4.2.1 Single‐Band PIFA

Shown in Figure 4.10 is a planar IFA (PIFA). PIFAs can be thought of as a mutation of IFAs. Both IFA and PIFA have a ground strip and a feeding strip. By replacing the radiating strip of an IFA with a patch, we get a PIFA. In most cases, the patch of a PIFA is above the ground plane.

In all the following samples in this section, the width of both feeding and grounding strip is 1 mm; the dimension of PCB, which is used as the ground, is 50 mm × 120 mm. If the PCB size is changed, the antenna response will change accordingly. The ground’s impact on a PIFA is the same as that on an external antenna. Refer to Section 3.4 for further details.

In Figure 4.10, the dimension H is the distance between a patch and a ground. The S is the edge‐to‐edge distance between feeding and grounding strips. Similar to IFAs, the matching of antennas can be tuned by adjusting S.

Let’s start with single‐band PIFAs. Shown in Figure 4.11 are simulated results of three PIFAs. The dimension L of all three antennas is 50 mm, which is the same as the PCB width. The dimension W of three antennas is 5, 15, and 25 mm, respectively. To best match individual antennas, the dimension S, the distance between feeding and ground strips, is tuned case by case. The optimal S for 5, 15, and 25 mm patches are 1.5, 4, and 4 mm, respectively. When the W is 5 mm, the antenna is more like an IFA than a PIFA. With the increase in W, the resonant frequency of a PIFA decreases. This phenomenon gives us a hint that, unlike an IFA, the resonant frequency of a PIFA is not solely decided by the patch’s length.

For a PIFA, the resonant frequency is proportional to the summary of its length and width. As shown in Figure 4.12, a PIFA’s width plus its length is roughly a quarter of a wavelength at its resonant frequency.

When discussing a single‐band PIFA, most books prefer to give a current distribution which shows two clear current flows. Both of them start from the upper‐left corner and end at the bottom‐right corner. One of them passes the bottom‐left corner and the other one passes the opposite top‐right corner. That is a quite intuitive but coarse simplification. The real current distribution is a little more complex than that. In many cases, it is difficult to see a clear current pattern when you simulate a PIFA.

Shown in Table 4.1 is a comparison between three antennas. The second column in Table 4.1 is the calculated resonant frequencies based on the assumption that W + L equals to a quarter of wavelength. The third column is the actual frequency. The fourth column is the ratio between the calculated frequencies and the actual ones. The actual resonant frequency of a patch is normally lower than the theoretical one. That is due to the loading effect of the edge capacitance between patch and ground.

The critical parameter, which controls the bandwidth of a PIFA, is H, the distance between a patch and a ground. Shown in Figure 4.13 are the reflection coefficients of three antennas with different heights. All patches have the same L and W, which are 50 and 25 mm, respectively. The H of three patches is 5, 7, and 9 mm, respectively. All three antennas are individually tuned, their S are 3, 4, and 7 mm accordingly. By increasing the H from 5 to 9 mm, an antenna’s −10 dB bandwidth is almost tripled.

As a summary, there are three design guidelines for single‐band PIFA:

1. The dimension H is the most critical parameter which decides an antenna’s overall performance. However, because H is directly related to a device’s thickness, it is a tough task to bargain about it with industry design engineers or mechanical engineers. If there is only a 5 mm height clearance and a quad‐band GSM antenna is required, then PIFA is not an appropriate candidate. A folded monopole antenna, discussed in Section 4.3, is a better choice.
2. By adjusting the total length W + L, the antenna’s working frequency can be tuned. In practice, the L should be similar to the PCB’s width that results in wider bandwidth.
3. The dimension S is the tuning parameter which optimizes an antenna’s matching. One can either use a wider strip with a larger S or a narrower strip with a smaller S to achieve the same matching. A larger S is better for manufacturing, because a fixed process tolerance causes less electrical variance in the antenna. From a mechanical point of view, a narrower strip is better for spring finger type of designs, because it is more flexible and it is easier to achieve the required spring force.

4.2.2 Multiband PIFA Antenna with Slits

If you have already discovered PIFAs, then you might find that the following explanation is somewhat different from other books. All versions of PIFA explanations are partially correct, including the following version. As a PIFA is a distributed radiating system, too many parameters play a role in deciding the antenna’s characteristic. To explain such a complex matter, one has to substantially simplify it and that is the reason for the different explanations.

There are several ways to design a multiband PIFA. The most popular way is to cut slits on the patch, as shown in Figure 4.14. Similar to a single‐band PIFA, a dual‐band PIFA also has a grounding strip and a feeding strip. Two critical dimensions, D and C, are used here to describe the slit. The dimension D is the distance between the patch’s corner and the opening of the slit. The dimension C is the slit’s length. The dimension P is the distance between the horizontal part of slit and the edge of patch. The P is not really a critical dimension; however, it decides the shape of the slit.

Shown in Figure 4.15 are the decisive paths of different bands. Figure 4.15 only serves the purpose of predicting the trends of frequency shifting when different dimensions are adjusted. One should not try to use the illustrated critical path shown in Figure 4.15 as a way of calculating a patch’s resonant frequency. In reality, trying to predict a patch’s resonant frequency with a calculator is not really an efficient design approach. There are always so many unknown parameters, such as the permittivity of both the antenna support and the phone cover, the electrical property of nearby objects, and so on. Usually, all these unknowns can make the theoretical calculation far removed from the correct answer. A better approach is to memorize the effect of each parameter first, then make a PIFA with a piece of copper tape, install it on a mock‐up phone, cut a slit on the PIFA, and tune it by trial and error.

The impacts of the two critical dimensions are listed below:

1. The distance D can affect the resonant frequency of both the lower and the higher bands. It always shifts two bands in the opposite directions. For example, if we decrease D, the critical path of the lower band will be shorter, thus increasing its resonant frequency. In the meantime, the critical path of the higher band is increased, thus inducing a lower resonant frequency.
2. The slit’s length C only influences the higher band. The slit is about half the total critical path of the higher band. Increase C can lower the high band.

Of course, in the real world, the effect of any dimension change cannot be exclusively constrained in a single band. In fact, it is safer to claim that it has more impact on some bands than others. To help understand the design rules, the following design examples are provided. In all the following examples, the area of patch is 50 mm × 25 mm; the area of ground is 120 mm × 50 mm. H, which is the distance between ground and patch as shown in Figure 4.10, is 7 mm. The impact of D is illustrated in Figure 4.16. By adjusting D, both low and high resonant frequencies can be tuned simultaneously and in opposite directions.

Illustrated in Figure 4.17 are the impacts of slot length C. With the increase of a slot’s length, the resonant frequency of the higher band decreases. It can be seen that the resonant frequency at the lower band still drifts a little bit. For all three cases shown in Figure 4.17, the dimension P is fixed, the modification of C is realized by only increasing or decreasing the length of the horizontal portion of the slit.

Shown in Figure 4.18 are the results when the dimension P is changed while keeping the slit length C fixed. Although not significantly, the dimension P also has some impact on the higher band. It can be used in some circumstance as a means to provide the “extra mile” for tuning.

Shown in Figure 4.19 are the radiation patterns of a PIFA. At the low band, which is 0.78 GHz in this case, the radiation pattern is similar to a dipole’s. The radiation is mostly generated by the vertical current along the edge of the ground plane. The vertical polarized field E_θ is the dominant field component. At the higher band, which is 1.75 GHz, the amplitudes of E_θ and E_φ are in the same order of magnitude. Shown in Figure 4.19e is the radiation pattern of the total E field at 1.75 GHz. It confirms again that it is the ground plane, instead of the antenna element itself, which definitively decides the radiation patterns.

When designing any kind of antenna, antenna engineers may find themselves in a situation where one band has more than enough bandwidth but the other band cannot meet the specification. Knowing how to exchange bandwidth between bands is an essential technique. For a PIFA, cutting the slit shape differently can have some effect on the bandwidth ratio between bands. The other way to change the ratio is by adjusting the slit’s width. Shown in Figure 4.20 are the simulated results of three PIFAs with the same dimensions but different slit widths. With the decrease in the slit’s width, the bandwidth increases at the lower band and decreases at the higher band.

If an antenna still cannot meet the specification after the performance of both bands have been balanced, the only way out is to attempt to increase the patch’s height, which is the dimension H shown in Figure 4.10a. This definitely is an uphill battle.

So far, we have shown how to adjust both bands simultaneously or modify the high band independently. By combining both methods, we can tune the resonances freely in a quite wide range. Now let’s discuss some matching techniques. The easiest way to match an antenna is to design a matching circuit. However, a matching circuit always has some inherent loss. For a patch antenna, some of the dimensions can be modified, so we should always try to achieve a good match by tweaking the patch itself first. By comparing Figures 4.16 and 4.17, we can see that there are different ways to cut the slit to achieve two required resonant frequencies, and their matching conditions are also different.

Besides the patch itself, another useful tuning mechanism is the feeding and grounding structure. Shown in Figure 4.21 is the equivalent circuit of feeding and grounding strips. The grounding strip is equivalent to a shunt inductor, so it has more impact on the lower band. The feeding strip is equivalent to a series inductor, and it can be used to tune the high band. To increase the inductance of grounding strip, the following methods can be used:

• Increase the length of grounding strip by bending the strip.
• Decrease the width of grounding strip, t₂.
• Increase the distance between feeding strip and grounding strip, S.

Similarly, the inductance of feeding strip can also be increased by increasing the length or decreasing the width of the strip. Shown in Figure 4.22 are two other ways of matching. Figure 4.22a shows how to use the slit to increase the shunt inductance. Figure 4.22b shows how to add some distributed shunt capacitance. Another effect of the stub shown in Figure 4.22b is to increase the total current length, thus decreasing the overall resonant frequency. If you have a brainstorming session on this topic, lots of matching structures can be invented. Where should we stop and hand over the remaining tasks to lumped matching element? There is no right or wrong answer to this question. Normally, it is a purely personal choice. Some engineers prefer a self‐matched design without requiring any lumped components. Many of those engineers treat antenna design as an art. Some engineers like to reuse previous designs and want to push dimension changes to the minimum. For them, a matching circuit is the favorite choice.

Putting the personal preference aside, there are some objective criteria that can be used to evaluate any antenna design. How sensitive is a design to manufacturing tolerance? What kind of manufacturing process can it use? Is a design easy to be tuned when frequency drifts? Does frequency tuning require expensive and time‐consuming tooling modifications? What is the efficiency of an antenna?

It is not difficult to imagine that the weight of each criterion varies from project to project. For a low‐end phone, cost might be the first consideration. For a high‐end phone, it is most likely that all physical constraints have been pushed to the limit; the only goal for an antenna engineer is to meet the specification at any cost. The duration of a project is also an important factor when selecting different approaches. In fact, most issues mentioned here are not academic at all, they are purely practical matters. Section 4.2.5 is dedicated to this topic.

Referring to Figure 4.15, if the patch size gets smaller, to maintain the resonant frequency at the lower band, the opening of the slit needs to be moved closer to the feeding point, and the slit must be routed through the other direction. Shown in Figure 4.23 is an example of such a small PIFA. The rules for tuning the small PIFA shown in Figure 4.23 are pretty much the same as those shown in Figure 4.15. The critical dimensions are the slit’s length and the opening position of a slit. The length of the slit has more impact on the higher band, while the slit’s open position affects both bands.

Shown in Figure 4.24 are the simulated results of a small PIFA when the slot length varies. By decreasing or increasing the slot length by 2 mm, the resonant frequency at a higher band is shifted up or down, respectively, by around 35 MHz. Meanwhile, the lower band response is relatively immune to those changes.

So far, we have discussed how to design and tune a PIFA. The topic of current modes on a patch has deliberately been avoided in order to minimize confusion in the first stage of learning. Although the current distribution on a patch can provide a very useful insight on how different patches work, it frequently makes a new engineer confused about how to tune a PIFA. Of course, if we want to make some innovation in the field of antennas, current modes is a topic we must discuss.

The current distributions of a “normal” PIFA at both low and high bands are illustrated in Figure 4.25. The dashed lines in Figure 4.25 are the significant paths. At the low band, the current distribution along the critical path is more like a quarter‐wavelength monopole. The current reaches its maximum at point A, which is around the feeding point. At point B, the current decays to zero.

In the high band, the current is more like a dipole along the critical path. It has the maximum in the middle, which is marked by point A. At both ends, which are marked as point B, the current decays to zero. As we know, a monopole antenna is not able to resonate when the antenna length is half of an effective wavelength, because its port impedance is too high to be matched. On the other hand, unlike a monopole, which is fed from one end of a radiator, a high‐band patch mode is accessed from the middle, which explains why it has a decent impedance matching.

Although with regard to antenna tuning, a “normal” PIFA and a “small” PIFA are quite similar, their current distribution is totally different. Shown in Figure 4.26 are current distributions of a “small” PIFA. The critical current paths of a “small” PIFA at both bands are the same. At the lower band, the current along the critical mode is similar to a monopole, which has the peak current at point A and zero at point B. At the higher band, the total critical path corresponds to the third‐order mode, which is three‐quarter wavelength. There are two peak current spots and two zero current spots. One might wonder whether the resonant frequency of the third‐order mode is three times its basic mode. To explain that, we need to recall the technique we discussed in Section 3.1.2.2. There, by adjusting the pitch of a helix, the resonant frequency of the third‐order mode can be tuned in a wide range. Similarly, the third‐order mode of a “small” PIFA can also be tuned to where we want.

Although mathematical formulas are given in many papers or books to calculate resonant frequencies of different bands, sometimes we do not need them. There are so many unknowns when designing an antenna for a real device; no formula can give an accurate prediction. The best way is to find out what the effect of each parameter is, then tune the antenna accordingly.

4.2.3 Multiband PIFA with Separate Branches

When the antenna area is big enough, a better design approach is to use two independent branches to cover low and high bands separately. The antenna shown in Figure 4.27a is such an example. The antenna’s size is 60 mm × 25 mm. The ground’s size is 60 mm × 120 mm. The antenna is divided into two branches by the grounding strip. In Figure 4.27a, the feeding strip is located on the side of longer branch. This arrangement is more favorable to the performance of the lower band.

As the two branches are far apart, the responses at the low and high band are decoupled. The tuning process of this antenna is quite easy. When designing a large antenna, the first step is to adjust the length of both branches to get the resonances at the expected bands. Then by tweaking the distance between the feeding strip and the grounding strip, the matching at the lower band can be tuned. The final step is to design a matching circuit as shown in Figure 4.27b to obtain a good matching at the higher band.

Shown in Figure 4.28 is the simulated reflection coefficient of the large PIFA. By comparing this result with results shown in Figures 4.16 and 4.24, it is clear that the antenna has a wider bandwidth at both bands. As a rule of thumb, a larger antenna area is always better, which gives more design options, and it is easier to achieve better performance.

4.2.4 Multiband PIFA with Parasitic Element

There are many ways to expand PIFA’s bandwidth, which include using one slit with branches or multiple slits. PIFA variants using those techniques are too many to be enumerated in the book. Interested readers can find some enlightening examples in Professor Wong’s book [1]. The technique we are going to discuss here uses the parasitic element.

Shown in Figure 4.29 is a PIFA with a parasitic element. The parasitic element is electrically connected to the ground through a metal strip. There is no direct connection between the main radiator and the parasitic radiator. However, they are electromagnetically coupled.

Shown in Figure 4.30 is the simulation result of a PIFA with a parasitic element. Similar to a normal PIFA, the main radiator generates two resonances, one at the lower band and one at the higher band. The parasitic element only resonates at the higher band. The parasitic elements can frequently be found in penta‐band 3G phones, which normally require frequency coverage over 1710–2170 MHz at the higher band. A quad‐band 2G phone only needs to cover 1710–1990 MHz.

In practice, the parasitic element is normally used to cover the highest band. As higher frequency means smaller parasitic radiator, this makes antenna design easier. The other reason for using parasitic elements at the highest band is an antenna’s overall performance. With the existence of parasitic element, the efficiency of the main radiator might degrade from one‐tenth of a dB to a few dB. By assigning the resonance of a parasitic element to the highest frequency, the adverse effect on the lower band can be mitigated.

4.2.5 Manufacturing PIFA Antenna

Regarding designing techniques for PIFAs, this is as far as the book will go. The aim of the book is to aid in jump‐starting an antenna project. The book mostly focuses on the core principles of antenna designing. However, there are so many phones on the market, that if you do some reverse engineering, you can discover various elegant tricks. The book does not touch upon those advanced techniques. On the academic side, Professor Kin‐Lu Wong is one of the most innovative scholars in the mobile antenna world. Professor Wong has written two books [1, 6]that cover advanced techniques.

In this section, we are going to discuss the various processes of manufacturing PIFAs. The cheapest technology used to make an internal antenna is metal stamping. In production, a metal stamping antenna can be tuned quite quickly if the parameter needing to be adjusted is known and already included in the tooling design. The minimum number of parts necessary for a metal stamping antenna is two: one plastic carrier and one metal radiator. Multiple heat stakes are used to assemble the two parts together. As heat stakes are designed as a part of the plastic carrier, this feature is actually free. An assembly fixture is required to melt all the heat stakes and deform them into a mushroom shape. Shown in Figure 4.31 are some production metal stamping antennas. Most of the time, the volume available for antenna designing is irregular. To optimize the performance, an antenna must take full advantage of an irregular space. However, it is not difficult to see that those antennas still follow the basic design principles. Spring fingers are frequently used in metal stamping antennas as the contact feature. As spring fingers are formed from the same metal sheet using which antennas are made of, they are also free. To meet the humidity and other environment requirements, spring fingers must be gold plated. To cut down on the manufacturing costs, a selective gold plating process can be used to minimize the plated area.

At the dawn of internal antennas, dedicated volumes were reserved for antenna designing. With the continuous shrinking of mobile devices, eventually antennas have had to coexist with other components, such the microphone, the speaker, the camera, and so on. Shown in Figure 4.32 are some samples of integrated antennas. If one is putting other components underneath an antenna, they must be RF isolated. Take the speaker as an example. There are two lead lines out of a speaker. Both of them must be isolated by inserting series inductors into signal paths.

Normal metal stamping process can only bend metal sheets, so all the antennas shown in the figure are not truly 3D. They are either composed of multiple flat surfaces, or combinations of flat and cylindrical surfaces. The deep draw process can manufacture a true 3D metal radiator. Shown in Figure 4.33 is an antenna made by the deep draw process. This antenna is also an integrated antenna. The exploded drawing is shown in Figure 4.33b. Only parts 1 and 2, which are the metal radiator and the plastic carrier respectively, are destined for the antenna functionality. All the other parts are intended for the functionality of the speaker.

Compared with metal stamping technologies, the flex circuit technology has better consistency, but it is little more expensive. In many state‐of‐the‐art designs, the antenna volume is quite tight, which means the performance margin is relatively slim. Because flex antennas have better consistency, they also have better yield rate. The flex itself cannot provide connecting features, some extra parts, such as metal spring fingers or pogo pins, are needed.

The flex technology is not a true 3D technology (see Figure 4.34). A flex can only wrap around a combination of 2D surfaces. The most advanced and also the most expensive technologies in antenna manufacturing are double‐shot molded interconnect device (DS‐MID) and/or laser direct structuring (LDS). They have the best consistency, because the antennas are part of the plastic structure instead of separate parts. Both of them are based on a technique called “selective metallization.” The DS‐MID process begins with the application of a shot of plateable thermoplastic resin into an injection mold cavity. Next, the cavity is changed and a second shot of nonplateable thermoplastic resin is molded around the first shot to create a circuit pattern from the plateable material. These parts are then plated with a layer of copper. The DS‐MID takes the longest lead time, because any modification to the antenna pattern requires tooling changes.

The LDS is now the standard process for high‐end phones. The thermoplastic resin used in LDS process is nonplateable after the molding process and can be transformed to plateable by using a laser beam to activate it. The LDS process literally draws the antenna pattern on to the plastic. The pattern can be adjusted quite easily by uploading a new pattern file to the laser. Similar to the DS‐MID, a plating process is required to deposit copper onto the part’s surface. Shown in Figure 4.35 is an antenna made by the DS‐MID process.

When choosing from different antenna technologies, an engineer needs to take into consideration cost, consistency, and lead time. It is always a good idea to consult experienced engineers and antenna manufacturers.

4.5.1 Monopole‐Type Ceramic Antenna

In a state‐of‐the‐art cellular phone, there is most likely more than one antenna. The antenna used for the voice communication is normally referred to as the primary antenna. All other antennas are referred to as auxiliary antennas, such as GPS, Bluetooth, or wireless LAN (WLAN) antennas. Almost all ceramic antennas are narrow‐band antennas. Due to their limited bandwidth, they are mostly used as the auxiliary antennas. Some companies have proposed using ceramic antenna as the main antenna, but this has not been popular due to the performance issue.

For the primary antenna, the performance is critical, because users perceive the quality of a phone mostly by its sound’s quality. If all other phones can make a call in one spot but one phone cannot, the reputation of the phone will be badly damaged. On the other hand, auxiliary antennas are used in applications which are not critical. People do not expect seamless coverage for those applications in the same way they do for voice communication. Thus, the performance requirements for auxiliary antennas are not that stringent. A primary antenna must be custom‐made as optimized for each phone model to squeeze the last drop of the performance from the available physical space. On the other hand, ceramic antennas are designed and built as off‐the‐shelf components. The main consideration is reuse and standardization.

Shown in Figure 4.51 are two types of monopole ceramic antennas. One is made by Panasonic, which is a coil‐type antenna. The other one is made by Johanson Technology, which adopts the multilayer technology. The plating and laser etching process are involved in making coil‐type ceramic antennas. A small piece of ceramic block is plated with metal first. Then a procedure, which is analogous to making a screw, is used to laser etch a spiral slot on the metal layer. The end product is a metal coil rounding around a ceramic block. The axis of the coil aligns with the longest axis of the ceramic block. This process is quite similar to the one used in making metal film resistors. The advantage of this process is that the resonant frequency can be adjusted quite easily. The number of turns of a coil, which corresponds to the resonant frequency, can be conveniently adjusted by reprogramming the laser fixture.

Multilayer‐type ceramic antennas can be made by either the standard chip capacitor process or low‐temperature co‐fired ceramic (LTCC) technology. From the electromagnetic point of view, both processes are the same. They stack multiple layers of meander lines and make an interconnection between layers. The end product is a 3D meander line. The relation between a coil ceramic antenna and a multilayer ceramic antenna is analogous to a helix–stubby antenna and a meander line stubby antenna.

A monopole ceramic antenna is still a monopole, which means its relative position to a ground plane must be correctly arranged. Shown in Figure 4.52 are two incorrect ways of placing monopole‐type ceramic antennas. Why they do not work well can be explained by the image theory. For both placements, the image current on the ground is in the direction opposite to the current on the antenna, and the separation between the two currents is too small. The radiation forms of the respective currents cancel each other in the far field, and that prevents the antenna from radiating. If measuring the reflection coefficient of either antenna shown in Figure 4.52a or b, we might not observe any resonance at all.

Shown in Figure 4.53 are three correct ways of placing monopole‐type ceramic antennas. The one shown in Figure 4.53a is the best scenario, where the antenna is vertically placed on the edge of ground plane. All the copper layers on the top portion of PCB are clearly outside. When using this kind of layout, the antenna is identical to an external stubby antenna except it is concealed from the cosmetic point of view. Many companies use this kind of antenna placement when measuring the antenna’s performance.

In Figure 4.53a, the keep‐out area means no components or traces can be routed in that area, and that is quite a waste, especially for a cellular phone, where it is always a struggle for space to fit in all the components. Shown in Figure 4.53b is a real scenario. The keep‐out area has significantly shrunk. In this kind of layout, the dimension D, which is the distance between the antenna and the vertical edge of the ground, is the critical parameter. By decreasing D, the keep‐out area can be smaller; however, the antenna’s bandwidth and efficiency are sacrificed. So it is an engineering trade‐off again: you have to pay something to gain something.

Shown in Figure 4.53c is another frequently used layout. The antenna is horizontally placed and a segment of metal trace is used to feed the antenna. In this arrangement, both dimensions D and E are critical parameters. The larger the gap, the better the performance. When using a ceramic antenna, it is always easier to shift an antenna’s resonant frequency down than shift it up. So some ceramic antennas are purposely made with a resonant frequency a little bit higher than the target band. When using this kind of antenna, a small segment metal trace on a PCB is required to tune the frequency down.

Shown in Figure 4.54 is the measure response of a typical WLAN chip antenna. The antenna’s model number is 2450AT43A100, and it is made by Johanson Technology. It has a 10 dB reflection coefficient bandwidth of 300 MHz. The antenna is 1.2 mm tall, 7 mm long, and 2 mm wide.

For all monopole ceramic antennas, the ground around and underneath the antenna must be removed. A monopole ceramic antenna itself can be quite small. It is normally around 1 mm tall, a few millimeters wide, and less than 10 mm long. Most companies are often proposing those antennas as small footprint solutions. As antenna designers, we need to be very careful here. If an antenna claims a footprint of 2 mm × 10 mm and we request an antenna area accordingly, we will soon find ourselves stuck in a bad situation. We must not forget to include the keep‐out area. As a rule of thumb, it is always much easier to get antenna space at the start of a project rather than at the end.

4.5.2 IFA‐Type Ceramic Antenna

Shown in Figure 4.55 is an IFA ceramic GPS antenna made by Murata Inc.©. To increase the effective length of the antenna, the metal trace is routed into a U shape at the top surface of the ceramic block. On the bottom of the antenna, there are three solder pads or terminals. The antenna is fed from terminal (1). Terminal (1) on the bottom is connected to the antenna on the top through the sidewall. However, there is no galvanic contact between terminal (1) and the antenna. There is a gap between them on the sidewall. The gap functions as a series capacitor which is a part of the matching network. The other matching component required is a shunt capacitor installed on a PCB. If we remove these two capacitors, it is easy to see that without any matching, the antenna impedance is inductive and is on top side of the Smith chart. Terminal (2) is the grounding pad, through which the antenna is connected to the system ground. Terminal (3) of the antenna is a tuning pad. This pad is DC floating but electromagnetically coupled to the antenna. By connecting a capacitor or an inductor, the effective loading of an antenna can be changed, thus fine‐tuning of the antenna can be achieved.

An IFA ceramic antenna is also marketed as an over‐ground chip antenna solution. Based on the basic principle of image theory, we know that an antenna must be adequately separated from the ground to radiate effectively. The antenna shown in Figure 4.55 is 4 mm tall, which is about four times taller than its counterparts of monopole ceramic antennas. As an example of such antennas, a Murata WLAN IFA ceramic antenna (ANCG12G44SAA145) has a 4.5 dB reflection coefficient bandwidth of 84 MHz. The antenna is 3.8 mm tall, 9.8 mm long, and 2 mm wide. If comparing with a monopole antenna by the absolute size of the ceramic block, an IFA ceramic antenna is significantly bigger and its bandwidth is a lot narrower. However, an IFA ceramic antenna does not require a keep‐out area on the PCB, so the effective antenna space it requires is actually less than a monopole.

In practice, the PCB area around antenna can be safely used to lay out small components. However, be cautious when placing large metal components, such as the shield box, battery, and so on, around the antenna. The antenna’s performance will be significantly degraded when it is immediately adjacent to a metal object.

4.5.3 Loop‐Type Ceramic Antenna

Shown in Figure 4.56 is a loop ceramic antenna. The antenna is 1.85 mm tall, 4 mm long, and 1.5 mm wide. As illustrated in Figure 4.56b, all gray areas are metalized surface. The two long sidewalls are cleared of metal. At the bottom of the antenna, two slits are cut to form two solder pads. On the top of the antenna, there is one slit. It might seem small, but be careful because the “antenna” is actually only a component of a loop antenna. Shown in Figure 4.56c is a PCB layout of this antenna. The effective antenna is the loop around the keep‐out area. The ceramic chip antenna functions more or less like a series capacitor. However, the ceramic chip is comparably big, so it does play a role in the antenna’s radiation. The ceramic chip is made of high dielectric material that effectively decreases the overall size of the loop antenna. The chip is composed of low loss dielectric and has a high concentration of electromagnetic field, which improves the radiation efficiency. By cutting slits on a ceramic block, the capacitance of the chip can be better controlled than a multilayer ceramic capacitor, which improves frequency stability.

The impedance of a loop‐type ceramic antenna is affected by the loop shape and position. In most circumstances, a shut matching component is needed to achieve good matching. For the WLAN antenna chip (ANCV12G44SAA127), the –5 dB reflection coefficient bandwidth given in the datasheet is 84 MHz.

In conclusion, of monopole, IFA and loop ceramic antennas, the one that occupies the least PCB area is ceramic IFAs. For monopole and loop ceramic antennas, there is a design trade‐off between required PCB area and the antenna’s performance. There are different variants of ceramic antennas in the market, but in reality they can be analyzed by the way we are familiar with. After all, the ceramic antenna is not a startling new innovation, but it is only a different way of manufacturing antennas.

4.9.1 Explaining Capacity Boost Effect Through the Antenna Point of View

The term “MIMO” means multiple input and multiple output. It refers to a system which has multiple antennas at the transmitter side and also multiple antennas at the receiver side. Of course, most communication systems are bidirectional, which means that each side has its own transmitters and receivers. In most mobile communication system, the base station has much more space to deploy antennas than mobile terminals; thus, the number of antennas at the base station side is normally more than or at least equal to the number of antennas at the terminal side.

Assume there are N antennas at the base station side and M antennas at the terminal side; this forms an N × M MIMO system. Based on communication theory, given fixed frequency bandwidth, if a system with one transmitter and one receiver can achieve a communication capacity of C, the maximum capacity an N × M (N ≱ M) MIMO system can achieve is M × C. At first glance, this conclusion is contradictory to our instinct. With a given frequency bandwidth, how can we transmit M times information than the channel can support? Why don’t those different information streams in the same frequency band interfere with each other? Communication society likes to explain the principle of MIMO system from the point of view of signal processing; the tools they use are the Shannon limit, singular value decomposition (SVD), eigenvector, and so on. If you want to read more, online resource [32]is a very good one; you can also refer to some classic journal papers [33].

In this book, we will try to look at the same problem from a different angle. Let’s first look at an M × M system. For such a system, if we can think out of box, it is not difficult at all to construct a scenario which M transmitters send M timers information in a given frequency bandwidth, as long as there is no constrain on the center frequency of the carrier.

Shown in Figure 4.83 is a 3 × 3 wireless communication system which can easily triple the communication capacity. It uses three lasers as transmitters and three photodetectors as receivers. We all know that visible or infrared light is also electromagnetic wave, which is essentially the same as radio waves used in mobile communications except the center frequency.

When we look at Figure 4.83, it is so obvious that the 3 × 3 system can transmit three independent data streams, thus has three times capacities. However, when we think about mobile communications, we instinctively worry about in‐band interference between different data streams. The reason is that in cellular frequency bands, which is around GHz, there is no way we can focus a beam as tight as a laser.

Shown in Figure 4.84 is a 3 × 3 MIMO antenna system in free space. There are three transmitting antennas. The distance between adjacent antennas is half wavelength. There are also three receiving antennas. The distance between transmitting array and receiving array is 10λ. Looking from the transmitting array side, the view angle between two adjacent receivers is 3°. If the distance is longer, which is always true in real‐world scenarios, the view angle is even smaller.

Communication society has drawn the conclusion that a MIMO system in free space doesn’t work. Let’s first introduce their explanation. For a 3 × 3 MIMO system, the input–output relation can be represented as Equation 4.1.

(4.1)

where x₁, x₂, and x₃ are signals transmitted by transmitters 1, 2, and 3, respectively. h_ij (i = 1, 2, 3; j = 1, 2, 3) is transfer coefficient from transmitter j to receiver i. For each transmitter j, its signal can reach all three receivers through transfer paths h_1j, h_2j, and h_3j, respectively. For each receiver i, it can receive signals from all three transmitters through transfer paths h_i1, h_i2, and h_i3, respectively.

Equation 4.1 can also be rewritten as follows:

(4.2)

Here, the H matrix is the famous channel matrix. Communication society has proven that the condition of the H matrix decides the channel capacities. For the 3 × 3 MIMO system in free space, as shown in Figure 4.84, the H matrix is an ill‐conditioned one and the total system capacity has little improvement.

Now, let’s look at the problem from the antenna point of view. Based on antenna theory, an N‐element antenna can generate N − 1 nulls. In theory [34], the three‐element transmitting array of a 3 × 3 MIMO system can form a radiation pattern with two nulls. Each null can point to a different receiver. Thus, two receivers out of three will not receive any signals; only the last one in the receiving array can acquire transmitted signal. Similarly, two more radiation patterns with different nulls can be generated. In theory, by using three nulled patterns, three individual signals can be sent to three receivers independently.

Using the scenario shown in Figure 4.84 as an example, to only send data to receiver 2 without interfere with receivers 1 and 3, a radiation pattern with two nulls at ±3° is required. The solid line shown in Figure 4.85 is the normalized radiation pattern generated by algorithm given in reference [34]. The radiation pattern has two deep nulls at ±3°. However, although the radiation pattern can deliver signal toward 0°, the most transmitted power is sent toward ±90°. The normalized gain at 0° is only −44 dB.

For comparison, the normalized pattern of a 3 × 1 system, which is optimized to achieve the highest gain at 0°, is also given in Figure 4.85 as the dashed line. Its normalized gain at 0° is 0 dB.

Using the nulled pattern method, although we can send independent information to individual receivers, the overall system capacity is not increased. To understand that, let’s look at the maximum channel capacity or the Shannon limit given in Equation 4.3,

(4.3)

where C is the maximum capacity, B is the occupied bandwidth, and SNR is signal‐to‐noise ratio (in linear scale, not in dB).

As shown in Figure 4.85, to generate two nulls, the normalized gain at 0° drops from 0 to −44 dB. Because the signal strength is linear to antenna gain, the SNR in Equation 4.4 will decrease accordingly. If compared with a 3 × 1 communication system, which doesn’t have nulling, the capacity of a nulled system is actually much less. Even considering the capability of sending three parallel data streams in a nulled system, the total capacity has little advantage over a 3 × 1 system.

Communication society has also proven that in a multipath‐rich environment, the H matrix of MIMO system is well conditioned and a 3 × 3 MIMO can improve the overall capacity close to three folds. Next, we will explain that from the antenna point of view. Shown in Figure 4.86 is a multipath environment. There are two reflecting objects, one at the top and the other at the bottom. There are three independent propagation paths. The line‐of‐sight (LOS) path is a direct path from transmitting array to receiver array. Multipath 1 leaves transmitting array with a direction‐of‐departure (DOD) angle of −30°. It goes by the top reflecting object and arrives at the receiving array with a direction‐of‐arrival (DOA) angle of −30°. Multipath 2 has a DOD angle of 30° and DOA angle of 30°. In this scenario, the distance between the transmitting array and the receiver array is not important, because DODs and ODAs are decided by the position of reflecting objects.

Nulled patterns are used in both transmitting and receiving sides. To independently send information through multipath 1, a nulled pattern is generated at the transmitting side. Two nulls correspond to the DOD angle of 0° and 30°, respectively. The nulled pattern used at the receiving side also has two nulls, which correspond to the DOA angle of 0° and 30°, respectively.

Shown in Figure 4.87 are transmitting and receiving patterns for multipath 1. The patterns are shown in linear polar coordinates. It is assumed that the radiation pattern of each antenna element is omnidirectional. Both the transmitting array and the receiving array are linear array, so both radiation patterns are symmetrical along the array axis. Due to the existence of reflecting objects, the DOD and DOA angles of the desired travel path can be well separated from other two interfering paths. The nulled pattern still has reasonable gain at signal’s direction while forms nulls to suppress other two interferences.

Shown in Figure 4.88 are transmitting and receiving patterns for the line of sight path. Similarly, because the wide separation between desired path and interfering paths, acceptable gain has been achieved at the desired angle.

Now it should be clear what the multipath‐rich environment means to the communication society and to the antenna society. In a multipath‐rich environment, there must be multiple widely separated traveling paths. From the signal processing point of view, these paths are statistical independent; thus, the H matrix is well conditioned and the overall capacity is improved. For our antenna society, widely separated paths allow antenna array to form nulled patterns to suppress unwanted paths while still achieving reasonable gain at the desired path. By using nulled patterns, several data stream can be independently transmitted simultaneously, thus significantly increasing the overall capacity.

One might wonder how one antenna array can transfer several different data streams simultaneously. Shown in Figure 4.89 is a conceptual explanation through the view of phased array. It is a 3 × 3 system. On the left side is the transmitting array. Data stream 1 is multiplied by weightings α₁₁, α₂₁, and α₃₁, and sent to transmitting antennas 1, 2, and 3, respectively. The weightings used here are the same as those used in phased arrays. Each weighting can modify both amplitude and phase of the original signal. With weightings α₁₁, α₂₁, and α₃₁, the transmitting array generates a radiation pattern as shown in Figure 4.87. Data stream 1 exists in the direction of multipath 1 and is suppressed in other two directions. Data streams 2 and 3 have their own sets of weights; two more radiation patterns can be generated.

On the receiver side, different sets of weightings are applied to signals received on antennas 1, 2, and 3. With weightings β₁₁, β₁₂, and β₁₃, a receiving pattern as shown in Figure 4.87 can be generated. Only data stream 1 in multipath 1 is received. Signals in the LOS and multipath 2 are suppressed. Similarly, data streams 2 and 3 can be individually picked up.

In traditional phased arrays, the waveforms in all transmitting antennas are almost identical. The only difference among signals on transmitting antennas is their amplitude and phase. However in a MIMO system, signal waveforms on different antennas are different. Each waveform is a combination of three distinct data streams, each with a different weighting.

On the receiver side of traditional phased arrays, each antenna receives multiple copies of the original signal. The waveform of combined signals is pretty much the same as the original one, except for its amplitude and phase. For a MIMO system, each receiving antenna can collect all three transmitted waveforms. On each receiving antenna, these three waveforms are combined again with different phase and amplitude. It is obvious that waveforms on all receiving antennas are also different.

Finally, let’s discuss how communication society sends multiple data streams still using a 3 × 3 system as an example. Based on Equation 4.1, it is clear that even if three data streams are separately sent to dedicated antennas, three data streams will all mix together at the receiver side. The way communication society handles this is through SVD method. Assuming the transfer matrix H is known, the H can be decomposed into three matrixes as follows:

(4.4)

Here U, D, and V^H are all 3 × 3 matrixes. ()^H is the conjugate transpose operator. Both U and V^H are unity matrixes. D is a diagonal matrix,

(4.5)

To successfully send three discrete data streams, two steps are required. First, before sending three data streams to antennas, they are left multiplied by matrix V. Second, before processing output data streams, all signals received by three antennas are left multiplied by matrix U^H. The process can be expressed as follows:

(4.6)

Substituting Equation 4.4 into Equation 4.6, we get

(4.7)

Because U^HU and V^HV are both identity matrixes, we have

(4.8)

Equation 4.8 means each data stream x_i can be independently sent to the receiver i. By comparing Figure 4.89 and Equation 4.6, it is clear that V and U^H in Equation 4.6 are in fact the transmitting and the receiving weighting matrixes of the 3 × 3 phased array shown in Figure 4.89.

(4.9)

(4.10)

4.9.2 Antenna Correlation and Antenna Isolation

In Section 4.9.1, we have discussed why MIMO system can boost system capacity. In fact, the H matrix shown in Equation 4.1 includes contributions from two parts. Besides propagation environment, there are also contributions from antenna array itself.

Let’s examine one extreme case, when three identical transmitting antennas are placed in the same spot. In this scenario, the H matrix degenerates into Equation 4.11. The 3 × 3 system degenerates into a 1 × 3 system.

(4.11)

To quantitatively evaluate the performance of an antenna array in a MIMO system, antenna correlation was introduced [35, 36]. In the following discussion, to simplify the problem it is assumed that multipath environment is isotropic in the sense of both power density and polarizations. Under this assumption, the envelope correlation between antenna 1 and antenna 2 is given in Equation 4.12.

(4.12)

where are given in Equations 4.13–4.15.

(4.13)

(4.14)

(4.15)

Here, ()^* is the conjugate operator. In a spherical coordinate system, there are three orthogonal vectors, respectively. In the far field, all antennas radiate spherical electromagnetic wave, which travels along direction. Based on the Poynting theorem [8, 9], the electronic field can only have and components. In Equations 4.13–4.15, E_θi(θ, φ) and E_φi(θ, φ) (i = 1, 2) are the 3D radiation patterns of the and components of the ith antenna. In practice, 3D radiation patterns are measured by a 3D antenna anechoic chamber. Detailed information about 3D chamber can be found in Section 5.1.3.

For the extreme case described in Equation 4.11, correlations among three antennas, , and , are all equal to 1. That means three antennas are totally correlated. One important goal of MIMO array design is to minimize correlations among antenna elements in an array.

In some published papers, the envelope correlation between any two antennas is calculated through S parameters, as given in Equation 4.16.

(4.16)

Detailed deduction can be found in the reference [37]. When an MIMO system is lossless, Equations 4.12 and 4.16 are equivalent. Based on Equation 4.16, it is clear that the better the isolation, the lower the envelope correlation. When S₁₂ and S₂₁ are 0, the ρ_e is also 0.

When discussing an MIMO array, people often don’t distinguish between antenna correlation and antenna isolation. However, these two are not always proportional related. Shown in Figure 4.90 is an extreme example.

In Figure 4.90, inside the dashed‐line rectangle is a “well‐matched and high isolated dual‐port antenna.” If measuring the antenna with a network analyzer and using Equation 4.16 to calculate its correlation, the antenna seems a perfect dual‐port MIMO antenna. Both S₁₁ and S₂₂ are −49.5 dB. The isolation between two ports is −43.5 dB. Using Equation 4.16, the correlation ρ_e between two ports is almost 0.

However, if we look into the dashed‐line rectangle, there is only one antenna. The antenna is connected to port 1 and port 2 through two 20 dB attenuators, respectively. Because there is only one antenna, signals received from port 1 and port 2 are identical. Obviously, 3D radiation patterns of both ports, which are measured in an antenna chamber, are also identical. Based on Equation 4.12, the correlation ρ_e between two ports is 1.

In this example, the wrong result given by Equation 4.16 is not its own fault. The precondition of using Equation 4.16 is that a multiport system under evaluation must be lossless. In antenna designs for real products, antenna efficiency is normally less than 50% or −3 dB. That means Equation 4.16 is always unreliable. We should measure radiation patterns in a 3D chamber and use Equation 4.12 to evaluate correlations between antennas.

4.9.3 Improve Isolation Between Antennas

In a MIMO system, there are mainly two problems which are related to poor isolation. First, as has been discussed in Section 4.9.2, poor isolation means strong correlation between antennas, which decreases the overall system capacity. The whole purpose of putting multiple antennas in a MIMO system is to increase capacity. Second, poor isolation also means low antenna efficiency, because some portion of output power from an antenna will be absorbed by nearby antennas.

The easiest way to improve isolation is increasing distance between antennas. However, it isn’t always practical when designing mobile devices. To achieve high isolation between antennas in devices with constrained space, several methods have been proposed. Some of them can be used when there is enough freedom on select antennas’ form factor. Antennas with different radiation patterns can achieve very good isolation even when they are placed together [38]. By exciting orthogonal modes on a single radiator, two antennas with high isolation can be realized [39]. Electric dipole and magnetic dipoles can also be placed together to generate independent omnidirectional patterns with orthogonal polarization [40, 41].

In many mobile devices, due to thickness constraints and chipset arrangement, there are little freedom on selecting antenna locations. Thus we need some handy technique to suppress mutual coupling between antennas. Most isolation improvement ideas fall into two categories. One is blocking leaking signal and the other is canceling out leaking signal by introducing a new signal path.

Shown in Figure 4.91a is a device with two closely placed monopole antennas. Both antennas are identical and resonate at around 3 GHz. The antenna length is 20 mm, which is shorter than a quarter wavelength; this phenomenon has been discussed in Section 3.1.1. Because isolation is the focus here, no matching network is applied to antennas. The simulated S parameters have been shown in Figure 4.91b. The isolation between antennas, which is S₂₁, is around −10 dB.

One way to improve isolation is using choke. By adding a 1 mm × 23 mm slot, as shown in Figure 4.92a, the isolation between two antennas, as shown in Figure 4.92b, has been improved from −10 dB to better than −25 dB. The slot functions as a quarter‐wavelength choke, which stops current leaking from one to another. Again, the slot’s length is actually 23 mm instead of the theoretical 25 mm. As has been discussed in Section 3.1.1, the ground plane is also part of the radiation structure. The added slot in the ground impacts not only isolation but also antenna matching. This impact can be easily compensated by various antenna matching techniques discussed in Chapter 2.

The other way to improve isolation is to use a neutral line. The physics behind this is out‐of‐phase cancellation. Instead of blocking leaking current, which is what a choke does, a neutral line introduces a new transmission path. Shown in Figure 4.93a is a conceptual diagram. Inside the dashed‐line rectangle is the original two‐port antenna and there is around −10 dB mutual coupling between two ports. To mitigate mutual coupling, two 50 Ω transmission lines, each has an electrical length of 26°, are added to the antenna port. Another 123 Ω “neutral” transmission line with electrical length of 241° is used to connect ports 1 and 2.

Although the characteristic impedance of two 26° long lines is 50 Ω, they do have impact on both isolation and matching. Signal passing through the new path has the same amplitude as the original mutual coupling signal, but it is out of phase. As shown in Figure 4.93b, due to the cancellation effect of the out‐of‐phase signal, the isolation between two ports has been significantly improved.

Figure 4.93 is only a simplified demonstration. In a real design, the neutral line can take any form factor. Shown in Figure 4.94a is such an example [42]. The reverse U‐shaped line between two antennas functions as a neutral line. By adopting the neutral line, the isolation in 1.6–2.3 GHz band is improved by around 5 dB.

4.10.1 Entry‐Level Phone

Shown in Figure 4.95 are front, side, and back views of Hongmi 2A. The price of this phone is 549 RMB, which is around 87 USD, when purchased without a contract. Hongmi 2A was released in March, 2015. It has a 4.7 inch 1280 × 720 IPS screen, 2GB ram, and 16GB flash memory.

The phone supports multiple cellular standards, such as TD‐LTE (4G), TD‐SCDMA (3G), and GSM (2G). TD‐LTE and TD‐SCDMA are 4G and 3G standards, respectively. China Mobile, which has 816 million users and around 70% Chinese market, is the leading adopter of those two standards. The phone also supports 2.4 GHz WLAN (802.11b/g/n), Bluetooth, and GPS.

As a low‐cost phone, the back cover and main frame of Hongmi 2A are all made of plastic. The back cover of Hongmi 2A is removable and its battery is exchangeable. There is a strange phenomenon. Before the birth of iPhone in 2007, every phone had a removable back cover and an exchangeable battery. When iPhone came out, unchangeable battery is a “big” bug people liked to laugh at. I am not sure since when most high‐end phones all adopted unchangeable battery. In 2015, only low‐cost phones had exchangeable batteries.

Because structure parts of low‐cost phones are mostly made of plastic, it is actually easier to design antennas for those phones. High‐end phones like to use metal frames or even metal back covers, which make antenna designs more challenging.

Shown in Figure 4.96 is a Hongmi 2A with back cover detached. Because users have access to the main plastic frame, to disguise funny parts from curious users, all three antennas are covered with black solder mask. The positions of three antennas are marked by dashed arrow lines. However, the black solder mask does a pretty good job, not too much can be seen from this photo.

There are three antennas on a Hongmi 2A. All three antennas are manufactured by using flexible print circuit (FPC) technologies, which has been discussed in Sections 3.3 and Section 4.2.5. The Hongmi 2A supports the following cellular bands:

The primary cellular antenna supports all 2G, 3G, and 4G bands. The secondary cellular antenna supports only 3G and 4G bands. As the 2G B8 band is the lowest frequency band, the primary cellular antenna is the biggest one among three antennas. At 4G bands, the primary and secondary antennas work together to form a MIMO system. Based on 3GPP specification 36.306 [43], devices belong to category 2–4 have two antennas. The Hongmi 2A is a category 4 device.

Shown in Figure 4.97a is the bottom view of the antennas. There are three sets of gold plated contacts. When a phone is put together, there is always some assemble variations. To mitigate that, three sets of spring fingers, as shown in Figure 4.97b, are used in Hongmi 2A. The primary antenna locates at the bottom of the phone. The cellular chipsets are at the top portion of the phone. A thin coaxial cable is used to transfer signal from the top board to the bottom board.

Shown in Figure 4.98a is a production primary antenna. It has not been attached to a phone’s main frame yet, so it is still a flat piece. Only two gold plated contacts can be clearly identified in Figure 4.98a. Most other features are masked by a layer of black paint.

To give us a better look at the antenna, a semifinished sample, which is kindly provided by friends at Amphenol, has been shown in Figure 4.98b. It is clear that the antenna adopts a multiband IFA design. There are three radiators. Radiator 1 has a folded trace and is the longest. Radiator 1 surely radiates at the 880–960 MHz and can also radiate at higher band with its higher order modes. Radiators 2 and 3 correspond to higher bands.

There are two alignment holes next to gold plated contacts. As contacts are where an antenna makes connection to spring fingers, assembly tolerance must be well controlled. Alignment posts on phone frames and alignment holes on flexes are commonly used feature.

Shown in Figure 4.99 is a semifinished primary antenna, which has been installed on a plastic frame. Radiators 1, 2, and 3 are all marked out. There are several eclipse‐shaped holes along the folded edges of the antenna. They are stress‐release holes. They have little impact on antenna performance but make bending antenna much easier. At the end of radiator 3, some portion of the flex doesn’t have any copper. It is called “adhesive tab.” As radiator 3 is quite small, this part of flex tends to raise up even with a stress‐release hole. By adding some extra adhesive area, this portion can be better secured.

It is always a good practice to leave some copper‐free area on a flex antenna. Because whenever the resonant frequency of an antenna needs to be tuned lower, those reserved area will be much appreciated. Asking for more antenna area in late stage of any project is always a huge headache for everyone.

Although the antenna occupies a 3D volume, it is not a true 3D antenna. To be manufactured by flex technologies, it must be able to be spread into a 2D surface. To accommodate this requirement, the plastic frame is recessed and the antenna is also designed as an irregular shape. As the volume of an antenna is proportional to its performance, a flex antenna has to scarify in some degree.

The antenna makes galvanic contact with the PCB on the back of the plastic frame. The portion of flex, where antenna contacts are, passes through a slit and folds back to stick itself onto the back side of the plastic frame.

Shown in Figure 4.100a is a zoom‐in photo of the matching circuit of the primary antenna. There are three spring fingers on the PCB. Spring fingers 1 and 3 are all connected to ground directly. Spring finger 2 is connected to the matching circuit.

When looking at Figure 4.98 or 4.99, it is clear that there are only two contacts on the antenna. Actually, spring finger 3 is a redundant one. One possible guess is that in some stages in the development, a parasitic element has been considered and later abandoned. The slit on the frame and contact area on the flex are all reserved some space for this dummy finger. It is also a good practice to leave the redundant area as it is. One never knows whether those features will be needed again later on.

The matching network adopts a dual‐band matching. Four matching components are used. C1 and L1 take care of the lower band. C2 and L2 handle the higher band. Detailed information of various considerations on dual‐band matching can be found in Section 2.3.

Shown in Figure 4.101 are photos of the secondary antenna. Figure 4.101a is a production antenna. Figure 4.101b is a semifinished antenna, which is provided for a better view of different radiators. The antenna is also a multiband IFA. The main IFA has two radiators: 1 and 2. Radiator 3 is a parasitic element, which is shorted to ground directly. A detailed discussion of parasitic element can be found in Section 4.2.4. The secondary antenna only needs to cover frequency bands between 1880 and 2655 MHz; thus, its dimension is much smaller than the primary antenna.

Shown in Figure 4.102 is a semifinished secondary antenna, which has been installed on the top‐left corner of a plastic frame. Some portion of the flex, which has three gold plated contacts, is passed through the slit on the plastic frame and wrapped to its back.

Shown in Figure 4.103 is a zoom‐in photo of the matching circuit of the secondary antenna. There are three spring fingers on the PCB. Spring finger 1 connects to the parasitic element. Spring fingers 2 and 3 connect to the IFA. Both spring fingers 1 and 3 are shorted to ground through 0  Ω resistors. Those two 0  Ω resistors are not necessary in the design and spring fingers can be grounded directly. However, those resistors can be replaced by inductors or capacitors to fine‐tune antenna, which might be helpful. Using the parasitic element as an example, by replacing the 0  Ω resistor connected to spring finger 1 with an inductor, the element’s resonant frequency can be tuned lower. An inductor with larger value shifts frequency more. On the other hand, a capacitor can shift frequency higher. A capacitor with smaller value shifts frequency more.

Unless it is an emergency, try to avoid using this tuning method. Either inductors or capacitors have inherent loss, which decreases the overall antenna efficiency. It is a better practice to tune antenna through adjusting radiator length.

Spring finger 2 connects to a π‐shaped matching network. There are three pairs of solder pads (one shunt, one series, and one shunt) on the PCB. It seems that the antenna is tuned pretty well and no matching is used in the final design. A 0 Ω resistor is placed on the series pad to simply make galvanic connection.

Shown in Figure 4.104 are photos of the WLAN and GPS antenna. Figure 4.104a is a production antenna. Figure 4.104b is a semifinished antenna, which is provided for a better view of different radiators. This antenna covers two frequency bands: 1.575 GHz (GPS) and 2.4−2.482 GHz (802.11b/g/n and Bluetooth). It adopts a dual‐band IFA design. Radiator 2 is the longer trace and responsible for the GPS band. Radiator 1 takes care of the higher band.

Shown in Figure 4.105 is a semifinished WLAN and GPS antenna, which has been installed on the top‐right corner of a plastic frame. Some portion of the flex, which has two gold plated contacts, is passed through the slit on the plastic frame and wrapped to its back.

Shown in Figure 4.106 is a zoom‐in photo of the matching circuit of the WLAN and GPS antenna. There are two spring fingers on the PCB. The spring finger 1 connects to ground directly. The spring finger 2 connects to a π‐shaped matching network, which is similar to the one used in the secondary antenna. It seems that the WLAN and GPS antenna is also pretty well tuned. No matching component is used and only a 0 Ω resistor is placed on the series pad to make galvanic connection.

Shown in Figure 4.107 are switch connectors of the primary and the secondary antennas. Detailed discussion about switch connector can be found in Section 5.1.2.1. Because the impedance of a switch connector is always 50 Ω, it is an ideal place to tap in a coaxial cable for passive antenna measurement. No matter whether it is a 2G, 3G, or 4G phone; all phones have switch connectors on the signal paths of cellular antennas. The switch connector is used to measure a phone’s RF circuit; it can’t be used to measure antenna directly.

4.10.2 Flagship Phone

Shown in Figure 4.108 are front, side, and back views of Xiaomi 4. The starting price of this phone is 1999 RMB, which is around 317 USD, when purchased without a contract. Xiaomi 4 was released in June, 2014. It has a 5 inch 1920 × 1080 IPS screen, 3GB ram, and 16GB flash memory.

To convince a customer to buy a high‐end phone, which costs several times more than an entry‐level phone, faster CPU and GPU, larger memory, higher resolution LCD, and better camera are all essential. However, the most important aspect above all is its appearance, because most customers can’t even tell what GPU stands for.

A high‐end phone must look like a luxury item. One iconic feature of an high‐end phone is its metallic frame. For antenna engineers, metal frames make designing job much more challenging. Apple has patented hybrid slot antenna solutions [44–46]by using the gap between a whole metal frame and PCB inside a phone. This technique has been used in iPhone 3G and 3Gs. Since the advent of iPhone 4, Apple adopted a new approach which is cutting slits on the metal frame. By cutting several slits on the phone frame, from antenna point of view, the frame is actually several isolated metal pieces, which gives antenna designing much more freedom. Recently, big screen phones become the new normal, which actually makes available area for antenna designing much larger. People start to revisit the idea of using the gap between a whole metal frame and inner PCB to designing antennas [47, 48].

Xiaomi 4 adopts the approach of cutting slits on the metal frame. It is obvious there are four slits: two on the top edge and two on the bottom edge.

Shown in Figure 4.109 is the front and back photos of Xiaomi 4’s antennas. Because the battery of Xiaomi 4 is unexchangeable, its back cover is also not supposed to be removed. Unlike Hongmi 2A, all three antennas are not masked by black paint. Xiaomi 4’s antennas are manufactured by using LDS technologies, which has been discussed in Section 4.2.5. The Xiaomi 4 supports both 2.4 and 5 GHz WLAN band. The cellular bands which are supported by Xiaomi 4 are listed in the following:

Shown in Figure 4.110 is a PCB of Xiaomi 4. Similar to Hongmi 2A, there are three sets of spring fingers for the primary, secondary, and WLAN antennas, respectively. There are also three switch connectors. Different from Hongmi 2A, Xiaomi 4’s WLAN antenna has its own switch connector. More discussion can be found in Section 6.1.4, the short version explanation is that the new SAR regulation now requires measurements on WLAN bands. To carry out a SAR evaluation, the conducted power, which is measured through a switch connector, must be recorded first. Xiaomi 4 also adopts the design of bottom primary antenna. A coaxial cable is used to connect the antenna to the main PCB on the top. Both the primary and secondary antennas have grounding springs, which are part of antenna structure and different from Hongmi 2A.

Shown in Figure 4.111 is the primary antenna. This antenna is a variation of loop antenna, which has been discussed in Section 4.4. Actually, the antenna on the plastic cover can’t function by itself. It might be difficult to observe, but there is a closed loop embedded in the structure. The loop path, marked as A‐A′‐B′‐B‐Metal Frame‐C‐D‐E in Figure 4.111, passes through the antenna, metal frame, and PCB.

Following a detailed description of the loop path. The bottom portion of Xiaomi 4’s metal frame is broken into three pieces by two slits. One slit is next to point B and the other is next to point C. Because of these two slits, the middle portion is isolated from the rest of the metal frame. Points B and C are exposed inner surfaces of the floating metal frame. Point B′ on the antenna connects to point B on the metal frame. Because the LDS antenna can’t provide required spring force, a separate spring fingers is installed at point B′.

Point D on the PCB connects to point C on the metal frame through a grounding spring. The grounding spring was designed to provide required spring force on both points C and D. Both points D and E are on the system ground and automatically connected.

Point A is the center spring fingers on the PCB. It is also the feeding point of the antenna. Point A connects the back of the antenna. There are two metalized strips on the back of the antenna, as shown in Figure 4.109. The right strip connects to point A′ through a metalized hole, which is similar to a via on PCB.

Besides the loop antenna, there is a parasitic antenna. The parasitic antenna connects to the right spring finger on the PCB through another metalized hole. The right spring finger is connected to system ground through a 0 Ω resistor. It can be seen that there are three though holes on the structure. The left one is not metalized and there is also no metal strip on the back. As has been discussed in Hongmi 2A, this is a redundant feature and is left there just in case.

The antenna has a five‐element matching network, but only has three components been populated. Starting from the center spring finger, there is a shunt inductor, a series inductor, and a 0 Ω series resistor. The coaxial cable that connects to the matching network is the cable shown in Figure 4.109. The left spring finger is also connected to system ground through a 0 Ω resistor; however, it is a floating finger and does not connect to anything.

Shown in Figure 4.112 is the secondary antenna. This antenna is also a variation of loop antenna. Similar to the main antenna, there is a loop path, which is marked as A‐A′‐B′‐B‐Metal Frame‐C‐D‐E in Figure 4.112, passes through the antenna, metal frame, and PCB. This loop is almost identical to the main antenna. Two slits on both sides separate the metal frame to three pieces. The middle portion of the metal frame becomes part of the loop.

Similar to the main antenna, there are also three spring fingers beneath the secondary antenna. This time, the right finger is the redundant one. The left finger is connected to system ground through a 0 Ω resistor. This generates two extra loops, marked as loop 1 and loop 2 in Figure 4.112, between feeding point and system ground. Loop 1 is relatively small, so it likely provides matching in the lower band, which is similar to what IFA’s grounding branch does. Loop 2 is reasonably large and might function as an extra loop antenna at the higher frequency band.

There is one interesting feature, which is marked as the 4th spring finger in Figure 4.112. This 4th finger is also connected to system ground. The finger connects to the metal frame of the back camera. If we look at Figure 4.109, there is significant overlap between antenna and camera’s metal frame. It seems that the metal frame has been used as a parasitic antenna.

Shown in Figure 4.113 is the WLAN, etc. antenna. Xiaomi 4 supports GPS, Bluetooth, 2.4, and 5 GHz WLAN. The third antenna needs to cover all three bands. This antenna is a variant of IFA design. There are two spring fingers on the PCB. The top spring finger is the feeding point and the bottom one is the grounding point. Radiator 1 can generate the lowest band, GPS band. Radiator 2 and higher order mode from radiator 1 can take care of 2.4 and 5 GHz bands. There are only two through hole on the antenna. Both of them are metalized. Although there is a preserved matching network on the PCB, no matching component is populated. Only two series 0  Ω resistors are used to make the connection.

Contents in Section 4.10 are the author’s observations. Xiaomi Inc. has kindly provided phones, and Shanghai Amphenol Airwave Inc. has kindly provided antenna parts. To avoid conflict of interest or unnecessary leaking of trade secrets, all materials are provided as it is. None of the observation has been confirmed by these two companies. On the other hand, the author has not verified any observation by experiment or simulation. It is recommended that readers only use this section as a general reference.