JAJSMM7A August 2022 – September 2023 TPSM8D6B24
PRODUCTION DATA
To allow multiple devices with a shared output to communicate through a single PMBus address and single PMBus loop follower, the TPSM8D6B24 uses a back-channel communication implemented through BCX_CLK and BCX_DAT pins. During POR, all of the devices connected to VSHARE must also be connected to BCX_CLK and BCX_DAT and have appropriate (ECh) MFR_SPECIFIC_28 (STACK_CONFIG) settings. Any programming error among the devices of a stack will result in a POR fault and prevent enabling of conversion.
During POR, the loop controller reads the programmed values from the loop followers to make sure all expected loop followers are present and correctly phase-shifted. Then, the loop controller loads critical operating parameters such as the following to the loop follower devices to ensure correct operation of the STACK:
During operation, the loop controller device receives and responds to all PMBus communication, and loop follower devices do not need to be connected to the PMBus. If the loop controller receives commands that require updates to the PMBus registers of the loop follower, the loop controller relays these commands to the loop followers. Additionally, the loop controller periodically polls loop follower devices for status and telemetry information to maintain an accurate record of the telemetry and STATUS information for the full stack of devices.
Most PMBus communication must be directed to all phases by leaving the (04h) PHASE PMBus command at its power-on reset default value of FFh. If a specific device must be communicated with, the (04h) PHASE command can be changed to address a specific device within the stack, as set by the order value of the (37h) INTERLEAVE command programmed during POR.
When commands are directed to individual loop followers, write commands are queued by the loop controller to be sent to the loop followers through the BCX if other BCX communication is in progress. Queued write commands are written to the loop followers in the order the loop controller receives them. To avoid unnecessary delays on the PMBus and excessive clock stretching, read transactions targeting individual loop followers are not queued, and are processed as soon as the BCX bus is available. As a result, it is possible for a read command targeting an individual loop follower immediately following a write command can be processed before the preceding write command. To ensure accurate read-back, users must allow a minimum of 4 ms between writing a value to an individual loop follower and reading that same value back from the same loop follower.