Antennas and RF Engineering
Though I’ve enjoyed it a lot I haven’t made much money from consulting, so I’ve given it up. Last week I started a full-time job as an RF engineer with a large electronics company. As usual for jobs of that sort I can’t reveal practically any of my work. So, I’ll be ending this blog. I may occasionally write posts that don’t contain any content that could be considered sensitive or related to my current job.
There are several different schools of thought on the question “How should PIFAs be analysed?”, or “How do PIFAs work?”. I’ve found that some of them are useful and some less so, one day I’ll do a series of posts about them.
Right now though, what I’m interested in is the matching, specifically how the ground and feed pins affect it. According to some theories the ground pin plays a role in determining various antenna characteristics. What I’m concerned with here though is it’s role in affecting the input impedance. This part of the theory of PIFA antennas is really quite simple: if the ground pin is close to the feed pin then it forms a shunt inductance to ground. The size of that shunt inductance depends on the thickness and length of the ground pin and the metal between the two pins. So, the ground pin is effectively a matching element, it can be used along with the matching circuit in order to solve the overall matching problem. This is a very useful technique.
PIFAs with different ground and feed points
The diagram shows two prototypes for handset PIFA antennas. The groundplane is in green, though I haven’t shown it we’ll suppose that there is a full covering of groundplane below the PIFA area. The PIFA is in yellow, slots cut into the PIFA are in black, the feed pin is marked in blue and the ground pin is marked in pink. These pins could be “pogo” pins from the PCB below up to the PIFA.
The only difference between these two antennas is the position of the feed and ground pins, this is a big difference though. The smith chart or return loss chart for these two types antennas would be quite different. This is a dual-band design and both bands would be different. Even the antenna patterns would be different. Those are all subjects for another time. From the matching viewpoint the most interesting thing is that the path from the feed pin to ground is longer on the antenna on the right, and shorter on the one on the left. This is a shunt circuit so we should think in terms of admittance. From that view the design on the left has a shunt inductor to ground with higher admittance than the one on the right.
This fact gives the PIFA designer an extra degree of freedom in matching. Let’s suppose the matching circuit is a pi network of three components. Let’s suppose a Smith Chart analysis indicates that a shunt inductor is needed on the side of the pi network connected to the antenna. In that case the designer should think about changing the distance between the feed and ground pin. It is often much easier to change that inductance by altering that distance than by using the matching circuit. But, changing the pin positions also changes the antenna behaviour, making the problem more complicated. This added complexity though produces extra options which can be useful in design.
Often the pin positions are fixed early in a project before the PIFA design is anything like complete. Even in this case though the inductance can be varied by cutting a slot between the ground and feed points. This change will also affect many other characteristics of the PIFA, but it can often be made without difficult mechanical changes. This method can be used to reduce the shunt admittance (or if you prefer increase the shunt reactance), unfortunately there is no similar easy way to raise the shunt admittance.
Cutting a slot between the ground and feed points
Cellphones and laptops need multi-band antennas. Practically every cellular device must support at least two bands and most must support more. WiFi devices often need to support the 2.4Ghz and 5Ghz bands. So, there’s great demand for multi-band antennas today.
A multi-band antenna must have good radiation characteristics at all of the relevant bands, that’s the first challenge. I’ll talk about that another time. The second problem is to impedance match the antenna to the front-end. Often a lot of performance is left on the table because designers don’t understand multi-band matching. I was taught the right approach by Devis Iellici.
One approach is to design a matching circuit appropriate for one particular band. For example, suppose the unmatched antenna has worse performance in one band. The designer could make a circuit for that band and simply see what happens to the other bands. This is a bad plan since the effect on the other bands is most often detrimental.
A better method is to design the matching circuit with all bands in mind. For any narrow band there are several simple circuits that can improve the match, its best to find them all. I measure the Smith chart of the unmatched antenna and find the worst performing band using a radiation measurement such as efficiency. I then use Smith chart software to find all reasonably good matching circuits that can be made with three elements. I then pick the circuit that has the best performance in the other bands, that can be estimated using the Smith chart software too.
For example, suppose we have a quad-band PIFA antenna for cellular. It must cover 824Mhz-960Mhz and 1710Mhz-1990Mhz. Suppose the lower band has the worst performance. I would then find every three element LC circuit that matches the low band reasonably using Smith chart software. I’d then try each one out in the high band and see which performs best there. Then I’d implement it, test it and tune further using trial and error.
For this to work best the designer must take into account the parasitics of the components. Sometimes that can be done using the Smith chart software or by using a full microwave circuit analysis package.
The iPhone 4 has shown that hand effect problems can be very serious. This is generally true of antenna problems, I know of several handsets that were developed at huge cost and never shipped because satisfactory antenna performance couldn’t be achieved.
There are five directions this problem can be approached from:
Industrial Design
If the industrial designers and the antenna designers are prepared to be cooperative then that can help a lot. The industrial designers should to be flexible about the type and layout of the antenna. The antenna designers for their part should not try to build the easiest type of antenna possible, they should aide the industrial designers in achieving a good design. Generally the more cubic centimeters of space that are apportioned to the antenna the better the performance that the antenna designers can achieve on every spec. But, there are often more elegant ways to solve the problems than throwing CCs of space at them.
In some situations the industrial designer can design a case that encourages the user to hold the handset in a particular way. The case can encourage the user not to hold the antenna. For example, if there’s a bump around the rear-facing camera then the user is less likely to wrap their hand around it and the area above it.
Antenna Type.
The main types of antenna used in handsets are shown below:
Externals are rarely an issue these days, but they teach important lessons for internal antenna design. A retractable whip antenna is an external monopole that is parallel to groundplane and above it. The other main type of external is the stubby, these are a small round cylinder of plastic containing a type of shortened monopole antenna. Often stubby antennas are called “helicals” in books because it’s assumed that the antenna inside is a normal mode helix. Often though designers used other shapes, such as meanders. Anyway, both meanders and normal mode helicals are shortened monopole antennas. So, the two main types of external antenna, the retractable and the stubby are respectively a monopole and a shortened monopole. Investigations into these antennas years ago showed that the stubby antenna suffers more from the effects of the nearby head than the retractable. The reason for this is quite clear if you imagine a person holding a phone with both types of antenna. Part of the retractable antenna is close to the head, but all of the stubby antenna is close to the head. By compressing the antenna into a small space the stubby antenna creates a small space where the nearby dielectric materials have a large effect. So, stubbys suffer seriously from detuning problems, this problem carries over into internal antennas.
Internal antennas are often all called “PIFAs”, but the word is used for types of antenna that are really quite different to each other. Since PIFA means “planar inverted-F antenna” some people use it to refer to any antenna that has a feed and short pin that is made out of flat conductors. There are really at least two types though, firstly there are the sort built on top of a groundplane, those are somewhat like size-reduced patch antennas. Then there are the sort where the groundplane is parallel to the antenna but not underneath it, those are more like size-reduced monopole antennas. The diagram below is a side view of the internals of two handsets. The PCB material is in green and the groundplane is in orange.
The type with the groundplane underneath is what’s normally called a PIFA. The type with no groundplane underneath is closer to a monopole, even if it has a ground pin that only really affects the matching. Some types of normal mode helix have ground pins too. These two types of antenna have their advantages and disadvantages, it’s worth mentioning them all not only those connected with the hand & head effect problems.
The main lesson here is if you want the antenna to perform well in the presence of the head and hand then use an antenna with a groundplane.
Antenna Placement.
The conventional type of handset antennas I’ve discussed here only work well if they’re at the top of the handset or the bottom. They don’t work well in the centre of the PCB. So, the choice is between the top and bottom.
Generally users hold their phones by the bottom part, so hand detuning is worse if the antenna is at a the bottom. Sometimes the ID designers can help here by creating a case that encourages the user to grip the phone in a different way, but that’s difficult. Similarly, head detuning is generally worse if the antenna is at the top of the handset.
PIFAs with groundplanes below can be put at the top or bottom of a handset. If they’re put at the bottom then the user is likely to cover the antenna with their hand and cause serious attenuation and detuning. If the antenna is put at the top then the head is on the groundplane side not the antenna side, so it causes less detuning. So, this type of antenna should be put at the top.
The situation is different for antennas with no groundplane below them. These will generally fail the SAR test if they’re put at the top, so they have to be put at the bottom for that reason. This means that these type of antennas have a two-fold problem. They are sensitive to detuning by dielectric objects and since they generally have to be put at the bottom of the handset they are more likely to be close to the hand or even covered by it.
The main lesson here is to use a PIFA with a groundplane and put it at the top if you can. The situation for slide handsets and clamshell handsets is quite complicated, I might discuss that in the future in another post.
Details of the Antenna Design.
The closer the antenna is to the outside of the case the more it will be affected by proximity detuning. But, if the designer leaves a gap between the case and the antenna then that’s forsaking space and it will probably decrease the antenna’s performance in empty space. This trade-off has to be made on a handset-by-handset basis. It’s difficult and can require a lot of prototyping and simulation.
Another problem is areas of high E-field. Generally the sensitive parts of an antenna follow the E-field. If the E-field is high in a particular region then a dielectric object close to that place will cause a large frequency shift. It’s best to avoid creating such regions as much as possible, that can be done by widening out conductors on the antenna close to the high E-field region. Another possibility is to move the region further inside the handset.
There have generally been two theories about the iPhone 4’s problem. Firstly, lots of people online have said that dampness on the users fingers causes a conductive bridge over the gap in the case. As I said in previous posts I don’t think that’s very likely. Secondly, when the user touchs the gap that causes detuning. I think the detuning explanation is more likely, but the conduction explanation could be right, or it could be a bit of both. In either case this demonstrates a problem with putting the antenna on the outside of the case.
If GPS, Bluetooth or WLAN antenna is needed then it’s best to use a separate antenna for those bands. That helps the detuning problem because it makes the main antenna design easier.
Other Influences such as the Radio.
The output amplifiers in the transmitter part of the radio suffer from the problem of “load-pull”. The amplifier is built to work when applied to a particular output impedance which is generally 50 ohms. If the antenna’s input impedance is different than 50 ohms then there is mismatch loss. But, there’s also an additional loss caused because the amplifier’s performance degrades, this is load-pull.
Suppose due to detuning the VSWR on a particular channel gets worse, it goes from 3:1 to 2:1. The amplifer may not be able to supply full power when mismatched 2:1, so it’s power output may drop by a dBm or two. That would cause a big drop in performance. So, an amplifier with better load pull characteristics gives better performance when there’s detuning. This was a great problem in the early days of digital cellular in the 90s and early 00s, amplifiers then had much worse load-pull characteristics than those we have today. But, the problem has never completely gone away. (Exactly why it’s never gone away is a story for another time).
I’ve seen a few folks try to redesign the matching circuit to cope with detuning better. I’ve never seen this enterprise succeed though. I don’t think there’s much that matching circuits can do to overcome this problem.
ESD protection circuits are important if metal parts of the antenna are exposed to the user. The radio must be protected from static shocks that can occur if the user accumulates a charge by walking across a carpet for example.
Cellphones today cover many bands. A cheap European handset will cover the European GSM/EDGE bands at 900MHz and 1800MHz, the US PCS band at 1900MHz and the European WCDMA/UMTS band at 2100MHz. A cheap US handset will cover the US 800MHz and 1900MHz bands, GPS at 1575MHz and often the European 1800MHz band too. But, these handsets don’t compare to high-end Smartphones, those often support all of the bands I’ve mentioned above, that’s six bands. Such phones are called “world phones” because they support so many bands that they can be used almost anywhere there is a cellular network. Supporting all those bands requires a large antenna especially if a PIFA with a groundplane below it is used. This often results in detuning taking a back-seat in antenna design because band support becomes the big issue.
This problem can be solved by regional variants. The makers of cheaper handsets tend to make a US version and a European version. Most countries use the US or European band layout, so doing that provides a large market. These phones will still work if taken abroad because there is provision for that. For example, a typical European handset will work on the 1900MHz PCS band in the US, assuming the carriers involved have agreements. If Smartphone designers were to take this approach then they would have to build more handset variants, but it would make the antenna less difficult. It would make with-groundplane PIFAs more practical and they were used that would help prevent detuning.
A Concluding Thought…
If handset designers paid a little bit more attention to the issues above then the user experience would be better. From the user’s point of view “network coverage” would magically increase. But, users mostly blame networks for poor coverage rather than handsets. That means the handset companies often have little incentive to improve. But, after the well-publicised iPhone 4 problems users may be more critical of handsets in the future.
Apple showed a group of journalists around their Anechoic Chamber facilities. There are articles about the visit at Engadget, Macworld, New York Times and Techcrunch. Normal folks can learn a little from those articles about how antenna design is done. Antenna folks can get jealous and wish they had access to so many chamber.
Most large wireless companies have similar facilities, though most are not anything like as big. The photos show an ETS tapered chamber and a Satimo SG chamber. I commissioned an ETS 8600 in a previous job.
I don’t know what type the enormous cylindrical chamber pictured at the start of the New York Times article is. I’ve never seen a chamber like that before. It seems to have a series of range antennas in a ring around the wall.
At a press conference today Apple have agreed to give iPhone 4 customers “bumpers” to ameliorate the antenna performance issue. They have also agreed that customers can return their phones and contracts if they aren’t happy. I think this is a good strategy it shows consideration for the customer.
Steve Jobs showed that other phones suffer from similar signal strength issues. He demonstrated some other phones losing bars when held, he used the Droid Eris (called the HTC Hero in Europe), Blackberry Bold 9700 and Samsung Omnia II. See this article. Certainly other phones do suffer from this problem, as I wrote earlier. But, the experiment shown doesn’t really demonstrate that all these other phones suffer from the problem to the extent that the iPhone does.
As I mentioned previously there is no standard for signal strength reporting. Some manufacturers are quite conservative about it, some less so. A signal strength one phone displays as three bars another may display as one bar even though the two phones have similar performance.
Secondly, it is unlikely that the demonstration relied on a normal base-station. It’s unlikely that the space for the press conference is in a low-signal area. So, I expect Apple simulated the condition by using a base-station simulator. That’s perfectly legitimate and I would have done it too. But, it allows the channel selected to be picked for the demo. Apple could have gone through the hundreds of channels for WCDMA and EDGE that those three phones support. They could then have picked the channels that are affected the worst by hand detuning and used those.
Lastly, the three phones could have been selected because they suffer particularly badly from hand detuning.
I don’t know that Apple have done any of this sort of gaming of the test. But, they could have done it. What Jobs has shown doesn’t prove much.
Spencer Webb at AntennaSys has written a good post on the iPhone antenna problems. What’s he’s said about detuning and attenuation is very similar to what I’ve said in other posts. I doubt this is because he reads my blog, it’s more likely because we’ve both had similar experience with handsets before and know many of the same things.
I have been thinking about this more though, and there is a possible explanation for why damp hands may cause the problem. As I said earlier at GHz frequencies damp hands don’t conduct well. But, what if the problem is at much lower frequencies? It could be that the WLAN/BT/GPS antenna is polluted with low frequency interference in the KHz or MHz region. Then, when the user puts a damp finger over the gap that interference is conducted to the cellular antenna. This would only cause a problem if the cellular antenna radio were sensitive to the low frequency interference, that shouldn’t happen but it could happen if mistakes in the design were made.
That would be a very odd cause though, I still think detuning is more likely.
Apple are claiming that the problem with the iPhone 4 is only down to the signal strength display. But, many users have demonstrated that the drop in signal strength affects both data throughput and can cause dropped calls. As I mentioned yesterday Anandtech have measured the signal strength drop and found that it’s very large.
Apple have said:
Upon investigation, we were stunned to find that the formula we use to calculate how many bars of signal strength to display is totally wrong.
Our formula, in many instances, mistakenly displays two more bars than it should for a given signal strength. For example, we sometimes display four bars when we should be displaying as few as two bars.
This does make some sense. The users affected may be in poor signal areas that are being reported as good signal areas. So, the degradation is exaggerated. However, that doesn’t show that Anandtech or the other testers are wrong. Anandtech’s test didn’t depend on the bar display, they used a piece of software that extracts the signal strength in dB from the radio.
Also, the problem is reported to affect both WCDMA(3G) and GSM EDGE. The signal levels involved in these two protocols are quite different. The algorithm or formula that takes signal strength information (and sometimes other info) and produces the display of signal strength bars is generally different. That is, WCDMA has an algorithm to produce the bar display and EDGE has a different algorithm.
The folks at Anandtech have hacked an iPhone 4 to produce a signal strength indication in dB. They’ve then experimented to find the signal strength drop when the phone is held in various orientations. See the article here. Since they’ve measured this with a normal network in a real-world environment I don’t expect that the results will be very accurate. The propagation path between the test site and the nearest base-station will sometimes change. However, the measurements give a ball-park idea of what the drop in signal is like. If the handset is “cupped tightly” the drop in signal-strength is 24.6dB, it it’s “held naturally” the drop is 19.8dB.
One of the handsets they compare it with is the Nexus One. Interestingly the Nexus One doesn’t do too well either.
