Description
As a complex integrated circuit, the design of the power supply for FPGA systems is more demanding compared to general electronic systems, requiring high precision, high density, controllability, efficiency, and miniaturization. This article systematically introduces the different characteristics of FPGA power supplies, while also providing examples to help engineers gain a deeper understanding of the significance of each characteristic, as well as the constraints of FPGA specifications and their impact on power supply design, in order to quickly complete the power supply design for FPGA systems.
FPGA (Field Programmable Gate Arrays) are among the most complex integrated circuits today. They utilize advanced transistor technologies and chip architectures to achieve high performance and compact high-end products, while the power supply for FPGA systems has higher requirements compared to general electronic systems.
With the growing popularity of FPGA system applications in the market, the demand for power supply solutions has also surged. In addition to meeting basic requirements, FPGA power systems need to have high precision, high density, controllability, efficiency, and miniaturization.
In reality, engineers prefer to spend most of their time programming rather than focusing too much on how to design suitable power supply materials. Therefore, in this article, we will introduce the different characteristics of FPGA power supplies, while providing examples to help engineers gain a deeper understanding of the significance of each characteristic, as well as the constraints of FPGA specifications and their impact on power supply design.
The core power supply voltage is one of the most critical factors in balancing FPGA power consumption and performance. The specifications usually list the acceptable voltage range, but this range does not fully describe the requirements. For FPGAs, the power supply voltage must meet operational requirements while also needing to be weighed and optimized. The following figure illustrates the voltage requirements of the Intel Arria 10 FPGA core, which also represents the voltage requirements of other FPGA cores. Typically, it will show the tolerance range for the rated voltage, for example, the Arria 10 FPGA is ±0.03V, and the FPGA will operate well within this voltage window, but the actual situation is much more complex than depicted in the image.
In fact, FPGAs can operate at different voltage levels, depending on their specific manufacturing tolerances and the specific logic design employed. Even for the same voltage requirement, the static voltage required for one FPGA may differ from that of another FPGA, so the power supply design must take into account the variations between the dynamic and static requirements of the corresponding FPGA and adjust accordingly.
The goal of designing an appropriate FPGA power supply solution is to generate the right performance level to operate programming functions while minimizing unnecessary power consumption. From the perspective of semiconductor physics, both dynamic and static power increase significantly with the increase of core VDD, so our goal is to provide the FPGA with sufficient voltage to operate normally and meet its timing requirements—excessive power consumption does not help improve performance; rather, it causes transistor leakage current to increase with temperature, consuming more unnecessary power. For these reasons, it is urgent to optimize the design and operating point voltage.
This optimization process requires very precise power supplies to succeed. If the core voltage is below the required level, the FPGA may fail due to timing errors. If the core voltage drifts beyond the maximum specification, it may damage the FPGA or cause hold time failures in the logic. Therefore, it is essential to consider the power supply tolerance range to prevent all these situations and only guarantee the instructed voltage remains within specification limits.
The problem is that most power regulators are not accurate enough. The regulated voltage can drift anywhere within the tolerance range around the instructed voltage, and it can vary with load conditions, temperature, and aging. A power supply with a ±2% tolerance implies it can output any value within a 4% voltage range. To compensate for the possibility that the voltage is 2% too low, the instructed voltage must be set 2% higher than the voltage required to meet timing. If the voltage after regulation drifts to 2% above the instructed voltage, it will be operating 4% higher than the minimum voltage required for that working point. This still meets the voltage requirements specified for the FPGA but wastes a significant amount of power, as shown in Figure 2.
Figure 2, Power Regulator Tolerance Trade-off
The solution to this problem is to choose power regulators that can operate with stricter voltage tolerances. Using a regulator with a ±0.5% tolerance allows operation closer to the required minimum specifications at the desired working frequency and ensures that the voltage difference from the required voltage is less than 1%. This enables the FPGA to operate normally with minimal power consumption.
Devices in FPGA systems typically require different regulated voltages, such as core voltage processors that may require voltages of 0.8V, 1.0V, 1.2V, 1.5V, or 1.8V, etc. Although they are low voltage supplies, their dense transistor structures and the need to maintain high-speed operation can require power supply solutions of 10A or more. The specific processor requirements usually dictate other power supply requirements, such as load transient recovery and standby modes, which necessitate Point-of-Load (PoL) voltage regulators specifically designed for core voltage. PoL voltage regulators are high-performance regulators where each Vout voltage rail is independent of its respective load settings. This helps meet high transient current demands and low noise requirements for high-performance semiconductor devices like FPGAs. For example, ADI’s LTM4678 series includes two sets capable of simultaneously providing high-density power supply outputs of 1V@25A and 1.8V@25V.
Figure 3, ADI’s LTM4678 Application Circuit
FPGA contains a large number of complexly arranged transistors, with a chip containing hundreds of millions of transistors, which are divided into kernel segments, module segments, and partitions that can be designed and independently managed. These specific arrangements result in many different power domains, which need to be properly managed in terms of voltage, current, ripple, and noise, as well as the sequence of operations during startup, shutdown, and fault conditions. Therefore, the controllability of FPGA power supplies requires careful management of the output order and its power.
Newer FPGAs in the market will provide specific requirements for the sequence of operations when powering on and off, ensuring that the FPGA powers on and resets correctly while maintaining minimal current consumption and keeping I/O in the correct tri-state configuration during power transitions. Taking Arria 10 as an example, its technical specifications divide the power into three sequence groups (1, 2, 3) and require them to be arranged in ascending order: 1, 2, 3, and then descending in the reverse order: 3, 2, 1.
Figure 4, Sequence Diagram of Arria 10 Power Groups
For example, ADI’sLTC2936 can provide six programmable threshold analog comparators to detect fast events and send digital states to logic. This device also has three programmable GPIO pins for additional functionality. The programmable IC includes EEPROM, which can work almost instantaneously at startup; it can also store fault telemetry data for debugging via its I2C/SMBus interface.
Figure 5, ADI’s LTC2936 Application Block Diagram
Engineers can utilize FPGA development kits to assist in development. For example, the Arria 10 SoC development kit (DK-SOC-10AS066S-A) demonstrates the ADI’s LTM4677 µmodule power solution for Arria 10 SoC power requirements.
In the kit, the core power supply operates at a voltage of 0.95V and a current of 30A. Since these power requirements are relatively relaxed, a single LTM4677 module can easily provide the required current (up to 36A). For applications requiring more current and more stringent conditions, up to four LTM4677 modules can be run in parallel to provide up to 144A of current, as shown in Figure 7.
Figure 6, Arria 10 SoC Development Kit
Figure 7, Application Circuit Diagram Utilizing Four LTM4677 Modules in Parallel to Provide Up to 144A of Current
After understanding the application requirements, engineers can visit Digi-Key’s official website and choose the “Power – Board Mount” category under the “DC-DC Converters” subcategory. In the “Application Filter”, engineers can find “POL” under “Type”, or directly enter “POL” in the “Search in Results” to filter PoL voltage regulators.
In FPGA systems, power supply solutions are one of the important topics engineers need to consider. Unlike general computer requirements, FPGA power supplies require high precision, programmable functionality, scheduling, and most importantly, “high energy density”, which means low voltage/high current. The industry has categorized the development of this type of product as Point-of-Load (PoL) voltage regulators.
To facilitate component selection for engineers, Digi-Key’s official website has listed PoL products under the “DC-DC Converters” subcategory in the “Power – Board Mount” category, allowing engineers to select components more quickly and accurately.
Please visit Digi-Key’s official website for more technical information
