Smart Hot-Swap Control: The Next Step in Inrush Current Limiting

Hot-Swap Issues

High-availability systems must operate continuously without interruption of service. To achieve this reliability in large systems with many circuit cards in a chassis, it is necessary to be able to add new cards and replace old or damaged cards without removing power from the system. However, removing and inserting cards on a live backplane can cause a host of problems. There is often significant inductance between card connectors on the backplane, and significant capacitance on the backplane power supply. Disconnecting a card that is still drawing load current can cause ringing and large voltage spikes on the backplane, which can interfere with or even damage nearby boards. In higher voltage applications, such as 48V power distribution systems, removing a card can cause arcing, resulting in damage to connector contacts. Even more serious is the fact that when a new card is inserted and its power rails are connected for the first time, the unrestricted backplane current can charge the card capacitance and cause catastrophic transient surges. To date, various circuits have gradually emerged to enable safe "hot" swapping of cards in such systems.

The Evolution of Hot-Swap Solutions

如图 1 所示，在背板和具有大电容的板卡之间放置开关 M1，以解决一些热插拔问题。以某种控制方式接通和断开这个开关可以最大限度地减小对背板电源的干扰。插入板卡时，这个开关起初靠 R1 保持关闭状态。电源轨接通以后，短引脚连接并通过 R2 给 C1 充电以接通开关。箝位电路 Z1 将 M1 的栅极电压限制在安全值上。当板卡拔出时，短引脚首先断接，允许 R1 慢慢关闭 M1。

这个简单的电路有几个缺点。首先，虽然 M1 栅极的波形是缓慢上升的，但是功率晶体管的高增益导致输出电压以快得多的速度上升。因此， 要实现满意的接通速度，所需的 R1、R2 和 C1 的值高到了不现实的程度。其次，通过 R1 断开 M1  的过程非常慢，也许比拔出板卡所花的时间还长。因此这个开关在长引脚断接时仍有可能允许电流流过。第三，如果板卡短路或过载，惟一的保护是慢速作用的保险丝 F1。最后，在正电压系统中，这个电路中的 M1 必须采用 P 型 MOSFET，这一般不如采用 N 型器件经济。

早期集成的热插拔控制器（如图 2 所示的 LT1640）解决了很多这类问题。当短引脚连接时，控制器以电流固定为 45μA 的电流源接通开关。起初，它会给 MOSFET 栅极充电，直到 MOSFET 接通为止。当输出电压开始变化时，该电流转而给栅漏之间的电容器 C2 充电。通过选择 C2 来准确设定输出电压转换率和浪涌电流。在拔出板卡时，该控制器在连接器接触点断开之前迅速断开 MOSFET 开关。此外，电流检测电阻 R _S 与比较器 A1 一起形成一个电子断路器，如果负载电流超过所设定的限度，那么这个断路器就断开开关。就正电压系统而言，像 LTC1422 这样的器件还提供一个充电泵以驱动不那么昂贵和用于高压侧电源轨的 N 型 MOSFET。

很多系统都有持续时间很短的电源瞬态，这是正常现象。例如，电信系统可能在冗余电源之间切换，从而在板卡输入电压中引起几伏的阶跃。在带有电动机的系统中（如磁盘驱动器），电动机启动时可能出现持续时间很短的大电流。这样的事件可能导致负载电流瞬间超过断路器门限，从而错误地关闭开关。第二代热插拔控制器发展到可以区别真的故障和工作瞬态，并提供恰当的故障保护。将图 2 所示的比较器变成主动限制电流的放大器就可以解决这个问题。LT4250 等器件将电流限制到一个安全值上，然后，如果故障持续时间超出了定时器的设定值，就关闭开关。

在出现浪涌和输出过载时主动限制电流这进一步的改进简化了热插拔电路。这允许去掉电容器 C2 并更快地升高输出电压。限流定时器常常是可调的，因此可以让过载持续时间与特定系统需求和 MOSFET 的坚固性相匹配。用于 -48V 系统的 LTC4252A 和用于正电压系统的 LT4256 就是这类器件。

Digital Monitoring—A New Era in Hot-Swap Controllers

Increasing Intelligence Improves Reliability

In complex high-availability applications, monitoring system power is becoming increasingly important for a number of reasons. First, over time, you can observe unusual trends that indicate abnormal operation and indicate impending failure. Second, it is common today for systems to allocate power between different types of cards to conserve total power capacity. Power monitoring ensures that the power consumed by the card remains within the allocated limit. Hot-swap circuits are placed near the power connector, making it a natural place to monitor the power entering the card. The latest generation of hot-swap controllers,

such as the LTC4260 and LTC4261, make power monitoring easy. In addition to the core hot-swap functions such as inrush current control and electronic circuit breakers, the LTC4261 (shown in Figure 3) includes a 10-bit analog-to-digital converter, multiplexer, preamplifier circuitry, and digital interface to enable unprecedented levels of monitoring and control. The device measures the input voltage and the voltage across the current sense resistor. Knowing the voltage and current allows the power drawn by the card to be determined and trends to be observed. Furthermore, the additional free data converter inputs can be used to implement additional measurements, such as measuring the voltage across a MOSFET or the input fuse status.

The added functionality of having a digital interface goes far beyond monitoring card power. You can also issue commands to the device to control power and configure and report various fault behaviors. For example, a switch may open due to a fault such as a backplane undervoltage condition. In this case, a flag can be set that determines the cause of the fault. This may also generate an "alert signal" that pulls the interrupt voltage low to notify the system that a fault has occurred. In addition, you can set which faults cause the card to permanently disconnect or automatically repower. For example, a remote or difficult-to-service system may prefer to allow the card to restart itself after the undervoltage condition ends.

This information can be obtained in two ways. First, you can use the common I2C ^bus to read and write registers. In -48V systems, this often means sending data across the isolation barrier to an I2C ^master . The LTC4261 separates the transmit and receive data pins (SDAO and SDAI) to simplify the isolation circuitry. Second, you can further reduce costs by taking advantage of single-wire broadcast mode. In this case, a single output pin and external optoisolator repeatedly send a data stream that provides register contents and measured voltage and current values.

Monitoring Multiple Faults

Modern hot-swap controllers monitor multiple faults in addition to overcurrent. In -48V systems, there are often strict requirements on the allowed operating voltage to protect the load. Two undervoltage inputs UVH and UVL allow the turn-on voltage (e.g., -43V) to be independently selected from the turn-on voltage (e.g., -38V). The overvoltage input OV disconnects the switch if the input voltage is too high.

The LTC4261 simplifies the power startup process and ensures proper power startup with respect to the load. It delays the generation of the power good signal, starting the load after the MOSFET switch raises the board supply voltage. Next, the power-on timer starts. The system monitor must come back to terminate the timer to indicate normal operation. If the timer eventually does not receive this signal due to a system startup failure, the power supply is shut down.

Soft-Start and Independent Inrush Current Limiting Improve Connectivity

In addition to many new monitoring functions, basic inrush control functions are also improving. In some applications with very large backplane inductance, it is necessary not only to control the inrush current, but also to limit the rate of rise (dI/dt) of the inrush current to minimize disturbances to the backplane. The current limiting circuit now incorporates a soft-start function to limit the dI/dt, which is controlled by the selection of the external capacitor C _SS . Another improvement is the separation of the inrush current limit when the card is inserted from the operating current limit used by the circuit breaker. In systems with large load currents, there may be large load capacitances. If such a large capacitance is charged to 48V, a considerable amount of energy is stored. The largest burden is the equivalent amount of energy consumed in the MOSFET switch at startup. The LTC4261 allows the circuit breaker current limit to be defined, which can be set with R _S , independent of the inrush current set with C2. Therefore, you can power up large load capacitances while minimizing the power dissipation and size of the MOSFET. Figure 4 shows the LTC4261 starting up a -48V card after it is plugged in.

Conclusion

Safely managing power during hot-plug insertion is a critical function in systems that require high availability. Over the years, many improvements have been made to circuits that ensure safe hot-plug insertion. Early hot-plug circuits roughly limited inrush current. Later improvements added electronic circuit breakers to isolate faults, comparators to monitor voltages, and power-good signals to safely enable loads. Today, these circuits include digital interfaces with on-chip data converters that allow the system to continuously monitor and manage power. With this information, systems can be designed to achieve higher reliability and durability.