Cryorefrigerator System Design and Test Results with COSL GatesC. J. Fourie and U. Büttner

Department of Electrical and Electronic Engineering, Stellenbosch University, Stellenbosch, South Africa

ABSTRACT: We present a cryorefrigerator system design for the testing of niobium-based COSL and RSFQ digital circuits at 4 Kelvin, and report observed output signals for COSL circuits. Thermal design, noise reduction and effective electronic interfacing are discussed, and although logic errors are observed, we show that the system allows easy readout of superconductive digital electronic signals to room temperature electronics at low frequencies.

IntroductionExamples of cryorefrigerator-cooled superconductive digital electronics (SDE) have been dem-onstrated [1]. However, cryocooling without liquid helium still remains an obstacle to the deploy-ment of SDE in industrial systems, primarily due to thermal and electrical noise. Furthermore, Complementary Output Switching Logic (COSL) [2] has not previously been demonstrated to function in a cryorefrigerator.

We present a cryorefrigerator system for the testing of COSL and RSFQ circuits, and highlight important design issues. A semiconductor controller is also presented, which allows testing au-tomation, as well as COSL outputs to be measured by computer. We also show test results.

References

Conclusion

A. M. Leese de Escobar, E. Wong, D. Gupta, A. Kirichenko, “Results from the first tests of a superconductin ADC integrated onto a cryo-1. cooler function out of its lab of origin,” Appl. Superconduct. Conf., Seattle, 2006.W. J. Perold, M. Jeffery, Z. Wang and T. van Duzer, “Complementary output switching logic - A new superconducting voltage-state logic 2. family,” IEEE Trans. Appl. Superconduct., vol 6, no 3, pp 125-131, September 1996.C. J. Foure, W. J. Perold, “A single-clock asynchronous input COSL Set-Reset Flip-Flop and SFQ to voltage state interface,” 3. IEEE Trans. Appl. Superconduct., vol 15, no 2, pp 263-266, June 2005.

The cryorefrigerator COSL/RSFQ test system is ready for circuit testing. Thermal and noise issues have been solved, but magnetic shielding still needs to be addressed. Low-frequency inputs can reliably be supplied to the device under test, and sub-millivolt outputs measured. However, correct circuit operation still needs to be observed. Furthermore, next-generation sys-tems must address high-frequency interfacing.

Cryorefrigerator setup

Interface electronics (controller)

Test circuits

Results

CRYOREFRIGERATORWe use a two-stage CryoMech ST405 pulse tube refrigerator (PTR) with 0.5 Watt cooling power at 4K (second stage), and a 65 K first stage. See Fig. 1 for a system diagram. Other specifac-tions are:

Double pump vacuum system - Diffusion and turbo molecular pumps used to pull 10• -5 atm vacuum.3 phase, 50 Hz ac compressor unit, water cooled.• Vacuum housing re-engineered to fit semiconductor controller motherboard inside vacuum, and reduce • electrical connections entering the housing.

A low cost (< €200) all-CMOS Atmel AVR-equipped controller was developed to operate from battery supplies. It operates correctly down to 70K, but due to good thermal shielding around the DUT, the controller remains practically at room temperature during testing. The controller is mounted inside the vacuum to reduce electrical noise pickup, as well as the number of wires leaving the vacuum housing (only 5: VCC, VEE, GND, UART-RX and UART-TX). Other specifica-tions are:

Interchangeable input/output boards to allow upgrades.• 12 bipolar current outputs (-25 mA to +25 mA, 1uA output resolution).• 4 differential voltage inputs (-125 mV to +125 mV, 4uV resolution).• 48 bit pattern generation.• Real-time impedance sensing on all outputs (to scan DUT for broken wirebonds).• Supply monitoring and frame capture.• 1 Mbps RS232 interface to computer via optical fibre and RS232-USB converter allows testing on • any USB-port equipped computer with no noise coupling.Frame-based uploads and downloads allow 20 frames per second to be executed, with an effec-• tive 3 kbps input rate into DUT.

CRYOREFRIGERATOR SYSTEMCold finger reaches 2.6 K under test, and die housing 3.3 K.• Thermal oscillation, with no thermal reservoir, less than 50 mK.•

NOISEVacuum pumps add noise - 2-3 mV (peak) spikes measured on outputs from DUT.•

Switching off pumps before testing, when vacuum already good, solves problem.• Plastic vacuum hose from turbo molecular pump to cryorefrigerator should also cut noise.•

Biggest source of noise is cryorefrigerator - 12-15 mV (peak) spikes observed on outputs from DUT when • system running without isolators on helium hoses, but less than 100 uV with isolators.

Remaining noise may originate from dc motor in cryorefrigerator. CryoMech offers motor separation • procedure, but noise already low enough to permit small circuits to operate.Dc motor noise could also be reduced by adding current return paths from first and second stages • to cryorefrigerator vacuum housing - at the expense of increased thermal leakage.

CIRCUITSMeasured results shown in Fig. 8 and Fig. 9. COSL outputs clearly observed with little noise, but incorrect output bits (outputs always trigger with clock) may be result of flux trapping (poor magnetic shielding).

Compressor

(3-phase,H20-cooled)

Isolators

Helium hoses

Cryorefrigeratorvacuum housing

Diffusion pump

Turbo molecular pump

Dc motor

1st stage (65K)

2nd stage (4K)

AC mains

1 atm

Thermal shield

Cold finger

Ferromagnetic shield

Nb die housing/shield

Interface wiring

Brass thermal shield

Controller RS232(electrical)

RS232(optical)

USB(electrical)

Computer

USB-RS232

Optical fibre

Fig. 1: Schematic diagram of cryorefrigerator test setup.

Fig. 2: IC mounted on Nb die holder, wirebonded to circuit board. During testing, all screws are non-magnetic

brass

Fig. 3: DUT mounted on cold finger. Twisted-pair

wires thermally grounded with aluminium duct tape

Fig. 4: Thermal shield and thermally grounded

interface wires.

Fig. 5: Controller board inside vacuum housing.

Fig. 7: System during testing.Fig. 6: Temperatures.

0 4 8 12 16 20 ms-0.5 mV

0

0.5 mV

-5 mA

0

5 mA0

-500 uA

0

500 uA

-200 uA

0

200 uA

OUTPUT

CLOCK

SET

RESET

Fig. 8: COSL SRFF test results.

NOISE REDUCTION, SHIELDING AND THERMAL STABILITYElectrical and thermal noise could cause circuits to malfunction. Thermal losses may also pre-vent the Device Under Test (DUT) from reaching a low enough operating temperature.

Isolators on helium hoses from compressor - eliminate ground loop noise (majority of electrical noise).• Optical fibre data interface to battery-powered controller prevents noise from computer or ac mains sup-• ply from coupling into controller and DUT.All signal wires are unshielded to prevent currents from flowing on the outside of grounded shields and • into DUT, and also because effectively connecting shields to DUT will result in large thermal sinks.Signal wires are twisted pair with ground wire to eleminate ac magnetic noise from coupling into low im-• pedance loops.Outputs use twisted pair wires grounded at DUT, and are measured differentially at controller - thereby • eliminating ground offset effects between controller and DUT.Thermal shields inside vacuum are connected to 1• st stage. Signal wires first grounded thermally to cold finger (4K), then after 100 mm in vacuum, grounded to 35K, and after another 100 mm grounded to 65K shield. Final grounding is to second shield (approximately 110 K). All grounding sections at least 1 inch in length, and aluminium duct tape is used.Ferromagnetic shield outside second thermal shield, but inside vacuum, to attenuate magnetic fields.•

TEST SYSTEMFig. 2 shows an SDE integrated circuit mounted on a Nb die housing, and wirebonded to a circuit board. Non-magnetic screws are used during testing. A Nb cover is clamped over the housing to form a superconducting magnetic shield. Fig. 3 show mounting on the cold finger. Aluminium duct tape is used for thermal grounding. Fig. 4 shows twisted pair signal wires taped to the thermal shield. Fig. 5 shows the controller with signal lines from the DUT connected. The 5-wire electrical interface connected to vacuum housing is also visible. Fig. 6 shows the tem-peratures of the cold finger (A) and the Nb housing of the DUT (B) during testing. Fig. 7 shows the entire system during a test.

The system was tested on a COSL set-reset flip-flop and an SFQ-to-COSL converter manufac-tured with the Hypres 3-um Nb process [3]. COSL output amplitudes should be < 1 mV.

0

0.5 mA

1 mA

0 4 8 12 16 20 ms-0.5 mV

0 mV

0.5 mV

1 mV

-5 mA

0

5 mA

OUTPUT

INPUT

CLOCK

Fig. 9: SFQ-COSL converter test results.