rite phase shifters that, together

with an HTS feed network, willform a fully functional supercon-ductive phased-array antenna.This antenna may offer a low-costalternative to conventional activearrays that are based on mono-

lithic microwave integrated cir-cuits and place an amplifier ateach element to overcome normalconductor distribution loss.

Other examples of passive mi-crowave applications under devel-opment include probe coils fornuclear-magnetic-resonance

0

10

20

30

Frequency (GHz)

Transmission(d

B)

40

50

Au (300 K)Ag (77 K)

100 MHz

YBa2Cu3O7 (77 K)

f0 0.2

f0~4 .8 GHz

f0 f0+0.2


8/35




(NMR) spectroscopy systems,switched filter banks, electricallysmall antennas, superdirective an-

tenna arrays, and tunable filters.HTS probe coils for NMR are of-fered as a commercial product,built by Conductus and availablethrough Varian. Low-noise re-ceiver front ends have been con-figured as hybrid systems that usecryocooled semiconductor de-vices with passive HTS filters.These low-noise front ends andsharp-skirted HTS filters may

prove feasible even in the com-petitive wireless communicationsarena. Field trials in actual wirelessbase-station networks are under

way with HTS-based receiverfront ends built by Conductus,Superconducting Core Technolo-gies, Superconductor Technolo-gies Incorporated, and IllinoisSuperconductor.

References1. H. Piel, H. Chaloupka, and G.Mueller, in Advances in Superconduc-tivity IV, H. Hayakawa and N.Koshizuka, eds. (Springer-Verlag, To-kyo, 1992), pp. 925930.

2. See Reference 23 in main text.3. G. Matthaei, L. Young, and E.M.T.

Jones, Microwave Filters, Impedance- Matching Networks, and CouplingStructures (Artech House, New York,1980).

4. D.E. Oates, W.G. Lyons, and A.C. Anderson, Superconducting Thin-Film YBa2Cu3O7x Resonators andFilters, Proc. 45th Annual Symp. onFrequency Control, Los Angeles, 2931

May 1991, pp. 460466.5. See Reference 27 in main text.6. J.S. Herd, D. Hayes, J.P. Kenney, L.D.

Poles, K.G. Herd, and W.G. Lyons, Experimental Results on a ScannedBeam Microstrip Antenna Array witha Proximity YBCO Feed Network,IEEE Trans. Appl. Supercond., vol. 3,no. 1, pp. 28402843 (1993).

FIGURE D.Calculated effect of feed-network distribution loss on antennagain for a 60-GHz phased-array antenna with various conductor materials

in the feed structure. The calculation assumed the feed configuration

shown in the inset with a 0/2 spacing between antenna elements (0 = 0.5

cm at 60 GHz), and a 50- microstrip transmission-line structure on a

10-mil-thick substrate with a dielectric constant of 10. The effect of the in-

sertion loss of phase shifter elements was neglected.

FIGURE E. Superconductive stacked-patch microstrip phased-array an-

tenna geometry. The stacked structure produces a bandwidth of about

10% with a center frequency of 12 GHz. The upper quartz wafer was used

as the dewar window of the vacuum chamber, and the lower LaAlO3 wafer

was thermally isolated from the quartz by a vacuum gap. Silver was used

as the upper antenna-radiator element while the feed network and lower

antenna element were patterned in YBCO.

0.01 0.1

90

70

50

30

101

Flat-plate array size, D (m)

Gain(dBi)

10

Cu (77 K)

Lossless, Nb (4.2 K)

YBCO (77 K)

Cu (300 K)

(2 0) (2000 0)

Inset feedline for 50-

match

Array assembly

Reactive-teepower combiners

Radiator

Elementcross section

Feed

Upper 3-in-diameter quartz

Lower 2-in-diameterLaAIO3


9/35




20-mil thickness of the brittle substrates limited thedelay to 12 nsec to avoid excessive line-to-line cou-pling in uncoupled sections. LaAlO3 has since be-come the substrate of choice for microwave applica-tions because of its chemical, structural, and

thermal-expansion match to YBCO, and its low mi-crowave loss tangent. This low loss tangent is unusualfor rare-earth perovskites.

An obvious discrepancy exists between the de-signed frequency response and the measured responseshown in Figure 3. One of the challenges the LaAlO3substrate presents is the variation of the relative di-electric constant r, due to the crystallographic twinsin the rhombohedral material. Measurements of nar-

rowband filters [1] and the variation in the time-do-main reflectometry response of a microstrip spiral lineare consistent with an rvariation in LaAlO3 of 1 to2% [1, 24]. Lower frequencies tend to average out the

variation, while high frequencies and lumped-ele-ment circuits see a larger variation. Additional sourcesof degraded chirp-filter performance are YBCO filmnonuniformity; wafer-thickness nonuniformity; im-pedance mismatch in the microwave transition fromcoaxial cable onto devices with a high-dielectric con-stant material (r is approximately 23.5 in LaAlO3);air gaps in the stripline caused by surface undulations(more severe because of twinning); and packaging ef-fects, such as feedthrough. Figure 3(b) shows feed-through in the time-domain response. Just past the

first tick mark on the time axis, prior to the responseof the first coupler, the signal jumps up slightly as aresult of input-port to output-port feedthrough.

Another large source of error is forward coupling,which is magnified by the length of the delay. Thiseffect is absent in an ideal stripline device, but in anactual stripline device both air gaps and rvariationscause the even- and odd-mode velocities to differslightly. This mode velocity difference results in non-ideal backward-wave couplers with a nonzero cou-pling coefficient in the forward direction, thereby

producing signals propagating in the wrong directionwithin the filter.

Throughout our work on HTS chirp filters, strip-line has been the preferred structure for the transmis-sion line, just as it was for niobium chirp filters.Microstrip has been an unacceptable structure forproximity-tapped chirp filters because of the unequaleven- and odd-mode velocities, which result in tre-mendous forward coupling. Coplanar delay lineshave the isolation and equal mode velocities requiredfor backward-wave couplers, but require smaller di-

mensions than stripline to avoid moding problems,and are therefore more lossy than stripline. Some suc-cess has been achieved with coplanar waveguides foranalog delay lines, but at the expense of insertingmany air bridges to tie the two ground planes to-gether [26]. Apparently the phase response of thatcoplanar structure is easily perturbed because of itssimilarity to a slow-wave filter, making a high-perfor-mance chirp filter difficult to achieve.

FIGURE 3. (a) Frequency-domain response of a Ham-

ming-weighted YBCO chirp filter at 77 K. The chirp filter

has a bandwidth of 3 GHz, a center frequency of 4.2 GHz,

a (designed) insertion loss of 5 dB, and a dispersive de-

lay of 12 nsec. (b) Downchirp time-domain response of a

Hamming-weighted YBCO chirp filter at 77 K with the

same parameters as in part a. The applied signal is a step

function, and the response of each of the Hamming-

weighted couplers can be discerned.

Transmission

(dB)

Design

YBa2Cu3O7 (77 K)

2 3 4 5 6

Frequency (GHz)

(a)

0

10

20

30

5

15

25

35

40

10

Outputvoltage(

mV)

5

0

Time (2 nsec/div)

(b)

T = 77 K


10/35




Despite the numerous burdens of cryogenic cool-ing, HTS chirp filters overcome many of the perfor-mance limitations of conventional filter technology.

A 100-nsec proximity-tapped chirp filter in room-

temperature copper stripline would exhibit 75 to 100dB of dissipation loss in the range of 5 to 10 GHz[22, 27, 28], while an equivalent HTS chirp filter

would produce negligible dissipation loss. In normalmetal, dispersion caused by the frequency-dependentskin depth would also be a tremendous problem. Thethick normal-metal layer required to achieve even the75 to 100 dB of loss would add a further complica-tion by introducing a large air gap into the striplinestructure. Compared to SAW chirp filters, which of-ten require ovens to generate a thermally stable envi-

ronment, HTS filters are already in a temperature-controlled cryogenic environment. When operated attemperatures below approximately 60 K, YBCOHTS filters have little temperature dependence be-cause the superconducting properties (order param-eter and superconductor gap) change little belowtwo-thirds of the transition temperature. SAW de-vices also produce at least 20 dB more insertion lossthan HTS filters. Furthermore, because of the slowSAW propagation velocity, SAW devices are difficultto build accurately at those high frequencies where

the structural dimensions become exceedingly small.Typical SAW wavelengths are on the order of 5 to 10m. Submicron lithography control must be appliedto the transducer structures. The situation is differentfor HTS chirp filters because they are based on elec-tromagnetic delay lines with wavelengths of manymillimeters. This larger wavelength relaxes the di-mensional control requirement somewhat, butlengths in the third dimension such as substratethickness do become an issue.

Concept of Compressive ReceiverThe concept of a compressive receiver based on chirpfilters dates back almost forty years. Early work in-cludes W.D. Whites patent on the compressive re-ceiver [29], as well as theoretical and experimental

work by W.E. Morrow et al. [30]. The chirp-trans-form algorithm [8, 3136], basis of the compressivereceiver, can be understood mathematically if we start

with a standard Fourier transform of a signal h(t),

H h i d ( ) ( ) exp( ) =

, (1)

and perform a linear mapping of frequency into timeby substituting equal to t,with the chirp slope(rate of linear frequency change with time). This lin-ear mapping permits the following substitution in thecomplex exponent of Equation 1:

=

i it

it

i

( )

.2 2 2

2 2 2

The expression for the Fourier transform becomes achirp transform,

H t it

h i it

d

( ) exp( )

( ) exp( ) exp( )

,

=

2

2 2

2

2 2

(2)

by using infinitely long chirp signals in the same waythe Fourier transform uses infinitely long sinusoids.In Equation 2, the expression inside the first set ofsquare brackets represents a multiplication of the sig-

nal h(t) with a chirp signal. A chirp signal varies lin-early in frequency over time and has a quadratic phaseas a function of time. A convolution with a chirp ofopposite slope is performed by the integration. Fi-nally, another multiplication is done with a chirp sig-nal of the same chirp slope as the first multiplication.Equation 2 is called a multiplication-convolution-multiplication (MCM) chirp transform and producesthe complete Fourier transform of the input signal,

where frequency, amplitude, and absolute phase areall mapped into time.

By taking the Fourier transform of Equation 2,and recalling that convolution in one domain be-comes multiplication in the other and that chirpstransform to chirps, we can demonstrate that a con-volution-multiplication-convolution (CMC) con-figuration implements the same chirp transform.These continuous chirp transforms are not to be con-fused with a chirp-Ztransform, which is a sampledversion of this analog chirp transform and is imple-


11/35




mented digitally or with CCDs [32]. In actual micro-wave implementations, multiplication with a chirp isperformed with a chirped local oscillator and a mixer.Convolution with a chirp is achieved by passing thesignal through a chirp filter. Actual implementationsof the chirp-transform algorithm, however, cannot

use the infinitely long chirp signals of the ideal math-ematical expression indicated by Equation 2.

The finite length and finite bandwidth of actualchirp signals and real filters lead to two possibleimplementable full chirp transforms, the M(s)-C(l)-M(s) and the C(s)-M(l)-C(s), where l stands for longand s stands for short. Typically the long chirp istwice the length and bandwidth of the short chirp.

Absolute phase information requires a full MCM or

CMC chirp transform. If only the frequency, ampli-tude, and relative phase between two channels are re-quired, the last M of the MCM or the first C of theCMC can be dropped. This requirement leaves twopossible algorithms for a compressive receiverM(l)-C(s) and M(s)-C(l). There are advantages to each

[37]. The M(l)-C(s) system is used most often, butrequires alternation between a pair of channels toachieve 100% time coverage. The M(s)-C(l) systemrequires only a single channel, but effectively halvesthe filter length for the convolutionC(l) must betwice the size of C(s) for the same frequency cover-ageand a single-channel implementation has diffi-culty with out-of-band signals and filter triple-transiteffects. An additional feature of the M(l)-C(s) system

FIGURE 4. System diagram of a compressive receiver M(l)-C(s) chirp-transform algorithm with receiver bandwidth BR,

chirp-filter bandwidth Bc, and chirp-filter dispersive delay T. This architecture is well suited to extract the frequency and

amplitude of input signals. The measured 77-K compressed-pulse response of a matched pair of YBCO chirp filters and

a photograph of a 12-nsec YBCO chirp filter are shown as insets.

Chirp

filterInput

Pulse-detection

and

signal-report

electronics

Filterresponse

Delay

Chirp

generator

2Bc

Chirp

signal

2T

Superconductive stripline chirp filter

(YBa2Cu3O7 on LaAIO3)

(Frequency

mapped into time)

Compressed-pulse

amplitude

Time (1 nsec/div)

Compressed-pulse

response

BR BR

2BcBc Bc

T Tt

t

f

f

t

f

t

f

2TT

tTT

Amplitude

(10mV/div)

T = 77 K

Stripline

package

Backward-wave

couplers

Tapered-line impedance transformer


12/35




is the ability to readily increase frequency coverage bylengthening the multiplying chirp signal to overscanthe bandwidth of the convolving chirp filter whilemaintaining the same chirp slope. This lengtheningof the multiplying chirp signal extends the bandwidthcoverage of the receiver at the expense of reducedtime coverage. The M(l)-C(s) is often referred to as amicroscan, or sliding-transform, receiver.

Figure 4 illustrates the operation of an M(l)-C(s)receiver. One of the four purple inputs (shown as fre-

quency-versus-time curves) is dashed so that an inputsignal can be followed through the entire chirp-trans-form process. Input signals over the receiver band-

width BRare multiplied with a chirp signal of length2Tand bandwidth 2Bc, with T the dispersive delayand Bcthe bandwidth of the chirp filter. The multipli-cation (mixing) process produces a set of frequency-offset chirp signals (the beige balloon in Figure 4), in

which each offset is determined by the input fre-quency. This set of chirp signals is convolved by thechirp filter, producing a compressed pulse at the out-

put of the filter for each input signal. The exit timeand amplitude of these compressed pulses are directlyrelated to the frequency and amplitude of the inputsignal. The chirp filter must have the same chirp-slope magnitude but opposite sign relative to thechirp-signal generator. As indicated in Figure 4, mix-ing input signals over a bandwidth BR with a sweptlocal oscillator (SLO) of bandwidth 2Bc generateschirp signals that cover a frequency range BR + 2Bc.

OnlyBcof that range lies within the bandwidth of thechirp filter, producing a compressed pulse. This rela-tionship causes the input analysis window to be fre-quency dependent or slide in time, as shown in Figure4 by the diagonal red-dotted lines superimposed overthe input frequency-versus-time curves. Overscan-ning simply extends the SLO sweep to cover morebandwidth, and percent time coverage is inverselyproportional to the overscan ratio. Referring to Figure4, and assuming two alternating M(l)-C(s) channels

for 100% time coverage of receiver bandwidth BRwith an SLO scanning over the scan bandwidth BS,we define the following:

overscan ratio

time coverageoverscan ratio

= +

=

=

( ),

.

B B B

B

B B

BS c c

c

S c

c

2

1

For example, to cover 10 GHz with a 3-GHz chirp

filter requires an SLO scan of 13 GHz for a 3.3 over-scan ratio and 30% time coverage.

Figure 4 highlights the HTS chirp filter as the en-abling technology for a 3-GHz-bandwidth chirptransform. Conventional technology can be used tobuild the other chirp-transform components quiteadequately, as is seen in later sections of this article.

Table 1 lists the frequency resolution f= k/Tofan M(l)-C(s) compressive receiver [6] that uses Ham-

Table 1. Frequency Resolution as a Function of Hamming-Weighted

Chirp-Filter Dispersive Delay for an M(l)-C(s) Compressive Receiver

Dispersive Delay Frequency Resolution Bins per GHz(nsec) (MHz)

8 166 6.0

12 111 9.0

24 55.4 18.0

40 33.3 30.1

100 13.3 75.2

200 6.7 150.4


13/35




ming-weighted chirp filters, for which k= 1.33 [38]and where Tis the dispersive delay of the filter. Therange of delays shown is consistent with present HTSchirp-filter capabilities described later in this article.The compressed-pulse mainlobe width (3-dB pulse-

width) is still k/Bc, as for the matched-filter examplein the previous section on superconductive chirp fil-ters. The frequency resolution fis determined bydividing the bandwidth Bc by the number of k/Bcpulsewidths (frequency bins) that fit into an analysis

window of length Tso that f= Bc/[T/(k/Bc)] = k/T, which is independent ofBc. Table 2 translates the3-dB pulsewidth of the compressed-pulse envelopeinto a logic speed required to capture samples sepa-rated in time by this pulsewidth. This logic speed pro-duces a 3.0-dB accuracy in determining compressed-pulse amplitudes.

Comparison of Receivers

A wide variety of receivers have been used in elec-

tronic warfare. The most common can be classified assuperheterodyne, compressive, channelized filter,acousto-optic channelized, instantaneous frequencymeasurement (IFM), and crystal video receivers [8,31]. Future receivers need to perform well in densesignal environments over many tens of GHz. The keyrequirements for future receivers are therefore excel-lent wideband simultaneous-signal performance and100% time coverage of the bands of interest.

These considerations quickly eliminate crystalvideo, IFM, and superheterodyne receivers, and limitfuture advanced electronic-warfare receiver choices tocompressive, channelized-filter, and acousto-opticchannelized receivers. Crystal video and IFM receiv-ers simply do not function well in the presence ofmore than a single emitter. Superheterodyne receivershave a poor probability of intercept (time coverage)because of their narrowband nature, despite their ex-

cellent dynamic range, sensitivity, and resolution [8].A superheterodyne intermediate-frequency filter witha bandwidth Bhas a response time of 1/B, and thefastest superheterodyne scan rate without degradingsensitivity is approximatelyB/(1/B) = B2. As an ex-ample to point out the time-coverage limitations ofsuperheterodyne receivers, if the intermediate-fre-quency filter bandwidth is 10 MHz, then the fastestscan rate is 100 MHz/sec. If the input bandwidth is10 GHz, then the superheterodyne receiver will takeat least 100 sec to scan across the entire band. There

is a finite probability that the receiver will miss anypulses that are shorter than 100 sec. In this example,at any one time the superheterodyne receiver is look-ing at only 0.1% (10 MHz out of 10 GHz) of the in-put bandwidth.

Among the remaining receiver candidates, chan-nelized-filter receivers are considered in more detail ina later section as part of a direct comparison to anHTS compressive receiver. A major issue for channel-

Table 2. Pulsewidth (3 dB) of Hamming-Weighted Compressed-Pulse Envelope

and Pulse-Detection Logic Speed for 3.0-dB Amplitude Accuracy

Chirp-Filter Bandwidth Compressed-Pulse Pulse-DetectionWidth (nsec) Logic Speed (gigasamples/sec)

2.0 0.67 1.5

2.5 0.53 1.9

3.0 0.44 2.3

4.0 0.33 3.0

5.0 0.27 3.8

10.0 0.13 7.5

20.0 0.07 15.0


14/35




ized-filter architectures is the large number of indi-vidual filters required. In contrast, acousto-opticchannelized receivers [7] achieve channelization in acompact Bragg cell [39]. This arrangement is an effi-

cient architecture, particularly for frequency activityindication. However, acousto-optic channelized re-ceivers do have some potential weaknesses. The fullparameterization of emitters is often slow because ofthe parallel nature of the receiver output and the needfor quickly sampling these outputs. This parallel na-ture has forced the use of high-speed analog multi-plexing circuits to serialize the channelizer output to aspeed compatible with processing on a monolithicchip and at a frame rate fast enough to determinetiming details (such as pulsewidth) of the intercepted

emitters. The most mature acousto-optic technology(power-spectrum channelizer) does not allow relativephase information to be extracted, although acousto-optic heterodyne techniques are rapidly improving.Finally, the bulk-acoustic wave technology is limitedto 2-GHz analysis bandwidths.

The significant challenge facing the developmentof HTS compressive receivers is the high-speed pulse-detection circuitry required because of the serial na-ture of the analog chirp-transform (compressed-pulse) output. As noted for acousto-optic receivers,

data in a serial form have significant advantages ifavailable circuits can achieve the required speed.Semiconductor technology is now producing circuits

well matched to the multigigahertz bandwidths ofHTS compressive receivers.

HTSSE II Compressive Cueing Receiver

Shortly after the discovery of superconductivity in theHTS material YBCO at temperatures near 90 K, en-gineers at the Naval Research Laboratory (NRL) be-came interested in the potential applications of using

HTS electronic devices in high-performance remotesensing and communications systems. The lowattenuation, wide bandwidth, low noise, and highspeed associated with high-frequency superconductorapplications are attractive attributes for these systems.In December 1988, NRL initiated the High-Tem-perature Superconductivity Space Experiment(HTSSE) program [40]. One goal of the HTSSE pro-gram was to accelerate the development of HTS into

a viable electronic technology. Another goal was tofocus HTS technology toward potential applicationsin space, which represents possibly the harshest envi-ronment in terms of reliability requirements, tem-

perature extremes, and radiation levels. The HTSSEprogram consisted of two experimental payloads. Thefirst experimental payload, known as HTSSE I, fo-cused on simple HTS electronic devices. HTSSE I

was completed in late 1992 and manifested on asatellite launch scheduled for 1993 that did notachieve orbit. HTSSE II addressed complex HTS de-vices and subsystems, and was shipped to RockwellInternational in 1996 for integration onto the Ad-vanced Research and Global Observation Satellite(ARGOS), scheduled for launch in 1997.

Lincoln Laboratory delivered both qualificationand flight versions of an HTS wideband compressivecueing receiveran example of a promising HTSsubsystemto NRL for HTSSE II [41, 42]. A cueingreceiver is a spectrum activity indicator, producingfrequency information on emitters that can be used tocue additional receiver assets onto active signals of in-terest [8]. This simplest form of a compressive re-ceiver was chosen for the space experiment. Thequalification and flight deliveries followed the pro-duction of a breadboard version of the receiver [43]

and the delivery to NRL of a prototype [44]. All ofthe systems combine an HTS chirp-transform sub-system with high-speed semiconductor compressed-pulse processing circuits.

Figure 5 illustrates the operation of this receiver. An M(l)-C(s) chirp-transform algorithm is utilizedwith a 3.0-GHz-bandwidth YBCO chirp filter and achirp generator consisting of a fast voltage-ramp gen-erator driving a voltage-controlled oscillator (VCO)to produce a flat-weighted chirp signal. The com-pressed-pulse-detection portion of the system latches

the value of a 2-GHz digital counter whenever a com-pressed pulse above a fixed threshold is detected com-ing out of the chirp-transform subsystem. This latch-ing records the time a compressed pulse exits thechirp-transform subsystem and therefore records thefrequency of the detected input signal via a lookuptable. A 2-GHz oscillator serves as the clock generatorthat drives an 8-bit silicon emitter-coupled logic(ECL) ripple counter, which runs continuously. The


15/35




most significant bit (MSB) is used as a reset trigger(TRIG) to the chirp generator, thereby setting the

chirp-transform analysis window equal to 28 (0.5nsec), or 128 nsec. Valid data are accepted only whenthe MSB is high. The 2-GHz counter rate is requiredbecause a frequency bin corresponds to the 3-dBpulsewidth of a 3-GHz-bandwidth Hamming-

weighted compressed pulse, approximately 0.5 nsec,as seen in Table 2.

A compressed pulse generated by a signal at the in-put to the receiver is passed through an envelope de-tector to remove the carrier frequency. This com-pressed-pulse envelope (negative portion of envelope)

is then passed through a threshold detector (acting asan inverter) that strobes a silicon ECL logic gate toproduce an appropriate logic level to latch the 8-bitcounter value into an 8-bit ECL latch. The output ofthe counter is passed on to a first-in first-out (FIFO)buffer register following a voltage level conversionfrom ECL to transistor-transistor logic (TTL). TheFIFO contents are then available to the satellite databus and memory.

A 10-GHz oscillator was included on the qualifica-tion and flight versions of the receiver to produce an

end-of-band marker for on-orbit receiver calibration.Figure 5 also indicates the power consumption of thevarious room-temperature components. The semi-conductor ECL components are clearly costly to thepower budget. The amplifier following the chirp filteris required to overcome the insertion loss of themixer, cryogenic cables, and chirp filter, and thendrive the envelope detector at a sufficient signal levelto ensure linear performance from the detector. Thecompressed pulse, envelope-detected compressedpulse, and logic-compatible pulse waveforms are all

shown as insets in Figure 5. Figure 6 shows a com-pressed pulse and compressed-pulse envelope typicalof those produced by all versions (breadboard, proto-type, qualification, and flight) of the compressivecueing receiver.

Projections of limited power available on board thesatellite forced the Navy to restrict the cueing-receiverpower budget to 20 W. Therefore, only the singleECL latch shown in Figure 5 could be included, lim-

FIGURE 5. System block diagram of the High-Temperature Superconductivity Space Experiment (HTSSE) II compres-

sive cueing receiver. Signal waveforms are shown as colored insets. The space-qualified version of the receiver covered

a frequency range of 7.0 to 10.0 GHz, while the prototype covered a frequency range of 9.4 to 12.4 GHz. A 10-GHz oscilla-

tor was added to the input of the space-qualified receiver as an end-of-band marker. The power consumption indicated

in the figure was measured on the space-qualified receiver.

2-GHzclock

generator

1 W

1 W

10-GHzmarker

oscillator3 W

8-bitcounter

Chirpgenerator

YBCOchirp filter

(77 K)

Envelope

detector

Cryocooler

4.5 W

7.010.0 GHz

1.5 W

ECLlatch

1.5 W

3.5 W

0.5 W

LatchMSB

TRIG

ECL TTLconverter

2 W

FIFObuffer

register

Thresholddetector

1.5 W

Data ready

TTL digitaloutput toonboardmemory

Output strobe

FIFOinputstrobe


16/35


17/35




FIGURE 7. Prototype HTSSE compressive cueing receiver. The HTS chirp filter, room-temperature elec-

tronics box, and power supply box are shown.

FIGURE 8. (a) Space-qualified cryogenic package for the final HTSSE compressive cueing receiver. This hermetically

sealed package contains the 12-nsec YBCO chirp filter in a stripline configuration. (b) Space-qualified package contain-

ing the ambient-temperature pulse-detection and frequency-report electronics, mixer, and chirp-generator portions of

the final HTSSE compressive cueing receiver.

(a) (b)


18/35




High Frequency (EHF) Package, otherwise known asFEP [45], and the joint Lincoln/COMSAT/AT&Tdelivery of a narrowband YBCO filter for HTSSE I[46]. The cryogenic package was leak checked with aresidual gas analyzer to establish a leak rate below4 109 Torr-liter/sec. The package has a base foot-print of seven square inches. Total package height isapproximately 0.5 in, with an aluminum package

base that is 0.13 in thick.The fabrication of the YBCO chirp filter followed

most of the standard procedures initiated prior toHTSSE I [1, 46]. A 4-m layer of silver preceded by a200- layer of titanium was used for both upper andlower ground planes. Patterning of the YBCO signallines was accomplished with standard photoresist anda spray etch of 0.25% H2PO4, which successfully pre-vents the residual film formation typically seen with

other wet-etching methods. Undercutting on the or-der of 1 m is observed with this etch. Several tech-niques have been used for ohmic contact formation.The most successful technique has been a standardphotoresist procedure with an in situ ion-beam etchfollowed by electron-beam evaporation of 1.5 m of

Ag. Following photoresist lift-off, the contacts are an-nealed in flowing O2 for one hour followed by a slow

ramp to room temperature. Final packaging for thespace-qualified HTSSE II devices was performed byusing ultrasonic wedge bonding of 0.5 3-mil Auribbon directly on the annealed Ag contacts. Theseprocedures yielded low contact resistances and goodbond-pull strengths. The electrical responses of thespace-qualified versions of the chirp filters were simi-lar to those shown in Figure 3, with a shift in centerfrequency to 6.7 GHz. The HTSSE devices utilized

FIGURE 9. Frequency-report bin number versus input-signal frequency for the final flight version of the HTSSE

compressive cueing receiver. The frequency midpoint of each frequency-report bin is indicated. The straight line

shows the ideal location of each midpoint, which corresponds to uniform-width frequency bins. The average bin

width (frequency resolution) is 110.0 MHz, equal to the ideal bin width for 12-nsec chirp filters. The maximum

deviation from this ideal width is 30 MHz, and 40% of the bin widths are equal to 110 MHz.

Frequency-reportbinnumber

Input-signal frequency (GHz)

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

7.0 8.0 9.0 10.0


19/35




off-axis sputtered YBCO thin films [47] with typical

best parameters of transition temperature Tc= 88 K,critical current densityJc (77 K) > 2 MA/cm

2, andsurface resistance RS (77 K, 10 GHz) = 500 /sq.The 12-nsec length did not require values quite thislow to achieve negligible dissipation loss. More recent

work with longer delays has used films grown by a cy-lindrical magnetron, achieving these excellent param-eters as standard performance [48].

Extensive work was performed to ensure that theambient-temperature electronics box and the cryo-genic YBCO chirp filter would survive an orbital

rocket launch and the subsequent space environment.After a qualification version of both the ambient boxand chirp filter were fabricated and tested, final flightversions were fabricated with any necessary modifica-tions. Details on the space-qualification procedureare found in Reference 42.

Performance of the Final Space-QualifiedHTSSE Cueing Receiver

Figure 9 shows a plot of frequency-report bin numberversus input-signal frequency for the space-qualified

HTSSE flight receiver. The number of frequency binsis determined by the width of the compressed pulsesand the length of the chirp filter. The 3-GHz band-

width and Hamming weighting of the chirp filterproduce compressed pulses that are 0.44 nsec wide.The dispersive length of the chirp filter is 12 nsec.Therefore, the analysis window of the compressive re-ceiver supports 28 frequency bins, providing the 110-MHz frequency resolution. Timing jitter on the order

of 30 psec limited the definition of a bin width to ap-

proximately 10-MHz increments.The chirp generator deviates from a linear fre-

quency-versus-time slope significantly more than thechirp filter [43], and thereby sets an error-sidelobelevel of 19 dB. These error sidelobes act just as spuri-ous signals would in a compressive receiver, limitingthe dynamic range of the system to 19 dB because ofthe single fixed threshold crossing used for com-pressed-pulse detection. A multiple-threshold re-ceiver using the same technology could support asingle-signal dynamic range of 60 dB and a two-signal

dynamic range of at least 19 dB. The amplitude of theenvelope-detected compressed pulse deviates by lessthan 3 dB across the 3-GHz analysis bandwidth.

However, this 3-dB pulse amplitude variation,which can be traced directly to nonlinearities in theSLO, has a significant effect on the width of each fre-quency-report bin. An increase in pulse amplitudecauses the pulse to be detected sooner than the ideal,and a decrease delays the detection. Figure 9 indicatesthe frequency midpoint of each bin and the straightline illustrates the ideal location of each midpoint.

While increases in pulse amplitude push the mid-point above the line and shorten the bin widths, de-creases in pulse amplitude have the opposite effect.The movement of the midpoint with respect to theline (and the bin widths) closely tracks the measuredcompressed-pulse amplitude variation across theband, and can account for the maximum deviation of30 MHz from the ideal bin width of 110 MHz.

Table 3 summarizes the operating characteristics

Table 3. Operating Characteristics of Final Space-Qualified

HTSSE Compressive Cueing Receiver

Analysis bandwidth 3.0 GHz (7.010.0 GHz)Frequency resolution 110 MHz

Frequency bins 28

Analysis time 128 nsec

Maximum cryogenic temperature 83 K

Cryogenic power consumption 5 mW (cables), plus radiative heat load

Total ambient power consumption 20 W (not including cryocooler power)


20/35




for the space-qualified compressive cueing receiver.The analysis bandwidth, frequency resolution, andnumber of frequency bins are readily evident fromFigure 9. The 128-nsec analysis time is limited by the

speed with which the chirp generator can reset itselfand begin a new frequency sweep. Above 83 K the

YBCO chirp filter is too close to the superconductingtransition temperature to function properly. The am-bient power consumption of 20 W is clearly domi-nated by the discrete high-speed semiconductor ECLlogic operating at 2 GHz. A future version of this re-ceiver would make use of rapidly emerging, commer-cially available, monolithic high-speed componentsthat provide far greater digital processing capability

with far less power consumption per gate.

Bonded/Thinned-Wafer HTS Chirp Filters

As indicated in Table 1, the frequency resolution of acompressive receiver is tied directly to the dispersivedelay of the HTS chirp filters. The chirp filters arebased on a stripline configuration that uses two sym-metrically placed ground planes on opposite sides of apair of wafers. If the line-to-line electromagnetic cou-pling is kept constant in a stripline configuration,then the packing density of the delay lines, and there-fore the total chirp-filter length for a given substrate

area, is inversely proportional to the thickness of thetwo wafers. Standard 20-mil-thick, 2-in-diameterLaAlO3 wafers limit the delay, with appropriate line-to-line isolation, to approximately 12 nsec, as used inthe HTSSE compressive cueing receiver. A bonded/thinned-wafer technique has been developed to in-crease the delay achieved on a 2-in-diameter LaAlO3

wafer first to 24 nsec [49] and then to 40 nsec [50],a refinement of a technique used to demonstrate44-nsec YBCO analog delay lines [28]. As the waferthickness is reduced to 10 mil and less to allow more

delay, a support wafer is required to prevent the thinwafer from breaking. Figure 10 illustrates the tech-nique used to bond and thin a 2-in-diameter LaAlO3

wafer, and shows a photograph of a 40-nsec YBCOchirp filter fabricated by using the technique.

The wafer-bonding process begins with a 20-mil-thick LaAlO3 upper wafer with a sputtered layer ofTi/Au (300 of Ti followed by 2 m of Au) on thebottom surface, a 20-mil-thick LaAlO3 base wafer

with a sputtered layer of Ti/Au on the top surface,and a 10-m-thick gold foil. The two wafers and thegold foil must be kept very clean throughout the en-tire process. The wafers are forced together against the

gold foil in a hot press inside an oxygen atmosphere.The top wafer is lapped to a thickness of 190m,

and then polished with chemical-mechanical polish-ing compound to a final thickness of 125 m. Thepolished surface must allow for epitaxial growth of

YBCO. After polishing, the bonded-wafer pair isplaced in a standard gas-pocket heater developed byLincoln Laboratory, and growth of YBCO is per-formed with our cylindrical magnetron on the topsurface of the thin wafer [48]. Standard YBCO pat-terning techniques can be used following the YBCO

growth. A layer of gold, electroplated onto the sidesof the wafer, contacts the edges of the gold foil tocomplete the contact to the ground plane on the bot-tom surface of the thin wafer. The upper groundplane of the stripline configuration requires a secondbonded-wafer pair.

We initially demonstrated 24-nsec YBCO chirpfilters by bonding existing 10-mil-thick LaAlO3 wa-fers to a 20-mil-thick LaAlO3 carrier wafer and omit-ting the wafer-thinning step shown in Figure 10. Forthe initial demonstration of the 40-nsec YBCO chirp

filters on 5-mil-thick LaAlO3, we used the entire pro-cedure indicated in Figure 10. The 24-nsec YBCOchirp filters with a modified HTSSE VCO-basedSLO produced error sidelobes of 18 dB, limited bythe frequency-slope linearity of the SLO [43]. TheSLO generated an upchirp waveform, which was thencompressed into a pulse by using the downchirp portsof the YBCO chirp filter. This setup is essentially anM(l)-C(s) receiver front end. However, the initial 40-nsec YBCO chirp filters produced error sidelobes ofonly 13 dB with a similar SLO [50]. The longer dis-

persive delay clearly made the device more susceptibleto device imperfections such as forward coupling andpoor microwave transitions. The 24-nsec filters con-sist of 96 backward-wave couplers, implemented in a100-m-wide 32- stripline. The 40-nsec filters con-sist of 160 backward-wave couplers, implemented ina 100-m-wide 24- stripline.

We made improvements to the 40-nsec chirp filterby saw-cutting notches in the edge of the wafer and


21/35




Hot

press

Thin and polish

top wafer

(125- m thick)

Grow YBCO

film

Establish

YBCO pattern

(2.5-m-long line

on 5-cm diameter)

LaAlO3 base wafer(gold upper side)

LaAlO3 top wafer

(gold bottom side)

10- m

gold foil

gold-plating the inside of the notches. These gold-plated notches reduce reflections at the microwavetransitions in and out of the filter. The YBCO film

stops short of the edge of the LaAlO3 wafer, requiringlong bond wires and a nonstandard launcher configu-ration for the initial 24- and 40-nsec chirp filters. Asshown in an inset to Figure 10, these saw-cut notchesgreatly reduce bond-wire length, allowing a more rea-sonable microwave transition to be made. The goldplating shortens the ground-plane contact path andtherefore reduces inductance at the transition. We ex-pect to improve the microwave transitions further.

The improved 40-nsec YBCO chirp filter pro-duced 18-dB error sidelobes, once again the limit ofthe modified HTSSE SLO. Figure 11 shows this

compressed-pulse performance for the combinationof the SLO and the improved 40-nsec chirp filter.The measurement is made by the repetitive samplingof a digital oscilloscope to capture the compressed-pulse envelope.

The bonded/thinned-wafer technique used to pro-duce 5-mil-thick substrates on 2-in-diameter LaAlO3

wafers will scale directly to 3-in-diameter LaAlO3 wa-fers, enabling dispersive delays of 90 nsec, or to 4-in-

FIGURE 10. Illustration of bonded/thinned-wafer technique used to fabricate 40-nsec YBCO chirp filters on 125-m-

thick, 2-in-diameter LaAlO3 substrates. The chirp filters are constructed in a stripline structure. A photograph of a Ham-

ming-weighted 40-nsec filter is shown as an inset. The impedance transformers are based on a Klopfenstein taper [51].

0.2-nsec

Klopfenstein

tapers

Saw-cut notches

50- inputs

100- m-wide

24- coupledstriplines


22/35




diameter LaAlO3 wafers, enabling delays of 160 nsec.In both cases, thinner bonded substrates will producelonger delays. For these longer chirp filters, YBCOground planes will be required to limit dissipation

loss. This YBCO ground plane replaces the goldground plane shown in Figure 10.

Demonstrations with ExistingCompressive-Receiver Hardware

With Hughes Aircraft Company, we performed ademonstration with existing compressive-receiverhardware to produce the complete signal reports thatare typically generated in a stand-alone electronic-

warfare receiver [49]. In this demonstration, we re-placed the 1-GHz, 200-nsec SAW chirp filters in the

Hughes receiver with a 2-GHz, 24-nsec HTS YBCOchirp filter and VCO-based SLO. Although the ana-log and digital electronics in the Hughes receiver arematched to the narrower bandwidth of the SAW fil-ters, the receiver was demonstrated to have full func-tionality with 2-GHz-bandwidth HTS chirp filters atthe front end. This demonstration doubled the in-stantaneous bandwidth coverage of the receiver. AnHTS chirp filter with at least 24 nsec of dispersive

delay was required to fill a significant portion of the200-nsec receiver analysis window to produce ameaningful demonstration.

The Hughes compressive receiver is a completely

self-contained electronic-warfare receiver, capable ofproducing pulse descriptor words on multiple emit-ters. The descriptor words describe the emitter fre-quency, amplitude, pulsewidth, pulse-repetition in-terval, and time of arrival (TOA). The inputfrequency range for the demonstration was 9.8 to11.8 GHz. A ramp generator and VCO combinationfunctioned as a chirp generator to produce an up-chirp, using the M(l)-C(s) chirp-transform algo-rithm. Table 4 lists results of the Lincoln LaboratoryHughes demonstration. The frequency-versus-time

characteristic of the HTS chirp filter is significantlybetter than the characteristic of the VCO-based chirpgenerator, as described in the last section. The error-sidelobe levels set by the chirp generator act as spuri-ous signals, limiting the single-tone dynamic range to30 dB for a given signal detection threshold. A 50-dBdynamic range was obtained by adjusting the detec-tion threshold. The receiver is limited to 200-nsecTOA resolution and only 50% probability of inter-cept for short pulses (100 to 400 nsec) because the re-ceiver was designed to operate with 200-nsec-long

SAW chirp filters and a 200-nsec analysis window.The frequency resolution of 83 MHz is limited be-cause the receivers 1-GHz log amplifiers elongate the2-GHz-bandwidth compressed pulses generated bythe HTS chirp filters. No more than three simulta-neous signals can be detected because the detectedcompressed pulses must be at least 10 nsec apart, andthe HTS chirp filter is only 24 nsec long.

Some preliminary demonstrations have also beendone with linearized VCO-based SLO technology de-veloped by AIL Systems [52, 53]. These demonstra-

tions have resulted in reasonable performance from acombination of a linearized SLO and an initial 24-nsec YBCO chirp filter [54]. Future efforts shouldeliminate the SLO limitations described here.

Novel Wideband HTS Compressive Cryoreceiver

Figure 12 outlines a novel compressive cryoreceiverarchitecture. There are several key features to this newarchitecture, particularly the use of digital technology

FIGURE 11. Compressed-pulse envelope measured at

77 K for the test setup combination of chirp generator

and improved 40-nsec YBCO chirp filter. This setup pro-

duced SLO-limited error-sidelobe levels of 18 dB.

Amplitude(200mV/div)

Time (1 nsec/div)

T = 77 K


23/35




to perform pulse detection and preprocessing. Thisarchitecture does not rule out the hybrid use of moretraditional analog pulse detection [8], but allows re-ceiver performance to be significantly improved. Thecryoreceiver aspects of the architecture are generic toother types of microwave receivers.

Figure 12(a) illustrates the overall receiver andmakes clear the potential for using additional cryo-electronic components to enhance performance. Theoverall receiver is a three-channel compressive re-ceiver, using a monopulse antenna to determine angleof arrival by measuring monopulse-channel signalamplitudes only. (Relative phase extraction for inter-ferometry could also be performed but would requiredouble the pulse-processing electronics [21].) Theconfiguration is an M(l)-C(s), and the signal reportsare used to cue narrowband receiver assets for signal

demodulation. These signal reports also represent acomplete parameterization of frequency, amplitude,pulsewidth, time of arrival, and angle of arrival. TheHTS delay line provides the local oscillator (LO) con-troller enough time to reconfigure. The expectedlength of HTS analog delay linesup to 200 nsec

would stress the speed of the LO controller. Fast-tuned LO technology has been demonstrated with100-nsec tuning times for smaller bandwidths [55],

but has not yet been demonstrated for multigigahertzbandwidths.

Although the HTS chirp filters make the wide-band compressive receiver possible, additional cryo-electronic components can significantly enhance re-ceiver sensitivity and dynamic range. Initial

demonstrations or investigations of many of thesecomponents have already been made. Cryocooledlow-noise amplifiers [56, 57] and mixers [58, 59] canimprove sensitivity by lowering the noise figure of theamplifier and reducing the conversion loss of themixer. Adaptive notch filters [60] and tunablepreselect filters [61, 62] can improve dynamic rangeby eliminating spurious signals or out-of-band noise.

Any downconversion process is performed at cryo-genic temperature [63] in conjunction with the SLOmixing with the input signals. An HTS delay line

provides low loss and corresponding low noise figureto enhance the sensitivity of the superheterodynereceivers [26, 64].

The advanced semiconductor pulse-processing cir-cuits shown in Figure 12(b) move the analog-to-digi-tal (A/D) interface as close to the analog chirp-trans-form process as possible. An envelope detector stripsthe RF carrier from the compressed pulse, reducingthe effective bandwidth of the pulse from that of the

Table 4. Summary of Lincoln LaboratoryHughes

Joint Compressive Receiver Demonstration 1, 2

Parameter Measured Performance

RF input bandwidth 2.0 GHz

RF frequency resolution 2 GHz/24 cells = 83 MHz

Time of arrival and pulsewidth resolution 200 nsec

Dynamic range, single tone 30 dB 3

Simultaneous signal detection Up to 3

Short-pulse (100400 nsec) probability of intercept 50%

Long-pulse (>400 nsec) probability of intercept 100%

Amplitude resolution 1 dB

1 Taken from Reference 49.2 Performed with a 24-nsec, 2-GHz-bandwidth YBCO chirp filter.3 Greater than 50 dB with manual threshold adjustment.


24/35




FIGURE 12. (a) Concept for compressive cryoreceiver, including cued narrowband cryoreceivers for signal demodula-

tion. The compressive cryoreceiver core will produce real-time signal reports of frequency, amplitude, pulsewidth, time

of arrival, and angle of arrival. These reports will then be used to cue the narrowband cryoreceivers. (b) Detailed sche-

matic of proposed advanced semiconductor pulse-processing electronics of the compressive cryoreceiver.

Microprocessor

Advancedsemiconductor

pulseprocessing

Real-timesignal reports:

Signaldemodulation

Cueing data

FrequencyAmplitudePulsewidthTime of arrivalAngle of arrival


pulseprocessing

Smart LOcontroller

HTSchirpfilter

HTSchirp

filter

HTSchirpfilter

LNA

LNA

LNA

LNA

LNA

LNASum channel

Elevation-differencechannel

Azimuth-differencechannel

Adaptivenotchfilters


Hybridjunction

Feedhorns



pulse

processing

Cryocooler

Cryocooler

Monopulse antenna

Omnidirectionalantenna

1

2

N

HTSdelay line

Log

AmplifierSOI CMOS ASIC for

pulse detectionand data thinning

Thinnedpulse-

detectiondata Pulse-

descriptorwords

Chirp-transform

output

LO1

LO2

LON

Signal analysischannels

Tunablepreselect

filter

Tunablepreselect

filter

Chirp generator

Tunable

preselectfilter

(a)

(b)

1:4GaAs

DEMUX

6-bit GaAsHBT A/D

(3.0 GS/sec)

Envelopedetector

Advanced semiconductor pulse processing

DSPboard

~~~/

~~~/

~~~/

~~~

/

/

~~~

/

/

~~~

/

/

Integration results

Detection threshold

SOI CMOS ASIC for binary integration


25/35


26/35




DSP circuitry requires only a small fraction of the re-ceiver size, weight, and power. Enhancement ofTOA through DSP operations to identify the amountof partial analysis-window filling at the leading and

trailing pulse edges would require an additional 50operations per pulse, producing a total requirementof 75 MFlops.

Some size, weight, and power estimates can bemade by accounting for the cryocooler and its controlelectronics. A typical cryocooler/controller combina-tion (Stirling cycle) will require 25 to 50 W, 0.5 to2.0 kg, and 0.1 to 0.2 ft3 for 0.5 to 1.5 W of coolingpower near 70 K. The receiver system described, mak-ing use of new semiconductor monolithic compo-nents, translates into a small number of ASIC chips

per channel, possibly on a single board. This receiversolution can readily be estimated to consume well un-der 200 W for a three-channel compressive receiver.The receiver architecture is also better suited tohandle wider bandwidths than a receiver approachbased on multiple analog pulse-detection and pro-cessing boards. Timing uncertainties between boards

would preclude their use for bandwidths of 20 GHz,in which timing accuracies of less than 10 to 20 psec

would be the norm.The proposed HTS compressive cryoreceiver ar-

chitecture is capable of unprecedented multigigahertzbandwidth coverage per channel, translating into areceiver of small size, weight, and power. Comparedto SAW receivers, the HTS receiver could have thesame or better sensitivity, similar frequency accuracyand resolution, the same or better amplitude accu-racy, greatly improved TOA and pulsewidth accuracy,improved dynamic range, and greatly improvedshort-pulse capability.

Comparison to Conventional Technology

HTS-Compressive versus All-Digital Receiver

The signal processing power of an analog Fourier-transform process becomes evident with a compari-son to the equivalent digital fast Fourier transform(FFT) process in floating-point operations per second(Flops) to achieve the same frequency accuracy andresolution. Recall that an N-point FFT has a fre-quency resolution of 1/(NT) Hz, where Tis the sam-

pling interval [73]. The maximum frequency that canbe sampled and satisfy the Nyquist criterion isfmax =1/(2T) Hz. As long as the signal of interest can bebrought down to baseband, fmax is equal to the re-

ceiver analysis bandwidth BR. Performing an FFT inreal time requires that the FFT be completed duringthe signal, which is a time equal to NT. Assuming Nis a power of two, ( / ) logN N2 2 butterfly operationsare required for an N-point FFT [74]. The modernSHARC DSP chip (Analog Devices ADSP-2106X)performs an N-point complex FFT in 2 2N Nlogclock cycles with 3 Flops/cycle (120 MFlops at a40-MHz clock rate) [74], resulting in the followingDSP rate to perform an N-point complex FFT:

DSP rate (Flops) =

=

19 92

2 19 92

10

10

. log

( . ) log .max

NTN N

f N

Table 5 lists examples of the digital equivalents inFlops for analog chirp-transform algorithms relevantto HTS and SAW compressive receivers.

We recognize that round-off noise will requireextra bits to be carried internally to support thedesired dynamic range in the DSP FFT [75]. For anN-point FFT based on internally scaled fixed-point

arithmetic, the power ratio of round-off noise to idealoutput is 4N(22b ), where b is the number of bitscarried in the computation. Fixed-point DSP chipsusually clock slightly faster than their floating-pointcounterparts. A dynamic range of 2b0 requiresb b N= +0 21 2 4( / )log , where ( / )log1 2 42 N is thenumber of additional bits needed internally. Forexample, a 60-dB dynamic range in a 1024-pointFFT internally requires 16 bits to provide an outputof 10 bits.

The success of very-large-scale integration (VLSI)/

ultra-large-scale integration (ULSI) CMOS andBiCMOS circuits has led to dedicated fixed-pointand floating-point DSP chips. As an example, theMeshSP-1 synchronous processor [74] is an 8 8 ar-ray of sixty-four SHARC DSP chips and is capable of7.7 GFlops throughput requiring approximately 100

W for two 7.25 13-in circuit boards. This is ap-proximately the equivalent of a 50-MHz-bandwidth40-sec-long analog chirp-transform algorithm in a


27/35




SAW compressive receiver. Scaling the MeshSP-1processor directly by a factor of 40 to 300 GFlops

would require a minimum 4 kW of power and over2500 DSP chips. This digital solution is clearly not asmall size, weight, and power solution, and requiresmore than ten times the power estimate for the HTS

compressive receiver.

HTS-Compressive versus Channelized-Filter Receivers

A comparison to channelized-filter architectures canalso be made by determining the minimum numberof filters required in a filter bank to achieve the samefrequency accuracy as the novel HTS compressive-re-ceiver architecture described in the last section. Table6 lists examples of this comparison for 3-GHz and10-GHz-bandwidth receivers. If the comparison fo-cuses on the core function of the receiver and assumes

that the signal-report electronics is either separate orconsumes a negligible fraction of the receiver, then acompressive receiver consisting of a single HTS chirpfilter in a cryostat and several ASIC chips for pulsedetection (and signal report) could reasonably con-sume ten times less size and weig

Documents

High-Tc Superconductive WB Compressive Receivers_10.1.1.73