Josephson Arbitrary Waveform Synthesizer (JAWS)

• JAWS is a pulse-driven Josephson voltage synthesizer that uses quantized phase-slip events to generate analog voltage waveforms with quantum accuracy based on the flux quantum.
• It utilizes sigma–delta modulation and multiple drive modalities—including electrical, optical, and cryogenic BiCMOS—to achieve high precision and fast data transfer rates.
• JAWS is instrumental in metrology and calibration, providing quantum-traceable voltage standards for superconducting circuit testing, cryogenic electro‐optic sampling, and advanced waveform synthesis.

Searching arXiv for recent JAWS-related papers to ground the article in published work. Josephson Arbitrary Waveform Synthesizer (JAWS) denotes a class of pulse-driven Josephson voltage synthesizers in which digitally defined sequences of quantized phase-slip events in a series array of Josephson junctions produce analog voltage waveforms whose amplitude scale is anchored to the flux quantum $\Phi_0 = h/(2e)$ and the Josephson constant $K_J = 2e/h$ (Kohopää et al., 5 Sep 2025). In the formulations reported for pulse-driven arrays, each properly shaped pulse advances the phase by $2\pi$, transferring one flux quantum of voltage–time area per junction, so that for an array of $N$ junctions on Shapiro step index $s(t)$ driven at pulse frequency $f_{\text{pulse}}(t)$, the synthesized voltage is written as $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$ (Kohopää et al., 5 Sep 2025). JAWS has been used for quantum-accurate arbitrary waveform generation, precision voltage standards, and, more recently, as a platform for cryogenic optical control and in-situ quantum-traceable microwave metrology (Priyadarshi et al., 2024).

1. Foundational operating principle

The physical basis of JAWS is the ac Josephson effect. The voltage across a Josephson junction is related to the superconducting phase difference $\phi(t)$ by

$V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$

or equivalently

$V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$

with $K_J = 2e/h$0 (Priyadarshi et al., 2024, Kohopää et al., 5 Sep 2025). A $K_J = 2e/h$1 phase slip corresponds to one flux quantum passing through the junction, and integration over the phase-slip interval gives

$K_J = 2e/h$2

for one slip, or $K_J = 2e/h$3 for $K_J = 2e/h$4 slips (Priyadarshi et al., 2024).

Under periodic drive, Shapiro locking quantizes the average voltage. For a single junction,

$K_J = 2e/h$5

and for a series array of $K_J = 2e/h$6 junctions driven on step $K_J = 2e/h$7 at stable frequency $K_J = 2e/h$8,

$K_J = 2e/h$9

(Kohopää et al., 5 Sep 2025, Priyadarshi et al., 2024). The same scaling underlies pulse-density waveform synthesis: a high-rate stream of ultrashort quantized voltage pulses is digitally modulated so that the local pulse density follows a target envelope $2\pi$0, and after low-pass filtering the time-averaged output approximates

$2\pi$1

where $2\pi$2 is the normalized $2\pi$3 waveform (Priyadarshi et al., 2024).

This construction makes JAWS a quantum-accurate digital-to-analog synthesizer in a strict metrological sense, because the voltage scale derives from $2\pi$4 and the clock frequency rather than from an analog transfer coefficient (Kohopää et al., 5 Sep 2025). The integral of the output is exactly the sum of flux quanta when digital pulse counting and locking are maintained (Priyadarshi et al., 2024). A plausible implication is that waveform fidelity is determined jointly by phase-lock robustness, pulse integrity, and the low-pass reconstruction chain rather than by conventional DAC element matching.

2. Junction dynamics, damping, and array architectures

The junction dynamics are described in the reported works by the RCSJ framework. For a shunted junction driven by an optical-induced current, the governing equation is

$2\pi$5

with

$2\pi$6

(Kohopää et al., 5 Sep 2025). In the overdamped limit, the capacitive term becomes negligible and the model reduces to RSJ-like behavior (Kohopää et al., 5 Sep 2025).

A key design parameter is the characteristic frequency

$2\pi$7

together with the Stewart–McCumber parameter

$2\pi$8

For $2\pi$9, the junction is overdamped, which favors clean and fast Shapiro response, suppresses ringing, stabilizes locking, and reduces timing jitter (Kohopää et al., 5 Sep 2025). This overdamped regime appears in several implementations summarized in the source material: externally shunted Nb–AlO$N$0–Nb SIS junctions fabricated with the SWAPS process (Kohopää et al., 5 Sep 2025), and Nb–NbSi–Nb or Nb/NbSi$N$1/Nb SNS junction arrays used in cryogenic EOS calibration and cryogenic BiCMOS drive experiments (Priyadarshi et al., 2024, Kudabay et al., 23 Dec 2025).

The reported architectures differ substantially in scale and embedding. In the optically driven 2025 demonstration, the array consists of $N$2 externally shunted SIS junctions with $N$3 and shunt resistance $N$4, giving $N$5 and total array resistance $N$6 (Kohopää et al., 5 Sep 2025). In the cryogenic EOS work, JAWS chips with 3,000 and 14,000 junctions are used for calibration and direct waveform capture at 4 K, while a separate embedded 100-junction array is used for reflection studies (Priyadarshi et al., 2024). In the cryogenic BiCMOS integration study, the implemented arrays contain $N$7 and $N$8 overdamped SNS junctions in a 50 $N$9 CPW, with typical normal resistance $s(t)$0, characteristic voltage $s(t)$1, and characteristic frequency $s(t)$2 (Kudabay et al., 23 Dec 2025).

The dependence of system behavior on the ratio of pulse separation $s(t)$3 to the characteristic time $s(t)$4 is made explicit in the optical pulse-drive study. For the 1000-junction low-$s(t)$5 array with $s(t)$6 and $s(t)$7, $s(t)$8; when $s(t)$9, only even plateaus are observed, while for $f_{\text{pulse}}(t)$0 odd and even plateaus both appear (Nissila et al., 2020). The same qualitative criterion recurs in the 2025 optical flip-chip experiment through the double-pulse crossover used to identify the maximum reliable data-transfer rate (Kohopää et al., 5 Sep 2025).

3. Pulse encoding, $f_{\text{pulse}}(t)$1 modulation, and waveform synthesis

JAWS synthesizes analog waveforms by controlling the temporal density and sign of quantized pulse events. In the EOS study, the pulse-density description is explicit: a high-rate “clock” stream of ultrashort, quantized voltage pulses is digitally modulated so that the local pulse rate follows a desired envelope $f_{\text{pulse}}(t)$2, and low-pass filtering recovers the waveform average (Priyadarshi et al., 2024). The notation “$f_{\text{pulse}}(t)$3” in the figures there denotes the sigma–delta modulator that converts target waveforms into pulse density (Priyadarshi et al., 2024). In the cryogenic BiCMOS implementation, this is written as

$f_{\text{pulse}}(t)$4

and

$f_{\text{pulse}}(t)$5

with a target sequence length $f_{\text{pulse}}(t)$6 bits (Kudabay et al., 23 Dec 2025).

Several normalizations recur across the cited works. In the optical-drive studies, the array voltage is normalized as

$f_{\text{pulse}}(t)$7

or, equivalently in the mode-locked-laser experiment,

$f_{\text{pulse}}(t)$8

so that Shapiro plateaus increment by one in the normalized axis (Kohopää et al., 5 Sep 2025, Nissila et al., 2020). For the 2025 optical flip-chip system, one Shapiro step at $f_{\text{pulse}}(t)$9 corresponds to

$V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$0

as reported (Kohopää et al., 5 Sep 2025). In the 2020 mode-locked-laser experiment, a plateau at $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$1 for $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$2 and $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$3 corresponds to approximately $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$4 (Nissila et al., 2020).

The source material distinguishes between the internal pulse rate and the smoothed analog output bandwidth. The EOS paper demonstrates JAWS-generated single-tone and two-tone outputs at 1 and 2 MHz for calibration and direct time-domain capture, while noting that the device technology is compatible with much higher internal pulse rates and that the analog bandwidth in that work is set by the modulation and low-pass filter cutoff (Priyadarshi et al., 2024). By contrast, the optical and cryogenic BiCMOS studies focus on the delivery of high-speed pulse data into the junction array, rather than on low-frequency envelope reconstruction per se (Kohopää et al., 5 Sep 2025, Kudabay et al., 23 Dec 2025).

A common misconception is to identify JAWS only with low-frequency metrological tone generation. The cited studies show that this is incomplete: the same quantized phase-slip mechanism underlies low-frequency quantum-traceable reference generation, high-rate optical pulse delivery into Josephson arrays, and return-to-zero pulse driving from a cryogenic serializer (Priyadarshi et al., 2024, Kohopää et al., 5 Sep 2025, Kudabay et al., 23 Dec 2025).

4. Drive modalities: electrical, optical, and cryogenic integrated sources

Three drive modalities are explicitly documented in the supplied literature.

The first is room-temperature or electrically generated pulse drive, which remains the basis for EOS calibration in the 2024 work. There, JAWS arrays at 4 K are driven by electrical pulse sequences implementing $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$5 pulse-density modulation to generate 1 MHz and 1+2 MHz outputs, with the 3,000-junction array producing a single-tone output of $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$6 at 1 MHz (Priyadarshi et al., 2024). The same work states that the single-tone output exhibits a single spectral line at 1 MHz, while quantitative SFDR and THD are not reported (Priyadarshi et al., 2024).

The second is optical pulse drive using a mode-locked laser and a photodiode located several centimeters from the JJA. In the 2020 study, a custom ring-cavity mode-locked laser centered at 1335 nm and adjustable between 2.0 and 2.5 GHz is used at $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$7, with transform-limited optical pulses of approximately 6 ps FWHM broadening to approximately 25 ps electrical pulses after the photodiode (Nissila et al., 2020). An optical time-division multiplexer multiplies the frequency by 4, and a polarization-based splitter doubles the effective pulse frequency again, so that the maximum achievable pulse rate in that optical chain is $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$8 (Nissila et al., 2020). The photodiode is a commercial InGaAs p-i-n device, biased at 5 V and operated in liquid helium at 4.2 K (Nissila et al., 2020). This proof-of-concept demonstrates multi-step quantization and motivates improved photodiode–JJA integration by flip-chip methods (Nissila et al., 2020).

The third is direct optical data transfer into a Josephson junction array using flip-chip photodiode integration. The 2025 work employs a commercial mode-locked frequency comb at $V(t)=N\,s(t)\,\Phi_0\,f_{\text{pulse}}(t)$9 with repetition rate $\phi(t)$0, optical pulses dispersion-managed to approximately 2 ps FWHM, and amplitude modulation by an EOM (Kohopää et al., 5 Sep 2025). Double pulses are generated by splitting each pulse into two equal-energy pulses with orthogonal polarizations and a controllable delay $\phi(t)$1, then recombining them into a polarization-maintaining single-mode fiber (Kohopää et al., 5 Sep 2025). The photodiode is an Albis ultrafast InGaAs device with nominal 60 GHz bandwidth, reverse-biased at 1 V and flip-chip bonded directly onto the JAWS chip, which operates at 4.2 K (Kohopää et al., 5 Sep 2025). Optical pulses generate carriers in the photodiode, which in turn produce ultrafast current pulses into the CPW and JJA (Kohopää et al., 5 Sep 2025). The work emphasizes that optical fibers drastically reduce passive thermal load compared with cryogenic coaxial cables (Kohopää et al., 5 Sep 2025).

A fourth, electrically distinct modality appears in the 2025 cryogenic BiCMOS integration study. There, a 0.13 $\phi(t)$2m SiGe BiCMOS pulse pattern generator at 4 K generates high-speed return-to-zero bipolar pulses via a 16:1 serializer and modulator (Kudabay et al., 23 Dec 2025). The circuit delivers data rates of 30 Gb/s while consuming 302 mW at 4 K in joint operation with the JJA; the full chip footprint is $\phi(t)$3, and the serializer core area is $\phi(t)$4 (Kudabay et al., 23 Dec 2025). Open eye diagrams were measured at 30 Gb/s at room temperature and 4 K, and despite overshoot and ringing at cryogenic temperature the serializer maintained sufficient integrity to drive the arrays and produce large, flat Shapiro steps (Kudabay et al., 23 Dec 2025).

These studies collectively show that JAWS is not tied to a single source technology. Electrical pulse-density modulation, room-temperature optical generation with cryogenic photodetection, direct flip-chip optical injection, and integrated cryogenic semiconductor pulse generation all preserve the same quantized relation between pulse events and average voltage (Priyadarshi et al., 2024, Nissila et al., 2020, Kohopää et al., 5 Sep 2025, Kudabay et al., 23 Dec 2025).

5. Measurement methods, calibration, and demonstrated performance

A major methodological development is the use of JAWS as a quantum-traceable calibration source for cryogenic electro-optic sampling. In the EOS setup, the measured signal is modeled as

$\phi(t)$5

or in frequency domain

$\phi(t)$6

where $\phi(t)$7 is the electro-optic transduction coefficient and $\phi(t)$8 is the impulse response of the detection chain (Priyadarshi et al., 2024). A 1 MHz JAWS reference tone with known peak-to-peak voltage $\phi(t)$9 is used to define the calibration factor

$V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$0

thereby making the EOS voltage axis traceable to $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$1 via JAWS (Priyadarshi et al., 2024). The same work reports direct EOS captures of single-tone and two-tone JAWS outputs at 4 K and time-domain measurement of the $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$2 scattering parameter of a superconducting transmission line (Priyadarshi et al., 2024).

In the optically driven 2025 study, performance is assessed with a double-pulse method. At large pulse separation $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$3, the array resolves two pulses as separate events per period and only even plateaus in $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$4 appear; at $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$5, the two pulses overlap and all integer plateaus appear (Kohopää et al., 5 Sep 2025). The crossover identifies the minimum resolvable separation and thus the maximum reliable data rate. The reported observations are: at $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$6 the pulses are indistinguishable and odd steps are prominent; at $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$7 the system is in crossover; at $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$8 odd plateaus are negligible and pulses are reliably distinguished (Kohopää et al., 5 Sep 2025). Beyond $V(t)=\frac{\hbar}{2e}\frac{d\phi}{dt},$9, odd-plateau widths become nonessential, yielding a maximum reliable data-transfer rate

$V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$0

which is reported as about a factor of 4 higher than typical JAWS with electrical drive, approximately 15 Gbit/s (Kohopää et al., 5 Sep 2025).

The 2020 optical-drive work uses pulse pairs with separations $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$1 and $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$2 at a period of approximately 430–435 ps. For $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$3, only even plateaus are observed; for $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$4, odd and even plateaus appear; and for intermediate separations both odd and even plateaus occur, with experimental odd-step widths wider than expected from a simple single-junction model (Nissila et al., 2020). The authors map $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$5 and report sharp transitions between steps, consistent with low drive noise, although timing jitter and amplitude noise are not quantified (Nissila et al., 2020).

The cryogenic BiCMOS work demonstrates pronounced Shapiro steps for $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$6 at 10, 20, and 30 Gb/s, corresponding to RTZ pulse repetition rates of 5, 10, and 15 GHz, and for $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$7 at 12, 16, 20, and 24 Gb/s, corresponding to 6, 8, 10, and 12 GHz (Kudabay et al., 23 Dec 2025). Using

$V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$8

the reported example voltages are approximately 1.55, 3.10, and 4.65 mV for the 150-junction array at 10, 20, and 30 Gb/s, and approximately 0.093, 0.124, 0.155, and 0.186 V for the 7494-junction array at 12, 16, 20, and 24 Gb/s (Kudabay et al., 23 Dec 2025). The work also reports clear observation of half-integer steps, attributed to interaction of the incident microwave pulse train with intrinsic phase oscillations in overdamped SNS junctions (Kudabay et al., 23 Dec 2025).

6. Non-idealities, engineering constraints, and prospective developments

Across the cited studies, the central technical challenge is not the Josephson quantization law itself but the delivery of pulses with sufficient fidelity to maintain stable step locking.

Transmission-line non-idealities and reflections are repeatedly identified. In the 2020 optical-drive study, significant reflections arise from nonideal impedance at SMA interfaces, CPW transitions, and the high-impedance photodiode reflection; simulations with a 10 mm passive line between photodiode and JJA show strong reflections and background current fluctuations at the first and last junction comparable to $V(t)=\frac{\Phi_0}{2\pi}\frac{d\phi}{dt},$9 (Nissila et al., 2020). The measured pulse-integral scaling required to match the onset of the $K_J = 2e/h$00 transition is 28% smaller than expected from photodiode current measurements, indicating approximately 30% amplitude loss, likely from reflections (Nissila et al., 2020). In the 2025 flip-chip optical study, damped oscillations with approximately 12 ps period in the odd plateau widths are consistent with reflections or filter-induced resonances, and simplified modeling implicates the first two low-pass filter inductors as a likely source (Kohopää et al., 5 Sep 2025). In the EOS study, a strong bipolar reflection around 140 ps is attributed to the end of the Nb CPW and bond pads or wires, and the embedded-array measurement produces a complex multi-peak pattern attributed to JJA dynamics and impedance mismatch (Priyadarshi et al., 2024).

Filter and packaging design are therefore integral to JAWS performance. The 2025 optical experiment uses low-pass filters that provide DC bias and readout while presenting high impedance to picosecond pulses, and capacitors galvanically isolate CPW grounds near the photodiode to stabilize reverse bias during ultrafast transients (Kohopää et al., 5 Sep 2025). The EOS work likewise notes that the JAWS chip includes additional low-pass filters across the array and that the analog output bandwidth in that study is deliberately limited by filtering (Priyadarshi et al., 2024). A plausible implication is that JAWS design requires simultaneous optimization of quantized junction dynamics and distributed microwave packaging.

Thermal and dissipation considerations also recur. In the optically driven 15-junction system, the small critical current $K_J = 2e/h$01 is chosen to minimize cryogenic dissipation, and the typical dissipated energy per Josephson event in RSJ operation is described as scaling as

$K_J = 2e/h$02

giving approximately $K_J = 2e/h$03 for $K_J = 2e/h$04, while the paper notes that system-level dissipation also includes the photodiode and resistive terminations (Kohopää et al., 5 Sep 2025). In the cryogenic BiCMOS implementation, the corresponding estimate for overdamped SNS arrays is

$K_J = 2e/h$05

with approximately $K_J = 2e/h$06 per junction for the 150-junction array and approximately $K_J = 2e/h$07 per junction for the 7494-junction array, while the serializer itself consumes 302 mW at 4 K in joint operation and up to approximately 600 mW was verified (Kudabay et al., 23 Dec 2025).

Future directions stated in the supplied works are technically consistent. The 2025 optical study proposes pushing beyond 60 Gbit/s by increasing $K_J = 2e/h$08 through optimized shunt resistance while keeping $K_J = 2e/h$09 low enough to maintain $K_J = 2e/h$10, lowering parasitic capacitance, redesigning low-pass filters with full EM/FEM modeling, using cryogenic electro-optic sampling for in-situ characterization, and shortening or eliminating the transmission line between photodiode and JJA through monolithic or co-fabricated integration (Kohopää et al., 5 Sep 2025). The 2020 optical-drive paper recommends placing the photodiode close to the JJA using flip-chip techniques and engineering source impedance and terminations to suppress code-dependent backgrounds (Nissila et al., 2020). The BiCMOS paper anticipates co-located cryogenic SRAM and clock generation, adaptive delay tuning through active synchronization lines, and extension to phase-coherent multi-channel operation (Kudabay et al., 23 Dec 2025).

The application space described in the source material includes quantum voltage metrology, calibration of cryogenic electro-optic sampling, characterization of superconducting transmission lines and embedded Josephson arrays, and control of superconducting quantum circuits (Priyadarshi et al., 2024, Kohopää et al., 5 Sep 2025, Kudabay et al., 23 Dec 2025). In particular, the 2025 optical flip-chip work identifies relevance for superconducting-qubit control through higher update rates for digitally synthesized microwave envelopes and low-dissipation DAC functionality at cryogenic temperature (Kohopää et al., 5 Sep 2025). This suggests that JAWS is evolving from a specialized quantum-voltage-standard architecture into a broader cryogenic waveform-synthesis and measurement platform, while retaining its defining feature: voltage synthesis whose scale is fixed by $K_J = 2e/h$11, $K_J = 2e/h$12, pulse counting, and stable locking.