3.1 On-chip PMU
The on-chip power management unit includes a low-drop out (LDO) regulator, a bandgap-stabilized voltage reference, and an integrated temperature sensor. The LDO regulator generates the internal supply voltage for the internal digital and analog blocks, including the gate drivers for P-channel and N-channel on-chip power FETs.
The main challenge for the PMU is to guarantee its functionality when the battery voltage falls down to 2.5 V. If such a voltage is applied, the driver is not able to turn the active element on and the functionality of the converter is lost. The flexibility of the PMU was improved by implementing, see Fig. 4, a diode-based circuit able to switch the supply voltage of the PMU to the higher voltage between the battery voltage (VBAT) and the output of the converter. During the regular operating condition, the input voltage from the battery supplies the PMU, but when an undervoltage condition occurs (e.g., due to cranking) and the battery voltage drops down to a few volts, the system keeps working by switching the supply source of the PMU to the DC/DC output.
The bandgap stabilized reference voltage generates the internal reference for the DC/DC controller, see Fig. 5. A particular attention was paid to this block because the accuracy and the temperature drift of the regulated output voltage are directly connected to the performance of the bandgap. For the high voltage integrated bandgap, according to an IP reuse strategy, the hardware macrocell we already designed and verified through experimental measurements in [17] has been adopted.
The temperature sensor monitors the junction temperature of the IC and triggers an interrupt to the diagnostic block every time an over-temperature condition occurs by turning off the converter, according to the safety strategy of the chip.
3.2 DC/DC controller
The characteristics of a DC/DC converter mostly depend on the feedback control loop. In this work, considering all the different programmable conditions of the output voltage, the main challenges for the design of the control were
-
to reach high performance in terms of efficiency, line regulation, and load regulation
-
to reduce the power consumption when the load sinks a very low current
Hence, we implemented the following double-control strategy:
-
i)
Current mode control with two nested loops (current and voltage feedbacks) in normal conditions for better regulated performances
-
ii)
Hysteretic control in case of low load current for low-power operation
During the regular operation of the converter, to achieve good performance in terms of step response and in terms of line and load regulations, the control acts like a current-mode control [7, 8], with a switching frequency that can be configured up to 888 kHz. This current-mode control includes two nested loops, as shown in Fig. 5: the outer voltage loop (red) and the inner current control loop (green). The nominal frequency of the loop is programmable at 444 or 888 kHz. As further discussed in Section 3.6, to spread the electromagnetic interference caused by a pure synchronous converter, a dithering technique of the switching frequency is applied. The nominal output voltage is set through the digitally programmable resistive divider (PRD block of orange color in Fig. 5) integrated on-chip in the DC/DC converter macrocell. The divider generates two different voltage thresholds for every programmable value of the output voltage. The first threshold is the FB_THR and it is used for the regulation loop. The second threshold, the OS_THR, is used to clamp the output voltage reducing the overshoot of the output voltage during the startup of the system or during transient responses to variations of the input voltage or load current. Finally, the control generates a PWM signal driving the gate of the power FET. A proportional and integral (P.I. in Fig. 5) control function is applied within the voltage control loop to achieve the duty cycle modulation that regulates the output voltage to the nominal value. The coefficients of the P.I. can be configured to perform a flexible design, allowing to have a good trade-off between the regulation performance and the stability margin of the system in different conditions of the programmable output voltage.
The step responses of systems controlled through a P.I. function are often affected by overshoot. Even if the overshoot can be minimized in typical conditions properly configuring the P.I. coefficients, the process-voltage-temperature (PVT) spread can cause a higher overshoot in other operating conditions. As the target of the DC/DC converter is to work in the harsh automotive environment, it has to operate properly in a wide ambient temperature range. To solve this issue, a dedicated structure was added vs. the state of the art to the regulation loop to keep the overshoot under control. In Fig. 5, the overshoot is limited by the OS_THR produced by the digitally programmable divider. When the output voltage exceeds this threshold, the power FET is switched off and the overshoot is limited. When the regulated voltage begins to converge to the nominal value, the PWM signal is applied to the gate of the power FET again, and the regular operation is restored.
Figures 6 and 7 show how the modulation of duty cycle allows to compensate variations of the output current, keeping the output voltage stable to the desired value. In fact, in Fig. 6, the duty cycle is low as the converter is regulating to 18 V from a battery voltage of 6 V sourcing just 20 mA. Instead, in Fig. 7, the duty cycle is increased to supply a ten times greater current, 200 mA. Similar results are obtained for the compensation of variation of the input voltage. In Figs. 6 and 7, the DC/DC converter is configured in step-up mode. On the contrary, Fig. 8 shows an example of the DC/DC converter when used in step-down mode to regulate from 25 V battery voltage to a 18 V output voltage (with 200 mA load current).
The configurations in Figs. 7 and 8 are representative of automotive LED lighting applications with a string of 6 LEDs (each with 3 V forward voltage) when the input voltage falls down to 6 V (cranking) or grows to 25 V, e.g., due to overvoltage. To be noted that the waveforms in Figs. 6, 7, and 8 have been experimentally measured using the fabricated IC prototype and the measuring test bed proposed in Section 4.
The low power hysteretic control was implemented to fulfill the challenging automotive requirements on power saving. This control by-passes the regular synchronous control, when the load current is low (e.g., during standby states or when the load is working with reduced performance). In this condition, it does not required a very accurate supply, and the power consumption is mainly low due to continuous charging/discharging the gate capacitance of the power FET while the conductive power losses are negligible (the load current is low). Since the hysteretic control is an asynchronous control, it avoids unnecessary switching activity and the converter can save power. Indeed, when the converter is in low power mode, the power FET is turned on only when the output voltage falls down the nominal output voltage value. Once the power FET is turned on, the converter keeps sourcing energy to the output until the output voltage reaches a value slightly greater than the nominal one. After that the switch is kept off until the load current has slowly discharged, the output capacitor behind the target value and the cycle starts again. Since in this condition, the current sunk from the DC/DC’s output port by the load is very low, the voltage on the output capacitor will take a long time to discharge, and hence, the hysteretic control allows to minimize the switching activity. In low power mode, the total current consumption of the IC is reduced at about 300 μA, saving about the 80 % of the current consumption vs. the normal mode, which implements the nested control loops of Fig. 5.
3.3 Slope controlling of the power FET gate driver
The gate driver is the final stage of the control loop. It manages the turn-on and turn-off of the FET switch. In many applications, the gate driver is a simple CMOS inverter (i.e., a couple of a P-MOS, acting as a pull-up transistor, and an N-MOS, acting as a pull-down transistor with common input and output signals) with a size to have a current capability high enough to drive the input gate capacitance of the power FET at the desired switching frequency. In automotive applications, instead, constraints on power dissipation and electromagnetic emissions make the design of this block much more challenging [18]. In our design, electromagnetic emissions are reduced by minimizing the cross conduction. Conventional CMOS inverter drivers are affected by cross conduction during the commutation of the output signal. During the transitions of the PWM signal, the cross conduction generates current spikes from the supply (VDD) to ground (GND) when the gate-source capacitor (Cgs) of the power transistor is charged and discharged. These spikes are periodical at twice the switching frequency of the converter. In the presented solution, the current spikes are minimized by minimizing the cross conduction of the gate driver. The concept behind the proposed gate driver is shown in Fig. 9. The input signal of the gate driver is Vin33. The signal Vin33 is generated by the digital core of the IC, and it is immediately shifted from 3.3 to 5 V (Vin) to turn on the power switch having a Vgson typically greater than 2.5 V. The first stage of the gate driver includes the two CMOS inverters I1 and I2. These inverters are low-power inverters, and their function is to shape the input signal of the power inverter I3 to minimize the cross conduction. The low-side and high-side stages of I1 and I2 have different current capability and generate an output signal with asymmetrical rising and falling edges. When a rising edge of input signal occurs, the strong pull-down stage of inverter I2 quickly forces the gate of the NMOS of the power inverter I3 to ground. The power NMOS is then switched off. At the same time, the inverter I1 begins to force to zero the gate of the power PMOS of the inverter I3, to turn it on. As the low side of I1 has a lower current capability than the low side of I2, the falling edge of the gate of the PMOS of I3 is smoother and when it is low enough to switch the power PMOS on, the power NMOS is already off. The cross conduction of the power inverter (I3 in Fig. 9) during the falling edge is minimized in the same way. In this structure, only the low power inverters I1 and I2 are affected by cross conduction, but as a low current flows through these devices, this effect has no impact on the performance of the DC/DC converter. The final stage of the gate driver, instead, is significantly improved because its cross conduction is strongly reduced and the correspondent dangerous current spikes are eliminated.
3.4 Digitally programmable output voltage
Most of state-of-the-art DC/DC converters [7–9] need an external resistive divider or a more complicated circuit to set the output voltage value. In most cases, once the output voltage is set, it cannot be changed anymore. In our design, instead, a digitally programmable resistive divider is integrated on-chip, see PRD block in Fig. 5. Using an integrated digitally programmable divider to generate the feedback signal for the regulation loop of the converter saves costs and area on the PCB and allows changing the output voltage level in a range between 1 and 48 V at any time.
3.5 Diagnostic block
Safety has a primary role in the design of automotive ICs. These devices, in fact, must be able to prevent harms to the outer systems if a fault condition occurs [19]. In our design, a dedicated diagnostic block was implemented to detect and manage the following failures: (i) overcurrent into the primary side; (ii) overtemperature; (iii) undervoltage on the output pin; and (iv) overvoltage on the output pin.
It is very important to monitor the current flowing into the primary side of the transformer because if it exceeds the maximum allowed value (dependent on the external parts used), the external switch may be damaged and the magnetic element’s core may saturate, irreversibly compromising the functionality of the whole system. When an overcurrent condition occurs, the power FET is then immediately switched off. An integrated temperature sensor generates a voltage proportional to the temperature of the die. The system manages the overtemperature by switching the converter off as soon as the output voltage of the sensor exceeds a threshold corresponding to the maximum operating temperature. Undervoltage and overvoltage conditions on the output pin are detected by comparing the reference voltage from the bandgap to two dedicated additional partitions of the output voltage. If such a condition occurs, the DC/DC is switched off.
3.6 Dithering of the switching frequency
The dithering of the switching frequency was implemented to improve the performance of the converter in terms of electromagnetic emissions. If the switching frequency is constant, the power spectrum of the system is concentrated at this frequency reaching a very high level, not compliant with automotive regulations. The technique of the dithering consists in slightly changing the switching frequency. This allows spreading the power spectrum on a wider range of frequency, thus reducing emission peak value, according to automotive regulations. Once the programmable switching frequency fSW is set, the dithering function periodically changes the switching period according to a triangular law, see Fig. 10, in a range from the nominal value fSW minus 12.5 % to the nominal value fSW plus 12.5 %. The nominal switching frequency fSW can be set at fSW1 = 444 kHz or fSW2 = 888 kHz and is obtained by dividing with a factor n1 = 16 or n2 = 8 a high frequency clock of about 7 MHz, used by the digital part of the integrated circuit. This keeps the mean switching frequency equal to the set nominal value, but the maximum value of the power density spectrum is reduced by 20 dB according to Eq. (1), already used in literature [20]. In Eq. (1), fSW is the nominal switching frequency of the controller, δ is the percentage dither about the fundamental switching frequency (between ±12.5 % in this design), fDITHER is the dither modulation rate (set at 17.76 kHz so that the dithering period is a multiple of the switching period being fSW1/fDITHER = 25 and fSW2/fDITHER = 50), n is the system clock frequency divider used by the regulator (8 or 16 in this design). By further applying a frequency hopping between 444 and 888 kHz, we further spread the emitted power whose maximum level is reduced by a factor of two, i.e., a total attenuation of 23 dB of the emitted peak power level is obtained.
$$ \mathrm{Spectral}\ \mathrm{attenuation}\ \left(\mathrm{dB}\right)=10* \log \left[\left({f}_{\mathrm{SW}}*\delta \right)/\left({f}_{\mathrm{DITHER}}/n\right)\right] $$
(1)