We had a chat session recently where we discussed the integration of RF devices as it pertains to analog integration.
We discussed how 10 years ago, it was difficult to integrate RF circuitry onto a chip. Now, it's a lot easier and far more practical. We talked about some of the more well-known and familiar subsystems that are part of the software defined radio (SDR). It's possible and practical to put most of the UHF receiver section on a chip.
This, of course, pertains directly to cellphone applications. There are certainly other receiver applications — radio equipment for the military or safety forces — but if you're looking at high production volumes, then you're looking at cellphones. These high volumes are the reason you would want to integrate as much as possible onto the chip to reduce manufacturing cost. That cost includes the parts themselves and PC board real estate (i.e., materials). Also included is the effort to build, and more importantly, test the board assembly (labor). So the more you can integrate, the lower the overall cost.
The receiver section is the easy part. For the transmitter, it's pretty easy to put most of the circuitry on a chip — until you get to the RF power output stage and the antenna. With existing technology, it's not practical. The high-power output stage will be problematic because, well, it's a high-power RF source (probably a couple of watts) very close to the receiver circuitry. There are a few tricks you can do to minimize the inevitable crosstalk — active noise cancellation techniques, for example, though that doesn't completely solve the problem.
Even with crosstalk mitigated, the other difficult portion of the circuitry to integrate is the antenna tuner. Ideally, you'd like to have voltage variable capacitors (and perhaps inductors) to allow matching the antenna's impedance to the RF output stage. The antenna's impedance changes depending on proximity to surrounding objects and to the transmitting frequency.
Using the existing technology — the voltage variable capacitor or varicap — won't work here. The varicap is actually just a reverse-biased diode. The higher the DC bias (fed to the diode through an inductor to provide isolation), the farther apart the areas of conduction in the P- and N-layers are. Therefore, the capacitance is lower. So, vary the voltage, vary the capacitance.
Unfortunately, the applied RF voltage is high enough that it will muck this up by affecting the capacitance dynamically. The result will be some very bad waveform distortion.
Another choice that seems promising at first is the MEMS capacitor devices. With these, you can actually have a capacitor is the classic sense — two conductive plates that can be moved with a control voltage to put them closer together or farther apart. Sadly, at the RF power levels, arcing may occur across the capacitor.
A related possible version of this uses the MEMS device as a relay — a set of contacts that can be opened or closed. Here, you could select various fixed capacitor and inductor values, switching them in and out to get the desired value. But you would have a similar problem as above — arcing across the contacts as they opened.
To date, we don't have these problems solved — but it's only a matter of time before these design issues are successfully addressed. We can then integrate the complete transmitter chain, including the final RF stage and the antenna tuner.
Are you working on any designs where you are integrating large portions of RF circuitry?