What Are FPGAs Used For? Real-World Applications in Telecom, Oil & Gas, Aerospace and Rail

FPGAs are used where a fixed processor cannot deliver: custom IP cores, high-speed and non-standard interfaces, parallel signal and data processing, and long-lifecycle industrial equipment where the device family has to stay in production for decades. A field programmable gate array is programmable logic: hardware you configure after manufacture, which gives low latency and true parallel processing where a processor would work through instructions one at a time. InTechHouse has delivered FPGA and SoC work in four industries: telecom, aerospace, oil & gas and rail. This article starts with how the device works, then what modern FPGAs contain, then real applications, and finally the kinds of task an FPGA is actually given to do.

Key Takeaways

• An FPGA is an integrated circuit made of configurable logic blocks and programmable routing, which becomes a different digital circuit every time it is configured.
• Modern FPGAs add hardened DSP blocks, memory blocks and, on SoC devices, hard processor cores next to the fabric.
• FPGAs get chosen for different reasons: precise timing close to an interface, a device family that stays in production, or the ability to keep a safety design alive on new hardware.
• Typical FPGA workloads are tasks that need parallelization (large data volumes, image and signal processing), unusual interface configurations, and headroom for future development.
• FPGAs cost more and draw more power than a processor doing the same job, so they belong where a processor cannot deliver.

What are FPGAs used for? The short answer

The question "what are FPGAs used for" has a long answer across various industries, but the uses cluster into four groups:

1. Custom IP cores: logic that does something no off-the-shelf chip does, such as time synchronization at the network interface.
2. High-speed and non-standard interfaces: connecting to external devices whose protocols or timing a standard processor cannot handle.
3. Parallel signal and data processing: high performance work on continuous data streams, where every stage runs at the same time.
4. Long-lifecycle industrial equipment: products built to stay in production for decades, where the device family's availability matters as much as its performance.

The sections below take the device apart first, then show each group in real projects.

How FPGAs work: logic blocks, programmable routing and programmable input

An FPGA is an integrated circuit built from three kinds of resource: an array of configurable logic blocks (CLBs), programmable routing between them, and programmable input and output blocks at the edge of the die.

Configurable logic blocks CLBs are the working units. Each one contains several logic cells, and a logic cell typically holds a small look up table, one or more flip flops and carry logic. The look up table is a tiny memory that stores the output of a logic function for every combination of its inputs, so it can implement any logic operations of that size without fixed logic gates. Flip flops hold state between clock cycles, which turns combinational logic into sequential behavior. Carry logic speeds up arithmetic.

Programmable interconnects join these programmable logic blocks together. The routing consists of wire segments running in channels between the blocks, with switch boxes where channels cross. Electrically programmable switches inside each switch box decide which wire segments connect, which creates the routing pathways a signal follows. At the edge, programmable input and output blocks set how each pin behaves: voltage standard, direction, drive strength, and the timing of how a signal travels between the input pad and the logic inside.

"Field programmable" means the configuration happens after manufacture, in the product. The same silicon is programmed to become one digital circuit, and later it can become a different one. Because most FPGAs are re programmable, the circuit inside a deployed product can be changed without changing the board. That property, a programmable fabric that stays programmable, is what every application in this article depends on.

What modern FPGAs contain: digital signal processing blocks, memory blocks and hard processor cores

Modern FPGAs carry much more than logic cells. Most current devices include hardened digital signal processing blocks: dedicated multiply-accumulate units for arithmetic-heavy work such as digital filtering and other signal processing, which would be slow and large if built from look up tables. They include memory blocks of static random access memory spread through the fabric, and controllers for external memory when on-chip memory is not enough. Some families add analog features such as analog-to-digital converters.

On SoC devices, hard processor cores sit next to the FPGA fabric on the same die. These embedded processors run an operating system and control software, while the fabric handles the work that needs parallel hardware. How work is split between the two is covered in SoC FPGA.

The configuration technology also shapes real designs:

• SRAM based FPGAs keep their configuration in volatile memory. They load it at every power-up, usually from an external read only memory or flash chip, so the board needs a configuration path.
• Flash based FPGAs hold the configuration on the device. A flash based part starts almost instantly and suits low power operation.
• One time programmable parts can be configured once and then keep that configuration permanently.

One honest limitation: a large FPGA is power hungry compared with a microcontroller doing the same simple job. The fabric pays for its flexibility in power and cost, which is why an FPGA belongs where a processor cannot deliver.

FPGAs against application specific integrated circuits, and why the fabric stays

Application specific integrated circuits are custom chips designed for one product, and they are fixed function once fabricated. An FPGA keeps design iterations open after the hardware exists. That changes the design process in kind, not just in tooling: an ASIC team must finish verification before committing to silicon, while an FPGA team can keep correcting and extending the design in the field.

The two technologies also meet in ASIC prototyping, where teams use FPGAs to validate a chip design before committing it to silicon; this is a market category that InTechHouse does not offer. ASICs make sense at large scale, where per-unit savings repay the custom chip. Below that volume, or where requirements may still change, the FPGA stays.

Standard functions usually arrive in an FPGA design as licensed intellectual property cores rather than being written from scratch. The full comparison, including SoC devices, is in FPGA vs ASIC vs SoC.

What FPGA work looks like in practice: four industries, three projects

The FPGA work below comes from InTechHouse projects. Each one shows a different reason an FPGA was chosen, which is more useful to a buyer than a list of industries.

Aerospace: a custom IP core for time synchronization

InTechHouse built a Precision Time Protocol v2 IP core for an aerospace customer: time synchronization, with frequency and time measurement. PTP v2 keeps clocks on a network aligned. Timestamping has to happen as close to the network interface as possible, because every layer of software between the wire and the timestamp adds uncertainty. That is why PTP v2 so often lands in programmable logic rather than in software on a processor.

"We built a Precision Time Protocol v2 module, things to do with frequency and time measurement. An IP core for time synchronization. The one thing I'd say is that although we did it for aerospace, this particular customer didn't require anything special of the code, and I'd expect larger companies to add more, around code formatting and testing."

‍

Adam Szychulec, Head of Hardware / Embedded, InTechHouse

The case study: Developing a High-Precision FPGA IP Core for Aerospace Navigation Systems.

Oil & gas: an FPGA chosen for a twenty-year life cycle

An oil & gas customer used an FPGA for work that a much less expensive processor could have done, purely so the device could stay in production for decades without a mid-life upgrade. Then everything around the FPGA went obsolete, and the upgrade was needed anyway. This is not a use of the technology's strengths; it is a use of the vendor's production commitment, and it is a real reason FPGAs appear in long-lifecycle control systems. More in FPGA obsolescence and mid-life upgrade.

Rail: a legacy Spartan-3 safety design brought onto a supported platform

For a UK rail supplier, InTechHouse migrated a legacy Spartan-3 safety design to a supported platform. The work covered reverse engineering of the existing design, target platform selection, porting, and verification across two hardware variants, with the safety documentation preserved. The customer set safety-integrity requirements, and InTechHouse delivered to them.

What makes this an application and not just a service: the FPGA in this product outlived the board around it. That is the pattern of rail and other long-lived embedded systems in one sentence, and it is why a mid-life upgrade (MLU) is a normal event in an FPGA's life.

Telecom, and one line on medical

InTechHouse has delivered FPGA and SoC work for a telecom customer: a custom board platform. Separately, engineers at InTechHouse have delivered FPGA work for the medical device industry.

What goes into the fabric: the task classes an FPGA is given

Across SoC designs, the work assigned to the FPGA fabric falls into a small number of task classes. If your workload matches one of them, it is a candidate for programmable logic.

Tasks requiring parallelization. These are large data volumes, image processing and signal processing, including digital filtering. A processor handles such work one instruction at a time; the fabric can process many samples or pixels in the same clock cycle. This is where FPGAs' ability to run separate circuits side by side pays off, and where its parallel processing abilities give low latency that does not vary with load.

Unusual interface configurations. These are external devices with non-standard protocols, unusual pin timing, or more channels than a processor's peripherals can serve. The fabric can implement an interface that no standard part offers, at the timing the device needs.

Headroom for future development. Spare fabric reserved so that a later feature, a protocol change or a new sensor can be added in logic without a new board.

The point is to recognize the shape of your workload. A high performance pipeline that must keep up with a continuous stream, or an interface that no off-the-shelf part supports, is FPGA territory. A control task that runs comfortably on a processor is not.

Where else FPGAs appear: data centers, instrumentation and the wider market

Beyond the projects above, this section describes the market, not InTechHouse work.

FPGAs appear across telecom infrastructure, industrial control systems, test and measurement instrumentation, aerospace and defense electronics, medical imaging, automotive and data centers. Image and video processing are standard application areas, typically where the pipeline is continuous and latency-bounded. In data centers and edge systems, FPGAs increasingly accelerate artificial intelligence and networking workloads. In most of these roles the FPGA sits next to a general purpose processor or embedded processors rather than replacing them: the processor runs the software, and the fabric does the work that needs hardware.

The main FPGA vendors are AMD/Xilinx, Intel/Altera, Lattice, Microchip and Gowin. Their families span low power devices for control and glue logic, mid range parts for most industrial work, and large devices for high-bandwidth processing. The rapid growth of these workloads means FPGAs play a crucial role in places where neither a processor nor a fixed chip fits.

As an aside on breadth, the InTechHouse hardware team has also built an HDMI analyzer and multi-channel logic analyzers of between 50 and 100 channels on development boards.

InTechHouse case study: a Precision Time Protocol v2 IP core for aerospace

An aerospace customer needed precise time synchronization implemented in hardware. InTechHouse designed a Precision Time Protocol v2 IP core in programmable logic, covering frequency and time measurement. Placing the protocol in the fabric kept timestamping close to the network interface, which software on a processor cannot match. The customer received the IP core as a deliverable for its own system.