Monthly Archives: May 2014

Current Trends

Besides research in ground – and foliage-penetrating radar technologies, significant research is also conducted in the development of, so-called, space-time adaptive processing (STAP) algorithms. STAP refers to multidi­mensional adaptive filtering algorithms that simultaneously combine the signals from the elements of an array antenna and the multiple pulses of a coherent radar waveform. STAP can improve the detection of low-velocity targets obscured by mainlobe clutter, detection of targets masked by sidelobe clutter, and detection in combined clutter and jamming environments.

Significant research is also conducted into the use of signal processing tools other than the traditional Fourier transform-based ones for target detection and recognition. Such tools are, for example, based on the theories of, so-called, wavelet-induced multiresolution analyses (WIMA) of signals. A WIMA allows for the decomposition and simultaneous representation of a signal in time and scale and, therefore, is capable of processing signals at different scales. WIMA-based radar target detection and recognition is being actively researched.

Standing-Wave Arrays

For the resonant array, standing-wave fields are created in the waveguide by terminating the waveguide with a metal wall to create a short circuit. The slot locations relative to the short circuit are chosen to maximize the array radiation coupling. For series elements, such as the angled series slot, the slots are located multiples of a half guide wavelength from the short circuit. For shunt elements, such as the offset – shunt slot, the maximum coupling locations are multiples of a half guide wavelength from an open circuit which is achieved by locating a short circuit at one-quarter guide wavelength from the nearest slot. For broadside operation the slots are located in the guide to achieve a uniform phase dis­tribution. A uniform phase distribution is achieved by locat­ing the slots at one guide-wavelength intervals. However this results in an array spacing greater than one free-space wave­length which creates multiple main beams and reduces the antenna gain and performance. The spacing can be reduced to one-half guide-wavelength intervals by alternating the slot offsets to compensate for the 180° phase reversal in the guide.

Circuit Model for Resonant Linear Array. A circuit model of the array provides a good starting point for predicting the array performance and for determining the slot dimensions. The array is modeled by using the lumped-circuit elements for the individual slots, transmission line sections represent­ing the waveguide sections, and a short circuit termination. An offset-shunt-slot linear array and its circuit model are

Figure 7. Waveguide linear array of offset longitudinal shunt-slot radiators and its transmission line circuit model.

d

Go

Gn

Gn

G,

0

Gm P

G Pr

G0 Pm

To achieve an input match for an end-fed standing-wave array, the slot parameters should be chosen so that the slot conductances or impedances sum to one. The slot conduc­tances must also be selected to obtain the amplitude distribu­tion required to achieve the desired radiation pattern. If the slots are all assumed to be resonant and the desired distribu­tion is all equiphase, the ratios of the slot conductances are proportional to the relative slot radiated powers. This gives a set of N equations that can be solved to give the required slot conductances for a shunt-slot array:

Gn ■,nr

pr

-=1r 1

(25)

shown in Fig. 7. The transmission matrix is defined as (25)

V

A

B

‘V+i

Ji_

C

D

M.

The offset-shunt slot is a commonly used element for linear arrays. It offers a wide range of conductance values and linear polarization. The slot offsets and length can be computed us­ing the equations for the offset shunt slot given earlier, or interpolated from a graph, such as Fig. 4.

A section of transmission line of length d with a propagation constant /3 and a characteristic impedance Z0 has a transmis­sion matrix

Array Mutual-Coupling Effects. The simple circuit model de­sign is useful for understanding the slot array or for de­termining a set of initial design values, but it is not accurate enough for most applications. The assumption was made that the slots in the array have the characteristics of an isolated slot. In the array environment the slots are affected by the presence of the other slots, and slots at the edge of the array exhibit behavior that differs significantly from that of slots in the center of the array. If neglected, these effects result in an array where the input reflection coefficient and radiation pattern depart significantly from the desired values. The characteristics of the isolated slot were determined by using a single source of excitation for the slot: the incident waveguide mode. Multiple sources in the array environment excite the slot: the incident field, the fields scattered in the waveguide by the other slots in the guide, and the radiated fields from all the other slots. The characteristics for each slot are influ­enced by all the other slots in the array to some extent, so a slot’s configuration cannot be determined independently. A series of simultaneous equations is required to perform the array design. Elliott (17) has developed a series of design equations that account for the internal and external mutual coupling effects:

cos(pd) sin (pd)

Z0

jZ0 sin(0d)’ cos(pd)

(26)

The transmission matrices for the shunt element with admit­tance Y and series element with impedance Z are

"1

0"

’1

Z

and

Y

1

0

1

(27)

To achieve a satisfactory design, the coupling values for the individual slots must be chosen so that the reflection coeffi­cient at the array input achieves the required value and so that the relative coupling from the slots generates the desired radiated phase and amplitude distribution for the array. The input match to the array and its variation with frequency is modeled by cascading the transmission matrices for the slots and the waveguide sections. At the array resonant frequency the normalized input conductance or resistance for an array of N shunt or series slots reduces to

Ya /G0 _ fP vp vn

(30)

Ya /G0

fnVsVp

2 f2

Y N Y zl. — _

Yr ~ 2-і y_

1 Y0

Ya 1 n

Gn

2 f2 N’ Vs

" +j(P10/k)(k0b)(a/xf

0 m=1 Vn

or

f, 1 , (п/l) cos(Pwln) . (nxn

fn(xn, In) = ——sm (

Zj _ ул Zn

Zo~hZo

(28)

(n/ln )2 — P?0

V a

Г lm/2 Сln/2

n (xn, ln, xm, ^m ) = I cos(nZm flm )

J-l

The input reflection coefficient is given by

_ 1-YJY0 _ZJZ0-1 1 1+YJY, ZJZ0 +1

‘-^m/2

cos (nzn /ln )

-ln/2 e-jk0 R

‘ d2 j2

,3?2+ 0

R

The YaJG0 are the active slot admittances in the array envi­ronment, the Vn, Vn are the slot and the waveguide mode volt­ages, and the gmn terms represent the external mutual-cou­pling effects between the slots. The set of equations is applied iteratively along with the equation for the required input im­pedance to determine the slot offsets and lengths. These equa­tions assume an ideal cosinusoidal distribution in a narrow slot, and may not be accurate enough for some applications. A more thorough analysis, where the field distribution in the slots are unknowns which must be determined, provides a more accurate model of the impact of element mutual cou­pling on array performance (26). The computational require­ments for a rigorous analysis may limit the size of the array for which it is applicable. A good approximation to the final design should be found first by using an approximate method to reduce the number of iterations required.

because the summation of the individual reflections creates a large reflection. The traveling wave generates a linear phase slope across the aperture which corresponds to a scanned beam with the beam pointing direction given by (27)

cos 90 = k/kg — k/2d (33)

for an array with alternating slot offsets. The beam scans with frequency due to the change in the guide wavelength. This frequency-scanning effect is often exploited by making the waveguide path length between the slots greater than the array spacing to enhance the amount of scan for a given fre­quency change (28). The beam position for this case is given by

cos 90 = dgk/dkg — k/2d (34)

(35)

-j2nfid

(36)

г =

Variation of Performance with Frequency. The design achieves the desired input match at the design frequency, but the input match of the linear array degrades as the frequency moves away from resonance. This effect is modeled using the transmission matrices by adjusting the propagation constant in the transmission line sections and including the variation of the slot admittance with frequency. The bandwidth is gen­erally defined as the frequency range over which the input reflection coefficient remains below a specified level. The bandwidth is determined primarily by the number of ele­ments in the array, and it degrades as the number of ele­ments in the array increases. Watson (9) derived an approxi­mate expression for the variation of the input reflection coefficient which neglects the multiple reflections:

2 — £ Yn/Yo — e2Ne dJ^Yn /Y0,

(32)

N

Circuit Model for Nonresonant Arrays. The circuit model for the traveling-wave array is similar to that used for the reso­nant array, but a matched-load termination replaces the ter­minating short, and the line lengths are adjusted to reflect the nonresonant spacing. The initial design is simplified by assuming that the scattering from each slot is small, and hence multiple reflections are neglected. With this assump­tion and a perfect load termination, the reflection coefficient at the waveguide input is given by

where the reflection coefficient for a resonant slot

—2G/Go

2 + G/G0

j2nfi d

Pn —

N

2 + £ Yn /Yo ) ej2Ned + £ Yn /Yoej2ned

Neglecting the slot reflections, the ratios of the slot radiated powers for a traveling-wave array of resonant shunt elements are given by

GJG0 "-1 Gj/G,

П

0 j—1

(37)

Pi

P

П

(38)

The array excitations also change away from the center fre­quency because the slot locations relative to the standing wave in the guide are no longer optimum and the slot radia­tion characteristics also exhibit frequency variation. This causes pattern degradation and main-beam distortion. The ef­fects are modeled by using the circuit model to determine the slot excitations and computing the far-field pattern generated by these excitations.

WDM Network Design Issues

WDM network design involves assigning sufficient re­sources in the network that would meet the projected traf­fic demand. Typically, network design problems consider a static traffic matrix and aim at designing a network that would be optimized based on certain performance metrics. Network design problems employing static traffic matrix are typically formulated as optimization problems. To for­mulate a network design problem as an optimization prob­lem, the inputs to the problem, in addition to a static traf­fic demand, are some specific reguirements, e. g., required network reliability and fault tolerance requirements, net­work performance in terms of blocking, and restoration time when a failure occurs. The objective of the optimiza­tion problem is to find a topology that would minimize the resources, including the number of links and fibers, the number of wavelengths on each fiber, and the number of cross-connect ports, to meet the given requirements. The outputs include the network configuration and the routes and wavelengths that are to be used for source-destination pairs. The network design problem can be formulated as an integer liner programming (ILP) or mixed integer linear – programming (MILP) problem. As the number of vari­ables and constraints can be very large in WDM networks, heuristics are usually used to find solutions faster.

If the traffic pattern in the network is dynamic, i. e., specific traffic is not known a priori, the design problem involves assigning resources based on a certain projected traffic distributions. In case of dynamic traffic, the net­work designer attempts to quantify certain performance metrics in the network based on the distribution of the traffic. The most commonly used metric in evaluating a network under dynamic traffic pattern is blocking proba­bility. The blocking probability is computed as the ratio of number ofrequests that cannot be assigned a connection to the total number of requests. With this metric, one makes decisions on the amount of resources that are needed to be deployed in a network, the operational policies such as routing and wavelength assignment algorithms, and call acceptance criteria.

MICROELECTRODES FOR CHEMICAL MEASUREMENTS

The small size of microelectrodes makes them ideal for measuring local concentrations of various chemical species, such as neurotransmitters, or the partial pressure of var­ious gases (see Biosensors). The carbon fiber microelec­trode is very popular. These fibers have diameters as small as a few microns and can be inserted into glass mi­cropipettes for insulation by exposing a short length of the electrode. Contact between the fiber and the electronic circuit is made by filling the electrode with mercury or silver paint and sealing it with epoxy. Then the assem­bly is inserted within physiological tissue and selective measurement of various compounds is carried out using a combination of selective membranes, enzyme coatings, sur­face modifications, and various electrochemical techniques. For example chronoamperometry has been used to de­tect the release of easily oxidized neurotransmitters, such as dopamine, serotonin, and epinephrine. In chronoam – persmetry, a constant voltage is applied between the car­bon fiber and a ground electrode. The current is measured and is directly related to the level of oxidation. Carbon fiber microdisks are typically more useful than microcylin­ders since they provide a better geometry and excellent resolution. They are also more difficult to fabricate and have lower current amplitudes. A three-electrode system is used to generate a constant voltage between the work­ing electrode and the medium around it. A reference elec­trode is located close to the working electrode to estimate the medium’s voltage. Using feedback, a current is applied between the working electrode and a ground electrode to maintain a constant voltage. For the very low currents of microelectrodes and microdisks, a two electrode system is sufficient since the ohmic drop and polarization of the ref­erence electrode are negligeable. The amplitude of the cur­rent is modulated by oxidation of the compound to be mea­sured and is proportional to its activity. To improve selec­tivity, a differential method is used whereby the current amplitudes obtained at two voltages are subtracted and the contribution of the interfering species, such as ascor­bic acid, can be reduced or eliminated. The electrode can also be coated with selective membranes. Nafion is a com­monly used membrane because it prevents various charged molecules from interfering with measurement. The surface of the carbon fiber can also be modified chemically or with a laser to improve the selectivity and sensitivity of the elec­trode (14). By depositing appropriate enzymes on the sur­face of the electrode, very selective electrodes can be made to measure glucose, nitric oxide, acetylcholine, etc.

Field effect transistors (FET) have also been adapted to allow measuring various ions or chemicals. Ion-sensitive field-effect transistors (ISFET), for example, are enhanced MOSFET transistors but use an ion-sensitive membrane instead of gate metallization (14). Then the transistor is immersed in a solution containing the ions to be measured. An electrochemical potential is established at the interface between the solution and the gate dielectric. This potential is established with respect to a reference electrode located in the solution and can modulate the conductance of the channel under the gate. The electrodes can be made very small but have drift and selectivity problems.

6 Micropipettes MICROELECTRODE ARRAYS

The silicon technology used to make integrated circuits can be adapted to manufacture arrays of microelectrodes. The activity of single cells can be recorded with the mi­cropipette technology discussed previously. Neuroscien­tists are now increasingly interested in recording simulta­neously from a large number of cells. Moreover, by stim­ulating a large number of cells in the spinal cord or in the brain selectively, it should be possible, in principle, to restore motor function in paralyzed patients or vision in blind patients, for example. Therefore, multiple arrays of electrodes capable of recording or stimulating the ner­vous system are clearly important to understanding ner­vous system function and to designing neural prostheses. Three silicon-based types of microelectrode arrays have been developed: (1) A 1-D beam electrode where a thin – film platinum-iridium is deposited on a thin layer of silicon substrate (15). This thin substrate provides a surprising amount of flexibility and can be utilized for the leads and the electrode pads. (2)A2-D array for recording the activity of neurons grown in cultures and axons in nerves. A thin film microelectrode array is made of gold electrodes covered with platinum black on a silicon substrate. The assembly is built into the bottom of a neuron culture dish. Neurons grow over these electrodes and make direct contact with them (16). In another design, micromachining of a silicon wafer generates a matrix of 64 square holes with a side dimension of 90 x-m. Gold pads and leads are deposited near each hole. Then the thin wafer is inserted between the two sides of a severed nerve. As the axons grow inside the holes, it is possible to record from a selectively small groups of axons (17). (3) A 3-D array for cortical recording and stimulation. The 1-D beam electrode discussed previ­ously can be assembled to form three-dimensional arrays. The longitudinal probes are inserted perpendicularly into a silicon platform. The leads from each probe are transferred to the silicon probe and are routed to a digital processing unit (18). Current work also involves including low noise amplification directly on the platform. Another implemen­tation involving micromachining and etching techniques was used to fabricate a 10 x 10 electrode array. One hun­dred conductive needlelike electrodes (80 x-m at the base and 1.5 mm long) are micromachined on a 4.2 mm x 4.2 mm substrate (19). Aluminum pads are deposited on the other side of the substrate and make contact with each needle electrode. The tips of the electrodes are coated with gold or platimum. A high-speed pneumatic device is used to place the array into cortical tissue because the high density of the electrodes makes insertion difficult. Then the micro­electrode arrays are available for recording from a large number of cortical sites.

LINEAR SLOT ARRAYS

A linear array of slots serves alone as an antenna or as a building block for a planar array. The far-field radiation pat-

cos2(fi^/2) sin (nx-^/a-^) cos2 (nl/2a2) cos2 (nx2/a2) (23)

tern for a linear array of N elements located along the z-axis is given by

N

S(4>,e) = A(4>,e)J2 Vnejkzncose

n = 1

The A(^>, Ф) term is the radiation pattern of a single radiating element, which is assumed to be the same for all elements in the array. The element pattern is similar to that given in Eq. (2), but it is modified by the array environment. The summa­tion term represents the array factor where Vn are the excita­tion values and zn are the element locations. The excitation values of the elements are chosen to generate a specified array radiation pattern. Slot-radiating elements allow con­trolling the excitations to achieve the desired pattern charac­teristics.

Linear arrays are divided into two categories: the stand­ing-wave (or resonant array) and the traveling-wave array (or nonresonant array). The antenna and system performance re­quirements dictate the choice between the two configurations. In the standing-wave array, the slot locations are chosen to maximize the coupling at the design frequency, and the wave­guide is terminated with a short circuit. This creates an effi­cient antenna, but the bandwidth is limited. In the traveling – wave array, the slot spacings are nonresonant, and the waveguide is terminated with an absorptive load. The travel­ing-wave array operates over a larger frequency bandwidth than the standing-wave array, but the power absorbed by the load reduces the radiating efficiency. It is difficult to generate a beam perpendicular to the guide with the traveling-wave array, and it also exhibits beam scan with frequency.