The mystery of the forgotten HART high speed physical layer and long form address, and why this is important in the era of Ethernet APL and HART-IP

The Spanish version of this article is available here .

Some changes coming

The original idea with this newsletter was that it would be issued in a weekly basis. But in that case, LinkedIn's newsletter tool allows me to send just one newsletter per week.

Now, as you know, I publish each article in an English and a Spanish version simultaneously. Therefore, LinkedIn's tool would force me to publish both versions as different newsletters with a week between them.

The solution is a compromise (as all solutions are). The newsletter is declared as daily. To be fair with all readers, this week's English version newsletter will appear on Monday, and the Spanish version on Tuesday. Next week this will be reversed: Spanish version will appear on Monday and English version on Tuesday. I hope this is not an issue.

Thanks to you all. Now let's get into today's subject.

Introduction: HART was always slow

The most common commentary people make when they find out that the data transmission speed possible in the HART communications protocol is that it is slow.

It may have seemed slow when it was first presented in 1984, when the choice for the computer networking standard protocol still was a race between various competitors and the Internet technology was computer geek’s stuff.

1 FSK HART encoding

At that time Hart’s protocol transfer rate was 1200 baud, or 1,2 kbps while the contemporary Ethernet networks were running at a comparatively blazing 3 Mbps.

But computers were also slow, ran single task operating systems and used text based graphical interfaces, so the 1,2-kbps data transfer of HART devices speed did not seem to be a fundamental problem.

But the world became faster

By 1995, the world’s relationship with computers and networking had changed dramatically. TCP/IP was the Internet’s protocol of choice and therefore a de facto standard. Ethernet had long ago ditched both coaxial cable and bus topology and had changed to switched networks and twisted pair cabling. And most important: it now supported data transfer speeds of 100 Mbps as a standard.

HART protocol’s networking functionality had been limited from its beginnings by the choice of the original physical layer'. It was based on the FSK Bell 202 frequency shift key encoding. Although a fully digital, multidrop capable working mode was implemented in the HART standard, due to the limitations of the physical layer, the 1,2 Kbps limit remained through time. And the maximum number of devices that could be daisy chained was stuck at 16.

2 Frequency Shift Key encoding waveform

A solution appears

To address these limitations, in 1995 the HART Foundation’s Technology Working Group started the development of a new complementary physical layer that could be compatible with the existing FSK physical layer and therefore could coexist with the ample installed base of HART devices all over the world.

The result of this efforts received the name of Coherent 8-way Phase Shift Keying (C8PSK) physical layer which it was presented by 1999 and included in the HART Communication Protocol Specification number 6.0.

Enter the C8PSK physical layer

The C8PSK Physical Layer is backwards compatible with the original FSK Physical Layer and with the current loop signaling concept. It employs current phase changes to encode the digital information in the 3200 Hz carrier. But, instead of using one symbol per baud, like FSK, it allows for the transmission of three bits per symbol, thus enabling the protocol to work at a data transfer speed of 9600 bps without the need of different wiring.

3 C8PSK encoding waveform

Most interesting is the fact that no modifications were required in the HART protocol, except for some extended commands created to address the specific requirements of C8PSK devices.?

C8PSK allowed HART devices to send 10 to 12 updates per second instead of 2 to 3 in standard Host/Device mode and from 15 to 18 updates per second instead of 3 to 4 in Burst mode.

Why is that most HART technology presentations and demonstrations do not mention this feature??

The easy answer is that the members of the HART community did not consider this technology enhancement significative enough. They may have seen that its adoption required too much work for too modest an improvement. Tha lack of C8PSK HART field devices gave little incentive to control system suppliers to incorporate support for the C8PSK physical ayer into their offerings.

I have an additional and personal point of view:?

It is a matter of simplicity. The FSK encoding can be easily represented in an XY graph that shows the current carrier signal and embedded into it, the two frequencies involved in the encoding. It is also obvious that the FSK signal does not affect the current signal because FSK is a sinusoidal wave, so its average value is 0.

A matter of perceived complexity

In comparison, C8PSK its much more complex, and therefore it cannot be easily read by just watching the signal form. C8PSK encoding uses 8 symbols, each carrying 3 bits of data and associates each symbol with 8 phase changes. Each phase is separated by 45 degrees starting at 22,5°.

Since this can be seen clearly in a constellation type graph, but cannot be seen in a 2d representation, my point is that C8PSK encoding required a deeper understanding of digital signal encoding methods than FSK. You are not willing to adopt something that you cannot understand.

For sure it was not a matter of costs since, according to the available documentation, the replacement of a FSK modem with a C8PSK modem had a cost of under 10 USD. Considering the widespread availability of HART devices, the most probable outcome would have been that the extra cost would either have been covered by the manufacturers once economies of scale had been achieved.

4 Table of Gray Code Phase Assignments in C8PSK encoding

The lack of adoption of the C8PSK physical layer can be demonstrated by the number of versions of the current specification: Just the original 1.0.

?5 C8PSK?Phase Amplitude Constellation graph

The C8PSK encoding standard has not changed since its initial release. And that is a bad sign, even more when you look at the current version of the FSK specification, which is 9,1.

It required too much work for a modest improvement: too little, too late.

And what about multidrop long form addressing

There is another better-known enhancement that the HART Protocol received in version 7, the maximum number of devices that can be connected in a single multidrop line.

This number, which was traditionally limited to 16 devices per multidrop line, became extended to 64 with the release of the Hart specification version 5.

This substantial increment was achieved by the incorporation of the long form address (5 bits) as an extension of the traditional short form address (4 bits). With the traditional short form address, since 4 bits are available and address 0 is not used for networking, addresses 1 to 15 are available. With the long form address, 5 bits are available for this purpose, therefore addresses 1 to 63 become possible.

Again, why is that the extended addressing range is not used in typical Hart applications?

The answer is the following:

Every HART device can manage 4 HART variables, named in a rather unimaginative way as primary, secondary, tertiary, and quaternary.

When a HART device address is set to 0, the device transmits the primary variable through the 4-20 mA loop, and the remainder field variables are available as digitally encoded information sent in either FSK or C8PSK encoding.

But when the device address is different from 0, the HART’s multidrop mode is enabled and the system’s current becomes fixed, usually at a value of 4 mA. Afterwards, the systems master starts polling the slave devices sequentially. In this mode, all four HART variables are transmitted digitally, and due to the daisy chain wiring employed in these applications, a significant amount of cabling can be saved.?

In multidrop mode, the availability of the devices included in the daisy chain depends on the integrity of all the connected devices and the corresponding cabling. If one fails, the daisy chain is gone.

This would be quite a problem for a 16 devices daisy chain, and a really serious one in installations using up to 63 devices. Timing issues may become problematic because the data transfer speed remains fixed at 1,2 kbps either when using FSK encoding or 9,6 kbps if there were any cases of C8PSK encoding applications.

Additional problems include grounding, cable length, and EMC issues, among others.

This is the reason most RS-485 HART multiplexers were designed to be used as point-to-point interfaces. Also, it is the explanation for the use of the traditional short form addressing scheme in these devices.

Why these stories matter in the HART-IP and Ethernet-APL era

The HART communication protocol has received a significant boost in importance due to the development of HART-IP technology and, more recently, from the upcoming Ethernet-APL (Advanced Physical Layer).

HART-IP offers the possibility of embedding the traditional HART protocol into an IP message, which can therefore be sent either via TCP or UDP to the corresponding addressee.

With HART-IP, Hart data can be transferred at up to 1Gpbs data transfer speeds. That level of improvement makes the 10x increase offered by C8PSK look definitively paltry.

HART-IP opens new applications for the use of HART devices, initially via the use of HART-IP capable multiplexers.

HART-IP capable multiplexers should really be considered as HART to Ethernet Gateways, because they do not execute sequential polls through the field devices connected to them. They can do simultaneous, in parallel, querying operations to their connected devices, thus offering a dramatically improved response time. And they set the HART protocol free from the limitations of RS-485 HART multiplexers by offering standard Ethernet data transfer speeds, which, in typical industrial applications range from 100 Mbps to 1 Gbps.

Ethernet-APL enables the use of TCP/IP networking to the field level, making the installation and commissioning of an Ethernet APL field device as easy as the same operations using a 4-20 mA field device.

In the presentations of Ethernet-APL technology, there has been an emphasis in stating that, to be successful, Ethernet-APL will have to be as simple as 4-20 mA.

The users concern

My concern is that the same argument was used in the PROFIBUS DP/FOUNDATION Fieldbus era, until it became obvious that physical layer field diagnostics were necessary, a least for commissioning and maintenance work.

The solution came with the development of fieldbus physical layer diagnostics devices, which became ubiquitous in medium to large IEC 61158-2 fieldbus installations.

The suppliers of these devices usually included a digital oscilloscope in their products since common physical layer disturbances could be detected by the waveform of the fieldbus signal.

Instrumentation engineers became familiar with this feature, and a significative amount of certified training time was very much focused on the understanding of the fieldbus physical layer and the Manchester encoding scheme.

Ethernet APL is different, even more if used with HART-IP

But Ethernet-APL employs a much more sophisticated and complex encoding method than FSK or C8PSK. And since it uses the standard TCP/IP stack, addressing works automatically.

Let me ask you a question: when was the last time you required a physical layer diagnostic of either your home or office (and most recently, your SOHO) Ethernet connection?

The answer will most probably be “never”.

That is one key advantage of Ethernet-APL. By extending the use of switched Ethernet to the field, its promise is to make the connection of an Ethernet-APL field device to a plant’s industrial network as easy as a 4-20 mA field device to a corresponding AI I/O port of your control system or connecting your laptop to an Ethernet network.

And you will not need to understand the underlying encoding method or the waveform of the Ethernet-APL signal, like you do not need to understand your trusty daily Ethernet connection.

Postscript

I will end this note with a question to all readers interested in these technologies: What expectations do you have from the latest HART, HART-IP and Ethernet-APL developments??

You can still make your voices heard and contribute to the upcoming “Ethernet to the field” revolution, which is expected to be in the main spot for the next years.

You can find additional information about the HART protocol and HART multiplexer technology following this link .

Mirko Torrez Contreras ?is a Process Automation consultant and trainer that started to HART a long time ago. After finishing this article, he felt identified with the classic Beatles song:

"It's been a HART day's night

And I've been workin' like a dog

It's been a HART day's night

I should be sleepin' like a log"

Anyway, he was able to deliver this note at the end.

The opinions exposed in this article are strictly personal. No affiliation exists between the author and the companies mentioned. All the information required for and employed in this article is of public knowledge.

Phoenix Contact ?sponsors this article. The opinions exposed in this article are strictly personal. All the information required for and employed in this article is of public knowledge.

?