Practical Challenges of Designing Complex Multi Rail Power Architectures
Many of the programmable logic, microcontroller and microprocessor IC’s on the market require multiple voltage rails for operation. Many interface standards also require various voltages to power transceivers, line drivers, sensors, data converters, etc. In a perfect world, all of these voltages are the same, have wide regulation accuracy requirements and draw very little current. We, however live in the real world.
Many of the programmable logic, microcontroller and microprocessor IC’s on the market require multiple voltage rails for operation. Many interface standards also require various voltages to power transceivers, line drivers, sensors, data converters, etc. In a perfect world, all of these voltages are the same, have wide regulation accuracy requirements and draw very little current. We, however live in the real world.
FPGAs not only require multiple rails, but have specific sequencing and start up requirements that in some cases must be met to avoid device damage. You may have different devices requiring the same voltages, but different startup timing or sequencing. In these cases multiple regulators often must be used. Careful consideration must be taken that I/O signals connected between devices don’t expose unpowered parts to voltages before their own core supplies are in regulation. Before a device is fully powered, I/O pins may not be guaranteed to be in a high impedance state. Leakage paths inside the devices could cause them to be back fed through I/O pins which can stress or breakdown paths within the device. Often devices will allow for supplies to start simultaneously, which can be a way to avoid conflicting sequencing requirements. The tradeoff in tracking the supplies during startup becomes the additional strain on the input supply that the larger inrush current from multiple supplies starting would cause. Consideration must be taken for this initial inrush current to guarantee that the input supply doesn’t collapse or dip during startup, which could cause the downstream supplies to restart or shutdown depending on their protection mode.
A bigger issue facing modern circuit design is meeting the tolerance and noise requirements of high performance devices as the operating voltage continues to decrease. With core voltages pushing below 1V and regulation requirements also dipping below 3%, the laws of physics begin to become a factor. For the sake of simple math, let’s take a 1V rail at 3A requiring 3% accuracy as an example. 3% gives us 30mV of total error to work with. DC losses alone can begin to become a huge factor when selecting your components such as filter beads and your pcb design when deciding on trace widths and copper weights. Power monitoring through the use of current sensing becomes more difficult as well. Take a 10miliohm current sense resistor in our given example and we’re already dropping 30mV across it, using up all the total error available to remain within spec.
The diagram above reflects the cumulative effect of all the error sources in a typical power supply. The feedback error percentage is based on using 0.1% accurate resistors. The equation for calculating the error contribution is given below.
This equation is significant because 1% accuracy is typical for chip resistors. Higher accuracy 0.1% resistors must be specifically called out to ensure minimizing this error contribution. The alternative would be using a fixed output regulator which uses precise internal resistance to achieve higher accuracy.
Thermal relief also poses a significant challenge to designing a reliable system. Temperature rises of not only key components, but of switching regulators, magnetics and the PCB itself can create significant performance degradation. Junction temperatures can increase quickly and without proper relief can go into thermal runaway, a condition during which your junction temperature begins to rise exponentially until the device fails. Removing heat from the device means minimizing all of the resistance paths outlined in the thermal impedance model shown here. Internal resistances are dictated by the device, however external resistances are dictated by PCB and heat sink design.
Thermal Impedance Model of an IC
The use of heat sinks and fans are often required for high performance devices. Active cooling solutions such as fans increase power supply load demands as well as potentially introducing noise issues on the supply rail used to power them.
So what do all these design considerations mean? It’s more important than ever to plan your power and PCB design at the beginning of your project and not at the end. Unfortunately there is an overwhelming majority of designs that save the power until last, concentrating on the 1’s and 0’s and viewing the power architecture as more of a necessary evil rather than the backbone of a successful product. Ask yourself, where does that logic 1 come from? Your 1’s and 0’s are only as good as your voltages and grounds.
Step one, understand the requirements of your components. Create a spreadsheet outlining the voltage, current, accuracy, sequencing, startup timing and noise requirements for each device. Note that supplies for high speed analog interfaces will usually require their own power supplies to ensure a proper low noise accurate voltage. When combining devices onto common rails keep in mind that each load you place on that voltage can introduce additional noise. Also remember that when using intermediate busses, for example regulating down to 3.3V from 5V, then using the 3.3V supply to feed a 1.8V supply, that 1.8V supply will introduce additional noise onto the 3.3V rail. For example, this situation can become an issue with using 5V sensors that also provide the 5V input to your regulators. The integrity of that sensor output can be compromised if care is not taken to minimize noise on the power to the sensor.
Step two, create a floor plan / block diagram of all your components and the outline of your expected PCB. Create a scale drawing to make sure that the sum of your IC / passive area does not exceed 50% of your PCB area, doing so will help to ensure that you have sufficient space for routing and optimal placement. Identify locations for your different power supplies to minimize the space between outputs and intended loads. Placing all your supplies in one corner of the board is rarely optimal. Even using plane floods, high currents can create significant voltage drops over extended distances. Also consider using devices capable of remote sense, which allows the regulator to adjust its output based on sampling the voltage at the load. Remote sense is not trivial so be sure to use caution when implementing it in your design.
Step three, plan for thermal relief. Identify components that may require heat sinks or additional fans. If you have a system fan, be sure that tall components such as film capacitors or magnetics do not block airflow across high temperature areas. Also keep in mind that the PCB itself is a primary path for removing heat from your devices. The more layers and more copper in your PCB, the larger the thermal benefit you will get.
Doing a full system analysis before you start your PCB design and floorplanning with care given to the design and placement of your power components will go a long way towards insuring the success and reliability of your product.
