Suggestions on choice of antenna in direction finding system using BLE

Thread Starter

andkr

Joined Jan 26, 2022
23
Hello!

I am currently doing a school project where we are going to design a direction finding system using antennas and an automated robot arm to adjust to angle.

From what we have read the most reasonable choice of antenna is either a yagi or parabolic antenna. The robor arm is fairly small, and the preferred size of the antenna is at max 40cm x 40cm. We have found some candidates such as these:


But we are unsure if these are compatible with BLE communication at 2.4GHz and if a high gain antenna is preferred. As for the theory we have read; we want a high gain directional antenna to scan for the signal from the Rx side, and use a omnidirectional beacon as Tx.

However, the question to you guys is if the listed antennas are suitable, which one we should chose, or if you have any suggestions at all for what choice of antenna we should use.

Thank you so much in advance!
 

Ya’akov

Joined Jan 27, 2019
9,069
In the case of a directional antenna, gain is a result of beamwidth. And in this case, gain is a side effect because directionality is your primary concern.

Another important figure is front-to-back ratio which describes the difference in sensitivity to signals in the desired direction to those 180° away from them.

This is effectively rejecting signals from behind the antenna which, in a mutlipath environment like indoors, will reduce false positions from strong reflections to the rear the way a narrow beamwidth will make the antenna insensitive to signals off axis from the pointing direction of the antenna.

Among the antennas you linked, the first one (Laird MD24-12) seems best suited both electrically and mechanically.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
In the case of a directional antenna, gain is a result of beamwidth. And in this case, gain is a side effect because directionality is your primary concern.

Another important figure is front-to-back ratio which describes the difference in sensitivity to signals in the desired direction to those 180° away from them.

This is effectively rejecting signals from behind the antenna which, in a mutlipath environment like indoors, will reduce false positions from strong reflections to the rear the way a narrow beamwidth will make the antenna insensitive to signals off axis from the pointing direction of the antenna.

Among the antennas you linked, the first one (Laird MD24-12) seems best suited both electrically and mechanically.
Thank you very much for the reply! The deeper antenna-theory is pretty much new to us. Do you think were on the right path with the yagi/parabolic antenna path? Its smaller scale robot arm about 40cm in height.
 

Ya’akov

Joined Jan 27, 2019
9,069
Thank you very much for the reply! The deeper antenna-theory is pretty much new to us. Do you think were on the right path with the yagi/parabolic antenna path? Its smaller scale robot arm about 40cm in height.
A Yagi-Uda antenna can be very light though there is a direct relationship between beamwidth and length. A parabolic antenna can be short but the diameter of the reflector is proportional to the beamwidth and it tends to be heavier.

Another option is a corner relfector like this one which can offer excellent performance with a smallish foot print and rugged construction.

If you are inclined to do it, you could build any of these, though the Yagi-Uda and corner reflector will be easier. One option would be a circularly polarized "Ring Yagi". Which might be helpful in the multipath environment.

In any case, I would avoid premature optimization and either calculate the gain and f-b parameters you will need, or more sensibly, make some tests. You could easily put together a corner reflector antenna with simple sheet metal and a basic 2.5GHz resonant ¼λ antenna (of the sort you can buy to connect to bluetooth modules).

If you take this route, be sure to test the antenna and make sure it is resonant. Many cheap antennas are simply not resonant on the labeled frequencies.

Mr.Uda invented the Yagi-Uda antenna but Mr. Yagi, contributing a little to development, also patented it. Why his name was first in the antenna's name isn't clear, but Mr. Uda was dropped, probably to make it simpler. Had it been an Uda-Yagi antenna originally, we might call is an Uda antenna today.

[EDIT: Typo Repair]
 

nsaspook

Joined Aug 27, 2009
13,081
I think you might find a single non-spinning direction antenna system will have marginal accuracy and direction performance. There are other pure RDF techniques that have much better possible accuracy but might be too complex.

https://www.alarisantennas.com/blog/an-introduction-to-radio-direction-finding/
https://edfuentetaja.github.io/sdr/rfdf_adcock_watson_watt/

For BLE 5.1 there are device stack enhancements for location capabilities.
https://www.nordicsemi.com/Products/Bluetooth-Direction-Finding
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8196737/
 

Ya’akov

Joined Jan 27, 2019
9,069
I think you might find a single non-spinning direction antenna system will have marginal accuracy and direction performance. There are other pure RDF techniques that have much better possible accuracy but might be too complex.

https://www.alarisantennas.com/blog/an-introduction-to-radio-direction-finding/
https://edfuentetaja.github.io/sdr/rfdf_adcock_watson_watt/

For BLE 5.1 there are device stack enhancements for location capabilities.
https://www.nordicsemi.com/Products/Bluetooth-Direction-Finding
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8196737/

I wonder if a very small scale Adcock array would be practical. In any case, being a school project i have the impression that controlling the arm is a part of the requirements.
 

nsaspook

Joined Aug 27, 2009
13,081
I wonder if a very small scale Adcock array would be practical. In any case, being a school project i have the impression that controlling the arm is a part of the requirements.
Not sure what the arm will be doing and the bearing accuracy requirements at what range. A RDF requirement with a single directional antenna is scanning (with a human arm or mechanical device) to detect the signal slope to maintain directional pointing accuracy. The signal slope detection can be audio that changes frequency with the signal strength and a set of ears.

 
Last edited:

Ya’akov

Joined Jan 27, 2019
9,069
Not sure what the arm will be doing and the bearing accuracy requirements at what range. A RDF requirement with a single directional antenna is scanning (with a human arm or mechanical device) to detect the signal slope to maintain directional pointing accuracy. The signal slope detection can be audio that changes frequency with the signal strength and a set of ears.

I used to be an avid foxhunter and built many TDOA RDF rigs with 555s and 1N4007 pin diodes. They were cheap enough to give away, and I was able to get a lot of people into radiosport, as our Eurofriends call it, and it's a heckuva lotta fun.

We also made doppler finders but they were pretty pricey, especially in the days before cheap and cheerful MCUs, etc.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
@Yaakov @nsaspook @LowQCab

Thank you so much for the answers guys!

The mechanical device is controlled by 2 stepper-motors, using a microcontroller that are supposed to be connected to the antennasystem.

We want to use a directional antenna and use the RSSI to aim towards the omnidirectional beacon as accurately as possible.
As the robot-arm is fairly "small", we intend our antenna to be no more than 40x40cm in w/h, as it would be too large of a construction to fit the robot arm. It would preferably be <500g in weight.

Also, do you guys have any opinion on using AoA instead of pure RSSI to determine the direction? What are the pros/cons of AoA vs RSSI in terms of accuracy and complexity?

We found an antenna earlier that we considered using [http://assets.lairdtech.com/home/brandworld/files/ANT-DS-MD24-12 1115.pdf], but the beam-width at the -3dB points are at +-30 degrees, and i can only guess that the measured signals are pretty inaccurate. Do you have any thoughts on such a antenna that are both light-weight and compact, but also have a pointy and small main-lobe?

It is also worth mentioning that we prefer not to build our own antenna, but buy one from the internet, as there are other aspects of our projects we also need to finish in time.

Thank you in advance!
 
Last edited:

Ya’akov

Joined Jan 27, 2019
9,069
There are a couple of things to consider concerning AoA vs. RSSI. To start off, there are hybrid systems being produced and refined and by integrating both methods you can get some of the benefits of AoA with simpler infrastructure. That said...

First consideration is complexity of the algorithms. AoA can be far more complex depending on your approach. It could take a long time to refine your algorithm, maybe more than you have. However, there are many papers concerning this, it's a very hot topic, so that might help.

Second is the complexity of the infrastructure. AoA requires at least two antennas. Once you've done that, it is not clear if you are gaining anything from the robot arm. In theory, you could use the one antenna on the arm, and if you have appropriate degrees of freedom, reposition the single antenna to simulate two by integrating the two measurements.

I am not actually sure that would work well. I don't know for sure, but I can see the potential for considerable interference caused by the repositioning itself in an indoor, multipath environment.

I think you should, as I suggested above, buy or quickly cobble together a directional antenna (cantenna, corner reflector, not a Yagi-Uda which is more complex and will cost you much more time) then do some empirical measurements to see what sort of discrimination you can manage.

Since you can move your antenna, the wider beam angle might not be as much of an issue. You could also arrange to block off axis signals in a way that is no benefit to an antenna in normal use since you just have to be able to receive the signal and the reason you want gain is to narrow the beamwidth rather than increasing the signal strength.

So you could use reflectors that would reduce sensitivity to off axis signals while having a relatively wide beamwidth from the antenna itself. Sort of like putting blinders on a horse instead of strapping binoculars to its head. The binoculars would effectively restrict the field of view but the magnification is useless. The blinders don't restrict the field of view, but there are no signals for the horse to see in that part of their visual field so extraneous sights don't disturb it.

Radio "blinders", on the antenna could do the same thing. You'd not have increased amplitude (but I don't think you need that) but you would have signals from a smaller region in front of the antenna, which you do want.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
@Yaakov

We have considered AoA as an option, using an antenna array as done in the documention we have read. Our robot arm has 2 degrees of freedom, which allows us to rotate around a fixed point in 3 dimensions(?).

I like the idea of implementing both RSSI and AoA, as you will benefit from both techniques, and we are only in the planning-phase of the project. We do have the whole semester for this project and we have enough resources to use for hardware and materials.

However, as the HPBW is around +-30 degrees as seen in the radiation pattern, are you saying that we can use blinders to get a smaller sector when were scanning for the beacon? Or is that only to neglect the side lobes of the antenna. As were only interesten in the signal strength with our current model, does the sidelobe even matters i the signals received are poor in camparison to the signal received from the main-lobe angels?

We are planning on making a "calibration" algorithm, which allows the antenna to search for signals, mainly around the azimuth plane and map the different locations where the signals gets stronger.

As we want the working range to be between 1-20m, does not the gain of the antenna matter of the accuracy? As for what we understand, that the accuracy of a given antenna, by its radition pattern, can only give you the accuracy in degrees, given there are minimal noise.

So our problems are now;
- Can we use the given antenna from my last post with blinders to make the search more accurate and directed?
- How does the side-lobes affect or intervene with our system, given that we only use the RSSI and also the search algorithm to map the different locations of the signals
 

Ya’akov

Joined Jan 27, 2019
9,069
My suggest about shielding the antenna works like this:

If you have a 30° beamwidth you can use sweeping in the horizontal plane to find the peak, and it will be no different than the peak amplitude if you shield the antenna. The benefit of shielding the antenna (which, by the way, may be unnecessary which is why I am strongly suggesting some empirical testing to get baseline parameters) would be two-fold.

The off axis signals would be much weaker, and so you will be able to discriminate the peak more easily/quickly. That is, you will have less signal you need to consider to detect the peak. Second, it will reduce the potential effects of off axis multipath signals causing false peak detections. Of course, if you point the antenna directly at a coincidental peak caused by multipath interference you won't know, at the time, if you have the real thing or a ghost.

I would guess the way to deal with it will be a full, "calibration" sweep to find the largest peak, then centering on that to narrow the location. I imagine you will make a few sweeps of diminishing arc to find the real center. I think, though, trying it yourself will help you understand if it is potentially workable.

One more thing that I was just reminded of. I used to spend a lot of time doing ham radio "foxhunts" where it is a race to find a hidden transmitter. One of the techniques involves just your body and a handheld radio. Rather than finding a peak, you use your body as a shield and look for a null. I wonder if there is some way to adapt that to this case.

Also, I might try (and I am not sure of the feasibility of it) to create a mask directly in front of the antenna so that only a small aperture is transparent to the frequency of interest. It might even be something you could deploy with a servo after you've done as well as you can with the full beam of the antenna. Just a thought.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
@Yaakov Thanks again for the answer!

We fully understand that testing it, hands-on, is the best way to find a solution. But we need to sort out the details of the antennas beforehand, which is really a pain as we wont be able to test them. First when we get the antennas we will do a full-on research of our antennas characteristics and such using our universities antenna lab and other equipment.

We will certainly use your tips about shielding the antenna to reduce the off-axis "noise". We are going to study an algorithm for the search itself when we begin our project, as we think we will need to tailor the algorithm based on the antennas characteristics.

For the calibration algorithm; you are absolutely correct. We thought about sweeping the room, and then center our search from the highest peak, eventually reducing the area of interest in the best possible way. As for additional information, we are aiming to get a precision-range within centimeters, if it is possible. Do you think with such an antenna that we can get into the range of precision that we want?

As for the last section of your comment, i do not quite understand what you mean about creating a mask infront of the antenna.
 

Ya’akov

Joined Jan 27, 2019
9,069
So, if you look at the commercial AoA systems (for Bluetooth) they are claiming centimeter resolutions. Of course, they are using multi-antenna, and even multi-receiver systems, so you have more of a challenge. I really don't have a way to judge what your resolution would be.

The masking idea is to reduce the the angle a signal can be incident to the antenna and still be heard well. It's like the off axis idea, but actually in front, to reduce the signals that can be received, in both these cases, it is not clear to me if this is premature optimization, or even practical in a way that would give you some meaningful dBs of discrimination.

I mentioned the null vs. peak because I wondered if a central reflector could be used to cause a null more easily than something could be worked out to increase of narrow the peak. A null is as good as a peak, possibly better. If it's easier to achieve (I am not sure it is but I feel it might be) than it could be a fruitful direction.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
@Yaakov

When talking about antennas, in specific directional antennas. Then one can observe the radiation pattern given in the azimuth and elevation plane. How does these correlate to accuracy of the measured signal in term of beam-width. I have recently contacted one of the professors of our university, and he told us that given the radion pattern of the antenna [http://assets.lairdtech.com/home/brandworld/files/ANT-DS-MD24-12 1115.pdf], we will get accuracy no more than 10 degrees, given almost perfect circumstances. Would you agree to that?

We are struggling to comprehend the relationship between the given parameters we can find online and the specific accuracy we can obtain using the antennas we find.

As for example:
1, We use this antenna with a shield to neglect the other signals around the given shield.
2, The robotarm swipes the area to map the different signal strengths to give us information about where the highest peak is, ideally the beacon we want to aim towards.
3, We use our search-algorithm to get closer to the target, preferably at centimeter accuracy.

When talking about the antenna itself, would not it be a physical limit on how accurate it can be, by for example looking at the beam width of the main lobe? As you said earlier, it is easier to reduce the sector we are searching, but dont we need higher gain the get a more accurate reading? As for the main lobe; isnt the HPBW directly correlated to how "accurate" we can get the readings. For what ive researched, we need a considerably change in gain to detect the difference in the RSSI to aim towards the beacon as accurate as possible.
 

Ya’akov

Joined Jan 27, 2019
9,069
The pointing accuracy of the antenna will be dependent on it's own beamwidth and also the characteristics of the radiation from the transmitter. If you need more than just the bearing to the transmitter, you are going to have to move the antenna to a new location as far as practical from the first one.

Optimizing the antenna is helpful but you are really going to be dependent on computationally deriving the data from different measurements. The antenna itself is not the limit, I think. It's something like trying to locate a light that is shining 360° around itself. Unfortunately, you don't have a good analog to a focusable system, though if you imagine using neutral density filters to turn down the signal until the target is the brightest thing you can see, that's somewhat applicable.

There are other cases that make this hard, though. For example, discriminating between two sources of similar RSSI close enough to be in the beamwidth. It's roughly like the Sparrow resolution limit for a telescope, you can't tell there are two things there. Now, this might not matter for you, since it's as good to point to them both, but as they start to separate in distance you might find that you can't pick one from the other and the overall size exceeds your design goals.

I mentioned premature optimization. I think it is very important that you figure out the needs of your algorithm and the possibilities for computationally derived discrimination before spending the time to make what may well be small improvements in the antenna. You need to optimize your development time so it focuses on the most effective components and at this point I think the algorithm might be more important. At least you need some sense of how it ranks.
 

Thread Starter

andkr

Joined Jan 26, 2022
23
The pointing accuracy of the antenna will be dependent on it's own beamwidth and also the characteristics of the radiation from the transmitter. If you need more than just the bearing to the transmitter, you are going to have to move the antenna to a new location as far as practical from the first one.
I think the transmission beacon are going to send an omni-direction signal and is stationary, whereas the RSSI-deviation is yet to be calculated. Can you go more in depth of what you mean by having to move the antenna? Is it wrong to assume that you can aim accurately towards the beacon while the beacon and the robot arm itself being stationary, whereas you can move the antenna from that point?
 

Thread Starter

andkr

Joined Jan 26, 2022
23
There are other cases that make this hard, though. For example, discriminating between two sources of similar RSSI close enough to be in the beamwidth. It's roughly like the Sparrow resolution limit for a telescope, you can't tell there are two things there. Now, this might not matter for you, since it's as good to point to them both, but as they start to separate in distance you might find that you can't pick one from the other and the overall size exceeds your design goals.
This is a practical problem only to be encountered in real world situations? Or is it also possible to get two sources og the same RSSI from the same source? We would certainly need to look further into noise and the losses during transmission. But how can we calculate these errors given only the tx and rx doing the communication in the room?
 
Top