Wednesday, April 3, 2013

What's in a Node? Putting Market Research in Perspective

Since I have a pretty extensive background in market research related to process automation, I was greatly interested in the new study results announced by IMS Research, (one of the few market research firms aside from ARC, Frost and Sullivan, and AMR/Gartner that serves the automation business).  The new study examines the worldwide market for Ethernet and fieldbus technologies, and the study results have caused quite a stir in the media because they seem to predict the doom of fieldbuses in general in favor of Ethernet.

As marketing manager for one of the leading fieldbus organizations, I can tell you without a doubt that we are still experiencing double digit growth in the sales of products and services related to FOUNDATION technology, and to me the results presented by IMS Research warrant further scrutiny, so here's what I learned about the study and their methodology.

IMS makes no distinction between fieldbuses or networks related to process automation versus discrete automation.  So, the study lumps together networks covering the full spectrum from device level networks in discrete manufacturing such as AS-i, ControlNet, and PROFIBUS DP, to process automation fieldbuses such as FOUNDATION fieldbus, HART, and PROFIBUS PA.  IMS also includes HART devices in the scope of the study, although it is unclear to me if they are including all HART devices sold or just those integrated with the DCS.  All of these networks are categorized as "Fieldbus" networks in the IMS report.

IMS also includes all the Ethernet-based protocols, such as Ethernet/IP, EtherCAT, PROFINET, and presumably FOUNDATION HSE, categorizing all of these as "Ethernet-based" networks.

The primary method of quantifying the market for IMS' purposes is nodes, not revenues from products and services sold.  IMS defines a node as a network connection, but that definition has always been a little unclear to me.  Is a node equal to a device or sensor, or does it include connections at the controller and I/O level?  Counting nodes also seems to tilt things in favor of discrete networks, since you quite often have many more sensor connections in discrete applications than you do in most process applications.  A discrete network needs to be able to handle a lot of not-so-complex messages very quickly from a large number of sensors.  A process fieldbus has its own requirements for speed and message complexity and process control that are very different from discrete applications.

Anyway, this is the measure that IMS has chosen to use.  According to this measure their research shows that fieldbuses still dominate the marketplace, accounting for a little under 75 percent of total installed nodes versus their counterparts in the Ethernet world.  IMS sees "fieldbuses" accounting for less of a share of the total number of installed nodes. In 2016, IMS predicts that fieldbuses will account for a little over 69 percent of all the nodes sold.

So what IMS is really saying is that fieldbus nodes will still account for almost 70 percent of the total nodes sold in 2016 versus just under 75 percent in 2011, when you combine both the process and discrete industry sales.  Shipments of both fieldbus and Ethernet networks will experience growth over the next five years, but Ethernet-based networks will experience more growth.

To me that sounds a little more reasonable than sounding the death knell for fieldbus, but you also have to keep in mind that Ethernet in and of itself does not do all the things that are required of a device network or control network.  Ethernet only goes so far, and you need some kind of a protocol on top of Ethernet to get the job done, that's why we have numerous Ethernet-based variants on the market today.  The terms Ethernet and fieldbus are also not mutually exclusive, as we can see with FOUNDATION fieldbus HSE, where HSE supports the entire range of fieldbus capabilities, including standard function blocks and Device Descriptions (DDs), as well as application-specific Flexible Function Blocks (FFBs) for advanced process and discrete/hybrid/batch applications.

It is the assertion of IMS, however, that fieldbus networks are not sustainable because it take too much overhead to maintain these networks.  From the press release:
"Simplifying the network can reduce company overheads through an integrated system. This is difficult to achieve with fieldbus technologies. Instead end users will usually have separate office IT divisions and a factory [sic] IT divisions. Ethernet adoption across a plant or factory provides a better environment for sharing information and a single division with responsibility for the overall network. The benefits of which are likely to be less downtime and lower overall cost."
If it is the assertion of IMS that Ethernet-based networks are intrinsically simpler than fieldbus networks, I am not sure this is entirely true.  Again, Ethernet itself does not comprise the entire network, and there are differences in Ethernet-based protocols that must be understood by the user.  Waving your magic wand and making everything Ethernet does not solve your problems.  Ethernet also has its limitations at the plant floor in process applications.  The reason we use twisted pair as a communication medium in H1 FOUNDATION fieldbus is because it is the de facto standard for the process industries.  I don't think you are going to see Cat 5 connectors on pressure transmitters any time soon as long as there are issues surrounding bus power to devices, etc.

If you reference international standards, you will see that Ethernet/IP, Profinet, EtherCAT, HSE and others are all consider "fieldbus profiles" under IEC 61158 and IEC 61784.  Ethernet is standardized as ISO 8802-3 and specify the physical layer (wire or fiber optic) and parts of the media access control (when you can transmit). And, that’s the extent of the specification.  Ethernet does not describe the application (e.g. data models, services like read/write, etc).

The term fieldbus is defined in IEC 61784-3 as "communication system based on serial data transfer and used in industrial automation or process control applications"  The communication system is more than just a physical layer specification.  It describes the data model and data is transferred between different communication entities on the fieldbus.  Some Ethernet based fieldbuses can use "off the shelf" Ethernet and other profiles require specialized equipment to achieve high performance.  There are lots of choices, but each has their own consequences.  And, in the end, you still need to do the engineering to see what solution best fits your application.

IMS has done some interesting research here and I can speak from personal experience that doing these studies is not easy, and someone will always want to pick apart your results.  I must say that it is refreshing to see new research in this area, and we wish IMS success in the future.  Just take this as constructive criticism for the future from someone who's been there that sometimes it helps to put things in perspective.


  1. Well said, Larry!

    The one thing you did not mention is the requirement to operate in hazardous areas.

    1. You are right, Maris, I thought it would get too lengthy if I started going on about that. I did consider it.

  2. Very nicely explained . Thanks