2142 IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 5, NO. 2, JUNE 1995

An Integrated Digital SQUID Magnetometer with High Sensitivity Input

Masoud Radparvar and Sergey Rylov HYPRES, Inc., 175 Clearbrook Rd., Elmsford, NY 10523

Abstract-A single chip SQUID magnetometer is described that integrates a SQUID-based pre-amplifier with a h i sensitivity comparator gate and feedback circuitries on the same chip. The compprstorgate is an asymmetric SQUID gate drMng two SQUID quantizem in series with the feedlwclt coil. The chip's sensitivity and noise level are primarily determined by the pre-amplifier SQUID. The pick up coil is in series with the feedback transformer. Since the current in the feedback coil is maintained close to zero, the dynamic range of the chip cpll be extremely wide and is independent of the SQUID p"pMi'ir orcmqmrafor architectures. The chip's slew rate Ss determined by the bipolar clock biasing the comparator gate. Clocks running in the tens of MIEZ result in a magnetometer systemwith slew rate exceeding le @As (ao = 2.07Xl(r7 Gauss-cm~. This chip simplifies room temperature electronics and, due to its digital output, can be easily multiplexed on-chip. A system based on this chip can be operated in a relatively h i magnetic field environment without extensive magnetic shielding. The details of the chip as well as preliminary measurement results for the pre- amplifier ps well as the digital circuit will be presented.

I. INTRODUCTION

In order to utilize a SQUID as an amplifier, its periodic transfer characteristic should be linearized by a feedback loop with high open loop gains. This linearization has the added benefit of substantially increasing the dynamic range of the SQUID circuit. The function of the feedback coil is to produce a field which is equal but of opposite polarity to the applied field. To simplify the sophisticated peripheral electronics, various types of digital and single chip SQUID magnetometers have been proposed and demonstrated[ 1-81. Many of these magnetometers either suffer i?om low dynamic range and/or poor energy sensitivity. The small dynamic range is due to the size of the on-chip feedback coil that can only accommodate a limited number of fluxons circulating in the loop. In aprevious publication[9], we demonstrated a single- chip SQUID magnetometer with practically unlimited dynamic range. This was achieved by combining the input and the feedback signals in the same superconducting loop, thus keeping the current in this feedback loop close to zero regardless of the magnitude of the input signal. However, the input SQUID inductance was limited to 1.5 pH, thus limiting the energy Sensitivity of the resulting magnetometer chip. To the best of our knowledge, none of the other single-chip magnetometers have dem-ated energy sensitivity suitable for practical applications. In this paper, we review the design of a high sensitivity single- chip magnetometer that combines the previously developed single-chip magnetometer with a SQUID pre-amplifier. Preliminary measurement results for the pre-amplifier as well as the digital SQUm circuit will be presented.

Manuscript received Oct. 18,1994.

II. CIRCUIT ARCHITECTURE

In order for single-chip (digital) SQUID magnetometers to be o f d a l value, they must, at least, have an energy sensitivity

puts the stringent criteria on the components of the single-chip magnetometer. The pre-amplifier determines the energy sensitivity. However, the compa"s current (or magnetic field) sensitivity should be also adequate to be able to exploit the pre- amplifier's sensitivity. This should be accomplished without sacrificing the margins on the bias c m t of the winparator. To meet these objatives, the SQUID pre-amplifier can be coupled to a DC array SQUID amplifier before integrating it to the comparator SQUID. To improve the sensitivity of the comparator, it may be designed to have a multi-loop SQUID washer to allow coupling ofmulti-turn coils to it. Figure 1 shows the circuit diagram for such a high sensitivity single chip SQUID magnetometer. It consists of an analog SQUID, a DC SQUID amy atnpl ik , a comparator gate and a feedback circuit with two write gates

spproachingtheir CCMII&-P& analog SQUIDS. This requirement

Lp PICKUPCOIL.

0.7

All currents in m.4 Allrabtorr.hOh

x 100

0.75 m0.7S

i=O.3 CLOCKBIAS COMPARATOR SQUID

Fig 1 Circuit d h p m for a very sensitive singlechip SQUID magnetometer with DC SQUID array pre-amplifier and feedback Circuitry. The comparator

. t h e

limit tothe dynamic range forthis type of SQUIDmagnetometer since the input flux is cancelled by the feedback aurent. Dynsmic range is only

of the input analog SQUID. The pre-amplifier d i t was developed and & " h i a t NIST, and fabricated and demonstrated at HYPRES. The pic* up coil is integrated with the feedback loop to improve the sensitivity

be 0.7.

consists of 8 washers to improve its sensitivity by coupling of large inductance to its loop indud" . Thereisnointrkic

limited bythe Current-carryingcapacity ofthe input coil and the sensitivity

ofthe over-all chip. couprig coefficient between inductors is assumedto

1051-8223/95$04.00 0 1995 IEEE

2143

The operation of the pre-amplifier circuit has been discussed in detail by Welty and Martinis in a separate paper[lO]. In this ckuit, an input signal couples flux into the input SQUID, which is voltage-biased with a 0.05 R resistor, so that the SQUID current is modulated by variations in the applied flux. The flux modulation coil of the output SQUID is connected in series with the input SQUID, so that variation in the input SQUID current changes the flux applied to the output array. The series array is biased at a constant current, so the output voltage is modulated by this applied flux. The extemal field sensed by the hgh sensitivity analog SQUID is converted to a current that is applied, through the series array of inductors, to the series DC SQUIDS. These DC SQUIDs have current-voltage characteristics of shunted Josephson junctions, but about 100 times larger dynamic resistance. The series array of DC SQUIDs can generate a DC voltage on the order of milli-volts, that can be even directly ufilized by room temperature electronics and feedback circuitry to read out and apply to the high sensitivity analog SQUID.

The operation of the single-chip magnetometer has been described in a separate paper[9] and briefly, is as follows. The comparator SQUID has an asymmetric threshold characteristic and is biased slightly over its critical current using a bipolar current source. In the absence of any external field, the output voltage is also a bipolar voltage. For a sufficiently large applied magnetic field, the comparator only generates pulses in response to either positive or the negative portion of the applied gate current depending on its polarity. The write gates are designed to have asymmetric threshold characteristics and are biased below their thresholds. The bipolar current induced in the control lines of the write gates will cause the right and the left write gates to cross their lobes only in the positive and negative directions, respectively and produce SFQ pulses upon lobe crossing. When the comparator SQUID pulses negatively, the left write gates launches fluxons into the storage loop. (Alternately, when the comparator pulses positively, the other write gate launches antifluxons into the storage loop.) The injection of only fluxons or only antifluxons would continue in each clock period as long as the gate current of the comparator is below comparator's threshold current for positive or negative currents. With proper polarity, the SFQ-induced current in the superconducting feedback loop can eventually cancel the applied current and restore the comparator SQUID close to its original state. When the current in the feedback loop is close to zero, both write gates, altemately, emit fluxons and antitluxons into the loop in each clock period, keeping the feedback current close to zero. One advantage of this scheme is that the size of the feedback loop can be very small and is actually determined by the desired signal slew rate and SQUID sensitivity. The polarity of the missing pulses determines the direction of the applied field, and the switching probability leads to a voltage across the comparator SQUID which is a measure of the strength of the input signal. The digital output is the difference between the number of negative and positive pulses across the comparator. A multi-bit up/down counter coupled to the output of the one-bit comparator can count the down pulses and subtract these from the up pulses to exhibit the output in digital form.

111. CIRCUIT DESIGN

A typical analog SQUID possesses a flux sensitivity better than 6x1 O-' (PJHz" which is more than adequate for many practical applications. Due to the complexity of single-chip magnetometers and the ease of multiplexing, these chips are best suited for multi- channel biomedical systems, such as encephalograms. Such systems require a field sensitivity of 10 ff /Hz" for an input coil of around 1 pH and pick up coil area of A=l cm2. This gives rise to a needed current sensitivity of 1 pA/Hz". Required dynamic range is determined by interference which has a slew rate of approximately 3 pT/s at 60 Hz for a gradiometer system in a typical environment. The field sensitivity of the input pre- amplifier SQUID (B,) should be better than 10 ff/Hz", or B, = a, n /A c 10 f f k " , where a, is the flux noise of the SQUID and n is the current transfer ratio between the transformer and the SQUID loop inductance (LJ. Assuming #,= 6x107 a&", then an upper bound for the turn ratio is about 830 turns for a pick up coilareaof1cm2. SinceL,-L, - 1 pH=n%,,,thenL,=106/n2 in pH. Typical values for n and L, are 150 turns and 45 pH, respectively.

A typical value for the h-ans-resistance gain for the DC SQUID array amplifier is about 10 kV/A. Thus, a 1 pA/Hz" at the input translates to 10 nV/Hz" at the output. Since the comparator is driven by a current source, the output of the DC SQUID array pre-amplifier should be terminated by a resistor. The output resistance of this pre-amplifier 100 Q . So, when the DC SQUID array is also terminated by 100 Q, it gives rise to a minimum current requirement of 50 pA/Hz" by the comparator. For a system with 500 Hz bandwidth, this current is about 1.1 nA which is the sensitivity needed from the comparator. This current sensitivity is, obviously, considerably higher than the intrinsic noise level of the comparator gate. However, it should also be larger than the comparator's hysteretic current. To ensure the latter, the comparator's loop inductance is made of 8 parallel washers, each coupled to a 10-turn transformer (Fig. 2). All of these transformers are then put in series with the output resistor of the DC SQUID array amplifier. The current transfer ratio of the transformer between the pre-amplifier and the comparator with this arrangement is approximately 80, giving rise to a minimum current requirement of about 88 nA for the comparator.

In order to be able to fully reconstruct the signal from the output of the single-chip magnetometer, the sampling frequency should be at least twice the signal bandwidth or about 1 kHz. On the other hand, a signal slew rate of 3 pT/s or 1.5 x 1OS@.,/s requires a clock frequency of at least 15 MHz to discriminate against spurious 60 Hz signals. At this clock fiequency, the noise floor is about 10 f f k " X (15 W 2 ) " = 27 pT, which is the noise floor of the digital chip. Consequently, the hardware least significant bit (LSB) is 27 pT. Since this scheme uses digital filtering by over-sampling at a rate of 15 W 1 kHz = 15,000, the minimum detectable magnetic field (software LSB) is 27 pT /15000" = 220 f f . The hardware LSB also determines the minimum value for the feedback loop inductance. Given the minimum magnetic field of 27 pT and pick up coil of 1 pH, the maximum current per fluxon in the feedback coil for proper

2144

operation should be I,,, = 2.7 nA. On the other hand, each a,, in the feedback loop corresponds to @J (Li + L,) - 1 nA which is less than I,,,, as required. Decreasing the size of the pick up coil to a smaller area and lower inductance value, will not affect the overall system sensitivity as long as @&it in the feedback coil is less than the 2.7 nA. In this design, the pick up coil is directly integrated in the feedback loop to obtain the desired mini“ current in the feedback loop for each emitted fluxon and/or antifluxon. In addition, 1 nA/bit of feedback current corresponds to about 50 nA ([ 1 nA x 10 kV/A] / 200 Q) at the input of the comparator, which is less than the hysteretic current of the comparator gate, as required.

Fig. 2 Layout sdxmmtic ofthe c u q ” circuit. It consists of 8 parallel washers to -coupling ofthe 10 nH inductor into the SQUID loop. This substantially increases the sensitivity of the comparator gate.

and c8n be improved further by increasing the size of the inductors in the feedback coil. The bottom waveform is the integrated output voltage and can be used to determine the slew rate of the circuit. Figure 5 shows a triangular input signal together with its integrated output. The clipping at the top of the output is due to the hysteresis in the comparator’s characteristics whch can be eliminated by the integration of a suitable pre-amplifier.

I

Fig. 3 &hematic layout ofthe “parator gate, write gates and the pick up coil of the singlechip magnetometer.

IV. EXPERIMENTAL RESULTS

Figure 3 shows a photograph of a fabricated comparator SQUID chip together with the input coil and the write gates without the hgh sensitivity pre-amplifier. In this case, the feedback coil is directly integrated with the comparator SQUID. The comparator also functions as an analog SQUID as well as a one-bit comparator. This chip was fabricated using HYPRES’ standard niobium process technology using 1 kA/cmz Josephson tunnel junctions and junction capacitance of 40 fF/pmz. This circuit process utilizes 10-layers with all niobium electrodes and wiring, aluminum oxide tunnel barrier, MO resistors, Au metallization and SiO, insulating layers[l 11. The digital SQUID, without the pre-amplifier, operated properly as shown in Fig. 4. In this figure, the top trace is the input signal. The middle signal is the voltage across the comparator with the number of missing negative pulses being the measure of the strength of the input signal. The comparator’s sensitivity is approximately 0.1 w i t

Fig. 4 E x p i t m h l results for a single-chip magnetometer (1-bit digital SQUID) without the high sensitivity SQUID pre-amplifier. Top trace is the input signal of approximately 2 p.4 Middle trace is the bipolar signal fiom the cunparator SQUID with 6equency of 4 kHz (2 mV/div.). The number of negative p u k that are missing exhibits the strength of the input signal and can b e d by a multi-bit counter attached to the output of the SQUID. The comparator SQUID is eventually restored to its original state by the f a indudar and pulses negatively and positively when the net current in the loop a p p c h e s zero. M o m trace is the reconstructed output and can be used to determine the slew rate of the single-chip magnetometer at clock fkquency of 4 kHz.

We also fabricated DC SQUID array amplifier chips using our standard niobium process. Figure 6 shows a photograph of part of the SQUID array fabricated at HYPRES using this niobium process. The input analog SQUID stage consists of a low- inductance double-loop SQUID with matched input transformer. Two 4-tum modulation coils are wound (in opposite directions) on the SQUID loops, one on each side, and connected in SerieS[lO]. This double coil is connected to the washer @rimary)

2145

of an input transformer having 40 input turns. The washer inductance is approximately equal to the input inductance of the modulation coil. While the use of a separate input transformer results in significant coupling losses compared to a coil wound directly on the SQUID washer, it allows the SQUID inductance to be made smaller and helps isolate the SQUID oscillation from input resonance. The input coils were made with 2.5 pm (nominal) lines and spaces. The overall dimensions of the SQUID array amplifier with modulation coil are about 800 X 2700 p2. The amplifier circuit uses 3.5 x 3.5 pmZ trilayer Josephson junctions with critical current density of 1 kA/cm2. The DC SQUID array amplifier functioned properly and exhlbited a trans- resistance of 11.25 kVIA which is suitable for'the single-chip magnetometer. Noise measurements of similar amplifier chips evaluated elsewhere have shown to exhibit current sensitivity of 8 p m " r i 2 1 . - -

lines. The comparator SQUID utilizes a multi-washer SQUID with improved sensitivity. The pre-amplifier utilizes a well- developed SQUID array amplifier with excellent current sensitivities. Due to its high dynamic range, this single-chip magnetometer can be also operated in unshielded environments. The slew rate of the system can be improved considerably by operating the comparator with a faster clock (several MHz) or with on-chip clocks running at GHz frequencies. The slew rate is given by QJn x F,, where n is the current transfer ratio of the input trandormer and F, is the clock frequency. Consequently, with ordmry electmnics and current transfer ratio of around 100, slew rates in excess of IO'QJs is easily possible. On-chip Josephson junction-based counters can easily improve the slew rate by two orders of magnitude.

ACKNOWLEDGEMENT

This project was supported by the Department of Energy under contract number DE-FGO2-92-ER8 1407. The authors would like to thank Richard Welty and John Martinis of NIST for providing the design for the pre-amplifier SQUID, and NASA for supportmg the DC SQUID array project (contract number NAS5- 38023).

REFERENCES

[l] J.P. Hurrell D.C. Pridmore-Brown and AH. Silver, "Analog-to-Digital Conversion with Unlatched SQUIDS," IEEE Trans. Electron Devices, vol. ED-27, no. 10, pp. 1887-1896, October 1980. C.A Hamiltas "100 GHz B k Counter Using SQUID Flip Flop%" IEEE Trans. on Mag., vol. MAG-19, no. 3, pp.1291-1292, May 1983. S.V. Rylov, "Measurements of Dynamic Range and Linearity of Flux Quantizing A/D Converters," IEEE Trans. on Applied Superconductivity Vol. 1 no. 3.w.2558-2563. March 1993.

Fig. 5 Top trace is the input signal to the single-chip magnetometer without the pampWier. Bottom trace is the integrated output for clock fresuency of 20 ktEt The horizontal scale is lnddiv. The integrated output is not low- pass filtered and, consequently, shows significance out-of-band and ouantization noise.

[2]

[3]

[41

[51

Fig. 6 Schematic of the part of the DC S Q a D array chip.

V. CONCLUSION

In summary, we have developed components of a high Sensitivity single-chip magnetometer that has a very wide dynamic range limited by the current-carrying capacity of superconducting

F. Kuo, H. .&g, S. Whiteley, and M. Radparvar, "A Superconducting T~acking A/D converter," IEEE Journal of Solid State Circuits, vol. 26, no. 2, pp. 142-145, February 1991. S.V. Rylov "Analysis of High Performance Counter-Type A/D Converters Using RSFQ Logic/Memory Elements", IEEE Trans. on Mag., vol. 27, no.

D. Drung, E. Crocoll, R. Herwig, M. Neuhas, and W. Jutzi, "Measured Performance Parameters of Gradiometers with Digital Output," IEEE Trans. on Mag. vol. MAG-25, no. 2, pp. 1034-1037 March 1989. N. Fujimaki, "Josephson Integrated Circuits 111, A Single-Chip SQUID Magnetometer," Fujitsu Sci. Tech. J., vol27, no. 1 pp.59-83, April 1991. Pemg-Fei Yuh and Sergey Rylov, "An Experimental Digital SQUID with Large Dynamic Range and Low Noise", this proceeding. M. Radparvar, "A Wide Dynamic Range Single-Chip SQUID M a g n e l o " : IEEE Trans. on Applied Superconductivity, vol. 4, pp.87- 9 1, June 94. R.P. Welty and J.M. Martinis, "Two-Staged Integrated SQUID Amplifer with Series Array output': IEEE Trans. on Applied Superconductivity, vol. 3, no.1, pp. 2605-2608, March 1993. L.S. Yu, C.J. Beny, R.E. Drake, K. Li, R.M. Patt, M. Radparvar, S.R Whiteley, and S.M. Faris, "An All-Niobium Eight Level Process for Small and Medium Scale Applications," IEEE Trans. on Mag., vol. MAG-23, no.

B. Cabrera, Stanford Univ., Private Communication.

3, pp.2431-2434, March 1991.

2, pp. 1476-1479, Much 1987.