The asics in the mcms have been designed by Agilent's High Frequency Technology Centre (HFTC). Daniel Thomasson is the centre's director and senior manager. "Our job is to be a world leader in high frequency technology," he noted. "We do this by innovating and manufacturing monolithic microwave ics (mmics) that allow Agilent to differentiate its products."
The 90000-X project started from a clean sheet. "This is the first time Agilent has aimed III-V technology at scopes in a major way and the first time we have applied III-V technology at this level in a classic frequency domain application," Thomasson recalled. "It took a couple of years; longer than usual, because we were doing something new." But the 'clean sheet' approach allowed HFTC to take a fresh approach to the architecture, working closely with the instrument's designers.
Thomasson said the HFTC designs and makes many of the compound semiconductor products that go into the front ends and other parts of Agilent's test and measurement equipment. "We have access to 11 core technologies on one process and test platform," he said. "Currently, 190 ics developed here feature in Agilent's instruments, with more designs in the R&D pipeline."
Activities at the HFTC span the full life cycle said Thomasson – from applied research to manufacturing. "We are an integrated centre which has always been a part of Agilent and, before that, HP. But we also collaborate with universities, including Simon Fraser University in Canada."
Amongst the III-V technologies available at HTFC are: GaAs/InGaP heterojunction bipolar transistors (HBT); InP HBT (HB2B); GaAs on 0.11 and 0.25µm, high electron mobility transistors, fets and diodes.
The asics are fabricated using Agilent's HB2B process. "This is a 200GHz process," Thomasson explained, "which is optimised for demanding requirements and is always used to get bandwidth. Because the process supports low noise and jitter, it is much easier to make compromises."
Transistors in the HB2B process are based on a proprietary epitaxial stack, with a quaternary emitter. The process also uses a GaAsSb base, which Thomasson describes as 'advantageous' compared to other systems. There is also a dual collector based on a 1µm emitter. "All of this allows us to get higher speed at larger geometries," Thomasson continued. "A 1µm device can operate at 200GHz, with an fmax of 240GHz and a breakdown voltage of 7V."
But why InP? "If you look at a plot of breakdown voltage against transistor frequency (ft), InP pushes the curve upwards dramatically. While there is more to it than ft alone, you have to use a high ft process in order to get the bandwidth."
According to Thomasson, InP doubles the useable breakdown voltage compared with processes such as SiGe and GaAs. "These swings are needed, especially for serious applications where, if you're dealing with 1.5V, it's a showstopper for the ics we want to design." He added that InP provides a road map that allows frequency to be pushed higher while maintaining what he called 'beneficial characteristics'. "Electrons move more quickly in InP at any geometry," Thomasson pointed out, "and this can be as much as an order of magnitude better. In turn, this will allow bigger, more robust devices to be made with lower development costs."
The baseline HB2B process was developed by 2004, with an enhancement in 2007. "We've been shipping ics for use in instruments since 2004," Thomasson noted, "and have 14 designs in seven instruments, with 25 more designs in development."
Yet the level of design complexity, in terms of transistor count, is far lower than might be expected. "We can do designs with up to 1000 transistors," said Thomasson. "It's a more limited level of integration that found with silicon based systems, but in most cases, 1000 transistors cover 95% of what we want to do. It's a good trade off and we can access foundries if we need a higher level of integration."
Despite an apparent lack of complexity, Thomasson said his designers were balancing 'about 24 top level parameters'. "Packaging – and making sure we design with packaging in mind from the start, is also important. We try to do as much as we can in qfn packages, but when these things go into microwave circuits, packaging can make a big difference."
The reason for HFTC, said Thomasson, is to hit the 'sweet spot' for ics, the point where high performance aligns with instrument grade reliability. "There are different sweet spots for commercial markets," he noted. "For example, there are some high performance high frequency parts that may not have instrument grade quality and reliability."
HFTC addresses requirements up to 110GHz, with its distinguishing characteristic being that many of its designs span from dc to 110GHz using one amplifier. "Our aim is high frequency, high reliability," Thomasson noted, "and the foundation is our fab process, which has everything you'd expect."
One particular strength is characterisation. "We put a lot of investment into this area," Thomasson noted, "and it's one of the things which sets our devices apart. We test them hard before using them and we believe that we do an order of magnitude more testing than other places."
Modelling is another area where a lot of effort is expended. Some of the approaches include: integrated analytic and empirical device models; behavioural circuit models; Monte Carlo; and X-parameters.
The centre is also exploring the application of MEMS technology in high frequency circuits. "For example, we have used a thermocouple power sensor, as well as more conventional MEMS switches," Thomasson noted. The aim is to construct simpler, less costly front ends. "We don't see MEMS as a centrepiece, but it is an important part of the portfolio. E-mech switches tend to be big blocks and have a poor lifetime. We want to push in the other direction and MEMS will help," he concluded.
Compound creativity
4 mins read
How compound semiconductor technology helped scope hit bandwidth target.
When Agilent launched its Infiniium 90000-X range of oscilloscopes in April, the headlines were stolen by the scope's real time analogue bandwidth of 32GHz, along with its low noise and jitter.
Part of the design philosophy for the 90000-X range was to provide a true 32GHz analogue bandwidth, rather than fall back on enhancement techniques, such as dsp boosting or frequency domain interleave. According to Agilent, while these approaches will boost bandwidth, they add noise and distortions so, while bandwidth is increased, it is at the expense of accuracy.
Robert Drollinger, product planner for the Infiniium range, said instrument accuracy was critical. "Users want to make sure they have margins and noise floor is a key requirement in order to get the most accurate representation of voltage level."
Agilent has provided this level of performance through the development of a thick film multichip module (mcm) featuring five indium phosphide (InP) asics. Each module – there are two in each scope – includes two 32GHz preamplifiers, two edge trigger chips running at speeds in excess of 20GHz and a 32GHz sampler. And it is the use of InP that has allowed the 32GHz bandwidth to be achieved.
The asics in the mcms have been designed by Agilent's High Frequency Technology Centre (HFTC). Daniel Thomasson is the centre's director and senior manager. "Our job is to be a world leader in high frequency technology," he noted. "We do this by innovating and manufacturing monolithic microwave ics (mmics) that allow Agilent to differentiate its products."
The 90000-X project started from a clean sheet. "This is the first time Agilent has aimed III-V technology at scopes in a major way and the first time we have applied III-V technology at this level in a classic frequency domain application," Thomasson recalled. "It took a couple of years; longer than usual, because we were doing something new." But the 'clean sheet' approach allowed HFTC to take a fresh approach to the architecture, working closely with the instrument's designers.
Thomasson said the HFTC designs and makes many of the compound semiconductor products that go into the front ends and other parts of Agilent's test and measurement equipment. "We have access to 11 core technologies on one process and test platform," he said. "Currently, 190 ics developed here feature in Agilent's instruments, with more designs in the R&D pipeline."
Activities at the HFTC span the full life cycle said Thomasson – from applied research to manufacturing. "We are an integrated centre which has always been a part of Agilent and, before that, HP. But we also collaborate with universities, including Simon Fraser University in Canada."
Amongst the III-V technologies available at HTFC are: GaAs/InGaP heterojunction bipolar transistors (HBT); InP HBT (HB2B); GaAs on 0.11 and 0.25µm, high electron mobility transistors, fets and diodes.
The asics are fabricated using Agilent's HB2B process. "This is a 200GHz process," Thomasson explained, "which is optimised for demanding requirements and is always used to get bandwidth. Because the process supports low noise and jitter, it is much easier to make compromises."
Transistors in the HB2B process are based on a proprietary epitaxial stack, with a quaternary emitter. The process also uses a GaAsSb base, which Thomasson describes as 'advantageous' compared to other systems. There is also a dual collector based on a 1µm emitter. "All of this allows us to get higher speed at larger geometries," Thomasson continued. "A 1µm device can operate at 200GHz, with an fmax of 240GHz and a breakdown voltage of 7V."
But why InP? "If you look at a plot of breakdown voltage against transistor frequency (ft), InP pushes the curve upwards dramatically. While there is more to it than ft alone, you have to use a high ft process in order to get the bandwidth."
According to Thomasson, InP doubles the useable breakdown voltage compared with processes such as SiGe and GaAs. "These swings are needed, especially for serious applications where, if you're dealing with 1.5V, it's a showstopper for the ics we want to design." He added that InP provides a road map that allows frequency to be pushed higher while maintaining what he called 'beneficial characteristics'. "Electrons move more quickly in InP at any geometry," Thomasson pointed out, "and this can be as much as an order of magnitude better. In turn, this will allow bigger, more robust devices to be made with lower development costs."
The baseline HB2B process was developed by 2004, with an enhancement in 2007. "We've been shipping ics for use in instruments since 2004," Thomasson noted, "and have 14 designs in seven instruments, with 25 more designs in development."
Yet the level of design complexity, in terms of transistor count, is far lower than might be expected. "We can do designs with up to 1000 transistors," said Thomasson. "It's a more limited level of integration that found with silicon based systems, but in most cases, 1000 transistors cover 95% of what we want to do. It's a good trade off and we can access foundries if we need a higher level of integration."
Despite an apparent lack of complexity, Thomasson said his designers were balancing 'about 24 top level parameters'. "Packaging – and making sure we design with packaging in mind from the start, is also important. We try to do as much as we can in qfn packages, but when these things go into microwave circuits, packaging can make a big difference."
The reason for HFTC, said Thomasson, is to hit the 'sweet spot' for ics, the point where high performance aligns with instrument grade reliability. "There are different sweet spots for commercial markets," he noted. "For example, there are some high performance high frequency parts that may not have instrument grade quality and reliability."
HFTC addresses requirements up to 110GHz, with its distinguishing characteristic being that many of its designs span from dc to 110GHz using one amplifier. "Our aim is high frequency, high reliability," Thomasson noted, "and the foundation is our fab process, which has everything you'd expect."
One particular strength is characterisation. "We put a lot of investment into this area," Thomasson noted, "and it's one of the things which sets our devices apart. We test them hard before using them and we believe that we do an order of magnitude more testing than other places."
Modelling is another area where a lot of effort is expended. Some of the approaches include: integrated analytic and empirical device models; behavioural circuit models; Monte Carlo; and X-parameters.
The centre is also exploring the application of MEMS technology in high frequency circuits. "For example, we have used a thermocouple power sensor, as well as more conventional MEMS switches," Thomasson noted. The aim is to construct simpler, less costly front ends. "We don't see MEMS as a centrepiece, but it is an important part of the portfolio. E-mech switches tend to be big blocks and have a poor lifetime. We want to push in the other direction and MEMS will help," he concluded.
The asics in the mcms have been designed by Agilent's High Frequency Technology Centre (HFTC). Daniel Thomasson is the centre's director and senior manager. "Our job is to be a world leader in high frequency technology," he noted. "We do this by innovating and manufacturing monolithic microwave ics (mmics) that allow Agilent to differentiate its products."
The 90000-X project started from a clean sheet. "This is the first time Agilent has aimed III-V technology at scopes in a major way and the first time we have applied III-V technology at this level in a classic frequency domain application," Thomasson recalled. "It took a couple of years; longer than usual, because we were doing something new." But the 'clean sheet' approach allowed HFTC to take a fresh approach to the architecture, working closely with the instrument's designers.
Thomasson said the HFTC designs and makes many of the compound semiconductor products that go into the front ends and other parts of Agilent's test and measurement equipment. "We have access to 11 core technologies on one process and test platform," he said. "Currently, 190 ics developed here feature in Agilent's instruments, with more designs in the R&D pipeline."
Activities at the HFTC span the full life cycle said Thomasson – from applied research to manufacturing. "We are an integrated centre which has always been a part of Agilent and, before that, HP. But we also collaborate with universities, including Simon Fraser University in Canada."
Amongst the III-V technologies available at HTFC are: GaAs/InGaP heterojunction bipolar transistors (HBT); InP HBT (HB2B); GaAs on 0.11 and 0.25µm, high electron mobility transistors, fets and diodes.
The asics are fabricated using Agilent's HB2B process. "This is a 200GHz process," Thomasson explained, "which is optimised for demanding requirements and is always used to get bandwidth. Because the process supports low noise and jitter, it is much easier to make compromises."
Transistors in the HB2B process are based on a proprietary epitaxial stack, with a quaternary emitter. The process also uses a GaAsSb base, which Thomasson describes as 'advantageous' compared to other systems. There is also a dual collector based on a 1µm emitter. "All of this allows us to get higher speed at larger geometries," Thomasson continued. "A 1µm device can operate at 200GHz, with an fmax of 240GHz and a breakdown voltage of 7V."
But why InP? "If you look at a plot of breakdown voltage against transistor frequency (ft), InP pushes the curve upwards dramatically. While there is more to it than ft alone, you have to use a high ft process in order to get the bandwidth."
According to Thomasson, InP doubles the useable breakdown voltage compared with processes such as SiGe and GaAs. "These swings are needed, especially for serious applications where, if you're dealing with 1.5V, it's a showstopper for the ics we want to design." He added that InP provides a road map that allows frequency to be pushed higher while maintaining what he called 'beneficial characteristics'. "Electrons move more quickly in InP at any geometry," Thomasson pointed out, "and this can be as much as an order of magnitude better. In turn, this will allow bigger, more robust devices to be made with lower development costs."
The baseline HB2B process was developed by 2004, with an enhancement in 2007. "We've been shipping ics for use in instruments since 2004," Thomasson noted, "and have 14 designs in seven instruments, with 25 more designs in development."
Yet the level of design complexity, in terms of transistor count, is far lower than might be expected. "We can do designs with up to 1000 transistors," said Thomasson. "It's a more limited level of integration that found with silicon based systems, but in most cases, 1000 transistors cover 95% of what we want to do. It's a good trade off and we can access foundries if we need a higher level of integration."
Despite an apparent lack of complexity, Thomasson said his designers were balancing 'about 24 top level parameters'. "Packaging – and making sure we design with packaging in mind from the start, is also important. We try to do as much as we can in qfn packages, but when these things go into microwave circuits, packaging can make a big difference."
The reason for HFTC, said Thomasson, is to hit the 'sweet spot' for ics, the point where high performance aligns with instrument grade reliability. "There are different sweet spots for commercial markets," he noted. "For example, there are some high performance high frequency parts that may not have instrument grade quality and reliability."
HFTC addresses requirements up to 110GHz, with its distinguishing characteristic being that many of its designs span from dc to 110GHz using one amplifier. "Our aim is high frequency, high reliability," Thomasson noted, "and the foundation is our fab process, which has everything you'd expect."
One particular strength is characterisation. "We put a lot of investment into this area," Thomasson noted, "and it's one of the things which sets our devices apart. We test them hard before using them and we believe that we do an order of magnitude more testing than other places."
Modelling is another area where a lot of effort is expended. Some of the approaches include: integrated analytic and empirical device models; behavioural circuit models; Monte Carlo; and X-parameters.
The centre is also exploring the application of MEMS technology in high frequency circuits. "For example, we have used a thermocouple power sensor, as well as more conventional MEMS switches," Thomasson noted. The aim is to construct simpler, less costly front ends. "We don't see MEMS as a centrepiece, but it is an important part of the portfolio. E-mech switches tend to be big blocks and have a poor lifetime. We want to push in the other direction and MEMS will help," he concluded.
