Next Step in PMBus Evolution

Several years ago, my family took a trip to Australia. This was our longest vacation ever — 17 days. It was also the best vacation we took as a family, according to my daughters. One of the highlights for me was visiting the rainforest of northeast Australia. Our guide pointed out much of the unique flora and fauna. I never thought much about diversity until the guide pointed out vegetation and referred to it as primitive flowering plants. Up until then, I did not think of living things actively evolving and on their way to the next stage of differentiation. Another point the guide made is that flowering plants are the most successful and dominate the plant world.

I recalled this trip during my recent contemplation of Power Electronics 2020. While I know power management will continue to evolve, I couldn't help but wonder what attributes will cause one power technology to dominate over others. You could think about the evolution of power electronics from its beginning until today. An interesting paper by Thomas G. Wilson titled “The Evolution of Power Electronics” discusses how power electronics has evolved up to the year 2000.1 This paper does a very good job of defining power electronics and tracking its evolution. Dr Wilson's Figure 12 shows a power electronics timeline from 1880 to 1980. It starts with the Edison effect (thermionic emission) in 1883 and ends with the silicon-controlled rectifier (SCR) in 1957. Along the way, there are many references to magnetic amplifiers and vacuum tubes. As interesting as these technologies are, the really interesting stuff started after 1980 — at least in my opinion. As new silicon-based devices were applied to power, size and weight went down, and efficiencies went up.

This brings me to the topic for this blog: PMBus 1.3. The first public rollout for this update was in the Darnell Energy Summit in September.2 PMBus 1.2 is currently the dominant standard for power management, and 1.3 seeks to evolve this standard to secure its future domination. As loads become more dynamic, there is a need to increase the PMBus throughput, so 1.3 extends the clock speed to 1 MHz from the previous option of 400 kHz. In order to accommodate previous slower speeds, clock stretching has become part of the specification. I still have bruises from the original debate about clock stretching and the PMBus. Many adopters considered this to be a level of complexity that just was on the wrong side of too much.

Another area of evolution is the data format. PMBus 1.3 adopts the IEEE 754 floating point industry standard as one of the accepted data formats. Data format was another one of those hotly contested issues. The concern again was too much complexity. But just like in nature, evolution seems to handle complexity.

The biggest addition to the PMBus specification is that of adaptive voltage scaling (AVS). The original PMBus broke the specification into two parts. Without applying any imagination, we call them parts 1 and 2. Part 1 addresses the physical layer, and part 2 addresses the data layer. I think this is a critical addition in order to secure future dominance of this particular protocol. It does come at a price in that it adds an additional bus-type derived from SPI and a part 3 to the specification. The difference from the standard multi-drop SPI bus is that it does not use a select pin, so it is a point-to-point communication. This means that there is one AVSBus for each slave. This connectivity is shown in Figure 1.

The AVSBus can be either a two- or three-wire bus, depending on the need for bi-directional information flow. The goal for bus is speed and close collaboration between the power management and load devices. To address the base speed requirements for today's very dynamic loads, the maximum clock speed on the AVSBus is 50 MHz.

PMBus 1.3 adds functionality needed to ensure dominance. AVS is just one example of the evolution. This evolutionary step will help the system to achieve better efficiency by providing the voltage the load needs to meet its performance requirements. Additionally, PMBus 1.3 certainly adds more agility to the specification, which serves the spec well.

For early adopters and those waiting for a reason to adopt, you should visit for more information and details around the proposed specification. The adopters need to get their feedback into the specification working group.

Evolution is a powerful thing — whether in the organic world or the world of power management. It should drive each generation to dominance. What is your view of Power Electronics 2020? The year 2020 is not that far off, given how fast this decade is progressing. Will the loads gobble up power management? Will silicon devices still dominate power? Or will we become just part of The Matrix ?

Figure 1

Adaptive voltage scaling PMBus 1.3 proposal (see reference 2).

Adaptive voltage scaling PMBus 1.3 proposal (see reference 2).

For more information about this and other power topics, visit TI's Power House blog.


  1. Thomas G. Wilson, “The Evolution of Power Electronics,” IEEE Transactions On Power Electronics, Vol. 15, No. 3, May 2000
  2. Michael Jones, Travis Summerlin, “Introduction to the Preliminary Specifications for PMBus™ v1.3 Including new AVS Features,” 2013 Darnell Power Forum, Session 11.

This article was originally published on EDN.

3 comments on “Next Step in PMBus Evolution

  1. t.alex
    November 14, 2013

    I have seen this bus in action at several power management ICs, and it was amazing. Load can be shared or optimized and this effectively can help manage the system power at the most efficient way.

  2. Hailey Lynne McKeefry
    November 16, 2013

    @t.alex, do you see this being particuarly applicable to certain types of applicaions or verticals?

  3. t.alex
    November 16, 2013

    Maybe i am wrong but is it correct that PMBus is mainly used for inter-ic communication (on a same board)? If that is the case, the use is limited to single PCB design I believe.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.