Most modern dc-dc converters are still configured and controlled via analog signals derived from simple passive components. Even sophisticated high functionality converters, with state-of-the-art power conversion topologies, are likely to use external trim resistors and capacitors for defining values such as startup time, setpoint value and switching frequency. And of course, none of these parameters can easily be changed on the fly, making it virtually impossible to implement adaptive – let alone predictive – power management schemes.

With the exception of a few specialist converters for microprocessors (which offer limited digital programmability in the form of VID codes for output voltage control), most brick, intermediate bus and point-of-load (POL) converters on the market still operate in the analog control domain. The most urgent need is for digitally controlled non-isolated POL converters, because these are used extensively to provide the final voltages for devices on a board. However, the requirement also already embraces isolated converters, and designers will doubtless shortly be adding other digitally programmable power sources to their wish lists.

PMBus demonstrator kit, which includes a USB-driven board with eight digitally programmable POL converters and a PC-based graphical user interface

The reason for this seemingly odd scenario is simple: until now, there has been no industry-wide consensus on digital power management. A number of power supply manufacturers have launched digitally programmable POL converters which go some way towards addressing the issue, but these are based on proprietary architectures and silicon. What exactly is PMBus?
So, what exactly is PMBus?
PMBus is an open-standard digital power management protocol. It facilitates communication with a power converter or other device by fully defining the transport and physical interface, and the command language, needed to accomplish this. The protocol was founded by a coalition of power supply and semiconductor manufacturers, who recognized that lack of a suitable standard was inhibiting adoption of an all-digital power management solution, and it is now rapidly gaining wide industry acceptance. In March this year, Revision 1.0 of the protocol was placed in the public domain, and ownership transferred to an independent special interest group (SIG) known as the System Management Interface Forum, which is now responsible for further developing and promoting the standard.

It is worth noting that PMBus is not a standard for ac-dc power supplies or dc-dc converters. It does not specify attributes such as form factor, pin-out, etc, which are better served by industry alliances such as POLA and DOSA, nor does it address communication between one power source and another – this remains the responsibility of semiconductor and power supply manufacturers.

Low cost real-time control
The PMBus transport layer is based on version 1.1 of the low-cost SMBus (System Management Bus), which is a more robust version of the industry-standard I²C serial bus with packet error checking and host notification features. The I²C bus was originally developed by Philips Electronics for Inter-IC communication, while the SMBus was defined by an Intel led consortium for system management communications in PCs and servers.

SMBus features a third signal line – SMBALERT – which allows slave devices such as POL converters to interrupt the system host/bus master. This arrangement is inherently more flexible than a system employing the master to constantly poll slave devices and it imposes far less of a burden on the host processor, making it easier for designers to implement event-driven closed-loop control schemes. Furthermore, the PMBus protocol dictates that all slave devices must either store their default configuration data in non-volatile memory, or use pin programming, so that they power-up without any bus communication. System start-up times are consequently significantly shorter than with other digital control solutions on the market, which demand that the bus master configures all slave devices as part of the power-up initialization routine.

The physical address of each slave device is defined via dedicated pins – silicon manufacturers are certain to offer a variety of innovative approaches, such as tri-state pins and resistor value programming. In addition to the SMBus’ clock, data and interrupt lines, the PMBus protocol also specifies two hardwired signals for use with power conversion devices: one is a control signal used in conjunction with commands received over the bus to turn individual slave devices on and off; the other is an optional ‘write protect’ signal, which can be used to prevent any changes to memory-held data. A typical implementation is shown in Figure 1. SMBus uses the wired AND connection of all devices on the bus to provide arbitration in the event of bus contention, and is electrically similar to the I²C bus.

Figure 1:Typical connections for PMBus-based digital power management

PMBus has the distinct advantage that the master device is not based on proprietary silicon, nor does it act as a translator – all communications between the host and the power sources are conducted entirely via the bus. This saves on implementation costs and provides a much more flexible control approach. The host can be the system’s existing processor, a low cost general-purpose microcontroller, or even some gates in an FPGA. It can also, of course, be different things at different times. For example, during the board design phase a laptop PC can be used as the host, and then during production testing this role can be assumed by automatic test equipment, to comprehensively verify board performance – if necessary, dynamically changing the operating parameters of individual power conversion devices to accommodate the needs of the silicon on that board. The final select-on-test values can then be stored in the slaves’ non-volatile memory. PMBus, a simple command language
Simple command language
PMBus communications are based on a simple command set. Every packet contains an address byte; followed by a command byte; zero, one or more data bytes; and an optional packet error code (PEC) byte. Figure 2 shows a typical host-to-slave information transfer; the master uses single ‘start’ and ‘stop’ conditions to indicate the beginning and end of the process, and the addressed slave device uses a single bit to acknowledge reception of each byte. To minimize response times and processor overheads, the slave processes and executes a command immediately after it receives the ‘stop’ bit – unlike many other bus protocols, it is not forced to wait for a separate ‘execute’ command.

Figure 2: Standard master-to-slave communication sequence

While the protocol’s one-byte command code implies that as many as 256 commands are potentially available, it is not a condition of compliance that PMBus devices have to support all commands, and in fact many will use only a small subset to achieve their intended purpose. Considerable attention has been paid to ‘future proofing’ the standard; there is provision for two command extensions which effectively permit two-byte commands. One extension is reserved for PMBus device manufacturers’ own use, the other for subsequent revisions of the protocol.

Flexibility
The PMBus provides a great deal of flexibility. It is not possible in this short article to list the more than 100 commands available to power converter and power system developers. There are several commands for setting the output voltage, commands for setting warning and fault thresholds for input and output voltages, input and output currents, temperature and other parameters. In addition, how the unit responds to each fault, such as immediate shutdown or hiccup mode, can be programmed. There are commands for retrieving status bits and reading data like output voltage or the unit’s temperature. There are also commands for setting up the interleaving of paralleled converters, providing a software lock against accidental changes of data, configuring how the unit responds to on/off commands received from the SMBus port and CONTROL pin, and turn on and turn off sequencing. Commands are also available for storing and retrieving inventory information like a manufacturer ID, model number, serial number and date code.

It is important to note that the PMBus can be used with any type of power supply or dc-dc converter. The wide range of commands allow the configuring, control and monitoring of general purpose point-of-load converters, microprocessor powering VRMs, standard isolated dc-dc converters, bus converters and ac-dc power supplies. The PMBus even has the flexibility and capability for use with rectifiers in telecommunications battery plants!

To illustrate the flexibility of the PMBus specification, let us take a look at the commands that are available to control the output voltage, which include:

• VOUT_COMMAND
• VOUT_MARGIN_HIGH
• VOUT_MARGIN_LOW
• VOUT_TRIM
• VOUT_CAL
• VOUT_DROOP (as a function of IOUT)
• VOUT_MAX
• VOUT_SCALE_LOOP

Figure 3 shows a conceptual view of how these commands are used; the actual implementation is left to the manufacturer of the POL converter – but the overall behavior of the device must be as shown.

Figure 3: Conceptual view of how output voltage related commands are applied

The process of setting the output voltage starts with three basic commands: VOUT_COMMAND, VOUT_MARGIN_HIGH and VOUT_MARGIN_LOW. Each of these commands sends a value that is stored in a register in the PMBus device’s memory. One of these three values is selected by the OPERATION command and passed on to the rest of the output voltage command processing. The next step is to add the value in the VOUT_TRIM register to the output of the conceptual multiplexer. The value in the VOUT_TRIM register is a two’s complement number that can either add to or subtract from the value from the conceptual multiplexer. The VOUT_TRIM register will typically be used by the end user to adjust the output voltage once the POL converter is assembled into the end user’s system. This might be done, for example, to adjust the voltage at the pins of a critical IC to optimize its performance.

Next, the value from the VOUT_CAL register is added. This is also a two’s complement number and can add to or subtract from the voltage command value. The VOUT_CAL register will typically be used by the POL converter manufacturer to adjust the output voltage in the factory.

If the POL converter has an output voltage droop characteristic, it is applied now. The VOUT_DROOP coefficients are always greater than or equal to zero, and droop is only applied if the output current is greater than zero. The value of the VOUT_DROOP coefficient and the value of output current are multiplied and the result is always subtracted from the voltage command. VOUT_DROOP can only act to reduce the output voltage – never to increase it.

The next step is compare the commanded voltage developed so far with the maximum permissible output voltage set by the VOUT_MAX command. If the calculated voltage command would create an output voltage greater than the VOUT_MAX value, the POL converter limits the command voltage passed to the controller to the VOUT_MAX value, and also sets an alarm.

The same scaling factor that is applied to the external output voltage by a resistive divider is now applied to the calculated voltage command. This is done by multiplying the calculated voltage command by VOUT_SCALE_LOOP. At this point, the converter has a calculated value that is used as the equivalent to the reference voltage in a standard analog controller. This is the value to which the converter’s duty cycle.

For setting the output voltage, and related commands like setting the output overvoltage fault threshold, the PMBus allows up to sixteen bits of resolution. For other data, like reading back the input voltage or the output current, the PMBus specification allows up to ten bits of resolution. Ten bits of resolution corresponds to a resolution of approximately ±0.05% which is more than sufficient for the vast majority of the market.

Ease of implementation
Ease of implementation
The rich command set of the PMBus protocol enables designers to write lean and efficient power management programs – and implement the scheme – very easily and quickly. Voltage sequencing of POL converters provides an ideal example. Until now, many designers have chosen to use some of the excellent special-purpose controller ICs that are on the market to handle this task, accepting the fact that this involves developing programs using software provided by the IC manufacturer and using up valuable board space for the devices. Converters that can be directly controlled by PMBus potentially offer a more cost-effective and flexible solution, enabling a wide range of operating parameters to be changed at any point during the product’s life cycle, to accommodate engineering changes.

Only two PMBus commands are required for controlling the start-up sequence of a POL converter, as shown in Figure 4. TON_DELAY programs a time for the converter to wait until starting to produce an output, and TON_RISE programs the time for the output to increase from zero to the final programmed value. The user simply programs each converter with its own turn-on delay time and turn-on rise-time. Similarly, only two commands (TOFF_DELAY and TOFF_FALL) are required for turn-off sequencing.

Figure 4: Power-up/down sequencing typically involves just four PMBus commands

Voltage margining is another area where digitally programmable converters will make life considerably easier for designers – and production test personnel. Many board manufacturers now use this technique to evaluate the performance of ICs in the face of minor variations in their supply voltages; any marginal or below-spec devices can then be replaced as part of the normal production test process, before they become expensive, difficult-to-rectify field failures. Until now, margin testing has been a highly iterative and time-consuming procedure, involving fitting different value resistors to dc-dc converters in order to vary their output voltage a few percent either side of nominal. Using PMBus-compliant POL converters, this process becomes extremely simple. Two commands, VOUT_MARGIN_HIGH and VOUT_MARGIN_LOW, are used to pre-program the margin voltages. Then, as shown in Figure 5, using just the OPERATION command, each converter can be instructed to deliver tightly-controlled test voltages, while the effect on board performance is monitored. This can significantly reduce production test times, help eliminate ambiguity, and produce clearly documented test results.

Figure 5: Voltage margining is controlled with one PMBus command

The T1 and T2 time periods are determined by the VOUT_TRANSITION_RATE command, and the rate is defined in mV/s. Hence:

T1 = (VOUT_MARGIN_HIGH - VOUT) / VOUT_TRANSITION_RATE

And T2 = (VOUT - VOUT_MARGIN_LOW) / VOUT_TRANSITION_RATE

Setting the VOUT_TRANSITION_RATE to a value of FF FFh implies transition as quickly as possible.

More information
Further information about the open-standard PMBus digital power management protocol, and the SMBus transport layer, can be obtained from the System Management Interface Forum’s web site at www.powerSIG.com

About the author
Bob White is a Staff Engineer with Artesyn’s Worldwide Technology Development Group in Westminster, CO, responsible for supporting Strategic Marketing with the definition and specification of new products for future power system architectures. Bob White has actively supported the IEEE Power Electronics Society (PELS) for the past 15 years. He has regularly presented technical papers and seminars at APECs since 1987. Bob is also the author of the APEC Operating Agreement that brought together PELS, the IAS and the Power Source’s Manufacturer’s Association as co-sponsors. Bob has an SBEE from MIT (1980) and an MSEE from the Worcester Polytechnic Institute (1990), and is the holder of two US patents. He was also awarded the IEEE Third Millennium Medal in 2000.