IEEE Power Electronics Magazine - September 2014 - 34

A digital power management protocol that operates
over a serial bus must have two basic elements. First, a
transport mechanism is needed to move a message from a
sending device to one or more receiving devices. The transport mechanism must specify the details of the electrical
interface, the timing of the signals on the bus, the format
of the packet structure and the data within the packets, the
flow control with the means to signal the start and completion of a transmission, and an addressing scheme that controls which devices respond to a message and which ignore
the mechanism. Standards like the I²C interface, created by
Philips, describe how to get a message from one device to
another but give no meaning to the transmitted bits. Also
needed is a command language to convey the meaning of
the message, such as a command to "set the output voltage
to 3.32 Vdc" or a command to "turn off the output voltage."

How PMBus Came to Be
By the early 2000s, most every server and telecom manufacturer was using serial bus communication with the power
supplies embedded in their systems. Each equipment manufacturer had its own proprietary protocol or protocols.
Many of the manufacturers were also changing, or even creating all new protocols, with each new product family.
This made it difficult for power supply manufacturers
to develop and manufacture power supplies quickly and at
low cost. The power supply manufacturers were spending
a significant amount of time and energy on developing and
maintaining multiple digital power management interfaces.
In addition, the equipment manufacturers were having
problems with qualifying products in their systems. With
each new development program, they would often have to
go through a painful cycle of testing and helping each of
their power supply manufacturers debug the power management interface. The market situation was further complicated as power supply manufacturers, especially those
offering off-the-shelf product, had their own proprietary
digital power management protocols.
In early 2004, I was working in the Worldwide Technology Development Group of Artesyn Technologies. I was
asked to develop a digital power management protocol for
Artesyn because of my experience developing a protocol
in a previous job at a computer company. I worked on this
protocol for several weeks, choosing the SMBus for the
transport layer and working out the details of a command
language. As the specification for this protocol matured, I
realized that it was not really going to be useful. Artesyn's
original equipment manufacturer customers had their own
protocols and would never adopt one from a power supply
manufacturer-no matter how good.
I realized that it is not really important which transport
mechanism was used or what pattern of bits means "turn
on" or "turn off." What is important is that equipment manufacturers and power supply manufacturers all use the same
protocol. If we could standardize the digital power management protocol, everyone would win. Power supply manufac-

IEEE PowEr ElEctronIcs MagazInE

z September 2014

turers could develop the interface circuits and controllers
once and use them in all of their products. This would save
time, lower cost, and reduce risk on a new power supply
development program. Equipment manufacturers could develop the power management part of their system management software once-again saving time, reducing risk, and
lowering development cost. In addition, equipment manufacturers could simplify the process to test and qualify a
new power supply in their system. Furthermore, integrated
chip (IC) manufacturers would now have sufficient production volume to make developing power management interface circuits that could either be sold as stand-alone parts
or embedded in their power supply control ICs-reducing
cost, development time, and risk.
In June 2004, I went to my manager, Trey Burns, vice
president of technology, and the division president, Ken
Blake. I explained to them the idea of an open digital
power management standard. They quickly understood,
and Ken directed the vice president of marketing, Todd
Hendrix, to talk with other power supply companies and
IC manufacturers to gauge their interest. The idea caught
on quickly, and, by early August, a specification working
group was formed.
We agreed on the name "PMBus," which stands for Power Management Bus. We wanted the standard to be open.
This meant that the standard would not be owned by any
one company or group of companies but by an independent
nonprofit organization. We did not want to go through the
time and expense of setting up a new nonprofit corporation to own the PMBus specification. We approached the
System Management Interface Forum (SMIF), which developed and owned the System Management Bus (SMBus) and
Smart Battery standards, asking if they would be interested
in taking on PMBus as one of their standards. The SMIF
agreed, and the PMBus development effort was under way.
The PMBus specification working group took the draft
standard I developed as the starting point. The PMBus organization and intent of the specification were announced
in September 2004. The first PMBus specifications were
released a little more than six months later in March 2005.
The original PMBus specification working group was a
remarkable organization. Dave Freeman of Texas Instruments (TI) and I cochaired. Dave brought a wealth of experience as he had been involved in the development of the
SMBus and Smart Battery specifications. Several others
made contributions of particular note. Gabe Suranyi, then
with Tyco Power, contributed much from the point of view
of a power supply manufacturer and a focus on telecomm
power. Ken Fernald, then with Zilker Labs, contributed significantly by assuring that the command language could
be implemented with efficient algorithms. Keith Curtis of
Microchip Technology assured that specifications could be
easily implemented in general-purpose microcontrollers.
Bob Carroll, then with Primarion, provided guidance to assure that the protocol could be implemented cost-effectively in silicon. That the PMBus specification has become the

Table of Contents for the Digital Edition of IEEE Power Electronics Magazine - September 2014