Friday, January 29, 2010

The Key to SDR is the RF Front End

Software Defined Radio will come of age in the next five years. From a niche application in the military world to an academic research topic today, SDR will become main stream in five years throughout the commercial industry. Mobility will shift beyond the handful of standards that defines its role today into an open playing field where anyone can play. The key to SDR is the RF Front End. The solution must

--be ultrawideband to handle 20 MHz to 60 GHz
--work with any softrware
--work with any waveform
--use tunable filters

Some solutions are already coming to market such as

The IMEC research center in Belgium proposes a CMOS solution to this problem. The solution promises a low-power, low-cost solution.

Microtune offers a chip for car radio applications called the MT3511 that provides a generic hardware platform to support ­AM/FM, digital radio, CD audio, MP3 playback, navigation, as well as connectivity for iPod®, SD card, USB and Bluetooth.

Best regards,

Hall T.

Friday, January 22, 2010

Portable Waveforms--Some Design Considerations

Waveforms are just waveforms, right? Well, in the world of software defined radio a waveform is more than just what meets the eye. In the recent SDR Forum, I learned that "Portable Waveforms" consist of more than just the signal. For the JTRS program, a portable waveform consists of

--Detailed design docs
--Test code and data sets

Guidelines are in place for how to create portable waveforms because as this paper points out it's easier to rewrite code than to import a non-portable piece of code.

Part of the challenge in making a waveform portable is partitioning the waveform into real-time and non real-time components and then matching each component to a processor resource such as an FPGA, GPU, CPU, etc. Since hardware platforms vary widely, it's not reasonable to expect a waveform to run on any platform without some modification. Documentation including code, code hierarchy, multi thread uses, etc is required. Also, emulators and debuggers are necessary to work out how the waveform is generated and can be modified.

It's also interesting to note that IP cores specific to FPGA vendors should not be used as target platforms may not have those specific FPGA components. Clocking is another concern in that the use of multiple clocks to achieve some objective -- say lower power consumption -- may cause problems due to the target platform not having sufficient clocking resources.

Portability brings more design concerns and issues than originally meets the eye.

Best regards,

Hall T.

Friday, January 15, 2010

Smart Radio Challenge

The Smart Radio Challenge is a worldwide competition in which student engineering teams design, develop and test software defined radio (SDR) that address relevant in the wireless market. In the recent SDR Forum, several competitors described their projects and the challenges they encountered.

California State University talked about unmanned aero vehicle telemetry link. Their challenge was the lack of available spectrum since an air force base was nearby. They used the Ettus Research USRP with GNU Radio software. The students developed their packet data handling protocol in Python as GNU Radio didn't provide it. They could tell the UAV to take images, generate FFTs, etc. The system had a phone home function in the event it lost communication with the ground. They used an amateur radio transceiver so it could display spectral parameters and used SDR for all signal processing. The challenges in using SDR for this project is the students had DSP but no SDR background. They started from ground zero. No one had experience with Python and no experience with real communication systems which required heavy faculty involvement. GNU Radio Grand Canyon was a good start but only goes so far. The project came out well with a successful demonstration at Edwards Air Force base. The next step will be to add target recognition and to implement the SDR on an FPGA by using the USRP.2

Virginia Tech talked about lessons from the 2008 and 2009 competitions. Their 2007 challenge was spectrum access for first responders. They chose the GNU Radio/USRP over the Lyrtech SFF because they already knew the USRP. The 2008 Challenge focused on first responders in areas with no comms infrastructure such as the Katrina disaster. They used a wireless ad hoc network approach for this one. The lessons from these two are:

Play to your strengths
Choices have unintended consequences such as system integration
Make a decision and follow through with it

For 2009 the challnege will be to locate a first rsponder based on a 406 MHz packetized beacon signal.

Carnegie Mellon presented on their Spectrum Sensing for Dynamic Spectrum Access. It's a Cognitive Radio network that takes a crowded network and sets up a secondary network within it. They first map the spectrum to find unused channels and then setup their own network. They used GNU Radio, a USRP, and Ubuntu to implement their sensing algorithm running on a series of "sensing nodes" which pass the results to a "learning center" which identifies the open spectrum slots. It then issues a beacon signal that other nodes can lock onto and then use a sub channel for creating the secondary network. They ran video, text messaging and voice messaging through the network.

Best regards,

Hall T.

Friday, January 08, 2010

Challenges for Software Defined Radio

At the recent SDR Forum in Washington DC (now called the Innovative Wireless Alliance), Dr. Christer Svensson of Linkoping University in Sweden gave a keynote presentation on the challenges facing software defined radio. In his address he listed the critical requirements for SDR as

Digital protocols
Digital baseband
Receiver analog/antenna frontends
Transmitter Power Amplifiers and antenna frontend

While the first two are progressing well, the last two are not. Frontends are insufficient and power amplifiers lack multiband solutions. What's more, there's little research going on for the last two. The primary challenge here is that the receiver needs the ability to receive a weak signal in spite of a strong interfering signal given that software defined radio applications are often targeted to operate in a crowded and noisy spectrum space. For antenna/RF front ends the problem relates to antenna basics not software or waveform. One potential solution is tunable passive filters or switched filter banks. There are examples of these in the market but they are under developed at this stage.

Best regards,

Hall T.