Using Analog ASICs in Automation Applications

  • August 04, 2014
  • Feature

By Bob Frostholm,  JVD

Mary Shelley’s 1818 novel, Frankenstein, tells the story of a monster created with parts collected from random cadavers. The creature stands eight feet tall due to an inability to integrate all the necessary components into a standard humanoid form factor. Additionally, this haphazard collection of organs lacks sufficient neural network connections, accounting for its awkward gate and general stiffness of its arms and shoulders as it walks with forearms extended. This is perhaps the first documented evidence of the problems that can occur when designing a system using ‘point-products’, parts selected for their unique special function without regard for their perfect interoperability. Clearly, Mary Shelley was a visionary. Nowhere is the concept of developing a system with ‘point-products’ more hazardous than in its sensory inputs. It is here that the performance of every other aspect of the system is determined. However, using ‘point products’ elsewhere can be equally dangerous. Every part needs to work in perfect harmony with those around it. Dr. Frankenstein’s creation never had a chance. The human body is not unlike an analog system. We have inputs, our five senses. We have outputs, our controlled motions, speech, etc. And we have a processor, our mind. We have a power supply, our heart. And we have a signal path, our central nervous system. Mary Shelley’s knowledge of anatomy aided her in vividly describing her grotesque monster, allowing readers to draw comparisons to their more perfect selves.  As we all strive to make our designs perfect, we need to learn from Mary Shelley. Like Mary Shelley, each part we select for our design involves a compromise of one sort or another. To compromise is to add error and errors accumulate. So, how do we avoid turning our design into a Frankenstein monster? The first logical step is to reduce the parts count. Mary Shelley’s monster was an amalgamation of cadaver parts, some human and some animal. Imagine the theoretical complications of interfacing such disparate elements. Designing an Analog system can sometimes feel just as daunting. In reality, our job, as system designers is far easier but no less important. In the world of automation, Analog architectures play a critical role at the inputs (sensors) and outputs (actuators) of systems.  Let’s delve into the input side.  A sensor measures some physical quantity and converts it to an electrical signal. Behind the sensor is a myriad of electrical components responsible for the purification of the sensed information as well as the transformation of that information into a form that can readily utilized to make decisions that will ultimately guide the output of the system, whether a mechanical actuator or visual display. The circuitry that performs the conversion is often responsible for maintaining the calibration of the sensor as well. The semiconductor industry has made marvelous strides toward developing discrete components that play well in this environment, offering products with exceptionally low noise, and low drift with precision accuracy. For many/most applications, a circuit constructed from these analog ICs is more than sufficient. However, there are many applications where the use of such discrete standard product Analog ICs simply can’t do the job. Herein lies the need for an Analog ASIC. In an unbounded environment like Dr. Frankenstein’s lab, size is not an issue. But these days, as sensors become ubiquitous, size does matter. More and more, applications simply don’t have sufficient room to accommodate a circuit board populated with a handful of ICs. The solution is obvious yet sometimes difficult to obtain: Combine all of it into a custom Analog ASIC. Large analog semiconductor companies rarely participate in the custom Analog IC business because it’s just not profitable enough. Imagine a company like TI or Maxim, dedicating a team of engineers to design one chip that will be used by one customer. It happens, but the criteria to get such a project approved are formidable.  The volumes must be huge in order for them to recover the investment…even if the customer pays all the development and tooling costs (which is an industry norm). Why? There is a global shortage of good, highly qualified analog IC engineers and like any high value commodity in short supply, they are expensive and companies need to maximize the return on their investment. Remember, that same dedicated team might otherwise be developing a standard product that will be sold to thousands of customers in smaller volumes and therefore at higher prices.  It’s all in the math. Return on Investment (ROI) drives the decision making processes. Fortunately there are a handful of boutique Analog ASIC companies that address this underserved demand. In selecting one, CAUTION is the operative word. You only want a full custom design, not one that was created from analog cell libraries. Cells are fine for applications that must meet some industry standard like USB 3 or I2C interfaces. When addressing critical design aspects like those required of many sensor conditioning circuits, using analog and mixed signal cell libraries can be the kiss of death. Analog ASICs perform mission critical functions and it is mandatory that all elements of the design fit together perfectly (physically, which analog cells rarely do, as well as electrically), unlike our Frankenstein monster. Analog ASICs designed using cells from a library provided by or designed for the foundry that will actually fabricate the silicon chip often under- perform the discrete board level solution they were designed to replace. Why? Limited selection. For example, in the discrete IC solution, the original designer was able to choose from many hundreds of instrumentation amplifiers, selecting the one with the optimal performance for her/his application. When farming out an ASIC design to a company that uses cell libraries, the number of instrumentation amplifiers to choose from in that library may be less than ten, forcing the IC designer to make a ‘best fit’ decision.  Each ‘best fit’ decision adds some amount of compromising error into the design, which is cumulative. (see figure 1)   Another candidate for Analog ASIC integration in Automation is Power Management. Today’s complex systems employ a wide variety of semiconductor technologies and may need a vast array of supply voltages for proper performance. It’s easy to see the need for power management devices for 1.0V, 1.2V, 1.5V, 1.8V, 2.2V, 2.5V, 2.8V, 3.0V, 3.3V and more, all in the same box. Data sheets, PDKs and application notes make implementation easier than ever. If your volume is high enough, chip company application engineers are more than willing to do the design work for you. Sit back, watch You Tube, follow friends on Facebook and wait for the circuit to arrive by email. It’s not quite that simple, but let’s be honest, there are a lot of free resources out there to assist you. Before you jump on that bandwagon, consider who is paying that resource and what their real motive might be. Power Management is more than developing solutions that run cool and conserve power. It’s also about managing costs. Your costs. With today’s plethora of fifteen and twenty cent chips populating your power board, it’s easy to assume your design is financially viable. But is it? You should ask yourself, who is really managing your power management. Is it you or your suppliers? Who really understands your power management needs and more importantly, the solution you’ve implemented? Is your 7Amp 1.2V solution overkill for your 2.9Amp requirement? Could a lower cost LDO be used instead of that switcher? Financial management is inextricably intertwined with power management. Often power management solutions transcend multiple product generations. It’s the most logical place to drive cost out of a system for greater long term savings. Yet, for some reason, it’s also the most overlooked. The following figure represents the power board for a typical application. Ten different integrated circuits were used to generate the individual outputs needed for the application. At first glance, this is not a bad approach, considering the ubiquitousness and low cost of the devices.  Individually, each chip appears to offer the right cost/performance benefit for the solution it provides.  But, when taken as a whole, analysis of the entire solution reveals waste.     Depending on total volume, the Bill of Materials for the ten chips may range from between $1.50-$2.00 at the low end, to perhaps as high as $2.50. However, integrating these seven chips into an Analog ASIC would dramatically reduce size and still yield a much lower cost single chip solution while retaining all the desired power saving functionality of the original designs.   The cost of the ASIC for the above set of requirements would be in the neighborhood of $0.80 each. The cost savings comes from many sources; elimination of unused ‘features’ found in the original ten chips, reduction in the size of the silicon die compared to the ten die previously required, plastic package encapsulation of one chip vs. ten, development of one test system vs. ten,  physically testing one chip vs. ten, etc. Regardless of where in the system analog ICs are found, integration into an ASIC almost always yields user benefits. The graph below clearly shows the financial economic benefits of integration using an Analog ASIC. These numbers are typical and include amortization of all NRE and tooling costs to develop and put the Analog ASIC into production. The vertical scale is the total estimated lifetime volume of the Analog ASIC and the horizontal axis is the approximate cost of the components being replaced (integrated). Where does your application fit on this graph?   Thanks to the boutique Analog ASIC companies , custom analog solutions needn’t require mega-dollars or mega-units for justification. Stop wasting money today with off the shelf solutions. Your unique power management needs should reflect a uniquely low cost solution. Bob Frostholm is Director of Sales and Marketing at Analog ASIC Semiconductor company, JVD Inc. (San Jose, CA.) Bob has held Sales, Marketing and CEO roles at established and startup Analog Semiconductor Companies for more than 40 years. Bob was one of the original marketers behind the ubiquitous 555 timer chip. After 12 years with Signetics-Phillips, Fairchild and National Semiconductor, he co-founded his first startup in 1984, Scottish based Integrated Power, which was sold to Seagate in 1987. He subsequently joined Sprague’s semiconductor operations in Massachusetts and helped spin off its semiconductor group, creating what is now known as Allegro Microsystems. Bob has also just completed a screenplay, Tags, a technology-mystery centered in Silicon Valley. Bob is the author of several technical articles and white papers. Email: 

Learn More

Did you enjoy this great article?

Check out our free e-newsletters to read more great articles..