Standards and High-speed Instrumentation

In the last 20 years superconductivity has played a major role in the measurement of fundamental constants and the development of new definitions for the basic units. The most important developments include voltage standards, current comparators, null detectors, RF power meters, frequency multipliers, and thermometers [I], [2]. More recently, superconducting integrated circuits, which make major advances in the speed and sensitivity of time domain measurements, have been developed. A new concept for a superconducting bolometer shows promise for microwave power measurement and infrared detection. Many of the metrological applications of superconductivity make use of the unique nonlinearity of Josephson junctions. These junctions are formed by separating two superconductors by a thin insulating barrier. The current through a Josephson junction has a component that oscillates at a frequency exactly proportional to the applied voltage, that i s f = 2eV/h, where h i s Planck’s constant and e is the electron charge. This result is known as the ac Josephson effect. Since the Josephson voltage-frequency relation i s independent of a l l other parameters and has no known corrections, it forms the basis for voltage standards throughout the world. In practice, these standards operate by applying a microwave current through the junction and allowing the Josephson frequency to phase lock to some harmonic of the applied microwave frequency. This produces constant-voltage “steps” in the I-Vcurve of the junction. The voltages of these steps have the exact values V,  = nhf/2e, where n is an integer. The recent application of very large-scale integration (VLSI) to superconducting integrated circuits has made possible voltage-standard chips


INTRODUCTION
In the last 20 years superconductivity has played a major role in the measurement of fundamental constants and the development of new definitions for the basic units.The most important developments include voltage standards, current comparators, null detectors, RF power meters, frequency multipliers, and thermometers [I], [2].More recently, superconducting integrated circuits, which make major advances in the speed and sensitivity of time domain measurements, have been developed.A new concept for a superconducting bolometer shows promise for microwave power measurement and infrared detection.
Many of the metrological applications of superconductivity make use of the unique nonlinearity of Josephson junctions.These junctions are formed by separating two superconductors by a thin insulating barrier.The current through a Josephson junction has a component that oscillates at a frequency exactly proportional t o the applied voltage, that is f = 2eV/h, where h is Planck's constant and e is the electron charge.This result is known as the ac Josephson effect.Since the Josephson voltage-frequency relation is independent of all other parameters and has n o known corrections, it forms the basis for voltage standards throughout the world.In practice, these standards operate by applying a microwave current through the junction and allowing the Josephson frequency to phase lock t o some harmonic of the applied microwave frequency.This produces constant-voltage "steps" in the I-Vcurve of the junction.The voltages of these steps have the exact values V, , = nhf/2e, where n is an integer.The recent application of very large-scale integration (VLSI) t o superconducting integrated circuits has made possible voltage-standard chips Manuscript received October 14, 1988; revised April 14, 1989.C. A. Hamilton that generate more than 100 000 quantized voltage levels spanning the range from +IO t o -10 V.The ease of use of these new devices has enabled an increasing number of military and industrial standards laboratories t o maintain a primary Josephson voltage standard.
Another application makes use of the nearly perfect conductivity of superconductors to fabricate current comparators with extremely accurate ratios.While these comparators have been in use for manyyears, their importance has increased recently because of the emerging use of the quantum Hall standard of resistance.Quantum Hall devices generate precise quantized resistances with the values R h = h/ie2 where i is an integer.Comparison of the Hall resistance values with traditional decade resistance standards requires the precise measurement of an unusual current ratio.For example, when i = 4, Rh = 6453.20fl.Thus a comparison with a 100-Q resistance standard requires the measurement of a current ratio of 6453.20:100.One of the most accurate ways t o accomplish this is t o balance the magnetic fields generated by two opposing coils.When the coils are enclosed in a superconducting shield and a superconducting quantum interference device (SQUID) is used t o null the magnetic field, the current ratio can be made t o match the turns ratio t o a few parts in 10".
In the last decade, the application of integrated-circuit methods t o superconducting devices has opened u p a number of new applications in high-speed measurements.
One of the most important of these is the development of a superconducting sampling oscilloscope.This device takes advantage of the Josephson junction being a threshold detector that can switch between its zero-and finite-voltage states in only a few picoseconds.This has made possible a novel sampling oscilloscope which uses a Josephson pulse generator and comparator to achieve time resolution of a few picoseconds and a sensitivity of about 50 pV.
O n the horizon is a design for an extremely sensitive bolometer that uses the temperature dependence of the kinetic inductance of a microstripline.This bolometer should make significant advances in the accuracyand sensitivity of microwave power measurements and may also be useful as an infrared detector.
The remainder of this paper will review these four applications of superconductivity in metrology.

JOSEPHSON VOLTAGE STANDARD
In 1972 the U.S. legal volt was defined in terms of the ac Josephson effect [3].For the purpose of maintaining the voltage standard, 2elh was assigned the value of 483 593.42 GHz/V.Early realizations of the Josephson volt used one or two junctions driven at 10 GHz to produce reference voltages between 5 and 10 mV.Such low voltages made the calibration of secondaryweston cell standardsat 1.018Vavery difficult and time-consuming task.
An obvious improvement to Josephson voltage standards is to connect many junctions in series to generate a large voltage.A simple series extension of the single-junction standard has been used t o achieve a level of 100 rnV using junction parameters.For currents near zero there are about 10 values of quantized voltage and there are no stable regions of the I-Vcurve between the constant-voltage steps.Thus a single-current bias near zero is sufficient to guarantee that every junction in a large array is on a constantvoltage step.The zero-bias arrangement produced quantized voltages first with a single junction [6], then with small arrays[7]-[9],and finallywitharraysof upto18 992junctions [IO]-[15], which achieve voltages in excess of 10 V.

Josephson Series-Array Design
The success of a series-array voltage standard relies on achieving very stable RF-induced steps which cross the zero-current axis of the junction I -V curve.The junction design must meet four conditions to achieve this stability [14].First, stabilityagainst noise requires that the steps have a large current amplitude and thus that the junctions have a large critical current.Second, the solutions of the circuit equations have been shown to be chaotic for some parameter ranges and therefore unstable unless the junction plasma frequency i s lessthanabout one-thirdoftheapplied microwave frequency.This condition requires the use of low-critical-current-density (low-plasma-frequency) junctionsand a high drivefrequency.The requirementofalarge critical current at low critical-current density implies the use of large-area junctions.However, if the junction length (in the direction of current flow) is too large, the magnetic field due to the RF current will suppress the Josephson steps.Large-area junctions also have geometric resonances, and if any of these resonances fall below the drive frequency, the steps will also be suppressed.These last two conditions place strict limitations on the maximum length and width of the junction.In order to satisfy the above conditions, the junction length, width, and critical-current density must fall within a narrow range of acceptable values.
A set of nearly optimum values is listed in Table 1.The restrictions o n the critical-current density and the dimensions of the junction can be combined to produce an expression for the maximum current amplitude I, of the nth step when parameters are chosen to optimize the stability of the n t h step, Here = h/2e is the flux quantum, f is the drive frequency, V, i s the voltage of the nth step, and d is the sum of the penetration depths of the superconductors on either side of the junction.
Using I, as a figure of merit for the junction design, Eq. (1) shows the effect of various parameters.For example, I , increases almost linearlywith drive frequency.Thus we have chosen a drive frequency in the range of 70-100 GHz because it i s economically impractical to produce sufficient power at higher frequencies.It is also desirable to choose superconductors with small penetration depths such as niobium or lead.Niobium nitride is at a disadvantage because of its large penetration depth.The capacitance of the junction does not appear in Eq. (1).Thus there should be no performance difference between an optimized design using junctions with a high-capacitance niobium oxide barrier and a design using a much lower capacitance artificial barrier.Experimental results confirm these conclusions.Arrays with large stable steps have been fabricated in niobium-niobium, niobium-lead and lead-lead technologies.The lead-lead devices seem to produce slightly more stable steps but degrade with time and temperature cycling, and are therefore not very reliable.Although very robust, niobium-niobium arrays are sometimes prone to trap magnetic flux.
Fig. 2 shows a layout for the 78 992-junction array used inthelo-Vstandard.Afinlineatoneendof thechipcollects 70-100-GHz radiation from a waveguide and directs it into a microstripline.The microstripline splits into 16 parallel paths, each of which passes through 1187 junctions and terminates in a matched load.A network of high-and low-pass filters allows the microwave power to be applied in parallel while the dc voltage across all 18 992 junctions is added in series.The RF drive i s applied by inserting the finline end of the chip into a slot parallel to the €field in a WR-12 waveguide.The length of each microstripline section is limited to about 1200 junctions by the attenuation in the microstripline, which is estimated to be 0.004 dB per junction.
After passing through about 1200 junctions the microwave power i s significantly below the optimum value for step generation.Since each junction generates from 0.5 to I mV, the total number of junctions required for IO-V operation i s about 20 000.Thus, the microwave power must be split into 16 sections.The microwave power required is about 10 mW at the finline input.The microwave circuit is designed to be broadband and works well from 70 to 100 GHz.The dc output appears across superconducting pads on the edges of the chip.The operation of the array i s illustrated in the I-V curve of Fig. 3, which shows a few of the 100 000 constant-voltage steps that span the range from -10 to +IO V.In this case the applied frequency i s 96 GHz and produces a step spacing of hfl2e = 198 pV.  of about 20 pA.If even one junction in the array fails to lock to a constant-voltage step, the oscilloscope display shows a marked slope.A high-precision digital voltmeter is used to monitor the array voltage directly or as a null meter in comparison measurements.The microwave counter and DVMareconnectedviathe IEEE488 bustoasmall computer that can do all of the data acquisition and analysis required for a variety of calibrations.
When an array is used to calibrate a secondary reference standard, it is desirable to match the array voltage asclosely as possible to the reference value.This is accomplished by selecting a frequency fand a step number n which yield the desired voltage.The rapid selection of a particular step i s accomplished using a bias source with a load line centered on the step of interest and steep enough to intersect only a few steps, as shown in Fig. 3.When a small sweep is applied to the bias, the array can generally be locked onto the desired step within a few seconds.Calibrations of a Zener referencestandard using900 null readingsand two reversals can be completed in about 15 minutes.The uncertainty i s dominated by the noise of the Zener reference and i s typically 0.004 ppm.
The flexibility of the array in generating a wide range of accurate voltages has also found application in making accurate ratio and linearity measurements.At least one recently introduced digital voltmeter has a linearity specification of 0.1 ppm, a level that is difficult to verify with conventional voltage divider techniques.This instrument was tested by comparing its reading with the array voltage at a set of arbitrarily chosen points.These data were then fitted Several experiments have been performed to test the precision of Josephson voltage standards.The most sensitive of these experiments measures the difference between the voltage generated by two different Josephson devices driven at the same frequency.An upper limit on this differenceof 3 parts in 10" has been measured for single junctions [I61 and 2 parts in 10l7 for series arrays [17].Thus, for practical purposes any correction to the Josephson voltage-frequency relation i s negligible.
The simplicity and reliability of Josephson array voltage standards have resulted in their adoption at a steadily increasing number of national, military, and industrial standards laboratories.At the time of this writing there are array standardsoperatingat about20 locations around theworld.losephson

Sampling
The inherent high speed of Josephson devices as electric switching elements suggests their use in high-performance waveform acquisition anddisplayequipment.If this switching speed could be fully used in an instrument, such an instrument would be capable of displaying any signal likely to be encountered in present semiconductor-based electronics with minimal distortion produced by bandwidth limitations.With this prospect in mind, Josephson electronics are being investigated for use in single-shot transient recorders via high-rate analog-to-digital conversion.A Josephson-based sampler for repetitive signals has also been developed and incorporated into a commercially available instrument.Analog-to-digital conversion technology based on Josephson junctions is described in another paper in this issue.Here we focus on Josephson sequential sampling and the issues faced in the practical application of this technique.
Although a Josephson junction is a very fast switching device, the high geometric capacitance across the device limits its potential as an analog switch.Therefore, Josephson junctions are not suitable substitutes for the Schottkybarrier diodes in conventional sampling topologies in spite of the suggestive similarityof the Schottkyforward bias and Josephson quasi-particle I-V curves.A different sampling scheme appropriate t o Josephson junctions had to be invented.One early method allowed the monotonically increasing portions of an unknown repetitive signal to be observed [18], [19].The signal and a bias current are applied to a Josephson junction.When their sum exceeds the junction critical current I, it switches to the voltage state.The signal level at the time of the switching event can be determined by measuring the critical current and the bias current.The rising edge of the signal can be plotted as a function of time by sweeping the bias.
A significant improvement to this technique was made by adding a Josephson pulse generator to strobe the signal [20], [21].As shown in Fig. 6(a), the output I,(t) of the pulse generator is summed with the unknown repetitive signal I,(t) and a variable dc bias I, and is applied to a Josephson threshold detector.If the peak amplitude of the pulse /,(T) i s larger than the peak-to-peak amplitude of the signal t o be sampled, it is possible to sample all portions of the signal waveform.As shown in Fig. 6(b), the bias current is adjusted so that the sum I, + /!,(TI + I,(T) just equals the threshold level Io.This requires many repetitions of the signal event and readjustments of I, to find the threshold.Since Ib,Ip(T), and I, are all known, it is a simple matter to find 15(T), the signal level at the time t of the sampling pulse, that is, I,(T) = I, -I,(T) -lb.This is the Josephson sampling technique commonly used as an internal diagnostic in experimental Josephson circuits and in a commercially available instrument.In practice, the threshold detection device and adjustable dc current are part of a feedback loop, which maintainstheswitching probabilityon agivenevent at50%.For example, the presence or absence of switching determines whether negative or positive current is applied to a capacitor in the feedback circuitry, changing its voltage slightly before the next event.This voltage determines the bias current applied to the Josephson threshold switching device.Thus, if no switching is detected, the bias current increases until the switching probability reaches 50%.The dc bias current is therefore proportional t o the negative of the signal amplitude at the time of the strobe pulse.The accuracy is afunction of the noise induced in the feedback For each signal event, a conventional sampler provides an analog output proportional to the signal at the sampling time.The Josephson sampler provides only the information that the signal amplitude i s above or below the reference bias current.The Josephson sampler therefore requires many more events t o reconstruct the signal waveform, reducing the updating rate for a given event rate.O n the other hand, the low internal noise level of cryogenic electronics and the high sensitivity of Josephson devices provides a clearer display of low-amplitude signals at comparable updating rates.
The threshold switching device used as the sampling element in Josephson technology can be either a single Josephson junction [22] or an interferometer containing two or more junctions.In either case the device responds to an input current that exceeds its threshold by switching from the zero-voltage state to the gap sum voltage, about 3 mV for niobium junctions.The advantage of using an interferometer instead of asingle junction as the sampling detector lies in the isolation provided bytransformer coupling of the inputs.The inputs are dc isolated from each other and from the switching element itself.This is an important consideration in some circuits.Interferometers have the potential disadvantage of being prone to internal resonances, which can produce aberrations on the displayed signal in some circumstances.Single junctions have the advantage of less complicated internal dynamics and perhaps lower aperture widening.The responseof a single junction operated in this mode is dominated by its plasma resonance, as analyzed by Van Zeghbroeck [23].
The effective aperture of a Josephson sampler depends upon the width of the strobe pulse and the response time of the Josephson threshold switch.In a properly designed circuit, these parameters are governed by the junction's plasma resonance, which is proportional to the square root of the junction's critical current divided by its capacitance.The critical current depends exponentially o n the reciprocal of the barrier thickness while the capacitance depends linearly o n this quantity.Higher current-density junctions therefore have a higher plasma resonance frequency.This translates into narrower effective apertures and better performance.The shortest rise time observed to date i s 2.1 ps [24], using junctions with the relatively high critical-current density of 30 kA/cm2.When a conventional sampling circuit is strobed, it often generates a large "kickout" signal o n the input.This signal can propagate back t o the circuitry being measured and potentially disrupt operation.Josephson samplers, particularly those with magnetically coupled inputs, produce kickout that is lower by orders of magnitude than typical Schottky diode samplers, effectively eliminating this problem.Kickout is further reduced by attenuation at the sampler input.
In order to make use of a Josephson sampler in conventional test and measurement applications, two major and somewhat related problems must be overcome.First, the Josephson circuitry must be cryogenically cooled.Second, the signals t o be measured, derived in a room-temperature loop.
environment, must be transmitted t o the Josephson circuitry without appreciable bandwidth degradation or other frequency response distortion.I n the commercially available instrument, these problems were addressed in a novel and successful way [25], [26].In this instrument, liquid helium is used as the refrigerant for the niobium-based Josephson circuitry.The sampling circuit is fabricated in the corner of a 1.1-by 1.6-cm fused-silica substrate, which has extremely low thermal conductivity.By encasing the chip in foam insulation that exposes only the circuit area, and directing a small, well-controlled jet of liquid helium at this area, it is possible to cool the circuit area to 4.2 K, while the other parts of the chip are maintained at room temperature.As only a small volume of material is cooled, efficiency is high and cool-down time is less than 1 minute.
This highly localized cooling allows the superconducting circuitry to be in close physical proximity to the input connector, which is necessary to preservewide bandwidth.O n the fused-silica substrate, a 0.8-cm length of gold 5 0 4 coplanarwaveguide transmission line runs between thecircuitareaand theedgeofthechip,whereit isdirectly bonded to a wide-bandwidth coaxial input connector.This connector, o n the front panel of the instrument, is at room temperature, but only 1 cm away from the 4.2 K environment.
The coaxial connector, which provides the widest bandwidth available, is a limiting factor in instrument bandwidth, which is specified at 70 GHz (5 ps rise time).
The ability to self trigger, an important feature of this instrument, is made possible by superconductivity.An internal 350-ps superconducting microstrip delay line placed between the sampling circuits and a Josephson trigger recognizer delays the signal to allow sampling of the portion of the signal producing the trigger.Since this circuitry i s entirely cryogenic, very little time jitter is introduced.At full sensitivity, the millivolt-level 50-GHz cavity resonances excited by a tunnel diode are easily displayed.Self-triggering also allows display of random repetitive events, such as those produced by particle detectors.Furthermore, the effective triggering bandwidth is that of the instrument itself.For example, the unit can trigger and display (amplitude uncalibrated) a sine wave produced by a 100-GHz Gunn source.
An on-chip 5-ps-rise-time Josephson step generator allows the instrument to function as a time-domain reflectometer.In this mode itwas possible to measure, for example, a 2-Q error in the impedance of the glass "bead" of a 46-GHz coaxial-to-microstrip launcher later found to be due to gas bubbles included in the glass.This error could not be resolved with a 40-GHz network analyzer in the timedomain mode.
The wide bandwidth and high sensitivity of Josephson sampling devices, coupled with other attributes of Josephson circuitry, allow sampling measurements that are not currently possible with conventional electronics.These attributes include the inherent low noise produced in cryogenic circuitry, the availabilityof extremely low dispersion transmission lines, and electrically variable pulse-delay circuits with extremely low time jitter [27].As performance increases and new features are incorporated into commercially available measurement systems based o n Josephson sampling, Josephson electronics should facilitate the rapid development of high-performance systems based on state-of-the-art semiconductor and other technologies.

CRYOGENIC CURRENT COMPARATOR
Current comparators have long been made by balancing the magnetic fields generated by multiple coils on a highpermeability core [28].The accuracy of these devices was originally limited to about 1 part in I O ' by the difficulty of making perfectly matched coils.In 1972 Harvey showed that shielding the coils in a superconducting sheath resulted in perfect matching of the coils [29].The original cryogenic current comparator was wound with a composite conductor consisting of a number of insulated strands of superconducting wire surrounded by a superconducting tube.A current in any one of the wires induces an equal and opposite current on the inside surface of the superconducting shield.This image current returns on the outside surface of the shield i n the same direction as the original current.The resulting magnetic field depends only o n the net current flowing inside the shield and iscompletely independent of the geometryof the individual wires.Thus, when the composite conductor is wound into a coil, the individual windings are perfectly matched.
A more convenient method of making these coils i s to wind them together on a single form and then cover them with an overlapped toroidal shield, as shown in Fig. 7. [30], Fig. 7. Arrangement of coils and superconducting shields in cryogenic current cornparator.Inset is a c r m s section showing a double layer of shielding around the internal windings.[31].The shield is insulated attheoverlapand takesthetorm of a "snake swallowing its tail."In this geometry the external field is generated by shield currents induced by the currents in the internal windings.Theoverlap in the shield prevents flux leakage resulting from imperfect matching ot the windings within the shield.This geometry makes it convenient to wind coils that can achieve a very wide range of ratios.A common procedure is to wind a set of coils with turns ratios that allow a self-check as well as the generation of a ratio of specific interest.For example, a comparator t o measure the; = 4 Hall resistance relative to a 100-R standard was made using 66 coils of 109 turns and a single coil of 58 turns [32].The 66 coils were matched against each other by connecting them i n series opposition in a binary buildup sequence.Sixty-four of the 109-turn coils were then connected in series with the 58-turn coil.The ratio between this group and one of the single remaining coils is 64.5321101, which is within a fraction of 1 ppm of the desired value.
I n order to establish an ampere-turns balance, it is nec-essary t o null the external magnetic field produced by the shield currents.Fortunately, superconductivity also provides a very sensitive magnetic field sensor.This is the SQUID, which is described i n detail elsewhere in this issue.TheSQUlD input is normallyconnected toasensecoil consisting ot a few turns of superconducting wire wrapped around the outside of the comparator shield.Typically, the sensitivity of the SQUID null sensor is on the order of 1 nA .turn.It is thus possible to detect ratio errors at the level of 1 part in I O " by using currents of about 100 mA and coils with 1000 or more turns.
INDUCIANCE BOLOMETERS AN11 MICKOLVAVL POLl'tK STANDAKDS Microwave power has traditionally been measured with bolometric detectors.These detectors are favored because their response to absorbed microwave power can be made very nearly equal to their response to absorbed power at direct current or low frequencies.A well-designed bolometer thus provides a means tor using highly refined dc metrology for the measurement of microwave power.
The accuracy of this method is limited primarily by the uncertainty in the determination of the power that is absorbed in the waveguide mounting structure of the bolometer.Since this loss is ordinarily no more than a few percent and can be measured separately by calorimetric methods, measurement systems using bolometers have tor many years been the basis tor the primary standards tor the measurement of microwave power.In recent years there have been increasing demands to improve the accuracyand sensitivityof the primarystandards.Sincedissipation in the bolometer mount is a major source of uncertainty, a substantial improvement in accuracy can be made by constructing the mount with superconducting materials that have very low microwave losses.
Fig. 8 shows the design for a proposed power standard based on this idea.The system has a room-temperature

standard.
Simplified drawing ot propowd microwave power input flange, a short length of normal waveguide, and a superconducting bolometer mount.The intent with this system is to determine the total coherent microwave power going into the room-temperature port.It is anticipated that more than 90% of the power will be absorbed in the bolometer and measured by dc substitution.However, some power will be absorbed i n the normal metal waveguide at the topof the power meter and some powerwill be retlected back out the input port.Since both ot these represent firstorder corrections for the measurement, they must be determined accurately.We believe that the power absorbed in the normal waveguide will be on the order of 1 % of the inci-dent power, but of course it will vary with the waveguide size.
We anticipate that most of the reflected power will come from mismatch atthe load of the bolometer.Our initial concept for the load is to fabricate it on a separate substrate from the thermometer part of the bolometer, since this approach provides greater flexibility in the design of the load.Good thermal contact between the load and the thermometer will, of course, be required.It will be interesting to experiment with tuning structures to reduce the reflection coefficient of the load, but such structures cannot be allowed to introduce significant power losses.A tuning element based on the temperature dependence of the inductance of a superconducting film (similar in concept to the thermometer we discuss below) might be appropriate.The thermometer part of the bolometer is based on the temperature dependence of the kinetic inductance of a superconducting thin film.Current research on this microwave power standard is concentrated on the design and performance of the thermometer [33], which will be the topic of the remainder of this section.radiation absorber and temperature transducer attached to some mass with composite heat capacity C(T).This mass is thermally linked t o a stable-temperature bath through thermal conductance G(T).Consider the entire device to be in equilibrium with the bath at a temperature Tb.When radiation is absorbed by the bolometer, it will cause an increase of its temperature to T. The absorbed power is given by Any mechanism yielding a measurable quantity that is sensitive to changes in temperature may be used as the temperature transducer in the bolometer.
Previous superconducting bolometers used the resistive transition of a superconductor as a temperature sensor [34].These devices produced Johnson noise and could use only very limited bias current because of self-heating.In contrast, the superconducting stripline device is operated below its transition temperature, so it has no Johnson noise, and it can accommodate large sensing currents without selfheating.Also, its very low impedance means that it is ideal as a signal source for a low-noise SQUID amplifier.A theoretical analysis suggests that the Johnson noise and preamplifier noise for an ideal kinetic inductance bolom-eter can be reduced toabout7 x W/&foroperation at 0.35 K [33].The practical limitations of a device designed for only modest sensitivity and operating around 8 K are now being explored.ducting ground plane.Any low-frequency current in the microstripline is essentially confined within a penetration depth h in each of the superconducting films.This current confinement is due to the Meissner effect, which implies that a magnetic field will decay exponentially t o zero in the interior of a superconductor.The currents and magnetic fieldsare negligibleatdistancesgreaterthan Xfrom the film surface.
For W >> to the inductance of the microstripline shown in Fig. 9 is The first term is due to the magnetic field in the dielectric.
The second and third terms include contributions from the kinetic and magnetic energy of the charge carriers in the superconducting films.This inductance depends on the penetration depths& and X , o f the microstripand groundplane films.For temperatures just less than the superconducting critical temperature T,, the penetration depth has a strong dependence on temperature, The temperature dependence of the penetration depth and, hence, of the inductance forms the basis for a new type of thermometer.The strongest dependence of inductance on temperature is obtained if the superconducting films have thicknesses smaller than their penetration depths, the kinetic inductance limit.
A method for sensing changes in the kinetic inductance of a superconducting microstripline, based on a bridge circuit, is shown in Fig. 11 [33], [35].If the four inductors in the figure are structurally identical, their inductances will be equal when they are all at the same temperature.In that case the bridge as a whole is not sensitive to temperature changes.However, the bridge is sensitive t o temperature differences between the inductors.To make a practical differential thermometer, two of the inductors are on silicon islands, which have weak thermal coupling to the remainder of the bridge.Having two such elements ensures that the bridge can be balanced even with the inevitable fab- The heart ot the experiment i s the silicon chip on which the superconducting inductance bridge is tabricated.The four meander lines ot the bridge are superconducting rnicrostriplines.Two ot the meander lines are on thermally isolated sections ot the chip so that their temperatures ran be controlled by adjacent resistive heating elements.The sil- icon chip i s mounted on acopper blockwith heater and thermometer for controlling the temperature of the chip as a whole.
rication variations among the inductors.As the power t o the island heaters is varied, the SQUID galvanometer senses the current imbalance of the bridge.Two of the microstrip lines, labeled 1 and 2 in Fig. 11, are thermally isolated from the remaining circuit by a 15-pmthick membrane of silicon.This geometry i s achieved by first patterning a boron-doped frame around the meander line and then anisotropically etching the back side of the wafer.This procedure leaves a silicon island suspended by a boron-doped silicon membrane [35].
The first experiments to measure the bridge sensitivity at various temperatures used the experimental setup of Fig.
11 [35].The silicon substrate, on which the bridge circuit is fabricated, is mounted on a copper platform using vacuum grease.This platform contains a resistive heater and a commercial germanium resistance thermometer.The SQUIDamplifierandcopper platformaremounted inavacuum can and immersed in a helium bath.The copper platform is thermally linked t o the bath through a flange at the topofthe probevacuum can.An audio-frequencyoscillator is used to bias the bridge circuit.The imbalance current of the superconducting bridgewasamplified bythe RF SQUID and detected with a lock-in amplifier.
The rms noise at null was measured to be about 136 pA with a bridge excitation current of0.39 mA.The ratio of these currents, which we call the depth ot the null, is approximately 0.4 x I O -" .Simple circuit analysis shows that the resolution of the bridge for observing changes in inductance i s directly proportional to the depth ot the null.For our bridge the implied sensitivity t o changes in inductance is about 2.8 pH.This first experiment demonstrates the basic operation of the kinetic inductance bolometer [34].An important question t o beanswered is how large can the bias current for the superconducting bridge be without inducing magnetic flux flow?The current used in this first experiment was just a value that was conveniently available; it is not maximum current in any fundamental physical sense.
In the future we expect to measure the onset of flux flow asafunctionof biascurrent and temperature by monitoring the noise at the galvanometer.

CONCLUSIONS
Superconductivity has made possible significant advances in metrology.This is true not only in such classical areas as voltage and resistance standards, but also in the areas of microwave and high-speed time-domain measurements.I n the past, many measurement techniques using superconductivity were so difficult that their use was confined t o national standards laboratories.Today, the development of reliable superconducting microcircuits is making these techniques available to a wide variety of military and industrial laboratories.The discovery of high-temperature superconductivity opens the possibility that all of these techniques can be used at much more convenient liquid nitrogen temperatures.This will require the development of an integrated-circuit process for high-T, materials similarto thatwhich now existsforconventional superconducting materials.

Fig. 1 .
Fig. 1.I-Vcurve of a Josephson junction driven at 96 C H z .The microwave current through the junction induces constant-voltage steps at multiples of hf/2e = 198 pV.

Fig. 4showsacomplete
Fig.4showsacomplete Josephson array voltage-standard system.The microwave source i s typically a 90-mW Gunn diode oscillator at 96 GHz, which is frequency stabilized by a feedback circuit in the microwave counter.An oscilloscope monitors the array I-V curve using a current sweep

Fig. 3 .
Fig. 3. Several constant-voltage steps generated b y a large series array.

Fig. 6 .
Fig. 6.(a) Schematic diagram of Josephson sampling circuit.(b) Signal and pulse waveforms and dc bias combine to trigger the threshold detector.

Fig. 10 Fig. 10 .
Fig. 10 depicts a superconducting microstrip of width W separated by a dielectric of thickness to from a supercon-

amp 1 i f i e r a u d i o RF souid a t t e n u a t o rFig. 11 .
Fig. 11.Schematic of silicon chip and experimental setup.The heart ot the experiment i s the silicon chip on which the superconducting inductance bridge is tabricated.The four meander lines ot the bridge are superconducting rnicrostriplines.Two ot the meander lines are on thermally isolated sections ot the chip so that their temperatures ran be controlled by adjacent resistive heating elements.The sil- icon chip i s mounted on acopper blockwith heater and thermometer for controlling the temperature of the chip as a whole.
, D. C. McDonald, and J. E. Sauvageau are with the Electromagnetic Technology Division, National Institute of Standards and Technology, Boulder, CO 80303, USA.S. R. Whiteley i s with Hypres, Inc., Elmsford, NY 10523, USA.
I EEE Log Number 8929378.