And its importance is underlined by the vast number of variants available. Tony Ochoa, director of Intersil's mixed signal and analogue group, said there are more than 3000 precision amplifiers on the market. Kevin Tretter, senior product marketing manager with Microchip's analogue and interface products group, added: "I like to think the market is a mile wide and an inch thick." Sales of op amps are growing by 5 to 6% a year and the market is pushing toward $3billion a year. It's no surprise that, far from seeing a shakeout, the op amp market is attracting new companies.
Those companies making op amps are responding to a number of drivers, such as noise, power consumption, bandwidth and cost. In order to meet those demands, manufacturers are exploring a range of process technologies, some exotic.
Steve Sockolov, product line director for Analog Devices' precision linear group noted: "While op amps can do almost anything, they can be a challenge for those who are not familiar with them." The reason? Sockolov said there isn't one 'stand out' product. "You can't say 'this is the best'. There are many parameters and each application needs to pick up on different parameters for optimisation."
So, in this age of integration, why is the op amp still a discrete part? Analog's Jim Doscher, director of high speed signal conditioning, said: "They are at the edge of the real world and have to deal with real world problems. They might have to handle high esd and emi, so they need to be robust and often use exotic processes to allow them to deal with this." Sockolov added: "The analogue signal may be ±10µV; an amp integrated on a chip may never see the analogue signal because, as digital uses smaller processes, it becomes noisier and misses these lower voltages."
Intersil's senior product line manager for high speed op amps, Duc Ngo, noted: "You need some sort of amp to feed an a/d converter. But you need flexibility in gain, noise, power and so on. That's why they aren't integrated in the converter; customers want flexibility."
Noise is one of the major issues with op amps, particularly with the increase in the use of sensors in all applications. Doscher noted: "Customers can't get enough noise performance. They are always trying to measure small signals on a low power budget and a lot of product differentiation in this market comes from noise performance."
Sockolov added that process technology is important. "Low noise for bipolar maybe unattainable in cmos, so you might have to settle for a different value."
Tretter said there are certain design trade offs which can be made to create lower noise devices. "For example, you can resize input transistors in order to lower the noise floor," he said. Now, he says, noise performance is 'three times better' than in Microchip's earlier op amps.
Ochoa pointed out there are different versions of noise. "There's 1/f noise, current noise and voltage noise. Each is important for different reasons, but the key question is 'can you achieve a good noise floor?'."
Ngo added: "Noise has changed over time in terms of requirements; it's more strategic than it was five years ago because the resolutions of a/d and d/a converters have increased, along with the sensitivity of sensors. Where we used to talk about 10nV/vHz, we now have to talk about 4nV/vHz. That poses a lot of challenges."
One way in which companies are looking to reduce noise is by returning to an older technology – chopper amps – and combining this with auto zero. "There's a million things that cause noise," Sockolov observed, "but we have done work with choppers in cmos to reduce noise."
"One of the reasons we're seeing choppers," said Ochoa, "is because they don't have the traditional 1/f asymptotic rise in noise as you get closer to dc; it's a flat equation all the way to dc."
Process technology is becoming an important differentiator. Tretter said Microchip had gone back to some of its older parts – its first op amp was introduced 12 years ago – and redesigned them. "Microchip's cmos processes for the digital world have always been linked with low power. If the gain bandwidth is less than 2MHz, then cmos is very effective. We've developed new compensation techniques and can now offer products with similar performance to older parts, but which consume up to 50% less power."
Ochoa said: "Not all processes are equivalent; Intersil has cmos, bicmos and bipolar processes, as well as SiGe for high speed. In fact, we're beginning to see trench isolation silicon on insulator processes. As an example, Intersil's PR40 process is beginning to support a better trade off between noise and power." Doscher believed: "Differentiation doesn't just come from design expertise, it also comes from process expertise and Analog is investing in new processes focused on amps and linear signal conditioning." Ngo pointed out: "If you don't have enough ft or a high voltage process, you can't do world class products."
Tretter added: "Bicmos has not been cost effective in the past, but advances in technology mean it's not so much of a concern and there are a lot more bicmos amps appearing from old and new companies alike."
Yet Tretter believes process technology is just one facet; he's also concerned with what he call 'specmanship'. "Microchip states amp specs at 60pF loads. Specifying at 20pF seems to be reasonable, but I have seen one part specified at zero loading. It makes the numbers look great, but users would never get the performance they need, so pay attention to the 'small print'."
Users are looking to source parts which consume less power, partly because of the rise of the portable device, where power is always important. "Low power used to be a differentiator," Doscher said, "but now it's a requirement; customers want half the power for the same performance."
One thing all companies are certain of is that op amps aren't going away. Sockolov concluded: "I can't say they will always be around, but we won't see the demise of the op amp in the near future."
