Generating a Higher Output Voltage than 1.525V Using the MAX77812

Abstract

The MAX77812 is a quad-phase, high-current, step-down (buck) converter for high-end gaming consoles, VR/AR headsets, DSLR cameras, drones, network switches and routers, and FPGA systems that use multicore processors. The maximum output voltage of the MAX77812 is 1.525V per register setting. However, the output voltage can be increased up to 2.7V with an external resistive voltage-divider. This document provides a formula that determines the external resistor values for a given output voltage and explains practical considerations for it.

Introduction

The MAX77812 supports programmable output voltage from 0.25V to 1.525V in 5mV steps through an I2C interface. For some applications, an output voltage higher than 1.525V is required. The MAX77812 supports the higher output voltage with the addition of an external voltage-divider network.

External Voltage-Divider Network

A buck converter regulates the output voltage to the target value by comparing the sensed output voltage (VSNSxP) to the internal reference. If VSNSxP is lower than the actual output voltage (VOUTx), VOUTx will be higher than the nominal output voltage set through the I2C interface.

As shown in Figure 1, the external voltage-divider network consists of feedback resistors (RFB1 and RFB2) and a feed-forward capacitor (CFF). The resistors divide VOUTx to the lower value VSNSxP at the remote sense input (SNSxP):

The internal sensing resistor at SNSxP is RSNS. Voltage VSNSxP is then compared to the internal reference set by the output voltage setting register (Mx_VOUT[7:0]). Therefore, the relation between the actual and the nominal output voltages is:

An output voltage higher than 1.525V can be achieved by adjusting Mx_VOUT[7:0], and the voltage-dividing ratio is:

Figure 1. External voltage-divider network.

Figure 1. External voltage-divider network.

Value Selection

The selection of RFB1 and RFB2 must guarantee the accuracy of output regulation and minimize the power loss on these resistors. The resistance of RFB2 should be significantly smaller than RSNS to be dominant. Because the resistance of RSNS is approximately 350kO, the recommended value for RFB2 is around 51.1kO. To minimize the difference between the actual and the nominal output voltages, Mx_VOUT[7:0] is selected as close to 1.525V as possible. Once RFB2 and Mx_VOUT[7:0] are fixed, RFB1 can be selected based on equation (2). The accuracy of the output voltage highly relies on the accuracy of the voltage-dividing ratio, thus ±1% or better resistors are recommended for RFB1 and RFB2.

The external voltage-divider network creates an additional pole and zero at (RFB2 ||RSNS˜RFB2) for simplified calculation:

To maintain the loop stability, the recommended value for CFF is around tens of picofarads and is determined by the values of RFB1 and RFB2.

Table 1 shows a few examples of the value selection recommendation for common output voltages.

Table 1. Value Selection Recommendation and Measured Maximum Load Current
VOUTx (V) RFB1 (k?) RFB2 (k?) CFF (pF) Mx_VOUT[7:0] Maximum Load Current (with a 0.22µH inductor)
1.8 9.09 51.1 100 0xF9 = 1.495V 4.0A at VIN = 3.8V
2.4 27.4 51.1 39 0xF8 = 1.490V 2.5A at VIN = 3.8V
2.7 34.8 51.1 27 0xFE = 1.520V 2.0A at VIN = 3.8V

Considerations

Although the output voltage can be higher than 1.525V, it is still limited by the input voltage and load current. Theoretically speaking, the constant on-time is proportional to the ratio of the actual output voltage (VOUTx) to the input voltage (VIN). The MAX77812 calculates the on-time by sensing VSNSxP and VIN. Because VSNSxP is lower than VOUTx, the on-time is insufficient. As a result, more switching cycles are needed and thus the switching frequency (fSW) increases. The constant on-time control architecture also exhibits higher fSW when the load current increases. Therefore, supporting a high VOUTx under the heavy load condition leads to a substantial increase of fSW, which may be ultimately limited by the control architecture. For the same VIN, the higher the actual output voltage is, the lower the maximum load current that can be supported. Table 1 provides examples of the maximum load current measured on the bench. To mitigate the switching frequency increase, Mx_VOUT[7:0] needs to be selected as close to 1.525V as possible for a longer on-time.