Loop speed clues in hot-swap component choices

Loop speed clues in hot-swap component choices
Most literature on hot swap focuses primarily on the controller. But the functionality, robustness and ease of a hot-swap design depend on proper selection of both the controller and the MOSFET under steady-state and short-circuit conditions. And key to that selection are trade-offs in the different kinds of available hot-swap controllers and selection of the three main hot-swap components: the analog controller, the power MOSFET and the power sense resistor.

The basic hot-swap circuit consists of a hot-swap controller, U1, a power n-channel MOSFET, Q1, and a power sense resistor, Rsense. A current-controlled hot-swap controller contains at least one linear current amplifier. The current amplifier senses the voltage across Rsense. It then is able to limit the peak inrush current by pulling back the Vgs voltage of Q1 to maintain the voltage across Rsense. With a selectable inrush current, the rise time of the voltage delivered to the dc/dc input is primarily determined by the input impedance.

During turn-on, the inrush current reflects several patterns. The overshoot depends primarily on the load impedance, the MOSFET chosen, the driver of the hot-swap controller and the speed of the hot-swap linear current amplifier.

Should you choose an integrated hot-swap or a discrete MOSFET-controller solution? Let's consider the analog IC approach with an integrated MOSFET; here, the first thing you might want to consider is an integrated system IC vs. a separate controller and power MOSFET. There is at least one integrated solution available on the market today for -48-volt telecom applications, the HV111 from Supertex. This device has an on-board lateral MOSFET switch integrated into its analog process. Typically, an analog process will have a higher Rds(on) x area product than a power MOSFET. This means that an analog switch will take up much more space for the same Rds(on) or it will have a much higher Rds(on) in the same area. This is especially true for an 80-V or 100-V switch. The HV111 has an 80-V-rated switch with a 1-ohm typical and 1.5-ohm maximum Rds(on).

If you have a 1-A load current, you will have a maximum 1.5-V drop across the switch, which may cause you to go into undervoltage lockout. The manufacturer rates the device up to 1.65 A.

Since you have a 72-W load with a possible peak of 2.8 A, you have narrowed your search down to an external current-controller hot swap and separate power MOSFET and Rsense.

So now what kind of current controller? Since your main concern is system robustness, look for the controller with the fastest response to an overcurrent or short circuit. Vendors list this as propagation delay from Vsense High to the MOSFET gate being pulled near threshold or Low tpHLsense. They may also list the control loop transconductance, dIgate/dVsense. This is the change in gate voltage with the change in Vsense voltage. Also look at the gate drive features to make sure it will be capable to drive on and off the parasitic capacitances of your MOSFET.

Compare speed response to fault and gate drive characteristics. Faster devices with higher pull-down current will result in less power dissipation in a fault condition. Many controllers address the time delay through the controller only to leave it to the designer to calculate the on-to-off time through the MOSFET.

ON Semiconductor's NIS5101 controller with bias to full voltage takes a different approach by integrating a linear current controller into a power MOSFET. Rather than identifying time delay through the controller only, this device specifies the total delay time to the MOSFET turnoff. The response to short circuit is 3 microseconds, typical turnoff (total through control and MOSFET gate pulldown).

Since many manufacturers specify the speed of their control loops differently, it may be best to compare a few samples in the same test condition on your bench. Also, the propagation delay is from the time a fault is sensed until the gate is pulled to the analog current set point. Except for the integrated power MOSFET and control, NIS5101, the others will have an additional time from gate drive low to actual MOSFET gate being discharged.

When using a floating controller you will need to add an additional resistor called Rshunt. For example, the UCCx921, ADM1070 and LTC4151 are "floating" controllers. This means that the device contains a voltage regulator that is in series with a shunt resistor between ground and the -48-V line. The UCCx921 has a 9 V minimum regulated voltage so its Rshunt will dissipate:

Power Rshunt = (Vinmax - Vregmin)2 /Rshunt. Application notes suggest providing at least 2.5 mA bias current to the regulator at Vin minimum. So, Rshunt = (Vinmin - Vrefmax)/2.5 mA = 35 V/2.5 mA =10.3 K-ohms (Use a 10 kohm resistor.) Pwrmax Rshunt = (75 V 9 V)2 / 10 kohms = 0.44 W. (Use at least a 1-watt resistor.)

Alternately, the controller can see the entire line voltage across it. For example, the ISL6141 sees the line voltage. It is rated for 100 V. To survive a possible 200 V, 1 s transient, a transient voltage suppressor must be added to protect this controller. Addition of such a suppressor is not always a simple task, since system engineers often do not know the maximum transient energy level of their system.

Value of Rsense

To choose the value of Rsense, you must know the peak current that will be required of your hot-swap circuit. What brick are you powering up? What peak power must it deliver? What input capacitor or filter does the manufacturer recommend?

The board is plugging into a -48-V backplane. You need a 1.2-V output and 60-A peak load or 72-W converter. The manufacturer, like most brick manufacturers, recommends adding a 220-F capacitor across the Vin(+) to Vin(-) pins of this brick. Your input voltage line can vary from -75 V to -35 V. This manufacturer specifies 2.8 A as maximum input current at 60-A load. Alternately, you can calculate this current. The maximum input current occurs at the maximum load current and the minimum input voltage:

Input max = Pout maximum/(Vin min x theta) where theta is the worst-case efficiency at that line and load condition. = 72 watts/(36 V x 0.72) = 2.8 A

You want to set the fault level at least 100 percent over your maximum input current. This will prevent nuisance tripping caused by transients on the load or the input bus. So you want to set the fault to trip at 2 x 2.8 A = 5.6 A * Rsense. This Vfaultmin/Ifault = 47.5 mV/5.6 A = 8.5 milliohm. (Use 8 milliohms, the standard value, of a Panasonic P8.0MCT-ND.

The wattage value, PwrRsense = Imax2 x R = (5.6)2 x .008 = 0.25 W. (Choose at least 1-W for Rsense.)

Choosing the MOSFET

Depending on the specification you are required to meet, either ETSI ETS300 132-2 (100-V transient for 100 milliseconds) or Bell Core Gr-513-CORE telecom specification stipulates immunity to power surges of -75 V for 10 ms, -100 V for 10 s and -200 V for 1 s. Your system's specification is to meet 100 V for 100 ms. Choose a 100-V MOSFET.

We want to minimize the drop across our hot-swap MOSFET and also the power loss. The maximum power loss in your MOSFET under nonfault conditions will be: maximum power loss MOSFET = (Imax2)*Rds(on). You could consider using an ON Semiconductor NTB52N10 MOSFET in a D2Pak package. It is rated for 100V BVDSS, 30 milliohms maximum Rds(on) at Vgs=10 V.

The maximum Pwr(on) Loss will be (Ipk)2x Rds(on)max = (2.8)2 * 0.03 = 0.24 W.

Rds(on) assumes that the junction temperature wull be approximately at Tj=25C. (Finding the exact value will be an iterative process on a simulator.) To verify junction temperature (Tj), assume Tcase + (Power Loss * temperature dissipation) Tj = 25C + (0.24 W * 0.7C/W). You'll get 25.16C. The ON Semiconductor device is rated for +150C, so you are well within safe operating junction temperature.

Suzanne Nee is strategic marketing manager, ON Semiconductor (Phoenix, Ariz.).

See related chart