Anyone who wants to develop a wireless IoT device must choose an antenna. Antenna datasheets are full of a lot of technical data, produced at the discretion of the manufacturer rather than to an established set of standards. This can make interpretation and comparison a challenge for even the more experienced engineer. in this blog post, we shed some light into the murky world of antenna specifications.
Table of contents
- Typical antenna selection decision process
- Antenna selection based on return loss
- Selection on basis of the antenna size
- Selection by the efficiency of the antenna
- Selection by size and price without considering the technical data
- Selection based on the 3D radiation diagrams of antennas
- Selection on the basis of 2D directional diagrams
- The honest antennas
- Summary of antenna selection
- Services around antennas
Typical antenna selection decision process
The following paragraphs are based on our 20 years plus of experience with wireless projects. 20 years ago it was called M2M. Today it is called IoT. But the antenna does not know or care whether it is installed in an M2M or an IoT device. There has not been a lot of change to the licensed and unlicensed frequency bands in recent years either. Back in the day there was still no LPWAN with LoRaWAN, Sigfox, Mioty, NB-IoT or LTE-M. The innovative Sub-GHz Mesh-Net NeoMesh from NeoCortec didn’t yet exist either. We simply had GSM 900, GSM 1800, GPS and ISM 868 bands for our design for a people tracker. We set up a home zone at 868 MHz to reduce energy consumption, but with few suppliers for integrated antennas we had to develop our own. Today, there are plenty of suppliers of integrated antennas, but we still choose to develop the antennas ourselves for some projects either because it is cheaper or technically better.
Antenna selection based on return loss
Most developers look at the return loss of the antenna when making their selection. This is a correct first step. It’s good idea to check the number of components in any matching circuit. If there are a large number of passive components you should be immediately suspicious, because more components mean more losses in the matching circuit affecting the radiated power. DIY PCB antennas such as those designed by the Crout team do not need a matching circuit at all and therefore have a very high radiated power. In addition, when considering the return loss, you should check whether the measurements are made with a source that has a constant 50 Ohm impedance, or one that has been matched to the actual impedance of the antenna which may be different.
Selection by the efficiency of the antenna
For the second step, we need to look at the efficiency in average dB or in percent. Here lies another pitfall. Some supplier datasheets state the average of the peak value over all frequencies in all directions in 3D. Somewhere on the spatial view compared to the idealised isotropic omnidirectional radiator peak values are found. All the peak values are then averaged and named the average. The average is thus impressively large compared to the other manufacturers who also name an average. These latter use a different and technically correct average. Picking the highest number without understanding how it was arrived at can result in the wrong antenna choice.
If the efficiency targets in percent or dB are met, then the next point to consider should always be the size of the ground plane used to arrive at these numbers. For the same efficiency at the same frequency, LoRaWAN, Sigfox, MIOTY, NB-IoT or LTEM chip antennas require a ground plane that is 100 mm long or 140 mm long. This means that the average efficiency in the table in the data-sheet is correct and attractive for a design, but may only be valid for a very large ground plane. If the IoT developer’s device is much smaller than the dimensions used for the efficiency measurement given in the datasheet, then they will not achieve the advertised efficiency. With a smaller ground plane, the frequency bandwidth of the antenna also decreases. This is not easy to measure in the developer’s laboratory. In the certified test laboratory, measurements are taken on the band edges. The small PCB leads to a high return loss and thus to reflected radio waves. These then lead to mixed products at the output amplifier of the radio module. Harmonics occur and RED / FCC certification is failed and the unit cannot be sold and a redesign is required that can be expensive in both development cost and lost time to market.
Selection on basis of the antenna size
Another criterion for selection is often the size of the antenna. The whole unit is planned, the designer of the enclosure does their best, the colour of the enclosure is trendy and the unit is small. Some arbitrary amount of room is left for the chip or PCB antenna. If you elect to develop in this manner, you have to face the fact that there’s a good chance that no antenna you can find can will do the job and you will have to start again.
Selection by size and price without considering the technical data
Sometimes people just select with size and price. The technical data are not taken into account at all. The datasheet tells an efficiency of 10%, the price is OK and so is the size. The generally accepted target is an efficiency of 50 % at the band edges. With 23 dBm output power, -3 dB efficiency means that 3 dB of the output power is lost and 20 dBm are radiated. At 6 dB, the power is reduced to 17 dBm but the efficiency is then 25%. At 9 dB loss, the efficiency is then 12.5 %. One manufacturer of an NB-IoT / LTEM Eval kit actually states 10% efficiency. This is then a 10 dB loss in the antenna or the matching network. There will be additional losses in the enclosure or objects in the vicinity that must be added to this. To compensate for the 10 dB loss, you have to transmit with 10 dB more power. 3 dB is a doubling of the power. Since the voltage at the radio module remains the same, the current must double. To achieve 10 dB, the current must become more than eight times as large. The battery is there discharged more than eight times faster, or the battery must be significantly larger to provide the same in service lifetime. The larger battery increases the price and the small unit becomes larger again to accommodate it.
Another manufacturer’s datasheet states that a frequency band should have at least 20 % efficiency. However it does not specify whether it is 20 % at the corners of the band or 20 % in the middle. 20 % at the corners or 20 % in the average over the whole band is a big difference. The same manufacturer uses terms are either inadvertently or intentionally ambiguous: dB is used for efficiency and not dBi. The „i“ at the end of dB indicates an isotropic radiator as a reference. It could also be stated in dBd using a d for the dipole radiator. A temperature range that is given in degrees with no reference to measurement scale tells us nothing. Only by specifying Kelvin or Celsius does it become a valid value. The question is only which normal zero is assumed. If we make an appointment at 8 o’clock it is important to name AM and PM. If we arrange a telephone call on 2 continents, then we need to name the time zone so that we end up together.
Selection based on the 3D radiation diagrams of antennas
More complicated are the 3D directional diagrams. There are two allowed methods for simulation. One method is to set the output impedance to 50 Ohm and feed the antenna across the entire frequency band. Since the antenna reaches its requested impedance of 50 Ohm only at one point in the middle of the band, 25 Ohm and 100 Ohm at the band edges are still good and the requested 50 % efficiency is achieved.
In the other method, the output impedance of the generator is matched to the input impedance. If the antenna only shows 25 Ohm, then the impedance of the generator is adjusted to 25 Ohm. The simulation looks 3 dB better at the band edges as a result.
Very often the colour scale is changed within a datasheet. The colour range is set to match the measured values for a particular set of parameters, so for example dark red in the first diagram stands for 3 dBi and a similar red for 0 dBi in the next with a declining scale of colours in both from red to blue. The 3 dB difference is not always easy to read, particularly if the numbers on the scale are small. The sleight of hand is continued in further diagrams. The same antenna shows -6 dBi in another frequency band. In this band, -6 dB is then set as red. The 6 dB worse frequency range looks the same as the good range at first glance. If you look closely, you will see that dark red is always selected for the maximum value of the respective band. The 3D diagrams therefore cannot be directly compared via the colours you must pay careful attention to the scales on each.
The crowning glory is colourful 3D diagrams without naming the colour scale in the datasheet. They look impressive but are completely useless. The creativity of the manufacturers knows no bounds.
Selection on the basis of 2D directional diagrams
Another popular trick is to change the reference point in the directional diagram for antennas. If you change the reference in the middle carefully, you can turn an antenna with clear nulls into one that shows almost omnidirectional radiation. The diagram above illustrates this and shows the same antenna three times with different middle references. The LinkedIn poll on the right tracks the readers‘ ability to spot this. The majority of 51% have considered the almost circular diagram to be the best antenna and have been misled by it. 26% have guessed correctly. If 26 % guessed it right, then 74% got it wrong, hopefully those getting it right are the designers working with antennas and not the larger group.
Honest antennas, of course, can also be found. One supplier of honest antennas, for example, is akorIoT www.akoriot.com. Why are the antennas from this company honest antennas? Well, the datasheets for these antennas were created by Harald Naumann. All parameters in the datasheet are briefly explained on the last page of the data sheet. Furthermore, a scale is chosen in all two-dimensional and three-dimensional diagrams which enables the IoT developer to read the values in the diagrams and compare them between diagrams. The peak value of 1.28 dBi antenna gain in the three-dimensional diagram is shown in the table on the first page. If you convert the antenna effectiveness in dB into the antenna efficiency in percent, you will find that the two curves are congruent. How to convert one curve into the other is described in the datasheet. The datasheet of the antennas is not only a data sheet but also a small document with help and references.
Even more honest is the study called „Do it yourself PCB antennas for IoT devices“ by Harald Naumann. This 80-page DIN-A4 study explains in detail how mechanical changes to a PCB antenna, the PCB itself or the enclosure influence performance. After reading the free study, every IoT developer has the freedom to copy and use well-documented antennas.
The result of the study, funded by NeoCortec in Denmark, is a semi-automatic antenna generator software model that generates a well-documented custom antenna for €600 based on four standard antennas. The generator is fed with the information from the customer’s antenna order form. Afterwards, an antenna with the desired parameters of the IoT developer is generated by the software with default settings. This first antenna is usually already very close to the goal of the designer. This first-pass design is then changed manually and an antenna is generated again by the software model. Because these manual changes influence each other (Parameter A influences parameter B and vice versa and so on) the antenna is changed several times until the optimal result of antenna performance is achieved. Such an antenna then has 50 Ohm at the feed point in the middle of the band and no longer needs a matching network. Eliminating the matching networks avoids the problem that the more components there are in a network, the higher the losses become. Furthermore, the tolerances of the components add up. If these tolerances run against each other, higher losses occur in the matching network than in an ideal network without tolerances. Therefore, antennas without a matching network or with only one component in the matching network are clearly superior to chip antennas with 3-7 components in the matching network. The honest antenna is not only honest through its truthful data sheet but also technically better.
Summary of antenna selection
The examples for selection are not complete. However, the examples show that there are many ways to select the wrong antenna. Choosing the wrong antenna is an expensive mistake. Sometimes the error can be corrected by replacing the antenna. In the case of a chip antenna, this means redesigning the PCB. Sometimes the error can be fixed by a custom antenna. In some cases, the device cannot go into mass production because the antenna concept chosen was completely wrong and no working design in possible.
The developer of the wireless IoT device should be aware that there is no standard for antenna datasheets. It is therefore not possible to compare the data without experience. In part, the comparison is impossible because the ways the data you are using for comparison is derived are not even mentioned. The data always applies only to the manufacturer’s reference board. In your own IoT device, the supposedly worse performing antenna may end up being the better one. To find out, only a test setup with measurement of the return loss and, most importantly, the radiated power can help.
It is cheaper to get consulting from an external expert when choosing antennas. Crout offers this as a service. We have reference customers who selected the antenna together with us and developed the customised enclosure so that the antenna works perfectly. The antennas were measured and approved in the enclosure and on an empty PCB. In only 15 months of development, the IoT device with LTE, WiFi and LoRa of our customers received radio approval without any problems. We attended the radio approval test on behalf of the customer. In order to save costs and surprises, we make control measurements at the critical points before the radio approval.
We have reference customers who are very grateful for the rescue of their projects and have been requesting all custom antennas from us since. Other customers take advantage of our proven 0 USD PCB antennas wherever they can.
Services around antennas
- Consulting on antennas and analysis of the data sheets of all manufacturers
- Test set-ups and measurements with the selected antennas in the enclosure with an empty PCB
- Customised antennas in different manufacturing processes
- Antenna matching of antennas from all manufacturers
- Webinars, seminars and workshops on wireless IoT
- Support for IoT developers in the development of PCBs
- Pre-tests for RED / FCC and execution of measurements in the test laboratory
- Hardware-related consulting for the development of firmware („mini“ BIOS for IoT devices)
On request, we offer personal training on the selection of antennas and show in more detail known errors in the datasheets such as those we have discussed here. We have to do this face to face because the graphics are the intellectual property of the respective manufacturers and they certainly do not give their permission to discuss the errors.