Multiplexer breaks through 1MHz frequency limit

    Microprocessors and peripherals are increasingly demanding power, but obtaining power directly from an AC-DC converter is unlikely. At the same time, voltage requirements are becoming more and more stringent, and in order to obtain fast transient response, it is also required to minimize the distance between the power supply and the load. In this way, distributed power supply structures (see Figure 1) have become popular. The distributed power structure allows designers to group a small number of standard voltage lines in the entire system and use DC-DC converters to gradually increase or decrease the voltage, and finally obtain the required output voltage.

    In the past, since the operating current was usually less than 30A, the DC-DC converter powering the microprocessor generally consisted of a single-channel standard or synchronous compensation converter. However, the operating current of today's processors exceeds 30A and will increase exponentially. The single-channel compensation converter can no longer effectively power the new generation of processors because:
 
    1) Controlling the output ripple current requires a higher inductance 

    2) And Adding an inductor to reduce the ripple current will prolong the transient response time.

    3) In order to avoid power waste, a heat sink needs to be set up to solve the problem of concentrated power dissipation.

    4) When connecting MOSFETs in parallel to control high current, it is necessary to overcome the problems of current sharing and sufficient supply current. Conflict

New Methods

    Multiplex converters are gradually replacing single converters. Figure 2 is a four-way circuit structure. By connecting multiple plug-in converters in parallel, the peak current of each channel can be reduced and the following parameters can be improved: 

    1) Minimization of ripple current

    ·Input ripple current

    ·Output ripple current

    2) Reduce the parameter values ​​of passive components

    ·Input capacitance

    ·Output Inductor

    ·Output Capacitance

    3) Reduce transient response time

    The input pulsating current of the multiplexer converter may be continuous or intermittent, mainly depending on the number of channels, phase shift and load cycle of the converter. Usually, the phase shift value is obtained by dividing the number of channels by 360o, and the load cycle is the ratio of VOUT/VIN. Figure 3 is the input ripple current waveform of a four-way converter. Whether it is continuous or intermittent, the input ripple current of a multiplex converter is always lower than that of a single-channel traditional converter. As long as the appropriate phase shift is chosen, the worst-case input ripple current will only approximate the single peak output current. Reducing the input ripple current allows the converter to use a smaller number of input capacitors. The multiplexer draws all pulse input current from the input capacitors, thereby increasing the input ripple frequency and further reducing the number of input capacitors.

    The output ripple current of the synchronous compensation circuit is usually set to 30% of the output peak current. Figure 4 is an example of the output ripple current of a 90o phase-shifted four-way converter. The output ripple current of each channel is summed to the output capacitor, thereby canceling the ripple. The phase shift and the number of paths determine the degree of cancellation. For the same output inductor design, ripple cancellation reduces the peak-to-peak output ripple current of the output capacitor, thereby reducing the parameter values ​​of passive output components. Designers can reduce the number of output capacitors while maintaining the original output inductance; they can also reduce the output inductance of each channel while maintaining the original output current ripple specifications. The pulsation frequency increased by the summed output pulsation current varies depending on the number of channels. 

    Reducing transient response time is critical for multiplexers. The output inductance of each channel is in parallel, which can reduce the effective output inductance. Different numbers of parallel channels have different effects. Therefore, the n-channel multiplexer can reduce the transient response time of the output circuit. The larger the value of n, the shorter the transient time. Here are two options for improving transient response. Solution one is to set a larger output voltage variable during load transients, because the reduced output ripple voltage only consumes a small part of the total allowable error value of the output voltage. The second option is to increase the output current ripple ratio by reducing the output inductance of each channel, because the output ripple current in a multi-channel circuit is lower. 

    Operating the multiplexer above 500kHz cancels the ripple frequency effects, allowing lower inductor values ​​and fewer capacitors to be used in the converter design. High operating frequencies also allow designers to use all surface mount components, thereby reducing the size of most components and shrinking printed circuit boards. However, increasing the operating frequency increases MOSFET switching losses, thereby reducing efficiency. Converters with size restrictions usually use an operating frequency of 200kHz-300 kHz. Beyond this range, the switching losses of the MOSFET will increase significantly, while the size and number of passive components will not be significantly reduced. To significantly reduce the size of passive components, the converter needs to operate at frequencies above 1MHz.

Breaking through the 1MHz operating frequency limit

    Advances in MOSFET technology have paved the way for converters to operate at 1MHz. Figure 5 is a simplified equation for the synchronous compensation circuit and the power losses in Q1 (driver FET) and Q2 (synchronous FET). Obviously, Vin or frequency has a decisive influence on the switching loss of Q1, while Iout has an important influence on the conduction loss of Q2. The best driver FET will have the lowest QSWITCH x RDS(ON) value. Qswitch is the gate critical value, which is the sum of gate-source charge and gate-drain charge (Qgs2 + Qgd). The best synchronous FETs require low RDS(ON) coupled Cdv/dt immunity. Since the drain of Q2 is connected to the converter switching node, it becomes the transition bridge between ground and Vin. As Q1 turns on and off, the change in drain voltage dV/dt will be capacitively coupled to the Q2 gate, inducing a voltage pulse value sufficient to turn on the MOSFET, forming a short-circuit current. Therefore, the ratio Qgd/Qgs1 (gate-drain charge/ unit to reduce the possibility of turn-on Cdv/dt.

    International Rectifier's DC-DC optimized IRLR8103 and IRLR8503 meet the specific standards for MOSFET chipsets listed above. With QSWITCH x RDS(ON) values ​​as low as 65 (D-Pak package), the IRLR8503 is the best high-frequency Q1 MOSFET. The IRLR8103 has extremely low RDS(ON) (typically 8mΩ at 4.5 Vgs and a Qgd/Qgs1 ratio of 0.8, making it an ideal Q2 MOSFET.
 
Superior to discrete solutions

.     Multiplexed systems often require more components and are thus different from traditional single Compared with the previous channel design, it requires more motherboard space and more complex design. International Rectifier solves the above problem by integrating all the power supplies, drivers and passive components required for each channel on a single multi-chip module (MCM). Problem. Compared with similar discrete solutions, integrating all components in a single package saves more than 50% of the motherboard space and eliminates all important layout design (see Figure 6), thereby reducing parasitic capacitance and inductance. Parasitic components are reduced Finally, multi-chip modules can achieve the same efficiency as discrete solutions at higher frequencies, or achieve higher efficiency at the same frequency. The smaller size and higher operating frequency also allow designers to achieve the same efficiency in the smallest circuits. Hundreds of amps of current can be obtained across the board area.

    Figure 6 compares a future high-current, multi-channel, simultaneous compensation system with an existing system, demonstrating the potential benefits of this technology.