Cavendish Kinetics recently announced the availability of production samples of its tunable RF capacitors to key strategic partners. Shipped as a chip scale package (CSP), the Cavendish digital variable capacitor (DVC) technology is used to tune antennas, power amplifiers and filters to improve RF connection quality and signal strength. Moreover, Cavendish leverages MEMS technology to manufacture the high-performance, tunable RF components.
During the Globalpress Electronics Summit 2013 in Santa Cruz earlier this spring, Digitimes had the opportunity to chat with Dennis Yost, president and CEO of Cavendish and Larry Morrell, executive vice president, marketing and business development, at Cavendish about the issues the company was addressing in the market, the technology Cavendish was looking to bring to market and the value proposition its MEMS solution provided.
Q: What is the issue in the market that you are trying to address?
A: The mobile handset market has continued to move forward, progressively going from 3G to 3.5G and now to 4G. And as the technology progresses to higher platforms, one challenge for system designers is to look at ways to improve connectivity, especially for transmitting data very quickly, because we all want to access more data, watch more video and do more things on the Internet. The thing is, the modulation schemes to do that require a higher signal-to-noise (SNR) ratio than just a plain voice call. So as data becomes more important, the quality of the radios becomes more important.
Unfortunately, the radio part of the phone is becoming less and less efficient compared with what the modulation scheme should be able to give. If you look at 4G, you should be able to transmit 80Mbs but users actually see only 10% of that on a good day. Moreover, users at the cell edge (between cells) see even worse performance than that.
Our focus is looking at improving that antenna from 5-10% or maybe 13% efficiency transmission of energy to being 30-40% or maybe even 50% efficient in transmitting energy.
The result of this improvement is that you can save power on the transmit side and improve sensitivity on the receive side, so users will have a better experience. Battery life can also be extended because users don't have to transmit as often at full power and the power amps will not heat up as much.
Q: Can you explain in more detail some of the issues facing front end module design and antenna design with the transition to LTE?
A: In the 3G and 3.5G markets, you traditionally have been able to get a pretty decent world phone that covers frequencies from 800MHz to about 2.2-3GHz. That is a pretty good phone for 3G and for that, the antennas used were just good enough.
With 4G, frequencies are being added to both ends of the spectrum, so basically you are expanding the frequency range you have to cover to 700MHz to 2.7GHz. Now add to that the white space that is available - which is the digital dividend that comes from moving terrestrial TV from analog to digital. The US and Europe are talking about adding the 600MHz bands, which is going to make things even more difficult. In terms of bands that have already been approved by 3GPP, they now span from 698MHz to 3.5GHz. And while there is no 3.5GHz deployed, there are companies out there seeing if they can make that work. So there is much more spectrum that the antennas need to cover.
Moreover, antenna makers are not consulted when new phones are designed, and there really is no interest in doing them any favors when it comes to improving RF design. In fact, the very exact opposite is being done. Consumers don't want an antenna sticking out of their phone and nobody wants a small screen so antennas are becoming smaller and required to deal with more noise. OEMs also sometimes simply stick a connector right in the middle of the antenna or add speakers or buttons that interfere with the workings of the antenna. Antenna makers are then given impossible specs to meet and are expected to deliver in a short time anyway.
The antenna makers are the tail end of the dog and they would be more than happy to change the way they approach the problem.
Q: Aside from the difficulties of having optimal RF design in mobile handsets, it seems you are arguing that there is a problem with tuning RF signals in general. Why does this occur and how does your technology address this issue compared with what is currently used in the market?
A: If you want to tune an RF signal. One way to do it, and people have been doing it this way in different forms for a number of years, is to have a multi-throw switch that is attached to different values of load - imagine a one-pole 32-throw switch with each of those 32 switch elements attached to a different RF load, be it an inductor or capacitor or something of different value. So you have the power loss of the switch and the loss of whatever the passive component is. But you get very good tuning out of that and very good tuning capability.
Unfortunately, the switch consumes some of the RF signal by virtue of the fact it has resistance in it. So the power loss of the switch frequently sucks up all of the efficiency gains you can achieve elsewhere, because the switch itself has 1 ohm or 1.5 ohms of resistance. That may sound like a good low value for a switch but if you lose an ohm in the switch, you lose 3dB overall, meaning about half your signal is going out the front door. So, 1 ohm resistance in your switch is basically a killer. And that is what handset makers have to live with.
Our solution is to take the switch and throw it away. Our device allows for the RF signal to connect directly across a shunt capacitor, which is one of the ways you can do a load. And if you have a capacitor where you can change its value, as opposed to needing a switch, the losses of the switch can disappear. Lose the switch, lose the loss.
Q: How does it work?
A: We use MEMS technology for RF. We make a movable component in our technology. Image a parallel plate capacitor and as the plates move closer together they have high capacitance and as they move apart they have low capacitance. It is kind of a bi-stable capacitor.
MEMS is ideal because you eliminate all the non-value added parasitics, meaning if you were to have a switch in a series - and just be switching capacitors in and out - the resistance loss in that switch is a parasitic that you have to live with. With our component you don't have that. It is just a capacitor that changes a capacitance state. So there is no parasitic, which in this case is called equivalent series resistance (ESR).
Q: You compared your platform to only one example of a switching solution, but there are other companies addressing this market. How do you compare with them?
A: Anyone who is doing switches can address this market, and there are a number of different ways it can be done. One solution is to use discrete switches like gallium arsenide (GaAs) and put in discrete components. You can also integrate these solutions using solid state switches. You can build up a relatively good solid state switch with SOI technology and some companies are doing that quite successfully to address the switch market as a stand-alone market.
However, you have the same issue in all of these cases. If you have a switch in there, even with a very efficient capacitor, you still have the loss related to the switch. There are a number of companies in the switch market trying to do the same kind of application. Unfortunately they always have a switch - because architecturally they can't get rid of it. The reason is that their capacitors are all fixed plate capacitors. They don't vary like a MEMS capacitor can, which is intrinsically a variable capacitor.
Q: You didn't mention any MEMS competitor in your comparison. Are you the only company addressing this issue with a MEMS solution?
A: Companies have been trying to implement this solution in MEMS for quite a while because MEMS gives the best performance. This idea is nothing new. Research has been going on for probably around 30 years now. The problem and challenge people have had with MEMS is whether can you make it reliable, can you make it in volume and can you make it at a cost and size that makes it a viable solution for a cell phone maker?
Fortunately, we can meet all those requirements. Other companies may talk about using MEMS in RF solutions, but with them you are talking about US$5 and US$10 parts. There is not enough BOM in cell phones to do a US5$ switch.
Q: So what is the pricing of your MEMS solution?
A: Let's just say that if you look at current designs in the market for LTE, one solution is to run multiple antennas connected with switches. If you eliminate those antennas and you eliminate the switches you can save a dollar or more with our solution.
Q: Application processor companies such as Qualcomm have also announced solutions that improve the performance of the RF front end. How do their solutions differ from your approach?
A: Companies that have access to what we call the interior of the radio, or the other side of the antenna - where you are talking about the switches and PAs and so forth - use something called an impedance matcher. The impedance matcher is designed to convert the 50 ohms that the RF uses inside the phone to the free space that the antenna sees. So you have this conversion zone, and that conversion is done by an impedance matcher, which can be re-tuned as you change frequencies so the antenna works as well as it was originally designed for at multiple frequencies.
Now, while that is all possible to do, you unfortunately haven't changed the efficiency of the antenna by doing that. You simply made it work the way it was designed to work - across multiple frequencies. This can deliver a performance improvement of 10-20% or even 30% for some extreme cases.
However, since it is in the signal path, the losses from the switch remain, so you still have to recover those losses, which would be in the 1-1.5dB range. Therefore, the gains have to be above that to show a net gain, which is turning out to be extremely difficult to demonstrate.
This type of solution can be done with a variety of different architectures but they require multiple components and end up being much more complex circuits to control, because you need several variable elements which have to be traded off against each other. There are literally hundreds of thousands of combinations that have to be evaluated. It is very complex design task with marginal results.
Our belief is that if you have a lossless component on the antenna, then let's make the antenna do the job of becoming more efficient itself. It actually improves all your cases. You can add an impedance match onto that if you want, but then you're still kind of working on the wrong end of the problem.
Q: What is the current status of your MEMS switch?
A: We just announced the availability of production samples. Before that we were making sure it was a highly efficient design and that it has a high Q factor (a higher Q indicates a lower rate of energy loss). Our measurements with antenna makers show that our Q in actual usage conditions over the normal usable range of the device is in excess of 200. This compares with a Q of 40-50 for devices that use switches, and that represents a good number for them. Those losses are just being tossed out the front door. That is why we are able to improve the antenna efficiency by a factor of two or more.
Q: How has the response been so far from potential customers?
A: The response to the technology has been overwhelmingly positive. The way we demonstrated our technology was by buying commercially available phones and working with antenna companies to retrofit the devices. We didn't pick any particular method but let them choose their own style of implementation. We provided them with some early parts and they reported an improvement of 1-2dB, and in some cases 3dB over existing antennas that were already in production. This was done without the benefit of going back and re-tuning the industrial design. It was a very quick and dirty retrofit.
These results were also well received by the handset makers who are now waiting for us to come back when we are in full volume production. That is the process we are in right now. We expect that later this year we will be announcing design wins and big vendors adopting the technology.
Q: It is not always easy for startups to receive funding, despite any amount of "Wow!" their technology may have. As a semiconductor startup are you finding it difficult or easy to find funding?
A: Finding investments for semiconductor startups over the past few years has really been a challenge. There is less and less money available for the traditional startup, meaning those with a business model of designing a better CMOS chip than everyone else and going to a foundry to build it. The investment community has been looking for startups that own a unique technology platform for offering differentiated products. Fortunately for us, this has been what we are able to do.
We have been focused on the RF component market since late 2008. For the first couple of years we were a technology development company. Now that we have the technology, we are focusing on going to market. From a business stance, this has made us very attractive to investors because it has taken us more than four years to get where we are now, with the main reason being that the technology barriers are so high, so it is not easy to copy. Our investors are extremely pleased with this direction and also that we targeted a mobile handset market that is big and still growing. Billions of these devices are going out the door so the market is pretty large for us.
Cavendish Kinetics: Dennis Yost (left), president and CEO; and Larry Morrell, executive vice president