Managing power supply sequencing and supervision, part 1
A power supply sequencer and supervisor circuit requires a number of features to make it ideal for power supply management. It must enable seamless cooperation between any number of rails, including timing and sequencing, fast voltage supervision, fault handling, and debugging. Its EEPROM memory should enable complete autonomy, and the I2C/SMBus interface needs to enable it to be interactive in real time. With flexibility, however, comes some complexity, and there are design choices that need to be made in order to achieve success. This article walks through common scenarios and how to successfully apply the features of the LTC2937 power supply sequencer and supervisor.
We begin with a most basic hardware configuration, then expand to cover various improvements that take advantage of the features of the LTC2937 and add flexibility.
A minimalist circuit
It is possible to achieve power supply management with very few external components (Figure 1). The LTC2937 power supply sequencer and supervisor is programmed to be completely autonomous, handling sequencing up and down, voltage supervision, and fault management. A minimal number of external pull-up resistors and bypass capacitors are needed, but nothing else is required.
Figure 1 A minimalist LTC2937 circuit
- Connect SHARE_CLK to VDD through a 3.3k pull-up resistor. The sequencer won’t advance if this open-drain pin doesn’t toggle.
- Connect the open-drain ENn lines through pull-up resistors to a suitable voltage (up to 16.5V). Some DC/DC converters contain their own pull-ups, and may not require an external resistor.
- Connect SCL and SDA to GND, or optionally to VDD through 10k pull-up resistors. We are not using them in this configuration, but they should not be left floating.
- All other pins can take care of themselves – leave them floating. ON, MARGB, WP, ALERTB, FAULTB, RSTB, and SPCLK all have their own built-in weak pull-ups. The ASEL1,2,3 pins will float to set the (7-bit) device I2C address to 0x44, and WP pulls itself to the write-protected state, but that doesn’t matter because the I2C bus is unused here.
In this minimalist configuration the LTC2937 will derive power from the 12V input and power-up its VDD LDO regulator. When VDD rises above the undervoltage lockout threshold the device will load settings from EEPROM into operating memory (a process that takes about 2ms), then, because the ON pin pulls itself high, immediately begin sequencing up. All supplies will execute a coordinated sequence according to the preprogrammed rules stored in EEPROM.1
After the supplies are up the LTC2937 monitors for voltage excursions outside of the programmed OV and UV limits, and will detect a fault within less than 35µs of any excursion (ignoring short glitches).
When the device detects a fault, it can (depending on configuration) automatically bring down all supplies and, after a delay, try to sequence up again. It does this with no external intervention, protecting the system from damage. Note that many control behaviors are fully configurable in the EEPROM memory.
Adding features to the minimalist circuit
The minimalist circuit is very functional, but it does not take advantage of all of the features that the LTC2937 offers.
In the minimalist circuit, the device is not connected to the I2C bus, so it will only follow the programming contained in its EEPROM. It is not configured to reprogram the parts on the board, so preprogrammed parts are required. If the LTC2937 contains factory default programming, it will do nothing (do no harm).
Connecting the LTC2937 to the I2C/SMBus means that advanced features, such as updating voltage and sequence rules, reprogramming the EEPROM, LTpowerPlay support, fault debugging, single-stepping, and firmware control become available. More on this later.
In the basic circuit, there is nothing to tell the device when to sequence down. Neither the ON pin nor the I2C bus command are used. When 12V falls the ENn outputs remain high because the device does not pull them low. The sequencer does not sequence down, so the supervised output regulators are on their own as far as shutting down.
Adding VPWR monitoring
Because the minimalist circuit simply sequences up as soon as it can, and remains up until the 12V supply collapses, the likely result is an un-sequenced, unpredictable ordering of supply power-down.
There are several ways to overcome this challenge. At a minimum, we can implement a control mechanism using one of the supervisor channels to monitor the 12V voltage through a resistor divider, and cause a fault when it falls too low (Figure 2). The fault condition, while not ideal, at least causes all ENn pins to pull low simultaneously, shutting off the supplies before the 12V voltage collapses. This is a (minimal) improvement on uncoordinated collapse.
Figure 2 LTC2937 VPWR detect
In this case, the LTC2937’s fault retry delay should be programmed longer than the expected time that it takes the 12V supply to collapse. This prevents the device from trying to start again while the supply is falling. Setting FAULT_RESPONSE[7:5] = 3'b111 (13.1 seconds) will accomplish this.
To implement a full down-sequence there are several options. The best option is to use the ON pin to command the sequencer up or down. The ON pin controls when the supplies should be on or off with its logic level. Alternatively, an I2C command can force sequencing up and down if there is a bus master to send that command. Both of these options require additional circuitry to command the device. More on this later.
Continue to page two on Embedded's sister site, EDN: "A step-by-step guide to power supply sequencing and supervision, part 1."