Wide Band Aerials For 1700 to 2400 MHz, M1 - M6 Band

The 1700 to 2400 MHz band (M1 to M6) is a good band for the secondary control channel of air and ground robotic vehicles.  Little radios made by Doodle Labs in Singapore are popular and this is an antenna solution specifically aimed at the Meshrider radios (https://doodlelabs.com/products/mesh-rider-radios/nano/).

I employed Ye Olde Fashioned technique of carving an antenna out of double sided board with a ruler and a scalpel, using copper tape for little optimization experiments and fixing the mistakes.  Once one has the hang of it, it is possible to carve an antenna by hand in a day, vs waiting two weeks for a PCB factory.  

Etching it is good, if you have the chemicals and safety paraphernalia on hand and know what you are doing.

This application requires an aerial system with a few important properties, notably:

• Wide bandwidth of 20% or more with a centre frequency of 2 GHz.
• Directivity of 8 to 12 dBi, to reduce clutter and interference from the surroundings, without making it difficult to point at the robotic vehicle.
• Circular polarization when used in a high gain multi radiator array, to reduce multipath interference due to reflections off the earth surface.
• Simple, repeatable and rugged construction, so it can be easily replicated and will not break during outdoor use in a harsh environment.

The above made me look at various PCB aerial topologies, including:

• Monopole
• Patch
• Tapered Slot 
• Log Periodic
  

A PCB monopole aerial works very well, but it is inherently omnidirectional and and not easily stackable, so it is not great for ground use.  It will be good as a blade antenna on an aircraft, so I will revisit this one.

A PCB patch aerial needs to be very thick, with air dielectric, assembled with nylon spacers and bolts, which makes it difficult to replicate.  I made one like this once before - not recommended - much too fiddly.

That left me with two well known ultra wide band antennas to explore for ground station use, the Tapered Slot and the Log Periodic.  Both certainly will work well, if only I can get them tuned properly to the required band, which will take a few tries.  Patience, young padawan...

Tapered Slot Aerial

A tapered slot traveling wave aerial proved to be quite amazing and it worked over 1 to 3 GHz.  These type of aerials were designed half a century ago for use in radar sets and work very well at higher frequencies of several gigahertz.  

It is in effect a two dimensional horn antenna.  The required band is at the lower end of its useful range and some experimentation was required to get a practical design.

The prototype aerial has a forward gain of about 8 dBi.  One can increase the gain by placing three antennas side by side (at 1.25 wavelength spacing).  In a simple array, only drive the centre one, the two parasites will squeeze the beam to 11 to 12 dBi - somewhat like a Yagi aerial.

Using a power divider to drive two antennas, doesn't work any better than the above, since the divider, cables and connectors are lossy and a power divider is much more expensive than a third antenna!

I ended up making a collection of test aerials of various shapes and sizes (seven!), measured them all and eventually settled on the following design rules related to the low end frequency f1, centre fc and high end f2 (or λ1, λc and λ2, for the chosen PCB substrate permittivity), with the tapered slot on the front and the microstripline feed on the back of the PCB and established the following rules of thumb:

1. Slot width (Ws), Infinitesimally small = 5 mil
   
2. Microstrip width (Wm), 100 Ohm strip = 0.7 mm 
   
3. Microstrip taper to 50 Ohm (Wc) for the connector = 3 mm
   
4. Slot stub radius (Rs), 0.19 of the centre wavelength = 13.6 mm
   
5. Microstrip stub radius (Rm), 0.23 of the centre wavelength = 16.4 mm
   
6. Slot opening width (Wo), 1.1 of the low end wavelength λ1 = 92.5 mm
   
7. Tapered length (Lt), 3.5 times the centre wavelength λc = 250 mm
   
8. Ground side width (Gs), 0.25 times the centre wavelength λc = 17.9 mm
   
9. Ground back width (Gb ), 0.25 times the centre wavelength λc = 17.9 mm
   
10. Substrate height (h), the design is not sensitive to the thickness, FR4 = 1.6 mm
    
11. Substrate relative permittivity (Er), a higher permittivity reduces the size of the aerial, FR4 = 4.4

The above simple straight tapered slot design will be good enough for most users.  The slot is on the front of the board and the microstrip feed is on the back side.  The electric field of the strip couples to the orthogonal magnetic field of the slot where they cross over - magic! 

There are a few more optimizations that can be done, to reduce the size of the side lobes and eek out another 1 or two dB in forward gain.  In general though, the main problem with this design is getting the low end frequency low enough, the high end takes care of itself. 

A low frequency design cannot be made small - it has to be electrically large, so that 3.5 times λc is the minimum length and the opening (mouth of the horn) must be larger than λ1.  The only way to reduce the size of the construction is to use a board with higher permittivity, which will also be slightly more lossy - there is no free lunch.

The quarter wave stubs are flared 90 degrees, to increase the bandwidth - these could also be circles - no performance difference. 

If the slot is electrically long, then the little opening will have an impedance of 100 to 130 Ohm.  For high power operation, one can solder a co-axial cable directly onto it and tune the radio for maximum smoke, or for low power operation, drive it with a matched impedance stripline on the other side of the board.

There are little optimizations that one can do with slots and cut-outs, but the rule of diminishing returns apply and getting the slots right is very time consuming, while getting them wrong can make it worse.

An exponential flared slot (Vivaldi) looks cute and can make the antenna shorter by 10 to 20%, but the low frequency response then suffers, so there is much to be said for keeping the taper straight and simple.

Log Periodic Aerial

A problem with an ultra wide band antenna such as the tapered slot above, is that it could pick up significant out of band noise, which may desensitize the radio receiver.  

 

It is not a problem if you are patrolling over the sea, or a desert, but over Europe, with cell phone towers galore, an antenna with less band width may be better, which lead me to a log periodic design.  

 

This type of antenna is easier to limit to the desired band.  The six dipole elements will provide a forward gain slightly over 6 dBi.

A stack of prototypes arrived from China and the  VSWR plot showed that they are nicely band limited from 1700 to 2400 MHz.

 

Antenna Patterns

I made a couple of 2 GHz dipole antennas for reference (72 mm of thick copper wire) and used them to verify the antenna radiation patterns and gains.  A dipole has a gain of 2 dB, so one can measure and convert dBd (relative to a dipole) to dBi (relative to isotropic).  These type of measurements are very hard to do, due to radio and thermal noise.  

Since I live in a little village outside the city, my test range is simply my back yard and a spectrum scan at 2 GHz confirmed that I do not need a shielded box - there is nothing out there.  I could therefore set things up on a big wooden table and confirm that the antennas performed as expected.

The log periodic antenna has a forward gain of about 6 dB and a front to back ratio of about 25 dB.  The 3 dB beam width appears to be around + and - 20 degrees.

The tapered slot antenna has a  forward gain of about 8 dB and a front to back ratio of about 25 dB.  The 3 dB beam width is around + and - 15 degrees.

Both of these antennas will certainly allow you to tune your Doodle Labs radio anywhere over the 1700 to 2400 MHz band without having to swap the antenna.  Compared to a regular little stubby antenna, you should be able to fly your drone twice as far before going out of range, causing it to return home.

Range Extension

For even longer range, a log periodic with 14 elements should be doable - if someone wants one, let me know.  Otherwise, you could stack three antennas side by side, 216 mm apart and drive only the middle one.  Leave the two outside parasitic antennas floating.  That would give you another 3 dB gain.  

Note that the more gain you have, the narrower the beam becomes and the more accurately you need to point the antenna array at the drone and therefore you may eventually want to look for an azimuth antenna rotator from SPID or Yaesu. To track a UAV over long distance, you do not really need to rotate in elevation, unless you use an enormous parabolic dish reflector and even then it is not really necessary, since when you are close, the side lobes will connect.

Normally, you should head for the hills with your ground station, but a five to ten meter mast, to elevate the antenna and radio above local obstructions, will do wonders.

Also, don't point the antenna exactly horizontally.  Rather slant it about 5 degrees up.  That will reduce the influence of multi-path reflections off the ground.  There is also an effect called tropo-scatter, so your radio may work over the horizon while pointed slightly up.

FR4 PCB Properties - Er vs Eeff

FR4 PCB is as tough as an old tractor tyre and it has a permittivity of about Er = 4.4 (or higher) at these frequencies, so it is a good practical choice for this aerial.

One formula describes almost everything: λ = c / f sqrt(Er)

There are stripline calculators online at Microwaves101 and Pasternack and of course as part of KiCAD, which may all be handy:

• https://www.microwaves101.com/calculators/1202-stripline-calculator
• https://www.pasternack.com/t-calculator-stripline.aspx

 Here are stripline formulas that describe the rest:

The last formula above is interesting, since it is possible to measure the effective permittivity Eeff with a VNA, to confirm that it actually is what it is thought to be and then convert it to the bulk relative permittivity Er after a bit of algebraic formula massaging you will find that Eeff = 0.64 x Er + 0.36 for a surface stripline.

Note that: 1 < Eeff < Er, because with a stripline, >half the electrical field goes through air and <half through the substrate.

If one would make a long stripline on a piece of scrap PCB and solder a connector on one end only, leaving the other end open, then carefully calibrate the VNA and go to the Phase plot, the energy will reflect back from the far end and one will see standing waves with a sequence of zero crossings, for each harmonic.  At the first crossing, the wavelength at that exact frequency is 2L and then Eeff = (c / 2 L f)^2, which for FR4 is about Eeff = 3.17

The problem with the straight stripline above, is that you don't know exactly where to measure the length from - the tip of the SMA connector is the best, but this test setup is bound to be inaccurate.  It will give a good idea about the permittivity of an unknown board, but a better way to do it, is to make a circle and excite that through a small capacitor (a tiny gap).  One can calculate the circumference of the circle easily enough and then one will have a more accurate measurement of Eeff and the Er of FR4 will be just about Er = 1.39 x Eff

Similarly the Velocity Factor for a surface track on FR4 is about Vf = 1/sqrt(Eeff) = 1/sqrt(3.17) = 0.56

A Matter of Scale

The Velocity Factor Vf is important to scale an air wire antenna design to a PCB with dielectric on one side of the tracks, then all the wire dipole elements become shorter by 0.56.  A more obscure effect is the velocity factor for the space between elements of an array antenna (where there is no copper, only dielectric) as for a PCB Yagi or Log Periodic made on the surface of a FR4 PCB, is about 0.8. 

Therefore a FR4 PCB antenna is much narrower (60%) and only somewhat shorter (80%), compared to an equivalent air wire antenna.  It took me ages to figure this out!

Production

I sketched the designs with the KiCAD footprint editor and since wide band antenna designs are not super sensitive to the exact PCB properties, I ordered prototypes from Dirtypcbs (https://dirtypcbs.com/store/pcbs) in Hong Kong.

Both designs work remarkably well.  I only need to improve the aesthetics of the solder masks and silk screens.

If you want one or two (or ten thousand!), send me a message and I will let you know when the designs are final.

Costing

Hobby users can buy one or two at cost, for 30 Euro each plus shipping - about another 30 Euro for courier charges.

Business users can buy the KiCAD design files for 3000 Euro, customize it, change the outline and mounting holes, all without involving me and go make as many as you want.

Invoicing and payment is through Paypal.

La voila!

Herman