DJJ's navigation page
Antennas main page
miscellaneous/antennasmeasurements/
A loop antenna has a continuous conducting path leading from one conductor of a two-wire transmission line, "the feed", to the other conductor. You may think of it as a "coil that radiates". The coil may have only a single turn. It may have arbitrary shaped perimeter, but the essence of a coil is that the defining wire encloses an area. Thus, a folded dipole is not a loop antenna in this sense, since the area inside the conductor path is vanishingly small. In a loop antenna, the magnetic field (generated by the loop current) threads the area of the loop and this provides the loop with inductance.
"Symmetric" loop antennas have a plane of symmetry running along the feed and through the loop. "Planar" loop antennas lie in a single plane which also contains the conductors of the feed.
"Three-dimensional" loop antennas have wire which runs in all of the x,y, and z directions (in a rectangular Cartesian system). By definition they are not planar. They may, however, be symmetric about planes which contain the feed.
It is possible for the loop antenna plane not to contain the run of the feed. This matters (as far as the radiation pattern is concerned) for situations where the feed currents are not perfectly balanced. In any case, whether or not the feed is in the plane of the loop, unbalanced loop antennas are not functioning as "true loops". A balanced loop antenna has equal and opposite current flows in the conductors of the feed. That means there is no net current in the feed conductors (no "common-mode" current) and therefore, the feed will not radiate. Another way of looking at it is that the voltages at the ends of the loop are in anti-phase with respect to the ground. The loop itself therefore does not radiate as a monopole with respect to the ground plane or ground radials, if fitted. Loops fed by coaxial cable are quite likely to be unbalanced, unless positive steps are taken to choke off the outer sheath currents with a balun.
Of course, if all one wants to do is to construct an effective radiator, one might argue that there can be some advantage in having an unbalanced loop. But then the loop part is to some extent vestigial, and not central to the functioning of the thing as a radiator.
There are at least two distances which define the "notion of size" in a loop antenna. These are, the total length of wire between the "go" and "return" of the feed, and the largest distance from one point on the loop conductor to another, measured in a straight line (as light would propagate). One might also think of another distance that "matters", namely the distance from the feed junction with the loop, to the most remote point on the loop conductor. All these distances need to be thought of in units of a wavelength at the carrier frequency handled by the antenna.
Of course, in the case of the unbalanced loop antenna, there is the total length of the feed to consider as well. For this reason, many people have found anomalous "super performance" of compact loops on the end of sensibly long feeds.
If one wants to couple radiation from a two-wire feed (possibly coax) to a waveguide, one commonly does this by means of a probe (which couples to the electric field in the guide, and is the equivalent of a monopole) or by means of a loop (which couples to the magnetic field in the guide; the maximum of magnetic field lines pass through the loop). Waveguide may be regarded as a microcosm of the great outside world.
Critical to the functioning of any loop antenna is the concept of the "phase delay" that occurs for electromagnetic radiation to get from one point on the loop to another, some distance away. In the case of vanishingly-small loops, the traditional calculation assumes that the current is the same everywhere around the loop perimeter,
as can be seen diagrammatically in this figure.
In this case, the radiation along any loop diameter arrives from an oppositely directed but parallel element of current after a short time delay, which puts in a phase shift so that the radiation contributions do not entirely cancel. Since the current is assumed constant around the loop circumference, the lengths of the phasors involved in this calculation are identical, and, were it not for the phase delay for the radiation to cross the diameter of the loop, the phasor contributions would be exactly opposite in direction, resulting in complete cancellation.
This traditional argument quickly leads to the result that the radiation resistance of a small circular loop rises as the (ratio of loop diameter to wavelength) raised to the fourth power. A small increase in loop diameter therefore results in a greatly increased radiation resistance.
Just as with a small rod antenna, where the radiation resistance rises as the square of the length of the exposed radiating wire (see radimp.html), so also in a loop antenna the radiation resistance, were it not for the cancellation effects, might be expected to rise as the square of the circumference, and therefore as the square of the loop radius or diameter in the case of a small circular loop. However, there are these additional cancellation effects, and this puts in an additional factor proportional to the square of the diameter, radius, or circumference.
One of the most significant attributes of a loop antenna is that "go" current in one part of the loop is offset by "return" current in another. It is only because these go and return paths are physically separated in space that a small loop antenna can radiate at all. Otherwise the radiation from one little current element would exactly cancel that from the other. In fact, this does happen for radiation directions normal to the plane of a vanishingly small planar loop. In such directions there is a deep radiation null.
Quantitatively, for a circular loop of radius R, when R/lambda = 0.25, the diameter is half a wavelength and the 180 degree phase shift, for the radiation to get from the "go" current at one end of the diameter to the oppositely-directed "return" current at the other end of the diameter, results in an enhancement factor of 2 over the radiation from just a single current element. This enhancement factor is a property of "large loops", and is not applicable to small or intermediate-sized loops. The radiation from one element arrives "in phase" with the contribution from the other element, half a wavelength away but with current opposite in sign. The perimeter is then (2 pi 0.25) wavelengths which is 1.57 wavelengths and so the assumption of constant current around the loop perimeter has broken down. This is what is meant by the term "large loop".
For R/lambda = 1/100 or 0.01, the field contributions nearly cancel. The expression for the "enhancement factor" is [2 sin(2 pi R/lambda)] which then evaluates to 0.126 very nearly. This is a lot less (actually, a factor of 1/16th) than that due to the quarter-wave radius loop (enhancement of 2) and will make (1/16)^2 = 1/256 difference to the contribution of these little elements to the radiated power and to the radiation resistance.
For R/lambda = 1/50 (or 0.02), the enhancement factor is 0.25, and for R/lambda = 1/20 (or 0.05) the enhancement factor is 0.61 and at this point the perimeter has got to 0.3142 of a wavelength. Again, the assumption of constant current around the loop has, by now, broken down.
Of course, in certain loop structures the size of the currents in different elements of length along the loop wire will vary. Thus, loop antennas which have a total wire length approaching or exceeding "an appreciable fraction of a wavelength" can be efficient radiators with radiation resistance that approaches a match to common feed-line impedances. It is only in vanishingly small loop antennas that we are justified in assuming that the current is the same at every point along the loop wire. In intermediate cases, this may sometimes be a justifiable approximation, but certain textbooks which treat a circular loop antenna of radius lambda/25 (which has a loop wire length of about lambda/4) as if the approximation were sufficiently valid, may be in serious error. It is partly for this reason that there is some controversy about the radiation resistance of intermediate-sized loop antennas.
If the current around the loop wire is not constant, we may apply Kirchoff's current law and identify "displacement current" flowing away from the loop wire section through the parasitic capacitance between different parts of the loop. The circuit model for the loop wire is therefore best seen as series resistance and inductance, and shunt capacitance. Non-constant current therefore implies current phase shift, in general.
In the case of a folded dipole, the current is definitely not constant around the loop.
For a loop where the perimeter is about a whole wavelength, the folded dipole analogy may be better. We imagine the loop as being formed from a "bulged-out" folded half-wave dipole for which the current distribution looks like this :-
The current in the element of wire diametrically opposite the feed is now directed from right to left, rather than from left to right as it would be in the vanishingly small loop. The currents up and down in the elements on the horizontal diameter have vanished, and it is for this reason that there is no radiation from this scenario in the plane of the loop in the horizontal direction.
The radiation resistance of this mode is quite high because the cancellation between currents on opposite ends of a diameter is no longer so complete. For a circular loop of perimeter one wavelength, the radius is 1/(2 pi) wavelengths, or about 0.16 wavelengths. We are still talking about a physically compact-sized loop therefore. For example, at a wavelength of 20 metres such a loop would have a diameter of 6.4 metres and would be an effective wideband radiator.
Now consider a loop which has a perimeter of just one half of a wavelength. At 20 metres wavelength this would be a loop of diameter 3.2 metres. We may consider a bulged-out length of transmission line having the same total wire length. A little thought shows that the transmission line model is a short circuited quarter-wave section of line. The input current is zero in the parallel line approximation, as the line presents an open circuit to the generator. Of course, this will not be exactly true when we have "bulged out" the line section; for one thing, the short circuit point will have moved physically closer to the feed.
What is clear, however, is that in this approximation the current in the element diametrically opposite the feed runs from left to right, not from right to left as it did in the folded dipole example. It is also going to be significantly larger than the current supplied by the feed.
As we increase the perimeter of the loop from a quarter wavelength to a half wavelength, there must therefore be a region where the current opposite the feed is smaller than the feed current, and indeed it must at some point pass through zero. Thus it is apparent that for all intermediate loops of diameter greater than 0.16 wavelength, the "small loop" approximation is not valid, and very significant radiation occurs. The Q will be reasonably low and the bandwidth and radiation resistance will have usefully large values. Those people with simulation packages may like to quantify these "general statements".
Now, it is known that the radiation resistance for a small loop antenna is often swamped by the loss resistance due to the current being confined to a small skin depth of conductor at the wire or tube surface. Thus we have to consider phase shifts between oppositely-directed currents on opposite ends of a loop diameter (for the special case of a circular loop) brought about by the distributed inductance, resistance, and capacitance of the loop line. Loop radiation is often (usually) measured with the loop mounted in a vertical plane, and one goes away a significant distance on a flat ground so that the radiation is measured on a horizontal (level) path in the plane of the loop. It is easy to see that for a symmetric loop (as discussed above) the current elements contributing to the radiation, up and down on opposite sides of the loop, have balanced amplitudes and phases. Thus we expect the traditional formula for the radiation from a vanishingly small loop to be approximately correct for intermediate loops, in this scenario of horizontal path radiation. For those loops of this kind of size where radiated field strengths have been measured and reported, it is said that this was the measurement geometry.
For other directions of the diameters of the loop, which are at a slant (intermediate between horizontal and vertical) angle to the ground, there will be phase shifts and amplitude differences between the little elements of current flow at the ends of this diameter. As stated, these phase shifts are due to the combined effects of distributed series inductance and shunt capacitance, and series skin-effect loss resistance. However, for loops where the phase shifts have a significant effect, the total wire perimeter will probably be long enough so that the amplitudes of the current elements change as well, and this will also generate a contribution to the radiation.
Quantifying these phase and amplitude shifts would appear to be quite a difficult problem. In terms of the current flow through a continuous conductor having distributed inductance and resistance per unit length, Kirchoff's current law indicates that the current is the same everywhere and that there are no phase shifts. However, if we then allow for the shunt capacitance, between elements of the loop tube or wire (which has non-vanishing surface area), then the phasing of current flow around the loop becomes a function of the loop wire diameter (or tube diameter) as well as the skin depth loss. A simulation might sort out some of these issues, but as it would return global values for the antenna properties, the local behaviour might not be transparent.
It is not unreasonable to expect, therefore, that intermediate-sized loops will radiate more strongly along such slant diameters than the traditional theory might predict. This effect is expected to be quite small, overall. For, the cancellation of the oppositely-directed current elements is no longer so complete: they have differing amplitudes and phases. This will put up the radiation resistance of the loop. Paradoxically, therefore, the presence of loss resistance in the loop due to joule heating in the skin depth where the current flows, may enhance the total radiation over what it would be for a lossless conductor having the same geometry.
The loop (intermediate size) will therefore radiate up and down preferentially. If we mount the loop with its plane horizontal, it should be possible to check on this effect by moving around the loop at constant range, measuring the fields radiated as we go. The prediction is that there will be some anisotropy in the radiation, symmetrically disposed with reference to the feed axis.
For the case of non-constant amplitudes and phases, there will also be radiation normal to the plane of the loop. This forms the basis of a simple and sensitive experimental method of deciding whether a loop antenna is functionally "small", or in the "intermediate size" range. In the case of a truly "small" antenna, there should be a very deep null in the far field region at directions on the axis of rotation of the loop. This null progressively fills in as one makes the loop diameter larger. By the time the loop diameter is about lambda/10 there should be appreciable radiation along the loop axis. As remarked above, this will be accompanied by anisotropy in the radiation in the plane of the loop.
The folded dipole approximation to an intermediate loop antenna has deep nulls along the horizontal diameter (if the feed runs in from underneath) and for this reason radiation in this mode is not detected in the standard loop field-strength measurements reported by some others.
Loop antennas have area, and generate magnetic fields which thread this area. These changing magnetic fields generate a back emf at the loop terminals which provides the loop with inductive impedance. Generally speaking, the larger the area, the larger the inductance. However, as the loop wire becomes longer, the phase shift between induced voltage and the current that gives rise to it changes. At a certain wire length, generally held for circular loops to be about 1/3 wavelength, the loop becomes "self resonant". Another way of looking at this phenomenon is to consider a loop to be a "bulged-out" length of parallel wire transmission line, shorted at the end remote from the feed point. In the case of a true parallel wire line, self resonance may be defined to occur when the total wire length (go and return) is 1/2 a wavelength plus the length of the short at the end. The line is then a quarter wave shorted stub.
The touchstone of antenna work is to make careful measurements. Accordingly, we constructed a loop antenna with a total wire length of 500mm and a diameter of about 160mm. The antenna is pictured
We see here that there is between 3 and 4 mm of parallel wire transmission line between the coax connector and the loop proper.
This antenna has wire diameter about 2mm and it is said that the properties of such loops are somewhat sensitive to the wire diameter. The loop was placed in an anechoic room, and isolated from the feed with a ferrite clamp. A network analyser was carefully calibrated and the
We also see that there is an antiresonance where the input is high impedance, at 171MHz which corresponds to 28.5% of a wavelength (1/3.5 factor); and in this plot we also observe the fundamental anechoic room resonance, heavily damped and lightly coupled, at about 38MHz, indicated by the marker on the plot.
We see that at the 680MHz resonance the real part of the impedance is large, about 210 ohms. We recall our model of the folded dipole analogy; we would expect that if the 1-wavelength-perimeter loop was working as a folded dipole we would see real input impedance about 300 ohms.
We also comment that the 171MHz antiresonance might have been expected to lie at nearly 300 MHz if our model of the shorted quarter wave line (above) was correct. This is a significant difference between simple theory and the measurements. Surprisingly, the NEC simulations are said to be more in accord with the model of the shorted quarter wave line than with our measurements. Here is a comment from a modeller...
At 1 MHz the antenna is almost a non-reactive dead short.
The inductive reactance increases, as expected with all the loops that I have worked with, up until the half wave antiresonance is reached at 295 MHz where the antenna is an open circuit. This is almost exactly where expected since the antenna is 1 wl at 600 MHz.
The only resonance I found was at 680 MHz with a Rin of 150+ ohms. This, again is exactly as expected since we expect the fw resonant frequency to increase with the wire diameter. The loop is 13% longer than a wl at 680 MHz. The SWR 2 bw is 610-760 MHz and this is also what I would expect given the large diameter. Of course, if you used thinner wire, the resonance will be slightly lower - but never 600 MHz.
Professor Underhill has suggested that intermediate sized loops work in a combination of two modes; "pure loop" and "folded dipole". In the case of the 500mm circumference loop with an antiresonance at 171MHz, we may deduce that the inductance of the loop, regarded as a single turn of non-radiating wire, is tuned or balanced by the capacitance of a folded dipole mode, for a total dipole length of lambda/4.
The combination of modes has been experimentally confirmed, and the radiation patterns of each mode combine in such a way as to give a varying and intricate pattern as the frequency is raised through the intermediate size region.
Underhill also makes the comment that the real part of the loop impedance is sensitive to environment.
Alternatively, Alan Boswell has run a NEC model which suggests that a stray shunt capacitance of a little over 0.5pF, placed at the junction of feed and loop, is sufficient to run the antiresonance down to around 190-200MHz. It is possible that this capacitance is intermediated by the bulk of the ferrite clamp. Using the formula capacitance = epsilon0 area/separation with epsilon0 = 8.854 pF/m we might estimate this capacitance at about 0.35pF for a separation of 1cm and an area 2cm by 2cm. This neatly gets us out of the contentious problem of NEC being inadequate; it is just that with antennas of this size one really needs to be very careful about the modelling.
But our final view of this situation is concerned with the adjustment of the "port extensions" on the network analyser, to take into account the very short (just a few mm) length of parallel wire connections between the BNC connector and the loop itself. See the discussion.
To settle this controversy, we went to some effort to suppress the room resonances. A ferrite cable suppressor was slipped over the coax at intervals along the cable and careful attention was then paid to the ends of the cable.
We were rewarded with this 1 MHz to 801 MHz trace, with no wibbles due to cable or room resonances, and with the antiresonance at 301 MHz and the resonance at 619 MHz, after adjusting the port extension (see below) to 14mm, which is what is expected for the 3-4mm of wire link shown in the picture above.
Theory, simulation, and experiment are all now in pleasant agreement, and there is no need any longer for arcane theories about "why NEC does not model loop antennas accurately". It was not found necessary to include the postulated stray capacitance, which, if we look carefully at the experimental setup, is difficult to identify in the amounts required. The comparison with an equivalent NEC simulation is shown here (sim courtesy of N2DT Dan Handelsman)
The antiresonance is at 295 MHz and the resonance at 653 MHz.
However, if we examine the comparison between these traces carefully, we see that the experimental plot begins to depart from the periphery of the SMITH chart at above about 40MHz (j100) whereas the NEC plot hugs the boundary all the way round to the antiresonance. This lends credence to Mike Underhill's observation about the unexpected radiation efficiency of intermediate loop antennas, as revealed by experiments on tuned loops. Now, it is possible that the additional absorption is provided by the ferrite damping, or by the damping in the screened room.
When the Network Analyser is calibrated, a kit of standards is connected to the BNC connector at the end of the Network Analyser cable. These consist of an open circuit, a short circuit, and a matched load. It is very important that the network analyser knows exactly where the reference planes are for each of these standards. To do this, it is pre-loaded with data (on floppy disk) which describes the actual standards used.
When the device under test is connected, the precise position at which the Network Analyser measures the s-parameters may be adusted in software to run up and down the cable. You will see in the SMITH chart picture above that the port extension is said to be between 4 and 5 mm. We constructed a test short on the same kind of connector that was used for the antenna proper. The antenna is connected between 1 and 2 mm further along than is the reference short sheet of metal. We adjust the port extensions for this short. The position of the short circuit, plotted on the SMITH chart, does not alter over the whole sweep range from 150-950MHz, as we see here.
The port extension may then be adjusted empirically for the last 1-2mm to produce a curve which just touches the real axis of the SMITH chart at resonance. Now, an equivalent plot of the antiresonance has no port extensions switched in and so we can see that the trajectory at higher frequencies looks somewhat different.
The choice between these calibration details is a matter of experimentalist's judgement, and we are still considering it. There are some equivalent pictures for a straight simple dipole in miscellaneous/antennasmeasurements/ which you may wish to consult.
The inductance of a length of 50 ohm line, velocity factor 2/3, is Zo/u = 50/2E8 = 0.250nH/mm. Arguably the extension of the centre pin of the coax connector, and its mate the wire on the rim, form a transmission line of impedance about 300 ohms and velocity factor 1. Therefore the inductance of this bit of wire is 300/3E8 = 1.0 nH/mm. We should therefore multiply the length of connecting wire by a factor 1/0.25 = 4 to arrive at the desired port extension.
Thus, a 2mm piece of wire translates into an additional port extension of 8mm, and together with the intrinsic port extension needed puts it at 11 or 12mm. We may see the result of adding a 14mm port extension in this picture:-
Considering a plot on the SMITH chart, the trajectory close to the open circuit point at gamma = 1+j0 is very sensitive to any stray shunt capacitance at the reference plane. Conversely, the trajectory close to the short circuit point, particularly at the higher frequencies, is very sensitive to series inductance (possible provided by a few mm of wire). Unexpected effects should be considered first from these points of view.
A self-resonant antenna might be thought of as being "optimally efficient". Smaller loops require additional series or shunt capacitance to tune them to resonance so that the impedance presented to the feed becomes real.
In a self-resonant loop, then, it is clear that the standard small-loop theory breaks down. As we have indicated, this happens for total wire length of between 1/3 and 1/2 of a wavelength. The current distribution around the loop will be very non-uniform; the radiation resistance will be significantly large, and will swamp the loss resistance in all likelihood. As we gradually increase the dimensions of the loop antenna, nothing suddenly happens to the radiation properties. Therefore, we propose that the small loop limit really needs the loop radius to be very much less than the reported 1/25 of a wavelength. The controversy about small loops, however, deals with loops of precisely this size. We are not surprised.
Recently there have been reports in antenneX magazine about designs for three-dimensional small loops. In such loops, the ratio of wire length to maximum linear dimension of the antenna may be made significantly larger. Therefore, the small-loop limit will apply only for even yet smaller overall dimensions. The phase shifts, as we travel along the loop conductor, will result in less cancellation for oppositely-directed current elements and there will be enhanced radiation resistance and efficiency.
In three-dimensional small loops, it becomes easier to make the total wire runs longer than a wavelength, and to make adjacently-placed wire runs carry currents which run in the same directions, whose radiation therefore reinforces rather than subtracts.
Also, by wrapping up the wire runs, into a folded structure, the total current-carrying (and therefore radiating) elements inside the compact antenna volume may be significantly increased in length. This may be done without endlessly increasing the loop inductance, and so reaching self resonance at too short a length of radiating wire. For, the wire in 3-d may be run in such a way that the local magnetic fields generated subtract, from the contributions of different parts of the wire run. It appears, therefore, that we are still in the early stages of finding out what may be achieved, in small linear dimensions, with this exciting new class of antenna structures.
