Ultra Low Power

ELF/VLF Receiver Project

Umran S. Inan





     1. Automatic Geophysical Observatories (AGOs)

     2. Manned Stations

     3. Autonomous Deployment and Interferometer Mode



     1. Background on Stanford Ultra-Low-Power Work

     2. VLF Radio Receiver Development





The ultra-low-power radio receiver will measure extremely low frequency (ELF, 30 Hz to 3 kHz) and 

very low frequency (VLF, 3 kHz to 30 kHz) electromagnetic waves in polar regions. The receiver will 

be compact and low cost enough to allow for deployment of multiple units in a ELF/VLF interferometer 

configuration (e.g., see Figure 6), to determine the spatial location, shape and extent of ionospheric exit 

regions illuminated (from above) by plasma waves of magnetospheric origin. It would provide substantial 

advances in our understanding of the mechanisms of generation and resultant ionospheric and magnetospheric

effects of plasma waves which permeate high latitude polar regions, examples of which are discrete

chorus emissions and wideband auroral hiss emissions discussed briefly in Section C. Interferometric

measurements must be conducted simultaneously at multiple electromagnetically quiet high latitude sites,

located at distances (from each other) of tens to~100 kilometers, depending on the particular application.

Currently available ELF/VLF receiver designs (see Figures 2 and 3) do not allow such measurements,

primarily due to their physical size (requiring expensive shelters) and high power consumption, which

prohibits sufficiently long duration (e.g., a few months or ideally year-round) observations.

While the interferometric measurement is one specific scientific objective that would be uniquely

facilitated by the new instrumentation, the ultra-low-power ELF/VLF receiver to

develop will also bring about major simplifications (and thus cost savings) in logistics and resource

demands for ELF/VLF measurements currently in progress at manned and unmanned polar sites, in

Antarctica as well as in Alaska and other northern polar regions. Once it is developed, the highly

complex and bulky receiver systems currently deployed at these sites can all be replaced with the 

ultra-low-power ELF/VLF receiver with no loss whatsoever in resolution, bandwidth, coverage or other

capabilities, and in fact with substantially improved flexibility of operation and data acquisition modes.

Since the new receiver will be integrated onto a single chip, fabrication of duplicate copies will require

minimal effort. Thus, we fully expect that many other users will use this new receiver to conduct ELF/VLF

observations at sites worldwide, using the flexible DSP capability of the unit to configure its operation

to their particular needs.



Fig. 1.  Ultra-low-powerELF/VLFreceiver system for use in polar regions. Depicted here is a single autonomous system, enclosed in a dewar (for temperature control) and buried ~2 to 3 ft deep in the snow immediately underneath the sensors, which consist of two crossed magnetic loop antennas. The ultra-low-power ELF/VLF receiver system consists of several Application Specific Integrated Circuit (ASIC) chips, which together with associated external circuitry and nonvolatile memory, fits into an enclosure no larger than a single paperback book, and operates continuously while consuming only a total of ~10 milliwatts. At the same time, the new receiver is much lower cost (<$2000 per unit, including antennas and battery), much more robust and easy to duplicate (due to integrated design and much fewer components), and has many times the capability and flexibility of previous systems, in terms of bandwidth and programmable on-board DSP processing power.


The unprecedented properties of the ultra-low-power ELF/VLF receiver are facilitated because

of recent work in the Ultra Low Power group of the Space, Telecommunications and Radioscience

(STAR) Laboratory of Stanford University to develop and test ultra-low-power circuits. Recent demand

for handheld communications devices has driven substantial new research efforts in ultra-low-power

electronics [e.g., Machado, 1996]. The line of research pursued at Stanford involves tunable-threshold,

low voltage CMOS integrated circuits, which consume ~50 to 100 times less power than conventional

low power CMOS electronics, while excelling in functional performance. Ultra-low-power chips, fabricated

in a tunable low threshold CMOS process, are not yet commercially available, but Stanford has

successfully designed and fabricated several chips, including a 32x32 bit ultra-low-power multiplier, a

128x36 bit low-power SRAM, and an image processing chip for motion estimation. Of particular interest

is a 2-GHz single-chip radio receiver RF front-end recently fabricated [Shanaa et al., 1999] in the Ultra

Low Power group of STAR Lab, in preparation for a radio science occultation experiment for a

spacecraft mission to Pluto. The ultra-low-power ELF/VLF chip can thus be uniquely produced

at Stanford by tailoring this existing single chip radio design experience to the ELF/VLF frequency range

and the dynamic range and sensor impedance requirements appropriate for ground-based measurements

in polar regions. While it is conceivable that other groups may undertake similar development, the expertise

for development of integrated radio frequency low noise amplifier systems is currently highly prized

commercially and is thus only available in industrial development laboratories. Our particular specialty

is in the simultaneous optimization of the antenna matching and the noise figure of the receiver, and the

programmability of the filters on the chip to afford configurable performance.



ELF/VLF observations have been a crucial component of ionospheric and magnetospheric research in the

polar regions for the last four decades and continue to be currently conducted at South Pole, Palmer, and

McMurdoStations, Antarctica, as well as five different Automatic (unmanned) Geophysical Observatories

(AGOs) within the context of the U. S. Antarctic Program. In addition, such measurements are also an

integral component of the polar research efforts of other nations, for example the United Kingdom (at

Halley, Antarctica and and at several unmanned observatories), Japan (at Swoya Station, Antarctica),

South Africa (at Sanae, Antarctica), Brazil, and China. Observations in the Arctic are currently in

progress in Sondrestromfjord and Thule, Greenland, as well as at several sites in Alaska.

The ultra-low-power ELF/VLF receiver system will be usable in a diverse range of observations

in the polar regions, including (i) at unmanned Automatic Geophysical Observatories, (ii) at manned

stations, and (iii) as an autonomous system on its own, deployed either singly or in an interferometer

mode at multiple remote sites. The design brings in major science improvements, logistics

simplifications, and cost savings in the context of the first two applications, while uniquely facilitating

the last one. In the following, we discuss each of these applications.


1. Automatic Geophysical Observatories (AGOs)

At present, one of the most compact and power efficient ELF/VLF receiver in use in the polar regions

is the Stanford University ELF/VLF system currently operational at five different unmanned Automatic

Geophysical Observatory sites at the high latitude Antarctic plateau. This receiver system is shown in

Figure 2, and should be contrasted with the ultra-low-power receiver depicted in Figure 1.

Deployment and operation of AGOs such as that shown in Figure 2 is a major logistics undertaking

with the resource demands being largely determined by the physical size and power requirements of

the instruments housed. At present, the five United States AGO units deployed in the Antarctic plateau

provide a total of ~50 W average power, supplied by a thermoelectric generator powered by ten tanks

of propane (each 30 inches in diameter and 53 inches in height and weighing ~700 lbs), flown in every

year by the ski-equipped LC-130 Hercules aircraft (see top left panel of Figure 2). These fuel carrying

flights are the single most costly and difficult part of the annual maintenance of the AGOs, since all other

aspects of the annual maintenance visits can be handled by much smaller Twin Otter aircraft.

The power budget allocation for the ELF/VLF receiver at an AGO site is ~9 W, of which only ~6 to

7 W are actually used. Deployment of an ultra-low-power ELF/VLF receiver would reduce

this requirement to ~10 mW, representing nearly a thousand fold savings in power for the ELF/VLF

system, and facilitating a ~20% reduction in the power requirements for the entire AGO. This means that

fewer propane tanks (8 instead of ten; a savings of 1400 lbs of cargo weight) need to be flown in every

year or that additional experiments can be accommodated in the same AGO. There is also no more need

to have an ELF/VLF receiver unit inside the AGO (like the unit with a blue front panel in Figure 2), since

all of the receiver functionality and A/D conversion will be handled within the ultra-low-power receiver

housing to be buried under the sensors (presumably at some distance of ~500 ft or more from the AGO

to reduce electromagnetic interference), with the data brought back to the AGO either by a serial data

cable or an optical fiber. The ultra-low-power ELF/VLF receiver can either be powered by one or more

batteries (see Section D) or by a low voltage line level cable from the AGO.


Fig. 2. Stanford ELF/VLF receiver at an unmanned AGO site. This system, currently operational at five different remote high latitude Antarctic sites [Salvati et al., 1999], was designed and build in 1990 specifically for use in unmanned sites, with minimal power demands on power. It consists of a magnetic loop antenna (either a 1.7x1.7m or 4.9x4.9m square – both shown in the photo above) connected to a preamplifier (grey box with sealed lid shown in lower left) buried immediately underneath, which is powered by and sends data over a~500 ft coaxial cable (shown rolled up) to the station (shown in upper left), in which resides the main receiver unit (the box with the blue front panel) consisting of a line receiver, various fixed band filter channels, and a broadband snapshot system, as depicted in the block diagram. This ‘low power’ ELF/VLF system specially designed for unmanned sites with limited power nevertheless consumes ~6 to 7W, depending on operational modes. The ultra-low-power ELF/VLF receiver system will provide the same functionality at a power level of ~10 mW, while eliminating the need for the preamp, and the cable, and reducing the physical size of the main receiver unit to a housing as large as a small paperback book, and allowing for substantial enhanced flexibility (e.g., bandwidth and channel selection under software control) of operation.




All of the functionality of the present AGO ELF/VLF receiver unit as described in the block diagram of

Figure 2 will be easily implemented by the DSP processor within the ultra-low-power ELF/VLF

receiver. Additional functionalities that were not implemented in the original AGO ELF/VLF receivers

due to perceived power and data storage limitations at the time (1990) will be facilitated, such as the

simulatenous availability of the wideband waveform from two orthogonal antennas, which can be crosscorrelated

to determine (with ~1. accuracy) direction of arrival of discrete emissions such as ELF/VLF

chorus. Triangulation using the arrival azimuths from two or more sites in turn allows the determination

of the ionospheric exit point of the waves (see Section C).


2. Manned Stations

ELF/VLF observations in the polar regions have been an integral part of geophysical investigations of the

near-earth space environment since the International Geophysical Year (1957). Extensive observations

have been carried out for over three decades at many different Antarctic and Arctic sites. At present,

ELF/VLF receiver systems are in operation at South Pole, Palmer, and McMurdo Stations, Antarctica, at

Sondrestromfjord and Thule Stations in Greenland, and at various sites in Alaska.

The systems at most of these sites consist of two large (~ 30 ft base) orthogonal (typically oriented in

the magnetic north-south (NS) and east-west (EW) directions) triangular loop antennas, a preamplifier

unit buried immediately underneath the antenna (in a container similar to the grey box with sealed lid

shown in lower left of Figure 2), which is powered by and sends data (as an analog broadband voltage)

over a~1000 ft coaxial cable to the station, in which is housed the line receivers, A/D interfaces connected

to a PC and a variety of tape recorders (analog reel-to-reel, BetaMax video recorders, as well as digital

recorders, such as Exabyte tape units and CD writers).


Fig. 3. The cusp lab at South Pole. Parts of the ELF/VLF Receiver equipment currently in use are shown. The PC computer which does the A/D conversion is now shown. Similar systems are also in operation at Palmer and McMurdo Stations, Antarctica, and in Greenland (at Thule and Sondrestromfjord).


Power consumption by passive instruments such as an ELF/VLF receiver at a manned polar station

constitutes a very small fraction of the total power available, much of which goes to the provision of

the working environment and various other support for the station personnel. Thus, on the surface, the

ultra-low-power feature of the proposed receiver would not appear bring about a significant improvement

in logistics from the point of view power consumption. However, a very serious and progressively

deteriorating problem at such sites is the increasing level of electromagnetic interference in the vicinity

of the station, coming about from the rapid expansions of computer usage and other infrastructure at the

sites. As examples, video and computer monitors have scan rates within the VLF range and thus produce

radiative interference, and increased power usage and infrastructure at the stations leads to increased

levels of power line ‘hum’, which often dominates the ELF spectrum.


The ELF/VLF antennas at manned sites are typically deployed at distances of few hundred up to~2000

ft from the station, the maximum length being determined by the need to maintain analog signal integrity

(dispersion, distortion, attenuation, and dynamic range) during long distance transmission over a coaxial

cable. At the present time, the ELF/VLF antennas at both South Pole and Palmer are at maximum distances

from the station, and still suffer very significantly from station-generated electromagnetic interference.

At South Pole, the signal reception on one of the antennas (oriented such that it receives maximum noise

from the station) is now extremely noisy to the point of being nearly unusable.

This emerging practical problem can be alleviated with the use of the ultra-low-power

ELF/VLF receiver, by locating the antennas at distances of ~1 to 3 km (~3000 to 9000 ft) from the

station. Even a 50% increase in distance brings about ~10 dB improvement since the intensity of

the station-produced electromagnetic interference falls of with distance r as r-3. The ultra-low-power

ELF/VLF receiver can either be powered by one or more batteries (see Section D) or by a low voltage

line level cable from the AGO, with the data brought back to the AGO either by a serial data cable or an

optical fiber.


The substantially improved signal-to-noise ratio realized by such use of the ultra-low-power

ELF/VLF receiver at a manned polar station comes about with absolutely no loss of capability or functionality,

since the digital samples of the broadband waveform from both magneic antennas would be

continuously available at the station to process or record in ways as required by the particular scientific

objectives. These digital signals can be received and processed on a PC at the station, or the required

processing can be implemented within the receiver using its DSP capabilities, in which case reduced

data can be transmitted to the station for recording. The latter is not likely at a manned site, since ample

resources for data recording are usually available and the tendency is to record raw (unselected) data.

Other obvious benefits of the use of the ultra-low-power ELF/VLF receiver at a manned site

include the substantial reduction in rack space needed for ELF/VLF equipment at the station (what is

shown in Figure 3 is essentially reduced to a single PC computer to receive digital data from the unit

equipped a CD or DVD writer), substantially increased robustness/reliability, ease of duplication, and

much lower cost (<$2000 per receiver unit), all brought about due to integrated design and much fewer



3. Autonomous Deployment and Interferometer Mode

The ultra-low-power and highly compact properties of the ELF/VLF receiver is likely to revolutionize

this field of research by allowing observations to be carried out at any polar location that is

accessible once a year (with ground transportation, by ship, by small aircraft, or other means) even if no

logistics resources (power, shelter, etc.) are available at that location. The unit will be enclosed

in a hermetically sealed dewar (See Figure 9) and will operate autonomously for an entire year when

powered by a relatievly small, ~10 amp-hour battery (see Section D). The data is written on nonvolatile

flash memory and is downloaded during annual visits for battery replacement. Depending on the application

and the cost of flash memory chips at the time several to tens of GBytes of data can be collected

in a given year. In comparison, an entire year’s worth of ELF/VLF data from an AGO site (Figure 2) is

currently <2 GBytes. Data storage requirements can be further mitigated by well targeted science data

acquisition, and with the use of on-board DSP processing for event-based acquisition.

One obvious application of autonomous deployment of the ultra-low-power ELF/VLF receiver

system is ELF/VLF interferometry, which is described in Section C. This application requires the

deployment of multiple autonomous units, the data from which is acquired in annual visits. The spacing

and configuration of sites depends on particular applications, a few of which are briefly described in the

next section.



The scientific justification of the instrument development is underscored by the fact that

ELF/VLF observations are currently conducted at a host of polar sites, some of which are shown in

Figure 4. The scientific objectives of these observations vary across the full spectrum of ionosospheric

and magnetospheric physics, ranging from investigations lightning-induced precipitation of energetic

radiation belt electrons at low latitudes (e.g., Palmer), to the studies of the microphysics of wave-particle

interactions at middle to subauroral latitudes (e.g., at Halley Bay (HA), A80, and A77), to investigations

of substorms, auroral activity and associated waves and ionospheric effects (at SP, A84, P2, and P3), and

to studies of polar cap phenomena and waves on open field lines at high latitudes (P1,P4, and P6). The

ultra-low-power ELF/VLF receiver will bring about major simplifications and new capabilities

in all of these studies.


Fig. 4. ELF/VLF observation sites. Marked in red are manned and unmanned sites operated by the United States, while those sites marked in green are operated by the British Antarctic Survey. Stanford University ELF/VLF receivers are currently deployed at South Pole (SP) and Palmer (PA), as well as at the AGO sites P1, P2, P3,P4, and P5.


Fig. 5. ELF/VLF chorus emissions observed at P2, Antarctica. The top panels show 2-sec snapshots illustrating the characteristic frequency-time signatures of rising chorus emissions. The lower panels show continous records of signal amplitude in selected narrow bands, illustrating the characteristic morning local time (magnetic noon is approximately 1500 UT) peak in chorus activity. Simultaneous acquisition of the narrowband data and wideband spectral snapshots represent the typical targeted data acquisition strategies that would be used with the ultra-low-power ELF/VLF receiver.



We now provide brief background on one particular type of plasma wave, namely discrete ELF/VLF

chorus emissions. ELF/VLF chorus, named for its charactericstic sequence of repeating, usually rising

and often overlapping coherent tones, ranks as the most intense of all naturally generated plasma wave

emissions observed in the Earth’s magnetosphere, occurring regularly in association with disturbed

magnetospheric conditions and seen frequently in association with microburst electron precipitation

[e.g., Rosenberg et al., 1981] and other auroral activity. The onset time of growth of chorus has in past

work been associated with arrival of magnetic disturbances [Gail and Inan, 1990]. However, the nature

of such association has in general been complex and variable from event to event. More recently, a much

more direct association of chorus onset and turn-off with sudden fluctuations in solar wind dynamic

pressure and southward turnings of the IMF have been observed, both in situ [Lauben et al., 1998]

and on the ground [Salvati et al., 2000]. The ground-based observation is particularly exciting, since

the relationship between the solar wind and this dominant plasma wave emission can be continuously

monitored. Understanding the circumstances of that are conducive to the generation of these intense

emissions is important especially in view of the dominant nature of chorus in terms of its intensity and

its known effects as the driver of energetic electron precipitation, pulsating aurorae, and maybe even the

morningside diffuse aurora [Inan et al., 1992].



One important challenge in ground-based observations of waves of magnetospheric origin is the determination

of the location and shape of ionospheric regions which are illuminated from above which then

serve as the source region for waves propagating to nearby or distant receivers in the earth-ionosphere

waveguide as depicted in Figure 6a. In general, the source region cannot be determined from measurements

at one location but can be determined with simultaneous measurements at multiple sites. Especially

at the present time, with the availability of <1 µsec timing accuracy (GPS) between distant sites, both

relative phase and amplitude can be measured, so that a true determination of the exit region and its extent

can be realized. Ideally, multiple sites should be spaced at distances less than the lower ionospheric height

of ~100 km, possibly in a configuration as shown in Figure 6.



The development and implementation of the single-chip ultra-low-power ELF/VLF receiver

concept has only recently become feasible as a result of work carried out by the Ultra Low Power (ULP)

radioscience group of the STAR Laboratory, under the direction of Professor L. Tyler and Dr. I. Linscott.

This group is completely independent from the Very Low Frequency (VLF) research group headed by

Professor U.S. Inan, which is involved in a wide range of ELF/VLF observations in the high latitude polar

regions. The main scientific thrust of the ULP group has been planetary radioscience studies, specifically

radio occultation studies [Tyler et al., 1987;1992], and the single-chip radio development was pursued

primarily in preparation for an upcoming NASA spacecraft mission to Pluto.


Technological developments pursued for the purpose of this planetary mission, and specifically the

work carried out in the last three years, has serendipitiously facilitated the realization of an ultra-lowpower

ELF/VLF receiver for use in the polar regions of the Earth. The ingredients that bring about this

capability are the technical expertise in RF microelectronics, access to high performance integrated circuit

fabrication technology, and proven technical competence for design and construction of RF integrated

circuits that function as single-chip 2GHzradios. Our program will build upon these ingredients,

to meet the technical challenges in producing an ultra-low-power ELF/VLF radio receiver on a single

chip and to integrate the receiver with a high performance signal processor that can extract selected bands,

cross correlate signals from multiple antennas, and detect critical structure of the frequency dependence

of a VLF signal. In the following, we first provide a brief background of the past Stanford ULP work

and then describe the technical approach and methodology.


1. Background on Stanford Ultra-Low-Power Work

The STAR Laboratory ULP group was organized to bring together related interest and work with the

intent of affording opportunity for a synergism that would spark the innovation of a microminiaturized

receiver for uplink radioscience. The efforts within this group in addition to (i) the single chip radio

project, include (ii) development of ultra low power CMOS circuits, (iii) work on the optimization of

needed on-chip spiral inductors, (iv) theoretical investigation into the stability of small oscillators, and

(v) methods for building microminiaturized stable oscillators. Recent achievements in each of these five

areas are summarized below.


Single Chip Radio Project

The poor performance of single chip radio receivers with traditional heterodyne downconversion is due

for the most part to the high (typically 3 dB) noise figure (defined as NF=10logF, where F is the ratio

of the signal-to-noise ratio at the input versusthat at the output) of the Low Noise Amplifier (LNA),

and of the low power Gilbert cell mixers, typically 14 dB, that could be integrated on-chip. The more

familiar double-balanced mixers, although having a lower noise figure, use much more power and are

thus seen as unsatisfactory. A major milestone achieved during the past year is the development of a

2-GHz radio receiver front-end chip [Shana’a et al., 1999] (see Figure 7), realized by an ingenious and

insightful simultaneous optimization of the minimum possible noise figure for a low noise amplifier

(LNA), in MOS technology, while matching the input impedance of the LNA to an external antenna.

This discovery was based on the realization that noise figure of the LNA would be minimal by choosing

the optimum current density in the LNA transistors. Shana’a et al. [1999] further realized that the

input impedance was governed only by the device size and by making the devices large enough the input

impedance could be exactly matched to 50 ohms, the output impedance of an antenna. Surprisingly,

these design features were apparently not understood within the semiconductor industry. This discovery

indicated that an on-chip LNA could be designed with a noise figure as low as 1 dB, while being matched

to 50 ohms. This method for simultaneous noise optimization and impedance matching was presented

in a paper to the 1999 Solid States Circuits Conference in San Francisco. Within the same year, it was

speculated that the same design methodology could be applied to the Gilbert cell mixer and may result in

as much as an 8 dB reduction in the noise figure of that mixer. This approach led to the implementation of

traditional heterodyne down conversion in a single chip fully integrated 2 GHz receiver. Several months

of design and simulation followed, leading to an LNA design with estimated performance of +14 dB gain,

matched to 50 ohms, and a 1.2 dB (or 50 Kelvin) noise figure. The LNA was integrated with the Gilbert

cell mixer. The estimated noise figure of the mixer as 8 dB, and in combination with the LNA resulted

in an estimated total noise figure for the receiver of just 1.3 dB. This new design made it possible for a

single chip radio to be designed that meets the needs of planetary radioscience missions and is shown in

Figure 7a.


The micro radio receiver front end prototype was tested by first converting to a sub-micron BiCMOS

layout and then fabricating the chip in a premier, 0.25 micron BiCMOS process through a cooperative

agreement negotiated with MAXIM Inc., a manufacturer of RF microelectronics located in the Bay Area.

The fabricated micro-radio prototype chip was delivered in mid-July 1999, and subsequent tests have

demonstrated that the chip is fully functional and meets or exceeds the performance estimated from its

design. A photograph of the chip is shown in Figure 7b. In particular, the LNA of the chip has a gain

of 14 dB, and is matched at 50 ohms input impedance over the frequency range from 1.6 GHz to 1.8

GHz, with a noise figure of 1.2 dB at 1.6 GHz rising only to 1.35 dB at 1.8 GHz. This noise figure is a

world record for LNA performance in this technology and is a remarkable validation of the innovative

insight of our PhD student O. Shana’a for low noise optimization and of the chip design methodology.

After the front end prototype was completed and while the chip was in fabrication, the second stage for a

microreceiver (the IF stage) was designed. The completed design was also laid out in BiCMOS as well

and is currently in fabrication. The experience and expertise acquired in the microreceiver project will

be directly applied to the design and develeopment of a VLF micro receiver with particular emphasis on

incorporating the proven design methodologies to obtain a very low power (i.e, 10 mW), and a very low

noise figure (i.e, 2 dB), ELF/VLF receiver.


Fig. 7. 2-GHz single-chip radio developed at Stanford. (a) Layout of the 2 GHz, Integrated Front End CAD Tool layout specifying the design of an integrated LNA, with a Gilbert Cell mixer at 1.6 GHz. This design is optimized to match the impedance of a 50 ohm input to the LNA. The LNA is further designed to have a current density in transistor gate elements that produces the lowest noise figure for the CMOS technology. This is a noise figure of 1.1 dB. (b) Die Photo of 2 GHz Receiver Front End Photograph of a working chip fabricated at Maxim, Inc., incorporating the design shown in (a). Shown here is the LNA, with a noise figure of 1.2 dB, at 1.6 GHz, and a mixing stage that consists of a Gilbert Cell mixer with a noise figure of 6 dB. The mixer has a gain stage of 10 dB so that the combined noise figure of the receiver is 1.3 dB.



Optimization of On-Chip Spiral Inductors

Spiral planar inductors are an effective means of obtaining on-chip inductance for RF microelelctronic

designs. Spiral inductors have become very popular for this purpose. The Q of spiral inductors are low,

due to the resistivity of the traces and parasitic capacitances. While some design tools are available that

estimate spiral inductor performance, finding optimum choices for factors like number of turns and width

of traces has been limited to trial and error. An electromagnetic model for spiral inductors has recently

been developed by the ULP group [Post, 2000]. These spiral inductor designs are likely to be quite useful

in the post LNA programmable filters which will be required in the ELF/VLF receiver.


Oscillator Stability

A radio receiver requires a stable frequency. A typical frequency reference uses an ovenized crystal,

and together with its control and buffer electronics, the oscillator is relatively large and power hungry.

As we shrink the radio receiver to a single chip, the oscillator stands out as one of the last components

of a single chip radio that needs to be miniaturized. For this purpsoe, we have built close ties with

the Ultra Stable Oscillator Development Group at John Hopkins APL, where excellent capabilities are

available and recent progress in the miniaturization has been made. This association has stimulated our

interest in investigating alternatives for stable oscillators. Thus we have been looking into the behavior of

oscillators to understand the nature and limitations on stability. Two PhD students in the ULP group, Ed

Post and Mitch Oslick have collaborated to discover the role of waveform symmetry in the upconversion

of 1/f noise. The presence of 1/f noise in the waveform limits the ability to know the frequency of

the oscillator and thus its utility as a frequency reference. The students have found that components

of the oscillator waveform that depart from half wave symmetry are responsible for upconverting 1/f

noise into the fundamental frequency of the oscillator. The consequence is that the 1/f character of the

resultant waveform fluctuates the frequency producing the lower bound on the Allen deviation. These

findings are reported in [Post et al., 1998] and are likely to be quite important in our ultra-lowpower

ELF/VLF receiver design, especially in view of the dominance of 1/f noise in the ELF and VLF

frequency ranges [Fraser-Smith and Helliwell, 1985].


Microminiaturized Ultra Stable Oscillators

Our desire to miniaturize the oscillator has led to investigations into alternatives to ovenized crystal oscillators

as frequency references. We have been interested in arrays of coupled oscillators as a potential for

an oscillator that could be implemented on a single chip integrated circuit. Accordingly, arrays of Van der

Pol oscillators have been modeled and a software simulator written to study the collective properties, such

as modes, of the array. Weakly coupled oscillators exhibit a tendency to transition from a disorganized,

uncorrelated state of individual oscillators to a state of collective, phase locked oscillation much the same

way that a liquid undergoes a phase transition as it freezes [Ginzburg and Pustovoit, 1998]. Our simulation

model demonstrates this effect. We are now investigating the relationship between the frequency

distribution width of the individual oscillators, and coupling strengths, to the Luopinov coefficient which

measures the onset of collective oscillation to reveal the role they play in the upconversion of 1/f noise

in the oscillator waveform. A small (~4), group of coupled crystal resonators is an attractive method for

improving stability of a frequency and timing reference in a low power ELF/VLF receiver and will be

evaluated in the design.


2. VLF Radio Receiver Development

A typical ELF/VLF radio receiver first amplifies and filters the signal sensed by the antenna. In view

of the frequency range of interest, being 30 Hz to 30 kHz, no downconverting is needed, although this

can be implemented in cases where only a narrow band is targeted for observation. The baseband (either

original or downconverted signal) is directly sent to the A/D intefrace. An oscillator with good stability

is needed for the frequency references in the receiver as well as the clock reference in the digital interface.

The conditioned signal in the receiver is bandlimited by antialiasing filters with selectable lowpass cutoff

frequency and then sampled by the digital interface. Sampled sequences of the VLF waveform are

transformed by a digital signal processing subsystem into compact, high precision, large dynamic range

representations that are then stored in the instrument in high capacity memories. The configuration of

the receiver, as well as the digital interface, signal processing subsystem and memory store are managed

by an intelligent controller. The controller allows the instrument to be programmable and thus capable

of performing a wide range of ELF/VLF observational scenarios (broadband, narrowband, continuous,

synoptic) to match the scientific requirements of particular applications.


ELF/VLF receiver front end designs have been optimized based on many years of experience, with the

best designs using low noise front ends with discrete components [Paschal, 1988]. Numerous Stanfordbuilt

ELF/VLF receiver preampllifier systems currently in operation are based on this type of a design, and

utilize standardized magnetic loop antennas with 1./1 millihenry input impedances. Different antenna

sizes with the same input impedance are used depending on the sensitivity desired, ranging from 1.7x1.7

m square loops to triangular antennas with 30-ft baselength (known as the IGY antenna). A 4.9x4.9 m

square antenna/preamp combination provides a system sensitivity of ~410-5 V-Hz1/2-m-1, which is

below the atmospheric noise level in the terrestrial environment and is thus sufficient for most applications.

The linear dynamic range in typical ELF/VLF preamplifier systems is ~100 dB.

We will develop a miniaturized integrated ELF/VLF radio receiver that is low power, low

mass, and frequency stable, and consisting of (i) a single chip VLF receiver with an on-chip input LNA,

(ii) a stable frequency reference, and (iii) a microprocessor based control with DSP subsystem. The

connectivity of these four elements is illustrated schematically in Figure 8, below.


Fig. 8. Block diagram of the ultra-low-power ELF/VLF receiver. Schematic combining the single chip receiver with the single chip receiver and frequency reference oscillator and microprocessor and DSP control.



Antenna Impedance Match and Receiver Sensitivity

The ELF/VLF antenna has an input impedance consisting of a 1 mH inductance in series with a 1 .

resistance. Generally, matching the antenna impedance to the input impedance of a low noise amplifier

(LNA) often means sacrificing sensitivity because of the difficulty of simultaneously optimizing both the

matching impedance and the noise figure of the amplifier. For example, the minimum noise figure of

a good LNA might be 1 dB, with the LNA at room temperature and matched to an impedance that has

the lowest noise figure. When matched instead to the impedance of an antenna, the LNA noise figure

typically increases to 3 dB. However, discussed in the previous section, the Stanford ULP group has had

recent success at RF frequencies in simultaneously optimizing the LNA for both matching the antenna

impedance and minimizing the noise figure for a single chip receiver. In this case, antenna impedance

depends on device size while noise figure was discovered to depend on device current density. Since

device size and device current density are independant characteristics in an integrated circuit, they can be

simultaneously optimized. When this optimization was performed, a noise figure of 1.1 dB was obtained

for an LNA in a BiCMOs integrated circuit that was closely matched to the antenna impedance, with the

design shown in Figure 7.


At VLF frequencies, good receivers have a noise figure close to 3 dB, while the LNA in those receivers

have a noise figure closer to 1 dB. A similar opportunity is to reduce the receiver noise

figure by optimizing the noise figure and the impedance match. We will to integrate the VLF receiver

front end components in a single CMOS chip. By integrating the receiver components, the LNA will

be implemented in CMOS. Although CMOS devices tend to have higher noise figures than, for example

devices in bipolar technology, the trend to small device size offsets this difference to where CMOS

devices now have noise figures of ~2 dB. Thus, integrating an ELF/VLF receiver in CMOS can produce

performance that is at least as good as the current discrete component implementations.

We further will investigate the attractive prospect that the receiver noise figure can be improved

through the process of simultaneous optimization of the impedance match and noise figure minimization.

New elements appear in adapting the method from RF to the VLF and promise insight into the behavior

of low frequency amplifiers. For example, the low, almost purely reactive impedance of the VLF antenna,

together with 1/f character of noise at VLF frequencies differs from the 50 ohm, resistive impedance

and white noise distribution at higher radio frequencies. We will incorporate these differences in a

fresh approach that simultaneously optimizes the impedance matching while minimizing the noise figure

of the receiver.


Stable Time and Frequency Reference

The VLF receiver incorporates a miniature GPS receiver as the source of a stable time and

frequency reference. The power and size of GPS receivers has been shrinking rapidly, however a large

gap still remains between the ~100 mW chips now available and the ~10 mW goal of the

ultra-low-power ELF/VLF receiver. To mitigate this discrepancy, we are to achieve the ~1 µsec

time resolution needed in an ELF/VLF receiver (note that the maximum bandwidth is 100 kHz

so that 1 µsec resolution is well enough) by using the GPS receiver only occasionally, for example once or

twice per day, to synchronize a thermally isolated crystal resonator which is then the source for a timing

generator and frequency reference. A thermally isolated, good quality crystal resonator has an Allen

Deviation of typically better than 10-10, for 1000 to 10,000 sec, which is about ~1 µsec/day. The drift

due to aging is larger, about ~10 µsec/day. By synchronizing to GPS once or twice per day, the effects of

drift and variation could easily be compensated to the precision needed by the ELF/VLF receiver. GPS

satellite lock is achieved generally in at most ~1 minute, so that the duty cycle of the GPS receiver would

be 2 minutes every 24 hours, so that a ~100 mW GPS card would only consume an average power of

~0.1 mW, well within the 10 mW average consumption of the ELF/VLF receiver system.


Data Storage

The technology in FLASH memory has been leapfrogging its capacities with no sign of slowing down.

Present, commercially available memories have up to 64 Mbyte capacities, (see e.g. 16Mx8, Samsung

K9F2808U0M, 32Mx8 Toshiba TC58256FT, and 64Mx8 Toshiba TC58512FT) and a handful of these

chips would have the capacity to store 1 Gbyte of data. The great advantage of FLASH memory is

retention at zero power. The power used by these memories in writing to them is dependent on data

rates, so that for the small few kHz bandwidths of ELF/VLF applications only ~1 mW is needed. At

substantially higher data rates some compaction, or data compression would be necesssary. As part of

our development we will evaluate methods of compression and data abstraction using the DSP

capabilities of the receiver to optimize the power allocation between the signal processing and memory



When used at AGOs or manned stations, the data from the ultra-low-power receiver can be brought

back to the station over a serial data line or an optical fiber as discussed in Sections B.1 and B.2. Our

preliminary investigations of optical fiber links indicate that translators between twisted pair and ~2 km

fiber optic link are commercially available (e.g., Telebyte Model 373) and that the relatively low data

transmission rates (<1 Mbits/s) required for the ELF/VLF aplication in hand can be easily accomodated

within a 10 mW power budget. As we investigate these possibilities, we expect to significantly benefit

from discussions with Professor Leonid Kazovsky of Star Lab, an expert in optical fiber technology.


Digital Signal Processors

The announcement of comercially available very low power DSPs, (see e.g. TI TMS320LC545, and

TMS320C55x core), with power vs performancee in the range of 0.25 mW/MHz down to 0.05 mW/MHz,

opens the door to using DSP in a miniaturized VLF receiver operating at 10 mW. In addition, work in

the ULP group in STAR Lab has produced an ultra-low-power FFT processor with the lowest power

benchmark which has already been fabricated and is referred to as the SPIFFIE chip [Baas, 1999].

SPIFFIE is a 1024-point single-chip FFT processor designed for low power operation, operating at a

clock rate of 80 MHz, providing a 1024-point FFT every 60 microseconds while consuming only ~8

mW. This performance represents an energy efficiency of over 75 times better than any previous FFT

processor. By evaluating both commercially availiable DSP units and the continued development of ULP

DSP systems a digital signal processing capability can be included in the ELF/VLF receiver

within the power/performance goals.


Thermal Insulation of the Receiver System

The ultra low power consumption of the ELF/VLF receiver provides a particular challenge

when deployed for use in cold polar region environments (e.g., the Antarctic plateau) where the outside

temperature may vary between -80.C to +30. C. ELF/VLF preamplifiers currently in use (see the gray

box in Figure 2) are typically buried in ~2 ft of snow, so that they remain nearly uniformly at about

-50. C but are kept at acceptable temperatures (-10.C to +20. C) by the heat dissipation of their own

electronics. At a power level of only ~10 mW, the system cannot be kept in this temperature

range, without a more sophisticated thermal enclosure.


In this connection, we have consulted Professor R. Moffat of the Mechanical Engineering department

of Stanford, who is a world reknown expert in heat transfer, insulation and electronics cooling. A

preliminary design of a thermal insulation dewar by Prof. Moffat is shown in Figure 9 and

briefly described below.


The system consists of two small, rectangular, stainless steel capsules, one inside the other. The inner

one containing the ELF/VLF receiver and nonvolatile (flash) memory boards. Both are hermetically

sealed. Lead wires (6 are shown) stick out of one end of the inner capsule through gas tight seals. This

is important because they have to hold a fairly good vacuum in the space around this capsule. Any

volatile products that outgas from the receiver or its extension wires (or their insulation) will rapidly

raise the pressure in the supposed vacuum space. An increase in pressure will increase the heat loss to

an unacceptable level. The recommended pressure is 310-4 Torr in the chamber. The inner container

is suspended inside the outer capsule by its own lead wires, aided by some sort of a spider like, low

conductivity structure that is slipped into the system. It should not be pressed tightly in, since that would

increase its heat loss capability. Gold foil of at least 50 µm thickness should be wrapped loosely around

the inner container, and also fitted loosely around the inside of the outer pressure vessel. A third layer

of foil should be "built in" to the spider-like positioning and supporting structure. These three layers of

gold will reduce the radiant heat loss to a very low value. If the pressure inside the pressure vessel can be

kept below 310-4 Torr, the total heat loss should be less than ~10 mW watt in a -50. C environment,

and thus the VLF receivercan be kept at a temperature in the range of -10. C to +20. C. The pressure

inside the inner container is not important, hovever it must not leak. To acommodate a small amount of

outgassing a ’getter’ will be mounted in the vacuum space.


The dewar design will be further optimized during the development effort, which Professor

Moffat will participate as a Co-Investigator. Several dewar designs will be tested in the thermal insulation

laboratory facilities of Prof. Moffat’s research group, with an eye on the desire to keep the cost of the

dewar low, so that the total cost of an ultra-low-power ELF/VLF receiver together with antennas, battery,

and the dewar is ultimately <$2000.


Fig. 9. Thermal isolation dewar for the ultra-low-power ELF/VLF Receiver. Design consists of two hermetically sealed containers. Inner container encloses the VLF receiver pc boards, and is suspended inside the outer container by a spider filament network. One, or more layers of gold foil line the inside of the outer container to increase the thermal insulation. The outer container is hermetically sealed as well and is evacuated to a pressure of 10-4 torr. Six vacuum sealed feed-throughs allow electrical connection to the inside VLF receiv er electronics.




With a total continuous power consumption of only ~10 mW, the ELF/VLF receiver system

can be ideally powered by batteries, even when used at an AGO or manned site. For this purpose, we

are to investigate the potential use of batteries produced by Hawker, Inc., which are currently used

in Antarctica. Our preliminary investigation indicates that their Cyclon series of the Hawker batteries

works down to -65. C. To supply ~10 mW for one year, we need to more than 10 amp-hours. The

battery performance at -50. C is equal to 25% of its room temperature capacity. The Hawker J-cell

specifications, in the Cyclon series, is 8 A-hr, at 2 V, in an approximately 5 cm diameter, 14 cm high,

rigid plastic container that weighs 840 gm. To run a 10 mW, VLF receiver at 2V, for a year, would require

16 to 20 J-cells in parallel, assuming the batteries were buried in the snow at an ambient temberatureof

-50 C. The batteries would then weigh ~17 kg. The dewar containing the VLF receiver could be placed

inside the battery cluster to provide an a bit of extra thermal protection and simple access.

                Last Updated: June 2001.