1 : Comparison of a tri-plate with a waveguide-based circuit. [1] 

Contexts in source publication

Context 1

... theory has its roots deep in history going back to the formulation of modern electromagnetic theory by James Clerk Maxwell in 1873. However, the birth of microwave technology dates back to an unfortunate event in our history, World War II. Radar, owing its existence to this event, was the product that flashed the light of the microwave industry [ 1, 2]. Nonetheless, at the end of WW II there were still very few active components (microwave sources) available to engineers. Unlike today’s designs, which tend to implement all possible functions with active devices or even digital circuitry, majority of the early electronic circuits relied mostly on passive components with the rectangular waveguide as the most important building block. After WW II commercialization began and the ever-growing trend of miniaturization started to shape the future of microwave industry. With the introduction of semiconductor technology the number of available active devices and circuit functions started to increase. This led to the reduction of the passive component count in microwave circuits, and transformed their physical properties. Initially a microwave circuit would look like a combination of pipes and screws that could easily be mistaken for a part of a plumbing system (Figure 1.1). As semiconductor components were introduced, it became more and more difficult to integrate them into these pipes due to their miniscule sizes. The arrival of the planar printed circuit board (PCB) technology in the 1960’s offered a solution. The bulky waveguides were now replaced by PCB-compatible, planar transmission lines such as the tri-plate, and microstrip (Figure 1.1), [1] which formed the basis of various passive elements including 90 and 180 hybrids, directional couplers, power dividers, and filters. Active devices would now be soldered on PCB that contained the prefabricated passive circuitry (Figure 1.2). Prior to 1980’s the microwave industry was heavily dependent on military applications. Products were power hungry, expensive and hefty devices that mainly worked in the 2-18 GHz range and higher frequency bands. As the military market started to saturate, the industry experts were predicting the end of the golden age of microwaves. The rescue came with the explosive growth of the personal communication market in the 1990’s which created new opportunities for the industry. But the introduction of personal computers, internet, and mobile phones required a change of mindset. Now the driving force was not just high performance, but also reduction of cost, size, weight, and power consumption of microwave devices, which was considered less crucial for military applications. Besides, since most personal communication systems were operating at lower microwave frequencies (below 2 GHz), transmission line components were out of the game. Their length being comparable to the propagation wavelength (centimeters at frequencies below 2 GHz), these components were simply too large to be integrated on a miniaturized circuit board. Instead, lumped elements such as inductors, capacitors, and resistors became the key components (Figure 1.3). Those lumped elements, together with surface or bulk acoustic wave filters were fabricated independently and then integrated on the circuit board together with active components in a discrete manner. However, this approach was soon to be proven inadequate in view of the growing complexity of mobile wireless circuits. Introduction of different communication bands and digital communication methods, and the demand for multipurpose data transmission, imposed a multiband and multiprotocol structure on the mobile devices. Thus, the number of passive components increased tremendously reaching to a whopping 300 for an average mobile phone [4]. A fabrication error in just one component meant costly problems. Also performance tolerances became more and more stringent. In order to cope with the complexity and tight tolerances, and to increase repeatability, the RF industry had now to adapt the rules of microelectronics industry and had to shift most of its components from discrete to integrated devices. Initially, however, the integration of passive devices into the integrated circuit (IC) substrate was blocked by the textbook rule, which stated that it would be illogical to put inductors with meaningful inductance values on a substrate [6], since parasitics would have detrimental effects on the quality factor. Indeed this argument was correct for Si substrates, which were introducing significant amount of parasitics and substrate loss due to the very low substrate resistivity (2-3 Ω cm) required for active device integration. Yet, with the availability of larger GaAs wafers, a very good substrate for high-frequency active devices due to its high electron mobility, there was enough real estate on the wafer to accommodate the rather bulky inductors. Moreover, these devices were of sufficient quality thanks to the high resistivity of GaAs (in excess of 1 k Ω cm). This technological advancement paved the way for on-chip passive integration and, thus, the emergence of Monolithic Microwave Integrated Circuits (MMICs). Now lumped passive devices were fabricated together with the active components on a single chip. However, the high cost of GaAs wafers hindered its penetration into the commercial market. Since the commercial communication devices worked at relatively low frequencies (below 2 GHz), with a bit more relaxed requirements than military systems, it was a bit luxurious to implement passive devices on expensive GaAs substrates for commercial applications. Hence eyes were turned onto Si again. Silicon mainly offered bipolar junction transistors (BJT) and metal oxide semiconductor (MOS). BJTs were more favorable in radio frequency (RF) design due to their resistive type of operation. Unlike BJTs, MOS transistors relied on a capacitive gate control for their operation. As frequency increased MOS gates would start leaking current and performance would drop. This made complementary MOS (CMOS) a low pass process in nature, hence they were not suitable for the current circuit techniques [6], which used mostly discrete elements such as transmission lines or filters with 5-10 BJTs as the active circuit based on well established heterodyne receiver. Therefore MOS appropriate circuit techniques had to be developed. As an example first reported CMOS RF amplifiers had an integrated micromachined inductor to inductively tune the transistor capacitance to convert inherent low-pass behavior to band-pass [7]. Integration of the inductor was crucial in order to reduce parasitics which would shift the pass band, hence jeopardize the whole operation. Nevertheless, as time went by, CMOS components improved with mind boggling pace following Moore’s Law in feature size, hence gate capacitance, reduction. Soon active devices had sufficiently high cut-off frequencies to work in the low GHz range. In addition, researchers worked out new circuit topologies specifically suitable for CMOS [6]. Some of these innovative circuit designs required on-chip passive devices, hence came the first passives on Si. Eventually, the number of integrated passive devices increased with the research on RF CMOS circuits leading to an era of Si passive integration. The first examples of integrated passives in RF CMOS circuits came out in early 90’s and the efforts towards the development of single chip RF CMOS transceivers is going on ever since [7, 8]. However, industry acceptance of the concept is still away from completion. Current wireless communication devices employ a mixture of technologies forming a hybrid circuit. CMOS is used for baseband signal processing and for low intermediate frequency (IF) in general [9], GaAs heterojunction bipolar transistors (HBT) are preferred for power amplifiers (PA), and a mixture of on-chip and off-chip passives are used for various functions from biasing to matching. Considering the passives, there are several handicaps that are hampering fully-integrated, single-chip devices. The most pronounced problem is the effect of the Si substrate which was originally adopted to meet the needs of digital CMOS, not RF. The conductive nature of Si causes various problems for RF passive devices having detrimental effects on inductors in particular. It is instructive to explain these effects using a lumped element spiral inductor model given in Figure ...

Context 2

... theory has its roots deep in history going back to the formulation of modern electromagnetic theory by James Clerk Maxwell in 1873. However, the birth of microwave technology dates back to an unfortunate event in our history, World War II. Radar, owing its existence to this event, was the product that flashed the light of the microwave industry [ 1, 2]. Nonetheless, at the end of WW II there were still very few active components (microwave sources) available to engineers. Unlike today’s designs, which tend to implement all possible functions with active devices or even digital circuitry, majority of the early electronic circuits relied mostly on passive components with the rectangular waveguide as the most important building block. After WW II commercialization began and the ever-growing trend of miniaturization started to shape the future of microwave industry. With the introduction of semiconductor technology the number of available active devices and circuit functions started to increase. This led to the reduction of the passive component count in microwave circuits, and transformed their physical properties. Initially a microwave circuit would look like a combination of pipes and screws that could easily be mistaken for a part of a plumbing system (Figure 1.1). As semiconductor components were introduced, it became more and more difficult to integrate them into these pipes due to their miniscule sizes. The arrival of the planar printed circuit board (PCB) technology in the 1960’s offered a solution. The bulky waveguides were now replaced by PCB-compatible, planar transmission lines such as the tri-plate, and microstrip (Figure 1.1), [1] which formed the basis of various passive elements including 90 and 180 hybrids, directional couplers, power dividers, and filters. Active devices would now be soldered on PCB that contained the prefabricated passive circuitry (Figure 1.2). Prior to 1980’s the microwave industry was heavily dependent on military applications. Products were power hungry, expensive and hefty devices that mainly worked in the 2-18 GHz range and higher frequency bands. As the military market started to saturate, the industry experts were predicting the end of the golden age of microwaves. The rescue came with the explosive growth of the personal communication market in the 1990’s which created new opportunities for the industry. But the introduction of personal computers, internet, and mobile phones required a change of mindset. Now the driving force was not just high performance, but also reduction of cost, size, weight, and power consumption of microwave devices, which was considered less crucial for military applications. Besides, since most personal communication systems were operating at lower microwave frequencies (below 2 GHz), transmission line components were out of the game. Their length being comparable to the propagation wavelength (centimeters at frequencies below 2 GHz), these components were simply too large to be integrated on a miniaturized circuit board. Instead, lumped elements such as inductors, capacitors, and resistors became the key components (Figure 1.3). Those lumped elements, together with surface or bulk acoustic wave filters were fabricated independently and then integrated on the circuit board together with active components in a discrete manner. However, this approach was soon to be proven inadequate in view of the growing complexity of mobile wireless circuits. Introduction of different communication bands and digital communication methods, and the demand for multipurpose data transmission, imposed a multiband and multiprotocol structure on the mobile devices. Thus, the number of passive components increased tremendously reaching to a whopping 300 for an average mobile phone [4]. A fabrication error in just one component meant costly problems. Also performance tolerances became more and more stringent. In order to cope with the complexity and tight tolerances, and to increase repeatability, the RF industry had now to adapt the rules of microelectronics industry and had to shift most of its components from discrete to integrated devices. Initially, however, the integration of passive devices into the integrated circuit (IC) substrate was blocked by the textbook rule, which stated that it would be illogical to put inductors with meaningful inductance values on a substrate [6], since parasitics would have detrimental effects on the quality factor. Indeed this argument was correct for Si substrates, which were introducing significant amount of parasitics and substrate loss due to the very low substrate resistivity (2-3 Ω cm) required for active device integration. Yet, with the availability of larger GaAs wafers, a very good substrate for high-frequency active devices due to its high electron mobility, there was enough real estate on the wafer to accommodate the rather bulky inductors. Moreover, these devices were of sufficient quality thanks to the high resistivity of GaAs (in excess of 1 k Ω cm). This technological advancement paved the way for on-chip passive integration and, thus, the emergence of Monolithic Microwave Integrated Circuits (MMICs). Now lumped passive devices were fabricated together with the active components on a single chip. However, the high cost of GaAs wafers hindered its penetration into the commercial market. Since the commercial communication devices worked at relatively low frequencies (below 2 GHz), with a bit more relaxed requirements than military systems, it was a bit luxurious to implement passive devices on expensive GaAs substrates for commercial applications. Hence eyes were turned onto Si again. Silicon mainly offered bipolar junction transistors (BJT) and metal oxide semiconductor (MOS). BJTs were more favorable in radio frequency (RF) design due to their resistive type of operation. Unlike BJTs, MOS transistors relied on a capacitive gate control for their operation. As frequency increased MOS gates would start leaking current and performance would drop. This made complementary MOS (CMOS) a low pass process in nature, hence they were not suitable for the current circuit techniques [6], which used mostly discrete elements such as transmission lines or filters with 5-10 BJTs as the active circuit based on well established heterodyne receiver. Therefore MOS appropriate circuit techniques had to be developed. As an example first reported CMOS RF amplifiers had an integrated micromachined inductor to inductively tune the transistor capacitance to convert inherent low-pass behavior to band-pass [7]. Integration of the inductor was crucial in order to reduce parasitics which would shift the pass band, hence jeopardize the whole operation. Nevertheless, as time went by, CMOS components improved with mind boggling pace following Moore’s Law in feature size, hence gate capacitance, reduction. Soon active devices had sufficiently high cut-off frequencies to work in the low GHz range. In addition, researchers worked out new circuit topologies specifically suitable for CMOS [6]. Some of these innovative circuit designs required on-chip passive devices, hence came the first passives on Si. Eventually, the number of integrated passive devices increased with the research on RF CMOS circuits leading to an era of Si passive integration. The first examples of integrated passives in RF CMOS circuits came out in early 90’s and the efforts towards the development of single chip RF CMOS transceivers is going on ever since [7, 8]. However, industry acceptance of the concept is still away from completion. Current wireless communication devices employ a mixture of technologies forming a hybrid circuit. CMOS is used for baseband signal processing and for low intermediate frequency (IF) in general [9], GaAs heterojunction bipolar transistors (HBT) are preferred for power amplifiers (PA), and a mixture of on-chip and off-chip passives are used for various functions from biasing to matching. Considering the passives, there are several handicaps that are hampering fully-integrated, single-chip devices. The most pronounced problem is the effect of the Si substrate which was originally adopted to meet the needs of digital CMOS, not RF. The conductive nature of Si causes various problems for RF passive devices having detrimental effects on inductors in particular. It is instructive to explain these effects using a lumped element spiral inductor model given in Figure ...