Archive for category: Blogs

By Johan Fuhri, Product Owner

In the previous blog post we considered some of the ways in which we can express the performance of an DF antenna, and we noted that simply stating the ‘accuracy’ of a DF antenna is not quite that simple.  We also noted that there are various algorithms that can be used along with a typical DF antenna, and that the accuracy and sensitivity of DF system is dependent on both the quality of the antenna and the direction-finding algorithm.

In this blog post we will discuss the process of ‘characterizing’ a DF antenna array to generate a table of values which can be used in correlative algorithms. We will also look at a few real-life problems that cause inaccuracies in DF systems, specifically when using basic correlative algorithms.

What is a correlative algorithm?

Correlation is a mathematical and statistical tool that can be used to quantitively express how similar two or more sets of data are to each other.  It is basically a fancy way of playing ‘pick the closest match’.

To illustrate the process, we are going to use colour swatches (the type that adorns the walls of the paint department at your favourite hardware chain store) and create a simple direction-finding system.  Consider the image below, which is a circle divided into six sectors and a colour assigned to each sector.  Let’s suppose we have a DF antenna right in the center of this circle, and the lines/sectors represent the various directions from which a signal can arrive.  This, however, is a very special antenna, in the sense that it will tell you the colour of an incoming signal.

If we are clever, we can make sure that we install the antenna in such a way that a colour of our choosing, yellow in this case, is always pointing due north.  If we then receive a signal from our antenna (which will be a color, remember), we can compare the color that we received from the antenna with the image above to determine which direction the signal came from!  That’s not too hard now, is it?

On a very basic level, this is what correlative DF algorithms do.  The DF antenna is spun through 360⁰ and the incoming signal from each direction, on each of the elements on the antenna, is sampled and stored in a table.  This table is called the ‘characterization table’ (although different people might name it something else), and it effectively allows a DF system to calculate the exact direction from which a signal arrives by comparing the signal with those stored in the characterization table, just like we did with the colours.  The circle indicating which colour is in which direction is effectively our own ‘characterization table’.

 

Alaris Antennas offer characterization of our DF antennas and systems as a service to our clients.  These characterization measurements are most often done at an outdoor reflection range at the National Antenna Test Range just north of Pretoria, which we affectionately simply refer to as ‘Paardefontein’ (https://www.paardefontein.co.za/).

The measurements entail mounting the antenna on one of the turntable towers and setting up a signal source some distance from the tower.  The DF antenna is then turned through 360 degrees in 4⁰ steps (depending on client requirements) and a sample captured from each of the elements of the antenna at each step.  This data is then compiled into a characterization table which we make available to our clients.

When life hands you green lemons…

As you might expect though, in real life things are not quite as simple.  There are various factors that can influence the accuracy of a DF system in real life scenarios, and in this section we are going to consider just a few of these, using our colour-wheel to illustrate why they cause problems.

Resolution of the characterization table

In our colour wheel example, we divided the circle into 12 sections, each representing an angle of 30⁰.  It is thus (nearly) impossible for us to determine the direction of an incoming signal to anything more accurate than about 30⁰.  Note that the signal to noise ratio, or the strength of the electric field, or the amount of noise has very little influence in the accuracy we can achieve, since we are fundamentally limited by the number of colours we have in our characterization table.

We can significantly increase our accuracy by dividing the circle into 36 smaller sectors, giving us many more options to compare our incoming signal to!  We can now determine the direction of our incoming signal to within a 10⁰ sector.

 

The original orange signal we received is most similar to the third sector, indicating that our signal arrived from a direction of 25⁰±5⁰.

 

 

 

 

 

 

 

The same is true for DF systems in practice as well.  When the antenna is characterized, the angle steps, or angle resolution, can be freely chosen by the person doing the measurement.  While it is tempting to choose super small steps (let’s measure a sample every 0.1⁰), this implies that the DF processor has do deal with a table that has 3600 entries, which can potentially slow things down to a crawl when limited processing power is available.  In practice, steps of between 2⁰ and 5⁰ are typically used, depending on the specific implementation of the client.

It is also worth mentioning that certain more sophisticated algorithms can achieve resolutions smaller than the characterization table resolution by employing all kinds of clever mathematical techniques.  It is the equivalent of being able to look at a greenish-yellow signal, and inferring that it must be somewhere between the green and yellow sectors, and thus giving a better guess than simply choosing the green or yellow sectors purely by similarity.  The details of these ‘super-resolution techniques’ are, however, often part of the proprietary magic that system integrators implement as their own competitive advantage.

The effects of noise

Up to now we assumed that the incoming signal that we are evaluating is perfectly clear, and that our accuracy is only limited by the resolution of the table against which we can compare our incoming signal.  In antenna terms this is called the ‘large signal accuracy’ of the antenna, as it assumes that no noise is present in the measurements.

Real life signals, however, rarely qualify as large signals and often has noise superimposed on the signal of interest.  Let’s look at an example in our colour wheel example.

Adding noise to the signal makes it harder and harder to find the colour that best matches the characterization table.  If there is noise present on the signal, invariably the DF processor will look at the ‘colours’ it can choose from and make mistakes.  At best it can tell you that the signal belongs to one of these four sectors, and this reduces the RMS accuracy of the overall system.

Adding even more noise, at some point it becomes practically impossible for the DF processor to make any sense of the incoming signal, and it may start taking ‘random’ guesses.  These wild guesses can be in any direction whatsoever, and are called appropriately called ‘wild bearings’.

The main source of noise in a DF system is the internal electronic components of the system.  Every amplifier, transistor and resistor contribute a bit of noise, and all this noise is added to the incoming signal before the DF processor has the opportunity to compare this signal to the characterization table. We can thus see that it is important to ensure that the received signal is as strong as possible while keeping the noise as small as possible to get a good signal-to-noise ratio.

The design of the DF antenna, specifically the gain of the antenna elements, directly contributes to keeping the SNR as high as possible by ensuring that the incoming signal is converted to a large as possible voltage for the receiver to work with.  For a more detailed discussion on this topic, refer to the previous blog post in this series.

Multipath interference

In certain scenarios it might be possible for a signal to bounce off objects, typically large buildings, and the signal that arrives at the DF antenna may end up being a combination of copies of the same signal, arriving from different directions.

These multipath signals, which we will illustrate as a mix of two colours, are very difficult for DF processors to deal with, and is very likely to a reduction in accuracy at best, or wild bearings at worst.

 

Using our colour wheel example, we can see that a pure green signal leaves the source that we are trying to find, but along the way a signal bouncing off a building also makes its way into the DF antenna.  The ‘colour’ making its way into the DF processor is mostly green, with a dash of blue mixed in.  This mixed signal will certainly cause problems for most simple correlative algorithms, as the received signal does not match any of the signals in the correlation table, causing inaccuracies and wild bearings.

Having said that, there are a few clever algorithms, like MUSIC, which can actually use these strange signals to its advantage and can figure out that there are multiple paths in play.  How these clever algorithms work is beyond the scope of this blog post (and the authors current intellectual grasp…).

Conclusion

This blog post attempted to explain the principle behind most correlative algorithms while trying to stay away from the complicated maths that make it work.  We also looked at a few real-world effects that influence the accuracy of DF systems in general, and specifically the mechanism that causes problems in correlative algorithms.  Most of these inaccuracies are caused by external factors, i.e. they are not so much caused by the design of the DF antenna, but by the environments in which they are employed.

In the last post of the series we will discuss a few of the intricacies of DF antenna design and how they can cause inherent inaccuracies (i.e. non-environmental causes) in poor designs.  This will be a rare glimpse into the everyday challenges that antenna engineers at Alaris Antennas deal with on a daily basis!

By Eugene Botha and Kalinka Faul

Humans do audio direction finding to a remarkable precision of less than two degrees.  Shortly after the development of radio transmitters and receivers, radio direction finding (RDF or just DF) evolved for much the same reasons as human audio direction finding: firstly, for the location of possible threats and secondly for spatial awareness.

RF direction finding is used in several applications:

• Military: such as the direction of a threat, the location and movement of enemy transmitters and the direction of enemy jammers.
• Search and Rescue: the location of RF search and rescue beacons.
• Science: the tracking of animals in their environment.
• Radio monitoring: the location of sources of interference and of illicit transmitters

Terminology and abbreviations

RF: Radio Frequency

DF: Direction Finding

RDF: Radio Direction Finding

AOA: Angle of Arrival

 

Single Receiver DF Systems

Directional Antenna Technique

The simplest RF direction finding system consists of a directive antenna and a single receiver.  The antenna is pointed in different directions while the receiver indicates the received signal strength. Only the magnitude of the signal is used to determine the direction of a transmitter.

The accuracy of this technique is dependent on the width of the antenna radiation pattern. A narrow beam will improve the accuracy but will increase the time spent scanning all possible directions.  If the beam is too narrow the target may even be missed especially with intermittent transmission sources.  A broad beam will decrease the bearing accuracy.

Scientists use handheld RDF devices to follow and accurately locate wild animals once their general location is known.  Radio beacons can be made very small and consume very little power. This is ideal for long duration tracing of even very small animals.  The relative signal strength is used to determine the direction of the radio beacon and the absolute signal strength shows some measure of the distance to the radio beacon.

In military applications this RDF is called a Spinning DF system and is commonly used for broad band high frequency DF.  These systems are fully motorized and use knowledge of the antenna radiation pattern to improve accuracy.

The benefit of a single receiver RDF system is that they are cost effective, compact and have low power consumption.

The image below shows both a wide and narrow beam as they scan for signals transmitted by the target. The received signal level of both beams can be seen on the left. Note how the signal levels increase the closer the beams are to the direction of the target.

Doppler and Pseudo-Doppler Technique

The Doppler DF system uses a single receiver connected to an omni directional antenna that is physically rotated on the circumference of a circle.  As the antenna moves toward the radio source, doppler shift will increase the received frequency and the received frequency will decrease as the antenna moves away from the source.  The change in frequency (obtained by FM demodulation) is used to determine the direction of the radio source.

The modern approach is to successively sample each antenna in a circular array of antennas, removing the need for any moving parts.  This is referred to as Pseudo-Doppler DF.

The image in the left of the figure below shows the circular trajectory of a doppler DF antenna. The received signal is incident from the right.

The received frequency at various positions along the trajectory of the antenna is shown in the right image of the figure. As the antenna moves away from the incident signal, the frequency increases (red graph) and as the antenna moves towards the incident signal, the frequency increases (blue graph). The black curve shows the received frequency response at the position indicated by black circle on the antenna diagram.

 

Two Receiver Systems

Mono-pulse Technique

The mono-pulse or sum-difference RDF technique uses two antennas.  The antennas are connected to a four-port combiner 180° hybrid that generates a sum and difference signal.  Such sum and difference patterns are generated by means of closely spaced overlapping radiation patterns at boresight.  These signals form sum and difference radiation patterns.  The ratio of the sum and difference signals and knowledge of the sum and difference patterns are used to determine the direction of the transmitter.  Phase information is used to determine on which side of the sum pattern the transmitter is.  An advantage of this system is in its capability to determine the direction of a transmitter after receiving one pulse. Such pulse could be a mere few microseconds. Accuracies of 10meter over a 100Km distance has been reported.

 

 

Interferometer

The relative difference in the phase of the signal received by two omni directional antennas spaced a set distance apart can be used to determine direction or angle of arrival of a RF signal.  In the omni directional case, the interferometer does not have a way to determine if the signal arrives from the front or the back of the antennas.

As the frequency increases (or electrical separation increases) ambiguities appear as the phase differences wrap.  If the antennas are too close (electrical separation very small) then the resulting phase difference will be very small and the system will not be able to determine the AOA.  The frequency range of use is thus determined by the separation of the antennas and the noise figure of the receivers.

The image below shows two omni directional antennas with the incident signal circulating around them. The relative phase of the incident signal between the two antennas is shown in the image on the right. This phase difference is used to determine the direction of the incident signal.

 

Adcock Antenna, Watson-Watt Method

An Adcock antenna uses two crossed loop antennas. The bearing of the RF signal is determined using the level of the signals received at each antenna.  The method to process the information from a Adcock array is referred to as Watson-Watt. This is the best-known method of radio direction finding.

In a more general application four closely spaced omni directional antennas positioned in a square can be used to form an Adcock array.  The opposing antennas are combined using a 180° combiner to form figure-of-8 patterns, which creates a unique set of magnitudes for any bearing direction.  A Watson-Watt antenna cannot determine if a signal comes from the front or the back without the use of a third omni directional antenna to resolve the 180° ambiguity.

The figure below shows the typical radiation pattern of an Adcock array. The received signal rotates around the array. The relative amplitude of the signal received by the two crossed loop antennas (or combined omni directional antennas) is shown on the right. The + and – at the top indicates the relative phase of the two crossed loops with respect to the omni. It is this relative phase that resolves the ambiguity.

N Receiver Systems – Correlative RDF

With the technological improvements in receivers and digital processors, all the information produced by multiple antenna elements can be used to improve the performance of RDF systems.  Typically, the bearing is calculated using the phase differences of the signals received at the various antennas in the correlative array.  The correlative algorithm compares the phase differences of the incoming RF signals at each antenna to a set of calibration phases stored in the processor to determine the most likely AOA.  The correction function correlates the relative phases (and magnitudes for some correlators) of the received signals with the correlation table over all possible angles; the maximum of the correlation function indicates the AOA.  The calibration information can be obtained by calculation or by measurement of the antenna.

The most common implementation of a correlative array is to have several omni directional antennas (typically four to nine antennas) in a circular pattern.  In this configuration the differences in the phase of the incoming RF signal at each element is used by the correlative algorithm to determine the AOA.

Cutting edge systems can use both magnitude and phase of an arbitrary number of antennas arbitrarily positioned to determine the AOA in 3D, not just in a horizontal plane.  The spatial positioning of the antennas in the array is of critical importance to achieve good performance without introducing ambiguities.  3D calibration data (not just in azimuth but also in elevation) with elements spatially positioned in atypical configuration is used to determine the AOA in 3D.  In many practical applications the patterns of the antennas will not be omni directional, clever algorithms can incorporate the magnitude information to improve RDF performance.

Even an Adcock antenna can be characterized and used in conjunction with a corelative estimator to improve performance of the Watson-Watt method.

 

RDF Performance

The two critical performance measures for DF systems are accuracy and sensitivity.

Accuracy is the measure of how accurate the bearing direction can be determined.  The accuracy of a DF system is dependent on the DF processor, the specific design, the quality of the antenna elements used and the installation environment of the antenna.

Sensitivity is the measure of how well the DF system will perform in the presence of a small signal in a specified noise level.  The sensitivity is dependent on the receiver noise, losses in the antenna and even the topology of the antenna elements in the array.

A discussion of these performance measures will be the topic of a follow up blog.

Written by Johan Fuhri

Introduction

One of the questions we often get asked at Alaris Antennas, is the question of spacing between multiple antennas on the same mast. Often, this question is related to situations where the space on the mast is limited (mobile or tactical application) and where high-power transmitters are involved. Unfortunately, there are no easy formulas that will work in all circumstances, so the answer typically starts off with something along the lines of “it depends”. In addition to distance, there are also several other parameters that complicate the discussion.  This answer, obviously, is not all that useful to most clients, so this article will attempt to discuss some guiding principles that Systems Engineers can use to address this problem of spacing antennas on a mast.

What is the worst that could happen?

When two antennas are placed close to each other, they will inevitably couple to each other to some degree. ‘Coupling to each other’ is just a fancy way of saying that whenever one antenna transmits, some of that signal will be received by the other antenna. This stray signal may or may not be a problem, depending on the specific hardware, power of the signal being transmitted, gain of the two antennas, orientation, spacing and many other factors. A typical example may be where a high power transmit system is on the same mast or shelter as a monitoring system, or where multiple high power transmit antennas (with their associated transmitters) are co-located on the same mast. Typically, any receiver (or transmitter) can only handle so much power coming in from the antenna before being permanently damaged. When dealing with kilowatt-capable amplifiers somewhere in the system, other transmitters or receivers could easily be damaged if the design of the system does not specifically make provision to prevent this. The first and most important question is thus to determine how much isolation (the inverse of coupling) is really required between the various devices in the system. This will determine, in turn, how much isolation is required between the antennas.

Factors determining the isolation between co-located antennas

Let us consider some of the factors that will improve or deteriorate the isolation between two antennas. Since LPDAs are often used in wideband high power transmit systems, we will specifically use them to illustrate the principles, but they equally apply to most other types of antennas as well.

Polarisation

The one parameter that has a much larger impact on isolation than spacing, is the polarization of the two antennas.  As the name suggests, Log-Periodic Dipole Antennas consist of several dipoles that are fed in specific way to create forward gain.  Dipole elements, or any sort of “skinny” element, always have a null directly above or below the element when mounted vertically. When they are mounted horizontally, there is no longer a null in the radiation pattern aiming up or down where the other antenna may be located. If two LPDAs are mounted on a mast, one on top of the other, the coupling between them will be much higher when they are both horizontally polarised (HP) compared to when they are vertically polarized (VP). In a simulation of the LPDA-A0102 using the same spacing between the two antennas, we can see that the isolation between the two VP antennas at 100MHz is about 40.7dB, whereas the isolation between the two HP antennas is only 22.5dB. To put this into perspective, a 5kW (67dBm) signal transmitted from one of the antennas will induce only about 0.43W (26.3dBm) of power into the other antenna when both are VP. When both antennas are HP, a massive 28.2W (44.5dBm) signal is picked up by the second antenna.

The worst isolation will (almost) always be at the lowest frequency

From Figure 3 it can be seen that the highest coupling (lowest isolation) occurs at the lowest frequency point of the plot.  This will almost always be the case, with a few notable exceptions (like LPDAs with less efficient reduced length elements at the rear). It is important to keep in mind that many antennas are designed to be wide-band and LPDA’s are no exception. The isolation will be determined by both the distance between antennas and the frequency of the signal. For a given spacing, a higher frequency does not transfer as much power as a lower frequency signal, with the reduction in power proportional to the frequency squared! Hence with all other things equal (e.g. antenna gain), the lower frequencies will typically be where the problems lie. In addition to the propagation effect, it is also worth noting that all the other structures on the shelter, mast or building will be much closer (in terms of wavelengths) and is more likely to cause secondary effects that may (or may not) influence the isolation between the two antennas.  System Engineers will do well to tread carefully when trying to estimate the isolation between two antennas in the HF and even VHF frequency bands.

Physical spacing between antennas

Thus far we have seen that the spacing between the two antennas is only one of the many factors that determine the isolation between two antennas. Unfortunately, it is also one of the parameters that turns out to have a relatively small influence compared to things like polarisation. To illustrate, let us consider three scenarios where two identical antennas (LPDA-A0102) are horizontally polarised and spaced 2m, 4m and 8m apart.  While we are using them both in HP, the same principle and relative difference will apply for the VP case. Doubling the distance from 2m to 4m only increases the isolation by 6dB (and even less at 100MHz), which is not all that much since an additional 2m spacing on a mast is a practical impossibility under most situations.  As a rule of thumb, every time the spacing between two antennas is doubled (or halved), the isolation between them increases (or decreases) by about 6dB. This is true unless the antennas are really close to each other (in terms of wavelengths) and is true for both the VP and HP scenarios. It is thus clear that an increase in spacing between antennas is unlikely to be the magic bullet that will solve isolation issues in most practical scenarios!

Different type/models of antennas

While we have used two identical antennas to illustrate the various parameters that influence isolation between two antennas, we should not forget that most of these also apply to antennas that are not identical. It would be easy to assume that antennas that are not working in the same frequency bands will not couple to each other, and we would be very wrong. To make matters even worse, because of the physical smaller size, one would typically be tempted to space them closer to each other. Typically, antennas are designed to work well over a specific target bandwidth.  Often the performance of the antenna at other frequencies is considered not to be important but some antennas could have unintended resonant frequencies way outside of their specified bands of operation. For antennas like LPDAs (and many other wideband antennas), the performance outside of the specified bands of operation may roll off relatively slowly. When two LPDAs of different bands are mounted near each other, there may still be significant coupling between them, especially if they have partly overlapping frequency bands. To illustrate the potential pitfall, consider the scenario where an LPDA-A0102 (100-500MHz) is mounted close (2m) to another LPDA designed to work from 200-1000MHz. It can be seen that there is significant coupling between the two antennas from 500-750MHz, which is well outside of the frequency band where the LPDA-A0102 is designed to work.  The same is true for the frequencies below 200MHz, which is below the designed operating frequency of the smaller LPDA. Thus, just because two antennas do not have overlapping frequencies of operation, do not assume that the coupling will be negligible.

If you must know…

In some scenarios it might not be good enough to simply “do your best”, and a relatively accurate estimate of the isolation between antennas on a structure may be required.  Unfortunately this cannot easily be done on the back of an envelope but there are ways of getting a better estimate.

…measure the isolation

There is simply no better way of knowing than doing an actual measurement. One should still be careful in setting up the measurement as it is quite easy to get wrong answers if one is not careful. It is important to recreate the final setup as closely as possible and remove as many objects that may cause reflections (thereby falsely reducing isolation) or direct paths between the antennas. In many scenarios, measurements may not be practical or possible.

…simulate the system

With the advancement in computing power over the past decades it has become possible to simulate relatively complex and sophisticated systems involving multiple antennas and structures. Such simulation studies are extremely useful to point out unexpected problems and allow System Engineers to rapidly explore multiple configuration options while getting relatively accurate, or at least indicative isolation values for the various combinations. All this can be done without the investment in hardware and equipment that comes with measurements. But, as is the case with measurements, simulation studies have their own potential pitfalls and it is important to have a good understanding of the potential inaccuracies that may arise from assuming a ‘perfect’ world on a computer so as to avoid disappointment with the results. As is often the case, a combination of simulation-based studies and measurements are generally the best approach.

Conclusion

When considering the placement of multiple antennas on a mast or some other structure, it is important to understand the principles that contribute to the isolation between various antennas. While physical spacing between antennas is often the first and obvious parameter that is considered, it is by no means the only one, or even most significant contributor to the isolation between two antennas. Understanding the underlying principles will allow System Engineers to make educated decisions on how to co-locate antennas on a structure. To view the YouTube video, click here