FAQ on Hot Swap: from “don't dare do it” to “it’s now OK to do” part 1

It wasn’t that long ago that one of the strict rules when assembling a system or debugging it was simple: don’t plug or unplug anything — component, cable, board — with the power on. You were firmly instructed to shut off the power and only turn it back on when you were done with connecting /disconnecting or inserting/removing. The consequences of violating this rule would range from circuit malfunction to outright destruction of components.

For situations where it would be necessary to leave the system core up and running while removing or inserting a board, some vendors even put a small slide switch in series with the board’s primary power rail near the accessible edge of the board, so the board could be unpowered or powered up independent of the larger system.

Well, times have certainly changed. It’s now routine to plug/unplug boards and cables with power on, a process called “hot swapping” for circuit boards. The need for this has been driven by the requirement in high-availability systems, such as data centers and server racks, to remain fully operational while a board is replaced or an upgrade is installed.

In addition, mass-market consumers now expect to routinely plug and unplug cables such as the widely used USB ones. The ability to do so is especially important as many battery-operated and some AC-powered devices don’t even have a true “hard” power-off position, but instead have a “soft” on/off where some power is always applied in an idle or quiescent apparently “off” state.

How did circuits get from a strict “don’t even think of it” to “go ahead, it’s OK” situation? The solution to the hot-swap challenge is a combination of connector hardware as well as special controller circuitry. This FAQ will provide an overview of hot swapping issues, but it is not a detailed design guide (references at the end will provide that information).

Q: What is a hot swap?
A: If a board (or module) of a high-reliability system, which must have near 100% uptime but fails or needs updating, it must be removed and replaced while the overall system remains up and running. This process is known as hot swapping,

Q: What are the steps to implementing hot swapping?
A: To hot-swap safely, connectors with staggered pins are often used to ensure that grounds and local power are established before signal and data connections are established. In addition, each printed-circuit board (PCB) or plug-in module has an on-board hot-swap controller to ease the safe removal and insertion of the module from a live backplane. While in operation, the controller also offers continuous protection from short circuits and overcurrent faults.

Q: Is hot swapping the same as hot plugging?
A: There is some overlap, but there are differences as well, as indicated in Figure 1. In hot swapping, the new board is removed and replaced with the power on, but the system software does not get “involved.” In contrast, in hot plugging, the system acknowledges the board or device removal or connection, such as when a USB accessory or peripheral is connected/disconnected. The operating system automatically recognizes the changes that have been made. Many hot-plug situations are not mission-critical applications, and the hot-plug capability is for user simplicity.

Figure 1. Hot-plug and hot-swap have similarities as well as distinct differences. (Image: Geeks for Geeks)

Think of the neighbors

Q: If power and ground are connected before the signal pins, why is this a problem? Isn’t this what happens in normal use when you turn the power on?
A: That’s a good observation. Inserting a board into a live circuit with staggered pins is very similar to applying power. But that’s a limited perspective that does not take adjacent boards or the larger system into account, as seen in Figure 2.

Figure 2. The hot-swap issue is due to both the board powering up by itself as well as the sudden current-draw load it imposes on the system and adjacent boards. (Image: Data Center Knowledge)

Q: Why is this a problem?
A: Each of the hot-swapped modules usually presents a considerable amount of load capacitance to the supply rail source, typically on the order of 20 to 40 millifarads. This means when a module is first inserted, the inrush current due to the uncharged capacitors and overall capacitance can be significant, as the board demands as much current as is available to charge up that inserted load. If this inrush current is not limited, it could reduce terminal voltages, causing a significant brownout on the main backplane, resetting many of the adjacent modules on the system, and even possibly damaging the module’s connectors due to the high initial current.

Q: How is this problem managed?
A: The solution is to carefully control the current flow with a hot-swap controller on each card that may be swapped. The controller monitors and controls the inrush current to ensure the current flow is “throttled” during a safe power-up interval. In addition, the hot-swap controller continually monitors the supply current during normal operation after the power-up sequence for protection against short circuits and overcurrent conditions.

Q: What is the harder-to-satisfy challenge in devising a hot-swap controller?
A: Experienced designers know that the transient phase of any circuit or system operation presents potential difficulties, in contrast to a system that is running in steady-state mode. This applies to power-up/down modes as well as initialization of a communications link.

For hot swap control, the issues include soft start, current inrush, and power dissipation of any power-related current-control elements, such as the current-pass MOSFET. The next part looks at some of the design considerations related to hot-swap controllers.