In all areas of business, increased communication leads to more streamlined processes and greater potential for success. This is no less true in system design. Within large design organizations, baseband Field Programmable Gate Array (FPGA) and Radio Frequency (RF) signal processing communities have traditionally been separated. Today, however, many RF functions are moving into the algorithmic world and this is making communication between the two areas more crucial than ever.
Co-design, a process whereby all stakeholders actively participate in the design process, offers a solution to this dilemma. While organizations have certainly successfully designed electronic systems like wireless communication systems, smart phones and space systems without co-design, its’ use offers not only the potential for reduced risk and cost, but accelerated time-to-market as well; enviable advantages for any modern design organization. It’s not surprising then, that the co-design of baseband algorithms and RF distributed circuits has emerged as a critical and important trend. For modern design organizations, it’s now more imperative than ever that they simulate both Digital Signal Processer (DSP) and RF circuits together at the system level using mixed signals.
Without a doubt, co-design offers designers a number of important benefits. For example, by co-designing DSP and RF circuits, over design, inaccurate predictions and difficulties in assembly can be prevented. In contrast, organizations that utilize disparate design flows with little to no communication and don’t co-validate along the way can anticipate added cost and complications. The algorithmic engineers have a hard time accounting for RF signal impairments in the signal processing, while those working on RF circuits aren’t able to clearly see how their systems are being used in the overall
infrastructure. Co-validation during the design process allows for some of these problems to be accurately accounted for, and lets organizations find the cheapest technique to solve a problem.
Despite these benefits, DSP and RF co-design is not without its challenges. DSP and RF circuits are simulated with different techniques. DSP is typically done in the numeric and time domain whereas RF circuits are designed using frequency-domain techniques. Converting between the two domains takes a long time and even when RF is changed into the time domain, many elements still exist in the frequency domain. You either have a very detailed RF model that is trusted completely, but need to simplify the signals, or arbitrary signals on the baseband side with an extremely simplified RF. Co-design bridges the gap, offering the right level of behavioral modeling on both sides.
Interested in learning more about DSP and RF co-design? Watch this 13- minute video on YouTube - Dr. Murthy Upmaka's talk at IMS 2016.