The majority of non-synchronous, inductor-based boost converters exhibit a dc-current path between power source and load (Fig. 1). They're the step-up switching types, and their path can have two undesirable consequences.
The first problem occurs if a grounded output or other overload draws heavy output current for more than a few hundred milliseconds. Then the catch diode, which is usually a Schottky type, may exude that blended aroma of molten silicon and potting compound familiar to all true hackers. Or, say switching action is disabled for any reason, such as intentional shutdown. The load voltage remains just one diode drop below the supply voltage. If this residual voltage is outside the load circuit's expected steady-state operating range, the result can be indeterminate circuit behavior.
Both problems are neatly solved for relatively low-output-current applications—those less than 5 A. These employ monolithic current-mode controllers and high-side current sensing. The circuits replace the catch diode with a synchronous switching transistor that can be disabled by shutdown or the removal of input power.
Essentially, disabling this internal transistor or turning it off during shutdown removes the path for dc current flow. The load then sees a requisite high-impedance disconnect. When not in shutdown, the circuit's cycle-by-cycle current-sensing mechanism protects against catastrophic meltdown from internal current overloads. It does this through the use of an internal, high-side current-sense resistor. Finally, thermal-overload protection provides the circuit a safe area of operation.
For applications with higher output current, in which pricing makes synchronous switching impractical for monolithic devices, the load-disconnect function demands a high-side switch external to the controller die. A discrete current-mode topology using a high-side current-sense resistor and synchronous switching transistor is possible. But that approach suffers from pc-board parasitics and layout dependencies, especially at high switching frequencies. The result is a relatively complex design, particularly when system constraints mandate a low input voltage (less than 3.6 V).
A synchronous, high-side external switch becomes feasible at higher levels of peak inductor current—those greater than 5 A. But at the more moderate levels—approximately 1.5 to 5 A—cost and complexity override heat and efficiency considerations. A simple catch diode is again the most desirable solution. The challenge is to achieve the desired load disconnect while retaining use of the humble catch diode and the unadorned boost topology.
A Simple Solution
A simple and smart solution does exist. In it, a MAX668 controller illustrates the demanding task of boosting from low input voltages (Fig. 2). This current-mode boost controller drives a logic-level, n-channel, enhancement-mode MOSFET that's configured on the low side. That MOSFET is driven in series with a low-side current-sense resistor. The high-side switch is a Schottky catch diode chosen for its low forward-voltage drop, which is now standard. The simple boost topology remains intact. This application boosts 3.3 to 5 V and delivers load currents to 3 A. The MAX668 boosts only from 3 V or higher, but the MAX669 can accept inputs as low as 1.8 V.
The key element in implementing a smart load disconnect is the p-channel enhancement-mode MOSFET, Q1. The system can enable this boost circuit (SHDN) or shut it down. D1 conducts during shutdown. At the MAX810L supply terminal, it produces 3.3 V minus one diode drop.
The MAX810 is a tiny power-on-reset device with an SOT23-3 package. It draws about 24 µA of quiescent current, and guarantees operation at 1 V. In this case, the MAX810L output is high because its nominal reset threshold is 4.65 V. This forces Q1 off and disconnects the load from the main supply.
The controller's feedback resistors are set to produce a 5-V output when that device exits shutdown. When the rising output exceeds the MAX810L input threshold, an internal one-shot turns on for approximately 240 ms. After this timeout period, the output goes low and turns on Q1.
While Q1 is on, the MAX810 constantly monitors the supply line for overload currents. An overload causes the output to sag below the MAX810's internal threshold voltage. That output then goes high with a nominal 20-µs delay, turning off Q1 and disconnecting the load. Soon after, the MAX668's boosting action raises the MAX810 input voltage above its threshold. After timeout, the MAX810 automatically reconnects the load. This cycle repeats until the excessive load is removed or the boost circuit is disabled. Q1 and the MAX810 together act as a smart solid-state switch.
As a micropower device, the MAX810 has a rather wimpy push-pull output stage. It approximately resembles a 6-kΩ resistor when sourcing current and a 125-Ω resistor when sinking it. When the device turns off or on, these resistances slow things down by acting against Q1's Miller capacitance and the associated CGS. The associated RC time constant for a large pass transistor is about 0.6 µs. That's assuming a total effective capacitance of 5000 pF acting against the MAX810's 125-Ω sinking stage. So a full voltage transition can be approximated as 10RC = 6 µs.
Turning the same device fully off requires nearly 48 times more time, or about 290 µs. Though this approximation is workable, actual turn-offs occur when VOUT reaches the enhancement threshold voltage (VTH) well before 10 time constants elapse. Turn-off time is quite acceptable with the Q1/MAX810 combination acting as a solid-state fuse. But turn-on time may be a problem, depending on the startup load and the pass transistor's ratio of source-bypass to drain-bypass capacitances.
If the startup load is small and C1 is large compared to C2, a quick FET turn-on causes only a small voltage dip at the MAX810 input. It will actually be less than that required to trigger a reset. For these conditions, the least expensive implementation of this circuit topology is that of Figure 2.
Say the external load or the charging of C2 draws a heavy current at startup, such that a fast turn-on of Q1 could cause the MAX810 to issue a reset. An RC network can be added to slow the turn-on (Fig. 3). Properly selecting these components can apply the load over multiple switching cycles of the MAX668, enabling its output voltage to remain above the reset threshold. Slowing the Q1 turn-on may be desirable, but slowing the turn-off may not. Accordingly, the circuit includes a Schottky diode in parallel with the resistor to quickly disable Q1 in response to an excessive and unexpected load.
To fully enhance the channel and obtain a low RDS-ON, these circuits require a logic-level, p-channel MOSFET like Q1. If Q1's on-resistance is high enough to cause a significant voltage drop, it may be desirable to regulate the drain side of Q1. This is especially true in low-output-voltage applications or if the load is relatively far away. In regulating the drain side, be sure to minimize parasitics. Observe good circuit-layout techniques as well. This remote regulation can be implemented with a SPDT, low-voltage analog switch in a SOT23 package (MAX4544) that's controlled by the MAX810's output state (Fig. 4).
The MAX4544 operates within data-sheet limits for supply voltages as low as 2.7 V. With the input supply at 3.3 V and about 0.3 V across the Schottky diode, both it and the MAX810 remain operational even when the boost converter is shut down. The MAX810's output is high during shutdown, thus connecting the MAX4544 COM node to NO (the Q1 source). When the boost converter turns on, resistors connected to that COM pin provide feedback to the MAX668. The MAX4544's on-resistance is 60 Ω maximum with a 5-V supply. So feedback-resistor values should be much larger than that to minimize output-voltage errors. On-resistance is only 120 Ω at 3 V, so the MAX4544 switch errors are minimal even for lower output voltages.
When the boost circuit is enabled and the timeout period has elapsed, the MAX810 output goes low and connects the load via Q1. At the same time, the output makes the feedback resistors switch over to the drain side of Q1. This switchover enables output-voltage regulation at the load, remote to the main boost circuit.
That action also switches the MAX810 input to the drain side of Q1, where it can monitor overload conditions at the load. This arrangement is especially valuable if Q1's RDS-ON causes a voltage drop greater than 1% at the maximum load current. This can happen if RDS-ON is either greater than or equal to 50 mΩ at an output that's greater than or equal to 1 A on a 5-V supply.
If faced with a current overload, the MAX810 output goes high and quickly turns off Q1 through the Schottky diode. Simultaneously, it switches itself and the feedback resistors back to the source (input) side of Q1. This configuration gives the boost output a chance to get back into regulation, after which the MAX810 reconnects the load. The cycle repeats until the overload is removed.
The break-before-make switching action of the MAX4544 is rather quick at 10 ns. A small capacitor across the feedback resistors maintains the output voltage during the break period to avoid disrupting the MAX668 feedback loop. It also gives power to the MAX810. To avoid any appreciable effects on the MAX668's transient response, the capacitor should be large enough to avoid significant discharge during the break period. Yet it has to be small enough to ensure a small time constant with the MAX4544 on-resistances.
The MAX4544's switch-control input has no Schmitt trigger, but it can tolerate slowly moving logic-level signals. The actual switch action is quick, once the switching threshold is reached. Logic-level signals may, however, cause a 10−4 A order of current to flow during the transition from the supply node to ground.
When using the MAX669 to boost low output voltages of 2.5 V and below, a negative voltage may be required to fully enhance Q1. For example, an inexpensive, discrete charge pump attached to the LX node can generate −VOUT + VD (Fig. 5). For a 2.7-V output, it produces −2.0 V with a standard pn-junction diode or −2.4 V with a Schottky diode. This voltage is present whenever the boost converter is enabled. It provides a negative supply for the MAX4544, which tolerates supply voltages to 12 V, as well as biasing for Q1.
Sensing Output Voltage
Although Q1 turns on when the MAX810 output goes low, the MAX810's reset threshold cannot accurately sense the main output voltage. That's because its ground terminal is referred to the negative-charge pump output. Accordingly, the MAX810 ground terminal connects to ground. Its output drives a level shifter comprised of Q2 and Q3, such that Q1's gate is pulled to the negative rail for turn-on.
At light loads, the MAX668 features idle-mode pulse-frequency modulation (PFM). So it can skip charging pulses when load current from the main supply is low. When that occurs, Q2's emitter current (set by R1) discharges C3. This action may cause insufficient supply voltage to the MAX4544, even with the main output voltage in regulation. In turn, this effect can force the on-resistance to skyrocket in the internal analog switch. The feedback voltage at the MAX668 would then drop towards ground.
The MAX668 would try to compensate by raising its output voltage, possibly causing an overvoltage condition. As an antidote, watch the feedback resistors, which are the minimum dc load on the main output voltage. Make sure that they're small enough to discharge VOUT slightly faster than the discharge of C3 by Q2's emitter current. Whether Q1 conducts or not, the following inequality allows the sizing of C3:
(VOUT − VBE)/(R1 × C3) < VOUT/\[(RA + RB) × (C1 + C2)\]
If no charging pulses are skipped during normal PWM operation, C4 can be small compared to C3. But the more pulses that are skipped, the larger C4 has to be. Boosting action resumes after pulse skipping, while Q2 is held off. C4 should then be large enough to charge C3 before C1 becomes fully charged.
Many components and interconnects affect the MAX668 feedback path in Figures 3 and 4. A fault in these components can produce an overvoltage VOUT that destroys the load. For added safety, a zener diode (not shown) is connected from C1 to the MAX668 FB pin. Its anode is joined to FB. This zener diode can provide an overriding local feedback loop that clamps the output at (VZ + VFB). To prevent a large overvoltage, set VZ equal to the maximum regulated VOUT minus the maximum VFB.
If the system must control multiple loads individually while the boost converter remains on, replace the MAX810 with a MAX812 (in a 4-pin SOT143 package type). The MAX812's fourth pin is designed for manual-reset applications. But it can force a disconnect between the local load and main boosted output by acting as a logic-level signal that overrides each smart solid-state fuse. With this approach, the user is able to control each load on the main boost supply independently.
Note that this smart solid-state fusing technique auto-resets without power cycling and requires no replacement or field troubleshooting. Most importantly, it doesn't have to be limited to boost-converter outputs. It can replace a fuse on the dc-power bus of virtually any system, regardless of voltage.
Above 60 V
Of course, bus voltages higher than 60 V may require non-logic-level FETs and level shifters for the MAX810 output. Using just two precision resistors to set an appropriate external bias for the higher voltages, the solid-state fuse can be set to be triggered by a programmed sag in the bus supply voltage (Fig. 6).
Suppose that −48 V is to be protected against overcurrent. Interrupt the rail side instead of the ground side, because the voltage source is negative. Use an n-channel FET plus a MAX809T reset circuit, which has a reset-output polarity that's opposite to that of the MAX810. The supply voltage can range down to −36 V under normal operation (Fig. 7). Design equations are as follows:
The MAX809 quiescent current is about 100-µA maximum over temperature. The current through RH and RL should be about 100× higher to minimize the quiescent-current effect on trip voltage: 36/(RH + RL) = 10 µA. Therefore:
(RH + RL) = 3600 Ω
The MAX809 threshold is much lower than the supply trip voltage. So RL is smaller than RH, approximately by the ratio VTHRESHOLD/(VTHRESHOLD + VSUPPLY TRIP) = 3/(36 + 3) = 0.077. MAX809 IQ flows through approximately 93.3% of (RH + RL), causing a voltage-trip contribution of about 0.336 V. Taking this fact into account, set the initial trip voltage for calculating RH and RL at 36 V − 0.336 V = 35.664 V. Using 1% resistors for RH and RL, VSUPPLY TRIP = 35.664 V. This threshold occurs when the MAX809T threshold is at its minimum (3.15 V over the temperature range −40° to + 85°C):
Calculated values for RL and RH are 323.81 Ω and 3276.19 Ω, respectively. The closest 1% values are 320 and 3280. Consider these resistor values and the 100-µA IQ. The maximum supply-trip voltage becomes 36.09 V, slightly surpassing 36 V. This result also occurs only for simultaneous worst-case values for all errors. In practice, that's a rare scenario. For most applications, this design would be quite acceptable. The MAX809's nominal threshold voltage gives a nominal trip voltage of −34.65 V.
RH should have a power rating of 0.5 W. But the voltage across RL exceeds the MAX809's maximum input-voltage rating when VSUPPLY goes above its minimum limit. To deal with this issue, place a 5-V ±5% zener diode across RL, as shown in Figure 7.
Eugene Carey is a corporate field applications engineer for Maxim Integrated Products Inc., Sunnyvale, Calif.; (408) 737-7600. He holds a BSEE from the University of Notre Dame, South Bend, Ind.
Michael Hess is an applications engineer for the customer applications department at Maxim Integrated Products. He holds a BSEE from San Jose State University, Calif.