Integrated Submm Wave Receiver:
Development and Applications
Valery P. Koshelets, Manfred Birk, Dick Boersma, Johannes Dercksen,
Pavel Dmitriev, Andrey B. Ermakov, Lyudmila V. Filippenko, Hans Golstein,
Ruud W.M. Hoogeveen, Leo de Jong, Andrey V. Khudchenko,
Nickolay V. Kinev, Oleg S. Kiselev, Pavel V. Kudryashov, Bart van Kuik,
Arno de Lange, Gert de Lange, Irina L. Lapitsky, Sergey I. Pripolzin,
Joris van Rantwijk, Avri M. Selig, Alexander S. Sobolev,
Mikhail Yu Torgashin, Vladimir L. Vaks, Ed de Vries, Georg Wagner,
and Pavel A. Yagoubov
Abstract A superconducting integrated receiver (SIR) comprises in a single chip
a planar antenna combined with a superconductor-insulator-superconductor (SIS)
mixer, a superconducting Flux Flow Oscillator (FFO) acting as a Local Oscillator
(LO) and a second SIS harmonic mixer (HM) for the FFO phase locking. In
V.P. Koshelets (?) ? A.B. Ermakov ? A.V. Khudchenko ? N.V. Kinev ? O.S. Kiselev
Kotel’nikov Institute of Radio Engineering and Electronics, Russian Academy of Science,
Mokhovaya st. 11/7, 125009, Moscow, Russia
SRON Netherlands Institute for Space Research, 9700 AV, Groningen, The Netherlands
e-mail: firstname.lastname@example.org; Khudchenko@hitech.cplire.ru
P. Dmitriev ? P.V. Kudryashov ? A.S. Sobolev ? M. Yu Torgashin
Kotel’nikov Institute of Radio Engineering and Electronics, Russian Academy of Science,
Mokhovaya st. 11/7, 125009, Moscow, Russia
D. Boersma ? J. Dercksen ? L.V. Filippenko ? H. Golstein ? R.W.M. Hoogeveen ? L. de Jong ?
B. van Kuik ? A. de Lange ? G. de Lange ? I.L. Lapitsky ? J. van Rantwijk ? A.M. Selig ?
Ed de Vries
SRON Netherlands Institute for Space Research, 9700 AV, Groningen, The Netherlands
M. Birk ? G. Wagner
DLR German Aerospace Centre, Remote Sensing Technology Institute, 82234, Wessling,
S.I. Pripolzin ? V.L. Vaks
Institute for Physics of Microstructure, Russian Academy of Science, Ulyanova 46, GSP-105,
Nizhny Novgorod, Russia
SRON Netherlands Institute for Space Research, 9700 AV, Groningen, The Netherlands
European Organization for Astronomical Research in the Southern Hemisphere (ESO),
Karl-Schwarzschild-Strasse 2, 85748, Garching bei M¨ unchen, Germany
A. Sidorenko (ed.), Fundamentals of Superconducting Nanoelectronics,
NanoScience and Technology, DOI 10.1007/978-3-642-20158-5 10,
© Springer-Verlag Berlin Heidelberg 2011
264V.P. Koshelets et al.
this report, an overview of the SIR and FFO developments and optimizations is
presented. Improving on the fully Nb-based SIR we have developed and studied
Nb–AlN–NbN circuits, which exhibit an extended operation frequency range. Con-
tinuous tuningof the phase locked frequencyhas been experimentallydemonstrated
at any frequency in the range 350–750GHz. The FFO free-running linewidth has
been measured between 1 and 5MHz, which allows to phase lock up to 97% of
the emitted FFO power. The output power of the FFO is sufficient to pump the
matched SIS mixer. Therefore, it is concluded that the Nb–AlN–NbN FFOs are
mature enough for practical applications.
These achievements enabled the development of a 480–650GHz integrated
receiverforthe atmospheric-researchinstrumentTErahertzand submillimeterLImb
Sounder (TELIS). This balloon-borne instrument is a three-channel supercon-
ducting heterodyne spectrometer for the detection of spectral emission lines of
stratospheric trace gases that have their rotational transitions at THz frequencies.
One of the channels is based on the SIR technology. We demonstrate for the
first time the capabilities of the SIR technology for heterodyne spectroscopy in
general, and atmospheric limb sounding in particular. We also show that the
application of SIR technology is not limited to laboratory environments, but that
it is well suited for remote operation under harsh environmental conditions. Light
weight and low power consumption combined with broadbandoperation and nearly
quantum limited sensitivity make the SIR a perfect candidate for future airborne
and space-borne missions. The noise temperature of the SIR was measured to be
as low as 120K in double sideband operation, with an intermediate frequency
band of 4–8GHz. The spectral resolution is well below 1MHz, confirmed by
our measurements. Remote control of the SIR under flight conditions has been
demonstrated in a successful balloon flight in Kiruna, Sweden.
Capability of the SIR for high-resolution spectroscopy has been successfully
proven also in a laboratory environment by gas cell measurements. The possibility
to use SIR devices for the medical analysis of exhaled air will be discussed. Many
medically relevant gases have spectral lines in the sub-terahertz range and can be
detectedbyanSIR-basedspectrometer.TheSIR canbeconsideredas anoperational
device, ready for many applications.
A Superconducting Integrated Receiver (SIR) [1, 2] was proposed more than
10 years ago and has since then been developed up to the point of practical
applications [3–5]. Our approach consists in developing a single chip heterodyne
receiver, which is smaller and less complex than traditional devices. Typically, such
a receiver consists of a number of main components (local oscillator (LO), mixer,
antenna structure, phase lock circuit, etc.), which are usually built as separate units
and are complex (and thus costly). According to our concept (see Fig.10.1), we
have integrated all these components onto one single chip reducing overall system
complexityin changefor increasedon-chipandlithographicfabricationcomplexity.
10Integrated Submm Wave Receiver: Development and Applications265
Fig. 10.1 Block-diagram of the superconducting integrated receiver
An SIR comprises on one chip all key elements needed for heterodyne detection: a
low-noise superconductor-insulator-superconductor (SIS) mixer with quasi-optical
antenna,a flux-flow oscillator (FFO)  actingas an LO and a secondSIS harmonic
mixer (HM) for the FFO phase locking.The concept of the SIR is very attractive for
many practical applications because of the compactness and the wide tuning range
is limited by the tuning range of the LO, typically 10%–15% for a solid-state
structure and the matching circuitry between the SIS and the FFO. A bandwidth
up to 30%–40% may be achieved with a twin-junction SIS mixer design. Another
potential advantage is the use of arrays of SIR channels within a single cryostat that
could operate at the same or different LO frequencies.
One of the important practical application of the SIR is TErahertz and sub-
millimeter LImb Sounder (TELIS) [5, 9, 10] – a three-channel balloon-borne
heterodyne spectrometer for atmospheric research developed in a collaboration
of four institutes: Deutsches Zentrum f¨ ur Luft- und Raumfahrt (DLR), Germany,
Rutherford Appleton Laboratories (RAL), United Kingdom, and SRON – Nether-
lands Institute for Space Research, the Netherlands (in tight collaboration with
Kotel’nikov Institute of Radio Engineering and Electronics, IREE, Moscow). All
three receivers utilize state-of-the-art superconducting heterodyne technology and
operate at 500GHz (by RAL), at 480–650GHz (by SRON C IREE), and at 1.8THz
(by DLR). TELIS is a compact, lightweight instrument capable of providing broad
spectral coverage, high spectral resolution and long flight duration. The TELIS
instrument serves also as a test bed for many novel cryogenic technologies and as a
pathfinder for satellite-based instrumentation.
TELIS is mounted on the same balloon platform as the Fourier transform
spectrometer MIPAS-B , developed by IMK (Institute of Meteorology and
266V.P. Koshelets et al.
Climate research of the University of Karlsruhe, Germany) and is operated in the
mid-infrared.680–2;400cm?1/. Both instruments observesimultaneously the same
air mass, and together they yield an extensive set of stratospheric constituents that
can be used for detailed analysis of atmospheric chemical models, such as ozone
vertical profiles of ClO, BrO, O3and its rare isotopologues, O2, HCl, HOCl, H2O
and three rare isotopologues, HO2; NO; N2O; NO2; HNO3; CH3Cl, and HCN. In
this paper, the design and technology for the 480–650GHz channel as used in
the flight configuration are presented in conjunction with test results and the first
preliminary scientific results.
10.2 Flux Flow Oscillators
A Josephson Flux Flow Oscillator (FFO)  has proven [4, 5, 7] to be the most
developed superconducting LO for integration with an SIS mixer in a single-chip
submm-waveSIR [1–5]. TheFFO is a long Josephsontunneljunctionof the overlap
geometry (see Fig.10.2) in which an applied dc magnetic field and a dc bias
current, IB, drive a unidirectional flow of fluxons, each containing one magnetic
flux quantum, ˚0D h=2e ? 2 ? 10?15Wb. Symbol h is Planck’s constant and e is
the dc magnetic field applied to the FFO. According to the Josephson relation, the
junction oscillates with a frequency f D (1=˚0)V (about 483.6GHz/mV) if it is
biased at voltage V . The fluxons repel each other and form a chain that moves
along the junction. The velocity and density of the fluxon chain and thus the power
and frequency of the submm-wave signal emitted from the exit end of the junction
due to the collision with the boundary may be adjusted independently by proper
settings of IBand ICL. The FFO differs from the other members of the Josephson
oscillator family by the need for these two control currents, which in turn provides
the possibility of independent frequency and power tuning.
We experimentally investigated a large number of the FFO designs. The length,
L, and the width, W , of the FFO used in our study are 300–400?m and 4–28?m,
Fig. 10.2 Schematic view of a flux-flow oscillator
10Integrated Submm Wave Receiver: Development and Applications267
respectively.Thevalueofthecritical currentdensity,JC, is in therange4–8kA=cm2
giving a Josephson penetration depth, ?J ? 6–4?m. The corresponding value of
the specific resistance is Rn ? L ? W is
calculations, we use a typical value of the London penetration depth, ?L? 90nm
for all-Nb junctions, and a junction specific capacitance, Cs ? 0:08pF=?m2. The
activeareaofthe FFO (i.e.the AlOxorthe AlN tunnelbarrier)is usuallyformedas a
long window in the relatively thick .200–250nm/ SiO2insulation layer sandwiched
between the two superconducting films (base and wiring electrodes). The so-called
“idle” region consists of the thick SiO2layer adjacent to the junction (on both sides
of the tunnel region) between the overlapping electrodes. It forms a transmission
line parallel to the FFO (not shown in Fig.10.2). The width of the idle region .WID
2–14?m/ is comparable to the junction width. The idle region must be taken into
account when designing an FFO with the desired properties. In our design, it is
practical to use the flat bottom electrode of the FFO as a control line in which the
current ICLproduces the magnetic field, which mainly is applied perpendicular to
the long side of the junction.
There are a number of important requirements on the FFO properties to make it
suitable for application in the phase locked SIR. Obviously, the FFO should emit
enough power to pump an SIS mixer, taking into account a specially designed
mismatch of about 5–7dB between the FFO and the SIS mixer, introduced to avoid
leakage of the input signal to the LO path. It is a challenge to realize the ultimate
performance of the separate superconducting elements after their integration in
a single-chip device. Implementation of the improved matching circuits and the
submicron junctions for both the SIS and the HM allows delivering optimal FFO
power for their operation.
Even for ultra wideband room-temperaturePLL systems the effective regulation
bandwidth is limited by the length of the cables in the loop (about 10MHz for
typical loop length of two meters). It means that the free-running FFO linewidth
(LW) has to be well below 10MHz to ensure stable FFO phase locking with a
reasonably good spectral ratio (SR) – the ratio between the carrier and total power
emitted by the FFO . For example, only about 50% of the FFO power can be
phase locked by the present PLL system at a free-running FFO LW of 5MHz. A
lowspectralratioresultsina considerableerrorat resolvingthecomplicatedspectral
line shape . Thus, a sufficiently small free-runningFFO LW is vitally important
for the realization of the phase locked SIR for the TELIS.
? 50–25 ? ?m2. For the numerical
Earlier the Nb–AlOx–Nb or Nb–AlN–Nb trilayers were successfully used for the
FFO fabrication.Traditionalall-Nb circuits are being constantlyoptimizedbut there
seems to be a limit for LW optimizations at certain boundary frequencies due
to Josephson self-coupling (JSC) effect  as well as a high frequency limit,
imposed by Nb gap frequency .?700GHz/. That is the reason for novel types
268V.P. Koshelets et al.
Fig. 10.3 The dependencies of Rj/Rn ratio on critical current density Jc for SIS junctions of
different types fabricated at IREE
of junctions based on materials other than Nb to be developed. We reported on
development of the high quality Nb–AlN–NbN junction production technology
. The implementation of an AlN tunnel barrier in combination with an NbN
top superconducting electrode provides a significant improvement in SIS junction
quality. The gap voltage of the junction VgD 3:7mV. From this value, and the gap
voltage of the Nb film ?Nb=e D 1:4mV, we have estimated the gap voltage of our
NbN film as ?NbN=e D 2:3mV .
The dependency of the ratio of subgap to normal state resistance (Rj/Rn) vs.
critical current density .Jc/ for different types of the Nb-based junctions fabricated
at IREE is presented in Fig.10.3. One can see that the Nb–AlN–NbN junctions are
of very good quality at high current densities, important for implementation in THz
mixers. The same technique was further used to produce complicated integrated
circuits comprising SIS and FFO in one chip.
The use of Nb for top “wiring” layer is preferable due to lower losses of Nb
compared to NbN below 720GHz; furthermore, the matching structures developed
for the all-Nb SIRs can be used directly for the fabrication of receivers with
Nb–AlN–NbN junctions. The general behavior of the new devices is similar to the
all-Nb ones; even the control currents, necessary to provide magnetic bias for FFO,
were nearly the same for the FFOs of similar design.
A family of the Nb–AlN–NbN FFO IVCs measured at different magnetic fields
produced by the integrated control line is presented in Fig.10.4 .L D 300?m;
W D 14?m; WID 10?m/. A single SIS junction with an inductive tuning circuit
is employed as a HM for the LW measurements. The tuning and matching circuits
were designed to provide“uniform” coupling in the frequencyrange 400–700GHz.
10Integrated Submm Wave Receiver: Development and Applications269
Fig. 10.4 IVCs of the Nb–AlN–NbN FFO measured at different magnetic fields produced by the
integrated control line. The color scale shows the level of the DC current rise at the HM induced by
the FFO. Red area marks the region of the FFO parameters where the induced by FFO HM current
exceeds 25% of the Ig. This level is well above the optimal value for an SIS-mixer operation
Measured values of the HM current induced by the FFO oscillations (HM pumping)
are shown in Fig.10.4 by the color scale. The HM pumpingfor each FFO bias point
was measured at constant HM bias voltage of 3mV (pumping is normalized on the
current jump at the gap voltage, Ig D 140?A). From Fig.10.4, one can see that
an FFO can provide large enough power over the wide frequency range: limited
at higher frequencies only by the Nb superconducting gap in transmission line
electrodes (base and wiring layers) and below 400GHz by design of the matching
The Nb–AlN–NbN FFOs behave very similar to all-Nb ones. The feature at
about 600GHz where the curves get denser is a Josephson Self-Coupling (JSC)
boundary voltage. It was first observed for all-Nb FFOs . The JSC effect is
the absorption of the FFO-emitted radiation by the quasi-particles in the cavity
of the long junction. It considerably modifies the FFO properties at the voltages
V ? VJSC D 1=3Vg(VJSCcorresponds to 620GHz for the Nb–AlN–NbN FFO).
Just abovethis voltage,thedifferentialresistanceincreasesconsiderably;thatresults
in an FFO-LW broadening just above this point. This, in turn, makes it difficult
or impossible to phase lock the FFO in that region. For a Nb–AlOx–Nb FFO, the
transition corresponding to VJSCD Vg=3 occurs around 450GHz. So, by using the
Nb–AlN–NbN FFOs we can coverthe frequencygap from 450to 550GHz imposed
by the gap value of all-Nb junctions. The feature in Fig.10.4 around 1mV is very
likely dueto a singularityat the differenceof the superconductinggaps?NbN??Nb.
270V.P. Koshelets et al.
Continuous frequency tuning at frequencies below 600GHz for the Nb?AlN?
NbN FFOs of moderate length is possible, although the damping is not sufficient
to completely suppress the Fiske resonant structure at frequencies below Vg=3.
For short junctions with a small ’ (wave attenuation factor), the distance between
the steps in this resonant regime can be as large, that it is only possible to tune
the FFO at the certain set of frequencies. For a 300–400?m long Nb–AlN–NbN
junction, this is not the case – the quality factor of the resonator formed by a
long Nb–AlN–NbN Josephson junction is not so high at frequencies >350GHz.
Therefore, the resonance steps are slanting and the distance between them is not so
big (see Fig.10.4). This allows us to set any voltage (and any frequency) below
VJSC, but for each voltage only a certain set of currents should be used. So, in
this case we have the regions of forbidden bias-current values, specific for each
voltage below VJSC, instead of the forbidden voltage regions for the Fiske regime
in Nb–AlOx–Nb FFO . Special algorithms have been developed for automatic
working point selection in flight.
InFig.10.5,the typicalcurrent-voltagecharacteristics(IVCs)ofa Nb–AlN–NbN
SIS junctionof an area of about 1?m2is given, both the unpumpedIVC (solid line)
and the IVC when pumped by a Nb–AlN–NbN FFO at different frequencies(dotted
lines). One can see that the FFO provides more than enough power for the mixer
pumping.In this experiment,we use the test circuits with low-loss matchingcircuits
tuned between 400 and 700GHz. Even with the specially introduced 5dB FFO/SIS
mismatch (required for the SIR operation) the FFO delivers enough power for the
SIS mixer operation in the TELIS frequency range of 480–650GHz .
Fig. 10.5 The IVCs of the SIS mixer: unpumped – solid curve, pumped at different frequencies –
dashed and dotted lines
10Integrated Submm Wave Receiver: Development and Applications271
Fig. 10.6 The IVCs of the SIS mixer: unpumped – black solid curve, pumped at different FFO
bias currents (different powers) – lines with symbols; FFO frequency D 500GHz
Fig. 10.7 The pump current of the SIS mixer biased at 3mV as a function of the FFO bias current
at the fixed frequency 500GHz (see Figs.10.4 and 10.6)
An important issue for the SIR operation is a possibility to tune the FFO power,
while keeping the FFO frequencyconstant. This is demonstratedin Fig.10.6, where
the IVCs of an SIS mixer are shown, while being pumped at different FFO bias
currents (different powers). The dependence of the SIS pump current on the FFO
bias current is presented in Fig.10.7, showing that the FFO power can be tuned
more than 15dB, while keeping the same frequency by proper adjustment of the
control line current.
272V.P. Koshelets et al.
10.2.2 Spectral Properties of the FFO
The FFO LW has been measured in a wide frequency range from 300GHz
up to 750GHz using a well-developed experimental technique . A specially
designed integrated circuit incorporates the FFO junction, the SIS HM and the
microwave matching circuits. Generally, both junctions are fabricated from the
same Nb/AlN/NbN or Nb/AlOx/Nb trilayer. The FFO signal is fed to the SIS
HM together with a 17–20GHz reference signal from a stable synthesizer. The
required power level depends on the parameters of the HM; it is about of 1?W
for a typical junction area of 1?m2. The intermediate frequency (IF) mixer product
.fIFD ˙(fFFO?n?fSYN)at? 400MHzisfirst boostedbyacooledHEMTamplifier
.Tn? 5K; gain D 30dB/ and then by a high-gain room-temperatureamplifier.
To accurately measure the FFO line shape, the IF signal must be time-averaged
by the spectrum analyzer. To remove low-frequency drift and interference from
the bias supplies, temperature drift, etc., we use a narrow bandwidth .<10kHz/
Frequency Discriminator (FD) system with relatively low loop gain for frequency
locking of the FFO. With the FD narrow-band feedback system that stabilizes the
mean frequency of the FFO (but does not affect FFO line shape), we can accurately
measure the free-runningFFO LW, which is determined by the much faster internal
(“natural”) fluctuations (see Fig.10.8).
Fig. 10.8 Spectra of the Nb–AlN–NbN FFO operating at 515.2605GHz (blue dashed line –
frequency locked by FD; red solid line – phase-locked). Linewidth D 1:7MHz; spectral ratio D
10Integrated Submm Wave Receiver: Development and Applications273
The resulting IF signal is supplied also to the Phase Locking Loop (PLL)
system. The phase-difference signal of the PLL is fed to the FFO control line
current. Wideband operation of the PLL (10–15MHz full width) is obtained by
minimizingthe cableloop length.A part of the IF signal is deliveredto the spectrum
analyzer through a power splitter (see Fig.10.8). All instruments are synchronized
to harmonics of a common 10MHz reference oscillator.
The integrated HM may operate in two different regimes, either as a quasi-
particle mixer (SIS) or as a Josephson mixer. To exclude the noise from the
Josephson super-current fluctuations and thereby realize a pure quasi-particle
regime, the super current has to be suppressed by a relatively large magnetic field.
This requires a special control line placed near the SIS mixer. The quasi-particle
regime of the HM operation can also be realized with sufficient synthesizer power.
It has been shown  that the FFO LW and signal-to-noise ratio are almost the
same for these two regimes, although the phase noise might be somewhat lower in
the quasi-particle mode.
10.2.2.2Dependence of the FFO Linewidth on FFO’ Parameters
Detailed measurements of the FFO LW [18, 19] demonstrate a Lorentzian shape
of the free-running FFO line in a wide frequency range up to 750GHz, both at
higher voltages on the flux flow step (FFS) and at lower voltages in the resonant
regime on the Fiske steps (FSs). This implies that the free-running (“natural”) FFO
LW in all operational regimes is determined by the wideband thermal fluctuations
and the shot noise. This is different from many traditional microwave oscillators,
where the “natural” LW is very small and the observed LW can be attributed mainly
to external fluctuations. It was found [18,19] that the free-running FFO LW, ıf ,
exceeds theoretical estimations made for lumped tunnel Josephson junction. The
expression for the LW dependency on voltage and differential resistances found for
all-Nb FFOs [18,20] is valid for Nb–AlN–NbN junctions as well:
dC K ? RCL
where Si0 is the power density of low frequencycurrent fluctuations, RdBand RdCL
are differential resistances on bias and control line currents, respectively. Note that
ratio RdCL=RdBis constant for fixed FFO bias, so ıf D A(IB) (RdB)2Si0.
Earlier, a so-called Super Fine Resonance Structure (SFRS)  was observed
on the FFO IVCs, resulting in the jumps of the FFO between tiny steps (frequency
spacing is of about 10MHz, see Fig.10.9). The presence of the SFRS prohibits
phase locking at frequencies between the steps. This is unacceptable for practical
applications. Recently, we found that the SFRS is related to interference of the
acoustic waves created by the FFO (generation of the phonons by Josephson
junction, see ). A special technological procedure allows us to eliminate this
interference and to realize continuous FFO-frequency tuning in the SIR, being
274V.P. Koshelets et al.
Fig. 10.9 Down-converted spectra of the FFO: (a) free-running FFO; (b), (c) – the lines show the
maximum FFO signal level recorded in the MaxHold regime of the Spectrum Analyzer (the top
point of curve “a”) on the FFO frequency, measured before (b) and after (c) special Si substrate
Fig. 10.10 Linewidth dependency on frequency for two types of the FFO
vitally important for TELIS project (see Fig.10.9). Details of this study will be
In Fig.10.10, we present a comparative graph of the free-running FFO LW for
two types of the tri-layer. One can see that the LW of Nb–AlN–NbN FFO is twice
as small up to 600GHz. It should be emphasized that due to overlapping FSs
continuous tuning is possible and any desirable frequency can be realized. Several
10Integrated Submm Wave Receiver: Development and Applications275
“stacked” stars at certain frequenciesfor the NbN FFO mean that the best LW value
can be selected by adjusting FFO bias. Note that the spread in the LW values at
a selected frequency is small and all can actually be applied for measurements.
Each star corresponds to an “allowed” bias current at an FS (as described above in
Sect.2.1). Althoughthe FFO tuning on an FS is complicated,the benefit in LW (and
consequently the spectral ratio) is worth the effort. Linewidths below 3MHz can
be achieved in the whole range between 350 and 610GHz. An abrupt increase of
the FFO LW at some frequencies is caused by the Josephson self-coupling effect.
The JSC (absorption of the FFO-emitted radiation by the quasi-particles in the
cavity of the long junction, see above) considerably modifies the FFO properties
at the voltages V ? VJSC D 1=3Vg (VJSCcorresponds to 620GHz for the
Previous LW measurements have demonstrated[7,23] the essential dependences
of the free-running FFO LW on the FFO voltage, its current density and geometry
of the biasing electrodes. In this report, we summarize the results of the FFO study
and optimization of the FFO layout for both types of FFOs. Recently, it was shown
[4, 7] that the LW decreases considerably with increasing width, W , of the FFO
junction. This is valid for all frequencies of interest, and consequently, the spectral
ratio of the phase locked FFO for wide junctions is better. We have increased the
FFO width up to 28?m, which is more than five times the Josephson penetration
depth ?J. A number of FFOs with the same electrode layout, but different widths
of the FFO junction (W D 4, 8, 12, 16, 20 and 28?m) are fabricated using the
same technological procedure yielding the same junction parameters (normal state
resistance ? area, RnS D 30??m2). The results of the LW measurements of these
circuits at three frequencies are presented in Fig.10.11.
Fig. 10.11 Linewidth of free-running FFOs (left axis) and corresponding spectral ratio for the
phase-locked FFO (right axis) measured at different FFO frequencies as a function of FFO width.
All circuits are fabricated by the same technological procedure .RnS D 30??m2/
276V.P. Koshelets et al.
Even for the largest tested width .W D 28?m/, there is no evidence of deteri-
oration in the FFO behaviour. Furthermore, the power delivered to the SIS mixer
is getting higher and the LW lower at all frequencies. The decrease of the FFO LW
with increasingFFO width is in accordancewith existingtheoreticalmodels andour
expectations. The bias current differential resistance, Rd, decreases approximately
inversely proportional to the bias current IB. Since the FFO LW is proportional
to Rd2?IB, it scales down linearly with the junction width. Of course, one can
expect that the LW decrease will saturate and the FFO performance will deteriorate
with further increase of the width (e.g., due to appearance of transversal modes).
Without a reliable theory, the optimal value of the FFO width has to be determined
experimentally. Note that for a wider FFO the center line of the junction is shifted
away from the edge of the control line (the RdCLgoes down). This may result
in a considerable reduction of extraneous noise from external magnetic fields.
Furthermore, a wider FFO presumably will have a more uniform bias current
distribution . At the present state, the width of the FFO for TELIS is chosen
to be 16?m. This is a tradeoff between LW requirements and technical limitation
on the maximum bias and control line currents (both should not exceed 70mA).
In contrastto variationofthe FFO LW on theFFO width,previousmeasurements
 have demonstrated a considerable increase of the FFO LW with the FFO current
density. This contradicts the simplified consideration: the increase of the FFO
current density (as it is for increase of the FFO width) should result in the increase
of the total FFO bias current, IB, and reduce the FFO differential resistance on the
bias current Rd. Since the FFO LW is proportionalto Rd2?Ib, one should expect the
decrease of the measured FFO free-running LW for larger FFO current density. In
reality, Rddoes not decrease as much as this simple consideration predicts and the
LW increases. On the contrary, a high value of the current density .Jc? 8kA=cm2/
is important for wide-band operation of the SIS-mixer at the submm wave range.
The increase of the FFO LW with current density (as discussed above) creates a
serious problemin the designand developmentof SIR chips. Implementationof two
separate tri-layers with different current densities – one for the SIS mixer (high Jc)
and the other one for the FFO/HM (lower Jc) seems to be a solution. We have
successfully tested and verified this approach for the SIR microcircuits for TELIS.
Improvement of the FFO performance was obtained by enlarging the electrodes
overlapping area, the so-called “idle region”. Larger overlapping presumably
provides a more uniform bias-current distribution, due to reduced inductance of the
overlapping electrodes. Larger overlapping of the FFO electrodes also implies that
theFFO ofthe same widthis shiftedfromthe edgeof thebottomelectrode,resulting
in a considerable decrease of the RdCLvalue. Note that for a wide FFO also some
shift of the FFO center line appears due to increasing of the width. Experimentally,
we found that an idle region WID 10?m is the optimal value for the present FFO
design. Up to now, there is no adequate model that can quantitatively describe
both the processes in the FFO and a self-consistent distribution of the bias current.
Nevertheless, the presented results are very encouraging and these modifications of
the FFO were implemented in the TELIS SIRs.
10Integrated Submm Wave Receiver: Development and Applications 277
To further explore this approach, we have developed different designs of the
“self-shielded” FFO with a large ground plane in the base electrode. Such FFOs
are expected to be less sensitive for variations in the external magnetic field and
have to provide more uniform bias current distribution (since all bias leads are
laying over superconducting shield and have low inductance). Actually, the low-
inductive bias leads provide a possibility of optimal (rather than uniform) current
distribution, “regulated” by the FFO itself. The last feature optimizes the emitted
FFO power. Indeed, the IVCs of all shielded FFOs are much more reproducible;the
power delivered to HM is higher compared to a traditional design. Unfortunately,
the free-running LW for all variants of shielded FFOs with separate bias leads is
much larger than for FFOs of traditional design. It seems that injection of the bias
via separate leads results in some spatial modulation of bias current  despite
the additional triangular elements added for more uniform current injection. On
the contrary, designs that employed three superconducting electrodes provide both
perfect pumping and improved LW, details will be published elsewhere.
10.2.2.3Spectral Ratio, Phase Noise
As it was mentioned above, the free-runningFFO LW has to be well below 10MHz
to ensure stable FFO phase locking with a reasonably good spectral ratio (SR, the
ratio between the carrier and total FFO power). For example, only about 50% of
the FFO power can be phase locked by the present TELIS PLL system at free-
running FFO LW of 5MHz. A low spectral ratio results in a considerable error
at resolving of the complicated atmospheric line shapes . For the given PLL
system, the value of the SR is fully determined by the free-running FFO LW: these
two quantities are unambiguously related (see Fig.10.12, where data for FFOs of
different designs and types are presented). The theoretical curve, calculated in ,
coincides reasonably well with the experimental data. A possibility to considerably
increase the SR by application of the ultra-wideband cryogenic PLL system has
been recently demonstrated .
An important issue for TELIS operations is the possibility to tune the FFO
frequency and power independently, while providing the same spectral ratio of PL
FFO. The TELIS HM is pumped by a tunable reference frequency in the range
of 19–21GHz from the LO Source Unit (LSU), phase locked to the internal ultra
stable 10MHz Master Oscillator. The HM mixes the FFO signal with the n-th
harmonic of the 19–21GHz reference. The LW and SR of the TELIS FFO are
almost constant over a wide range of FFO bias current at fixed FFO frequency (see
Fig.10.13). From this figure, one can see that the SR is about 50% over the range
of bias current, Ib, 14–30mA, while the pumping level varies from 3:5?A at IbD
14mA up to 81?A at Ib D 30mA. Furthermore, the SR D 34% can be realized
at Ib D 12mA, where the HM pumping is below 0:5?A. It means that at proper
choice of the HM voltage and LSU power even moderate HM pumping by the FFO
is enough for efficient PLL operation (providing sufficient signal-to-noise ratio).