As the leader of the team that developed the AD9361 product, I'm often asked where the idea came from. Was it a "light bulb moment"? Divine inspiration? Uh, no. It was the result of a series of failures.
The AD9361 software-defined radio (SDR) is now one of ADI's most successful products. It is in hundreds of applications. Chances are you've used it and don't even know it. But how did we get here?
I was leading a fledgling RF transceiver development group at ADI. We had a family of WiMAX chips. (Remember WiMAX? Me neither.) When the WiMAX market didn't happen, we had to figure out what to do next. One thing I knew I'd never do again was to chase a market that may not develop.
This was when the industry was starting to migrate to the 65nm technology node. Along with the remarkable capabilities of this new technology, came what seemed at the time, an incredibly high development cost. (This is amusing in retrospect given where costs are now!) With these soaring costs, my management wasn't going to be very patient with more failure.
Anyway, the aha! moment came when I realized the implications of the really good switches that we could make in 65nm CMOS! (I'm picturing readers switching to some very important cat videos now - stay with me!)
By good switches, I mean switches that have low series resistance when closed and small parasitic capacitance when opened. It turns out that those good switches play a huge role in making SDR possible. Here's a short list of things these switches enabled:
- A "universal" local oscillator generator (LO) up to 6GHz. The core of the LO generator is a single-core LC VCO that tunes from 6GHz to 12GHz and is enabled by digitally-switched capacitors with a high Con/Coff ratio and pretty good quality factor.
- Inductor-less RF signal paths. Traditionally, RF signal paths have needed resonant circuits to tune out capacitance in order to keep power dissipation low. With the very low parasitic capacitance of deep-sub-micro CMOS, we were able to eliminate the inductors. This is important because nobody ever told inductors about Moore's Law so they're not only expensive but they get more expensive with Moore's Law because inductors don't scale but the cost per area does. And, this octave VCO is followed by cascaded divide-by-twos (made with switches, of course) from which we can generate any frequency up to 6GHz.
- Widely reconfigurable continuous-time sigma-delta (CTSD) analog-to-digital converters (ADCs). Aside from the bandwidth and the power reconfigurability, these are remarkable ADCs because they have inherent anti-aliasing which is good because ...
- ... after inductors, active analog filters probably scale most poorly with Moore's Law. With ADCs that don't need anti-aliasing filters we can use relatively low order filters and, because of those switches, they have 200:1 bandwidth tuning ratio.
Of course, it goes beyond just the switches. Lots of other things are enabling, including the direct conversion radio architecture with heavy use of "digitally-assisted-analog". Direct conversion is enabling because, compared to other radio architectures, it consumes the least power, eliminates impossible-to-integrate components such as IF filters and relaxes RF filters since there are no out-of-band images. Direct conversion is used in virtually every consumer wireless application and ADI is bringing it to the rest of the world.
That is just a taste of the benefits and challenges of direct conversion. Look for more on this topic and other
under-the-hood topics in future blog posts. Now, you may watch cat videos.
Click to Tweet: Was ADI’s successful AD9361 software-defined radio born of divine inspiration? ADI fellow Tony Montalvo: “Uh, no. ” @Adi_News #EngineerZone