“Cellular telephony technology utilizes a large number of frequency bands, providing ever-increasing bandwidth for mobile use. Each of these bands requires filters to separate the signal from the others, but the filter technology currently used in mobile phones may not scale to the full millimeter-wave (mmWave) range planned for 5G.
This article is reproduced from the article “The Search For 5G MmWave Filters” published on the semiengineering.com website.
There are plenty of new options, but no clear winner so far.
Cellular telephony technology utilizes a large number of frequency bands, providing ever-increasing bandwidth for mobile use. Each of these bands requires filters to separate the signal from the others, but the filter technology currently used in mobile phones may not scale to the full millimeter-wave (mmWave) range planned for 5G.
The mmWave era will come eventually, but not now. Earth exploration satellite services operate at 23.8GHz, which is only slightly below the millimeter-wave band deployed for 5G, so it must be effectively filtered. –Mike Eddy, Vice President, Corporate Development, Resonant
So far this hasn’t happened:
Surface acoustic wave (SAW) device or bulk acoustic wave (Bulk acoustic wave, BAW) device frequency will not exceed 10GHz. – Anthony Lord, Business Development Director, RF Division, FormFactor
There is also a challenge:
None of these filters can work in the millimeter range, they can only go up to 6GHz or 8GHz. In this regard, the industry has not found an effective solution. –Tim Cleary, Senior Marketing Director, RF Product Group, FormFactor
In current mobile phone products, SAW and BAW filters are most commonly used. While with further improvements, they may extend beyond the 6GHz range to some extent, there is still some distance from the 28GHz to 70GHz range that mmWave designs are aimed at. While there are currently solutions for devices with smaller space constraints, these solutions do not apply to mobile phones. Therefore, this field still needs to be vigorously developed.
1. The number of frequency bands has grown dramatically
With the advent of new mobile phone technologies, more frequency bands are opened up for use. The term “frequency band” can have different meanings, because broadband is allocated and auctioned, and a single channel refers to a subset of the broadband.
The number of small frequency bands is increasing dramatically. For channels utilizing frequency-domain duplexing (FDD), there are two adjacent sub-bands (one for transmit and one for receive) separated by a small gap to prevent interference. When using Time-domain duplexing (TDD), there is only one single frequency band for the entire channel.
Each of these bands or sub-bands requires a bandpass filter. As the number of frequency bands exploded, so did the number of filters required. Today’s mobile phones may have more than 60 filters. 5G technology will only increase that number, adding very high frequencies to the mmWave band.
In theory, a bandpass filter will pass all signals within that band and reject all frequencies outside that band. We can simply think of it as multiplying the in-band signal by 1 and the out-of-band signal by 0. However, the filter effect in practical applications is not ideal, which brings more challenges.
2. The function of the filter
A real filter does not “stop” at the band edges, instead, the band edges are arc-shaped and the attenuation is sloping rather than vertical. The center frequency, the upper and lower cutoff frequencies are the key properties of the filter, the cutoff frequency is the point at which the signal passing capability drops by 3dB (corresponding to the point where the signal power drops by half). The slope of the attenuation above 3dB is often called the skirt, and the downward trend needs to be very steep.
While it might be a good idea to design these three frequencies (center, upper, and lower) individually, in reality, the upper and lower cutoff frequencies are moved together so that the center frequency and overall width are designed, the latter moving with the center frequency. The width is usually a percentage of the center frequency.
Designing wider passbands can be a bigger challenge, and some 5G bands can be as wide as 20% of the center frequency. This puts a lot of pressure on filter design.
Figure 1: Simplified bandpass filter showing center frequency (f0), lower passband (fL), and upper passband (fH). The width of the passband is B. Source: Inductiveload – Own work, Public Domain
At the front end of the receiver, spurious signals need to be filtered out as early as possible to prevent them from entering the RF link. This means filtering occurs after the signal leaves the antenna. With massive multiple-in/multiple-out (MIMO) technology that supports beam steering, an array of antenna elements can be used. In this case, each element needs a filter.
Today’s component pitch is based on millimeter-wave pitch, which means that the pitch is about 5 mm. So you have to adapt to this spacing. –Mike Eddy, Vice President, Corporate Development, Resonant
Currently, this is not possible with mmWave, so it ends up being filtered after the signal has passed through a mixer.
Base stations have ample space to accommodate larger filters, but mobile phones have stringent size requirements. For the foreseeable future, the best frequency for a small filter is likely to be 28GHz, since that’s the millimeter-wave frequency that phones are likely to use. Higher frequencies are more likely to be used for tower-to-tower communications, as these systems are not as space-constrained as cell phones.
For use cases such as base stations, we will rely on ceramic dielectric filters and metal cavity filters, but they will never meet the space requirements inside mobile devices. –David Vye, Director of Software Technical Marketing, Cadence AWR
In the early days, the filtering requirements for the 28 GHz (or similar) band were more relaxed:
We often heard in the first few years that there would not be any mmWave filters in mobile phones. Because the frequency band was not decomposed at that time, the antenna was mainly used for filtering. – Jeb Flemming, CTO, 3D Glass
In this case, using the antenna as a filter can roughly meet the needs, but at some point, we need to prepare a real filter for the antenna element. So how exactly are these mmWave filters made?
3. Existing filter technology
Most filters in mobile phones today use sonic technology, which involves piezoelectric materials that deform slightly under the influence of an electric field, and then physically deform to produce an electric field. Therefore, electrical signals can be converted into mechanical vibrations, and mechanical vibrations can also be converted into electrical signals. These mechanical vibrations are equivalent to sound waves within the crystal.
By creating an acoustically resonant structure, the input signal can be applied to one end of the resonator. This input signal consists of many signals of different frequencies – some for other frequency bands, and some for ambient noise. The first job of the filter is to remove any signal outside the passband.
Signal frequency components within the passband will cause acoustic resonances, which are then detected by the acoustic wave filter and converted back into the electrical domain at the other end of the filter. Ideally, this output would consist of the input signal, with all unwanted frequencies in the signal removed.
These acoustic filters have many advantages, including a clean passband, very small size, and a favorable cost structure, which also reduces cost in high volume production.
At lower frequencies, the SAW filter dominates. When using these filters, waves at the surface of the material are excited and coupled to the output near the same surface.
Figure 2: A simplified SAW filter. Source: Matthias Buchmeier ― Own work, Public Domain
For higher frequencies, the BAW filter dominates. In contrast to SAWs at low frequencies, BAWs do not excite waves at the surface of the material, but use a bulk material to resonate from top to bottom with the output electrode below. This requires more complex processing, so they tend to be more expensive than SAW filters.
Figure 3: A simplified stand-alone BAW (FBAR) filter. Source: Khpsoi ― Own work, CC BY-SA 4.0
There are two basic versions of BAW filters that differ in how the internal standing waves are set up. One version requires reflection from bottom to top, and uses a free-standing resonator BAW (FBAR) filter and air cavity to do the job.
Another version uses a series of layers that look like acoustic mirrors (similar to a Bragg reflector of light) and is known as a solid mounted resonator (SMR) BAW filter.
Figure 4: A simplified solid-mount resonator (SMR) BAW filter. Source: Khpsoi ― Own work, CC BY-SA 4.0
Both SAW and BAW filters are fabricated using MEMS processing technology, but they appear to start to fail at higher frequencies, suggesting that the industry may need to find new filters for the mmWave band.
4. Selection of mmWave Filters
Millimeter-wave radio signals are not new. For example, radar and microwave devices already use them, but these tend to be large devices that can only handle one or two frequencies. For 5G, more frequency bands must be filtered more finely, and they must be able to fit into phones.
SAW and BAW have been left out of the equation, but Resonant has XBAR technology, which it claims could expand the availability of acoustic technology. The company redesigned the BAW filter from scratch, using a different piezoelectric material, lithium niobate, and placing both contacts on the tip, similar to a SAW.
However, the main difference from SAW is that with XBAR there is no physical movement of the contacts:
With SAW, the metal rods move physically, which means they lose power during the metal migration process. –Mike Eddy, Vice President, Corporate Development, Resonant
Figure 5: The XBAR prototype shown at Mobile World Congress (MWC) 2019, the small square in the middle is the filter. Source: Resonant
When we model this structure, XBAR provides the energy, bandwidth and power handling required for 5G — especially when we focus on 3 to 5 GHz. Right now we’re looking at WiFi from 5 to 7.1 GHz and then Ultra Wideband from 7 to 9 GHz. Can this model be used for mmWave? We think so. –Mike Eddy, Vice President, Corporate Development, Resonant
The XBAR filter looks promising, but the point is, it represents a new approach in this frequency range. Two other well-known mmWave filter technologies are waveguide and cavity filters. But unlike SAW and BAW filters, which use acoustic waves, they use electromagnetic waves to resonate, and both have a wide range of structural options and are typically used in microwave applications.
The size of these resonators is usually frequency-dependent, with the size or spacing being in the quarter wavelength range. The higher the frequency, the shorter the wavelength, and the smaller the filter. For 5G frequencies, the size of the resonator is shrinking — but it still doesn’t fit into a phone.
There is a medium called a ‘waveguide cavity’ whose height and width determine the energy that can propagate through it. Below that frequency, the energy does not propagate, and above a certain frequency, modulation problems arise. –David Vye, Director of Software Technical Marketing, Cadence AWR
The use of resonators (usually implemented as pillars) helps reduce unwanted modes:
The waveguide cavity filter has some pillars inside, which function the same as the ceramic filter, the characteristic is to stop or transfer the energy at a specific frequency according to the size of the pillar. The physical size between the resonators will affect the bandwidth, while the number of resonators will affect the attenuation, i.e. the more filters, the faster the attenuation. But in this way, the length of the filter is increased, and the material cost of the filter is also increased. –David Vye, Director of Software Technical Marketing, Cadence AWR
Figure 6: A simplified waveguide filter using pillars as resonators. Source: Wikipedia user SpinningSpark
For base stations, the technology is suitable because it can accommodate larger sizes; but for mobile phones, the size of the filter is still too large.
Microstrip filters are another option for frequencies up to 30 GHz. With this design, microstrip lines are created on the PCB to support electromagnetic resonance. There are still manufacturing variance issues, and PCB materials are generally considered to be of poor quality.
Changes in the thickness of the PCB, changes in the dielectric constant of the material, changes in the line width during printing, and temperature can all change the passband frequency. –Mike Eddy, Vice President, Corporate Development, Resonant
There are also other options to consider:
Material properties do drive performance, but there are only a handful of materials on the market. These very high-Q resonant ceramic materials are special and generally more expensive. Multi-layer ceramic caps (MLCCs) have been a reasonable material until now, but they start to fail around 25 GHz. – Jeb Flemming, CTO, 3D Glass
5. Substrate-integrated waveguides
Because of the shorter wavelengths at millimeter-wave frequencies, it is possible to make waveguides in silicon or other materials:
It’s almost like MEMS in that because these channels are being created, the microwave signal can go through the etched area and then metallize it on the silicon wafer. –David Vye, Director of Software Technical Marketing, Cadence AWR
3D Glass creates waveguides in glass rather than silicon through a photolithographic process that selectively converts amorphous glass into crystals through exposure to ultraviolet light. Converted crystallized glass (ceramic actually) is better for etching and easier to create through-hole features.
Ceramic etches 60 times faster in acid than glass, and we can make cavities, but with a timed etch because this ceramic layer has glass going through it. – Jeb Flemming, CTO, 3D Glass
Structures such as inductors can be fabricated in this way, as well as cavities with resonators for mmWave filtering:
If metal wires were used as resonators, and almost all the glass was etched away, the resonators would mostly float in the air. Since the limiting factor for 5G mmWave is the material, it is a success if you can remove the material and make it afloat and durable. Suspended striplines can reach around 40 to 50 GHz, and we show 10% to 15% bandwidth, which is quite a wide range. – Jeb Flemming, CTO, 3D Glass
These air-filled cavities can be extended to higher return frequencies:
We’re doing a lot of customer development in the 70 to 150 GHz range, some call it 5G, some call it 6G. – Jeb Flemming, CTO, 3D Glass
In the past, filter design involved multiple manufacturing steps to optimize performance, but there were too many variables and strict requirements, but now we can use simulation tools to optimize the structure of the filter before it is built. This helps with details, because details matter:
How it is packaged and how it is connected to the rest of the circuit is so important that people abandon empirical testing of the design and rely on EM (electromagnetic simulation) techniques to do it. –David Vye, Director of Software Technical Marketing, Cadence AWR
Cadence has previously worked with 3D Glass, using AWR® Microwave Office® for design and simulation, so is very familiar with 3D Glass’ work:
There are metal resonators inside a very low-loss structure that is suspended in the air by a small glass base, forming a very small filter—though not as small as an acoustic filter. –David Vye, Director of Software Technical Marketing, Cadence AWR
The economics of glass technology are attractive. Considering the need for volume, panels can be used instead of wafers. A 9’x9′ Panel can fit a lot of filters, so while today’s work is done on 6″ and 8″ wafers, and some customers want to move to 12″ wafers, they see a clear line of cost reduction path.
While there are some exciting possibilities on the horizon, these are not ready for commercial production, and a real winner in filter technology has yet to emerge. Millimeter wave in 5G phones is not yet fully realized, so there is still some time. But it’s worth noting that the problem the industry is facing right now is coming up with a solid plan and roadmap, not some idea that might work.