Power, Performance — Avionics Designers Want It All

Not long ago, the prevailing philosophy among chip designers for aviation systems could be summed up as, “I feel the need — the need for speed.” Today, aviation’s top guns have pulled back on the throttle a bit. There’s a more nuanced discussion balancing the need for performance versus power, with other factors coming into consideration such as safety, security certifications and overall product lifecycle.

Despite the pandemic and a collapse in commercial airline travel, the overall aviation market experienced steady growth. In its January 2021 report, Research and Markets expects semiconductor revenue for the military and aerospace market to grow by $632 million from 2020 to 2024 at a compound annual growth rate (CAGR) of 2%. Key drivers are the continued upgrading and modernization of aircraft, and rising military expenditures.

“In the old days, it was often performance, performance, performance,” said Ian Land, senior director, aerospace and defense vertical solutions at Synopsys. “The attitude was, ‘I’ve got a radar system to design, I need to crank that thing up as much as possible.’ Now, there’s a lot more discussion about performance versus power. They can go build the performance over the power requirements that they’re concerned about.”

One of the best ways to achieve better performance and lower power is to focus on the chip architecture, specifically the hardware-software partitioning. biggest impact is in the chip architecture, and specifically in hardware-software partitioning,

“Power consumption is still a beast, but it’s not always a 1:1 ratio,” said Russell Klein, HLS platform program director at Siemens EDA. “Faster designs consume more power, regardless of industry. Adding computational elements to a chip design that increase throughput will have a corresponding increase in power consumption. If data paths need to be expanded to support these additional computational elements, this will increase power, as well. However, the increase for the entire design usually will not be linear.”

While the extra computational elements use more power — and sometimes so do the wider buses — power consumption associated with data storage remains constant.

“In many designs, power used in storing and retrieving the data, including current leakage in memories, can be much greater than the power used in computation,” Klein said. “This is especially true for the smaller geometry technology nodes. For most designs, as additional logic is added, performance will increase at a greater rate than power consumption. So faster designs are more power-hungry, but are usually more energy-efficient. Since data storage and movement tend to dominate power consumption in today’s designs, anything that reduces the need to move or store data can improve both power and performance.”

Source: NASA Maxwell – Electric Propulsion Airplane

Aviation goes electric
As technology marches forward, one of the interesting moves is toward all-electric options. Smaller, lighter, higher-power electronics to facilitate electric actuation, generation, and propulsion functions eventually will help power the all-electric aircraft.

There are a few obstacles to overcome before the industry gets there, primarily the need for increased efficiency resulting in smaller, lighter form factors, for both the electronics and the cooling systems because of lower heat loss, said Paul Quintana, associate director of the Aerospace and Defense Group at Microchip Technology. Another challenge is the speed of adoption and use of new technologies that do not have decades of heritage for proving their reliability and useful life.

Microchip’s answer to this is wide-bandgap semiconductors. Silicon carbide (SiC) is used to reduce switching and DC power losses, thereby achieving higher efficiency in smaller and lighter aviation power systems. Incorporating SiC in the power system allows for extremely short switching times, which means similarly low switching losses, and allows for higher switching frequencies, resulting in much smaller systems.

This is important in existing avionics, but it’s critical in electric-powered airplanes, which currently are under development by startups and established companies across the globe. As with electric cars, one of the biggest challenges is range, and improvements in performance per watt can determine whether a plane reaches its destination safely.

Design issues and approaches
Much of this is brand new, and that has created some trepidation in an industry that has achieved safety through well-tested evolutionary technology. In this market, even being able to move more data through faster interfaces, which are commonplace in other markets, has created issues.

“High-speed interfaces like PCIe and Ethernet are outside compliance for DO-254,” said Louie De Luna, director of marketing at Aldec. “The problem is that you need to be able to capture the results, and there is no way to do that. When you look at debug and view the waveforms you get non-deterministic results. With PCIe, you need to learn UVM and PLM concepts, and that’s not easy to do for a culture that uses legacy technology.”

Not everything is lagging technology, though. The advent of digital twins, or using virtual prototypes, though not new, has gained momentum and can help root out errors in the design phase with speed, accuracy and while running simulations on multi-use modeling. With a digital twin, you create a digital representation of a physical system. It allows designers to quickly spin scenarios, learn from mistakes, and see what works well – all in a virtual environment. You quickly figure out, in a digital environment, proof-of-concept on a design’s power performance optimization, safety and security optimization, etc. And you can model multiple system versions at the same time. “This is reducing time-to-market, and time to deployment,” said Land.

For example, if a designer has a system targeting use in a commercial aircraft, and also wants to use it in a helicopter and unmanned aerial vehicles (UAVs), it means retargeting systems for different environments. “They all have different considerations – they fly at different heights in the air, which means they require different temperature performances, different safety requirements,” said Land. “From a business standpoint, it makes a lot of sense to have these digital representations that can be spun toward each separate design cleverly and safely.”

Land noted that he has discussions nearly every week regarding the use of digital twins to aid in the design process. He expects more to follow.

Using this digital transformation can eliminate unnecessary work from later phases in getting products to market, since you’re testing real-world scenarios until you find the winning combination, said Frank Schirrmeister, senior group director for solutions and ecosystems at Cadence. The future is now, said Schirrmeister, who also sees designers using virtual prototypes to gain competitive advantage in aerospace. He pointed to Will Roper’s digital acquisition whitepaper, There Is No Spoon, as an example of the game-changing effects of digital twins.

It’s a radical advance for the design phase. Comparing the digital transformation to the movie The Matrix, Schirrmeister said that Roper “talks about the digital transformation, where engineers are instead of physically doing this, build, testing, prototyping, failing — and then starting another — you are able to get to earlier integration of different technologies or different components and do it virtually right. Based on simulation, you know what your power consumption will be, what your thermal effects will be, as opposed to finding these things out much later in the design cycle.”

Planes trained by automobiles
That’s not to say avionics designers are without a blueprint. There are parallels with the automotive industry relative to safety and reliability, and many parallels exist. System designs for both also have more stringent design requirements, including longer operating lifetimes. It’s a good proving ground for many designs.

Helmut Puchner, vice president and fellow, Aerospace & Defense, at Infineon Technologies, has seen this scenario play out time and time again, and he sees a natural progression from automotive to aviation. It’s a great real-world test. “The way I see this whole environment playing out is a lot of the architectural features will be implemented first in the automotive world, and if it passes automotive certification, it will be elevated into avionics applications. If it’s successful in avionics, it might even be elevated into satellite or space applications.”

For example, TT Tech in Europe has developed a failsafe Ethernet switch for automotive. It has worked so well that the solution has been scaled up to avionics with Airbus, and it is now scaling up into space applications.

Because the safety factor is paramount in both automotive and aviation, it plays into design, and has led to a more systematic approach to verify safety and reliability. Designers have created “playbooks” that map workflows based on tools for designers to document and comprehend the entire full lifecycle verification — early life, normal life, and end of life – and comprehend the failure points of the design so they can mitigate the impact of those failures from a safety perspective, said Anand Thiruvengadam, director, product management & marketing, Custom Design & Physical Verification Group at Synopsys.

“It used to be more of an expert-level verification by designers, using a less formal starting point, saying ‘Okay, I know my design, I know what the working conditions are, what the safety requirements are, what the potential failure mechanisms are and based on my experience, so I am going to qualify this this particular design or IP or block as being compliant with X,Y,Z requirement,'” said Thiruvengadam. “Now, we’re seeing in the case of automotive — and I’m sure it’s happening in other mission-critical application areas, as well — where Tier 1companies are saying, ‘Prove to me these designs are compliant in a certain way,’” said And that’s where IC vendors are turning to a systematic analysis and using software to show they are compliant. It’s causing a paradigm shift in terms of how the chips for mission critical applications are being built.”

Maybe it’s the software
In looking at the power and performance question, Siemens’ Klein said anything that reduces the need to move or store data can improve both power and performance. Eliminating the need to store intermediate results in a computation can improve a design’s characteristics.

“This means restructuring an algorithm so that intermediate results are not stored, but immediately passed to the next stage of the computation,” he said. “When designing at the RTL level, this type of optimization can be hard to find, and even harder to implement. Moving the level of abstraction from RTL to C++ or SystemC can make it much easier for the designer to find and effectuate this type of change. High-level synthesis (HLS) enables developers to easily understand data flow and memory architecture of an algorithm by working on a much more abstract representation. The HLS tool allows the designer to try out a wide variety of different RTL implementations from a single algorithmic description, providing power and performance estimates for each one. This makes it far more practical to find the optimal power/performance point for the design. Exploring implementation alternatives like this at the RTL level is simply impractical.” (Siemens recently published a whitepaper on how HLS can be used in an avionics DO-254/ED-80 design flow.)

Klein suggested moving functions from software into hardware to increase performance and reduce power. Processors, and the software running on them, are comparatively slow and power-hungry in this case. Moving computationally complex functions from processors into hardware can boost speed for applications, while reducing the power needed for those applications — and Siemens has seen performance increases of more than 10X for many applications.

High-Level Synthesis (HLS) can aid in this type of migration of a function from software to hardware, as the C or C++ code from software can be used as a starting point for HLS. While it is not yet practical to take a software C/C++ function directly through HLS compilation, HLS does eliminate the need to describe, and subsequently interpret, what the function needs to do in a document or series of Zoom meetings. The original C/C++ can be used as a golden reference, eliminating any ambiguity of the expected results. Some verification can be performed at a more abstract level. And the RTL can be formally checked for equivalency against the C/C++, reducing the verification effort for the new hardware.

The long view
Designers also have to take the long view when they are working on new systems. It can cost time and money when they don’t. “It’s so important for designers to design at the leading edge of technology — and make sure they are selecting parts at the beginning of the product lifecycle, as well as the right partners,” said Bryan Brady, vice president of business development at contract manufacturer MicroBoard. Otherwise, they get stuck at the manufacturing phase with an obsolete part, or one nearing its end of life, he said.

Others agree. “Aviation system designers need to take into consideration all aspects of the system — not only the individual components, but how everything interacts with each other and is optimized for extreme reliability and the longest useful life,” said Microchip’s Quintana. “Managing and owning the supply chain is a critical success factor in today’s economic environment. Securing the supply chain is paramount to successfully reducing risk, minimizing time to market and keeping the production line up and running.”

U.S. defense contractors have expressed the desire for onshore manufacturing and mitigating supply chain risk on raw materials. As a result, Microchip recently announced plans to expand its Colorado Springs manufacturing facility to accommodate MIL-PRF 19500 discrete products. Others are expanding operations in the U.S. market, as well, in light of rocky U.S.-China relations and a growing number of cyberattacks.

Across the supply chain, trust and provenance and security are big issues. That’s why Cadence, for instance, has gone through the process of work with the Defense Microelectronics Activity (DMEA) to become certified as a trusted supplier for the U.S. Department of Defense.

That trust also involves a time element, and the ability to design for long lifecycles — preferably with companies that are likely to still be around throughout the technology’s lifetime. “Obsolescence is a big issue to us when it comes to the whole design chain, and being able to have designs for a long time,” said Cadence’s Schirrmeister. “You have to make sure that your designs are ‘future-proofed’ and that you can do custom integrations.”

Related
Design For Airborne Electronics
Safety, security and traceability are required in avionics, but more complex chips are exposing gaps in the tool chain and in engineer training.
Medical, Industrial & Aerospace IC Design Changes
Strict quality, safety and security requirements and increasing complexity are pushing companies to adopt some leading-edge commercial approaches.
Safety-Critical Coverage
Verification in automotive, medical and industrial designs.
Fluid Dynamics Of Sonic Booms From Supersonic Aircraft
As companies push a return to supersonic flight, CFD is helping to find ways to reduce the noise impact.