• Française
  • Deutsch
  • English
Although RF circuits are generally considered to be circuits that operate from tens of MHz up to 1GHz, and microwave circuits at frequencies beyond that, boundaries based purely on frequency are rarely appropriate. Analog integrated circuits based on lower-frequency design methodologies can now operate well into the microwave range, purely because of smaller feature sizes that are now available in CMOS and silicon-germanium technologies. BiCMOS integrated circuits that operate in the microwave frequency range, designed using low frequency architectures, are now abundant. However, classical microwave circuit design techniques are still important to model and understand problems arising from noise, mismatch, circuit losses, and limited bandwidth. We will focus on circuits that are differentiated from their historically lower-frequency counterparts by several features. In RF and microwave design, the phase shift of the component is significant because its size is comparable with a wavelength, its reactances and parasitics must be accounted for, and reflections occur between elements. We need to consider circuit losses that degrade the Q of an element as well as introduce noise, and nonlinearities that introduce distortion into the signal path. Electromagnetic radiation and capacitive coupling will also be features of such circuits. With integrated circuits, these 'RF and microwave' effects are most commonly observed when assembling circuits together at higher frequencies into systems, or when using discrete or custom devices.
Data Converters suffer from a wide range of non-ideal effects such as mismatch, parasitics, finite gain / bandwidth / slew rate, offset, noise and so on and so forth. Many of these effects are random effects of statistical nature. They also depend on first and second order characteristics of the process technology, layout, packaging, pinout etc. It is practically impossible to describe all these effects for all possible architectures, sizings, layouts on device (“SPICE”) level and to compare all the simulation results to decide on the best implementation. It is thus necessary to build higher level models which include all relevant effects, at least in the first order, but are still simple enough for quick modifications and efficient study of various layouts, topologies etc. Simulation times must not go beyond minutes. The relevant effects must, of course, be understood to be able to simplify and model them. When Data Converter models are coded and simulated, when first hardware samples come in, and finally during production, they have to be analyzed and their performance determined. Even if the converter should not be directly measured during production test, direct access is required for first characterization, to establish correlations between analysis and production test, and in case of problems. Time versus frequency domain, static versus dynamic characterization – the test methodology must yield the specified variables. Without thorough understanding of the underlying basics misinterpretations and large errors are possible.

CEI-Europe AB, Repslagaregatan 19, SE-582 22 Linköping, Sweden Phone +46-13-100 730 Fax +46-13-100 731 cei@cei.se