ROHM provides a wide variety of DC/DC converter ICs, among which are a lineup of devices that are suitable as the power supplies of FPGAs. The eight models described here can satisfy the power supply specifications required for FPGAs, and reference designs are also provided. We asked Mr.Takanobu Shibako at ROHM, about the performance and features of the step-down DC/DC converter series for FPGAs.
－ These are step-down DC/DC converter ICs for use with FPGAs. To begin with, does "for FPGAs" means that these ICs have some special functions?
In essence, they are general-purpose step-down DC/DC converter ICs, but they should be thought of as a series with performance and characteristics that satisfy the requirements for FPGA power supplies. For example, there are many DC/DC converters that can supply a 1.5 V/1 A output, but that does not mean that all of them are suitable for use as FPGA power supplies.
－ Well then, before we talk about DC/DC converters, please explain the demands made on power supplies for FPGAs.
Certainly. I think it's easier to understand when this DC/DC converter series is suited to FPGAs if this is clarified. First of all, there are a number of manufacturers of FPGAs, and the functions and configurations of these devices are many and diverse. Here, I'd like you to imagine an FPGA that is comparatively multifunctional, high-performance, and of mid-range scale. For example, the 7 Series of Xilinx, or Stratix devices from Altera, would be examples.
First, and what is probably the most important requirement, is that there are five or more power supply voltage types, including the power supply for the core, for I/O, and for different function blocks. Also, there are sequencing-related requirements, such as power supply rise times and sequencing order. In recent years there has been a tendency to relax these requirements, but if the required specs are not satisfied, the device may not operate normally, or may fail.
Next, when considering FPGAs, the finer design rules of wafer fabrication processes is an issue; but as finer design rules advance, for example the power supply voltage known as the core voltage has been lowered to around 1 V or so, and conversely, there is a tendency for power supply currents to increase. Other power supply voltages, such as 1.8 V or 2.5 V, are generally at or below 3.3 V or so.
－ Lower-voltage, higher-current power supplies are also being discussed for CPUs. Can this be regarded as the same trend?
In essence, I think it can be regarded as the same. However, the number of power supplies is much greater in the case of an FPGA.
－ What else is there?
In addition to multiple power supplies and low-voltage, high-current specs, there is a trend toward higher voltage accuracy. For example, the core voltage has reached the level of 1 V ±3%.
－ Is ±3% so very strict? Other system voltages, at 5 V or 3.3 V, typically have a tolerance of ±5%.
When considering actual tolerances, taking ±3% as an example, 3% of 1 V is ±30 mV. And, ±3% of 3.3 V is ±99 mV. Of course, when the voltage is lower, the absolute value of the allowable voltage difference is also smaller. A switching power supply has a ripple voltage in the output voltage. The output ripple voltage occurs as the product of the output ripple current and the output capacitor ESR, and is proportional to the magnitude of the output ripple current, regardless of the output voltage. The output voltage accuracy includes a ripple voltage, and so if the ripple voltage is high under low-voltage, high-current conditions, it becomes difficult to maintain the allowable difference.
In addition, there are elements that affect the output voltage accuracy. The output voltage can fluctuate considerably due to load transients, that is, sudden fluctuations in the output current. The load of an FPGA changes fairly dynamically. In particular, the load fluctuation when waking up from a sleep state can be considerable.
There are also other factors. If the load current (the output current of the power supply) becomes too high, the voltage drop due to the resistance of the board wiring is increased, and the voltage at the FPGA power supply pins is lower than the voltage at the power supply output pins. This is not a problem with the power supply performance, but from the standpoint of supplying the required voltage to the load device, it is a problem to be resolved on the power supply side.
Finally, a low noise level is also very important. Operation and communication in a frequency band in the hundreds of MHz range are performed using low voltages. Power supply noise not only degrades the S/N ratio, but is also a factor impeding normal operation.
－ So it seems there are various things about an FPGA power supply to consider. Could you summarize the important points?
The main requirements for an FPGA power supply are: 1) numerous power supply voltages, 2) power supply sequencing, 3) low voltages and high currents, 4) strict voltage accuracy (including such issues as ripples, fluctuations due to load transients, voltage drops due to board wiring resistance), and 5) low noise. Of course for separate devices and designs there are more detailed issues as well.
－ So we can state outright that meeting these requirements is a necessary condition for an FPGA power supply.
That's right. Such a power supply must provide low voltages and high currents, with high voltage accuracy and low noise. These are, frankly, pretty strict requirements.
Step-Down DC/DC Converter Series Compatible with Stringent FPGA Power Supply Requirements －Part 2－
Ask Direct to Engineers
- Coping with FPGA Power Supply Requirements
- You’ve explained the requirements for an FPGA power supply...