This issue of VME and Critical Systems magazine is dedicated to the 30th anniversary of the introduction of VMEbus. We reached out to the VME community to seek comments on the highlights of the past 30 years of VMEbus. To get the ball rolling, we asked the members of VITA the following questions:
- Which enhancement(s) to the VMEbus specification had the biggest impact?
- What was the most unusual or strangest application using VMEbus that you ever saw?
- What was your most difficult design challenge with VME products?
- Which time period in the history of VMEbus was most influential?
- What would your company have done differently regarding its strategy in the VME market?
- What was the most influential product over the past 30 years?
We received a mixed bag of responses with some very interesting insights. I had the pleasure of being in the middle of the VMEbus industry from a front-row seat at the Motorola Computing Group during this period. Motorola was the recognized leader in VME market share for most of the past 30 years, according to VDC Research. I had the honor of being responsible for the marketing of many very successful products during those years. My comments are peppered throughout the responses that I received.
Which enhancement(s) to the VMEbus specification had the biggest impact?
Consensus on which enhancements to the VMEbus specification had the biggest impact reached common ground with everyone. VME64 and VME64 derivatives were mentioned by nearly everyone. VME64 was the first major demonstration that VME could evolve and yet remain compatible with previous generations of products. Over time, this became the single most important requirement for any advancement to wear the VME badge. This same evolutionary migration was key in many of the decisions that drove the development of VPX in more recent times.
“Revision D, the addition of D64 (VME64), demonstrated the power of backwards compatibility and the ability of the architecture to innovate without creating discontinuities in the market or rendering existing products obsolete,” states Shlomo Pri-tal, Chief Technology and Strategy Officer at Emerson Network Power, Embedded Computing. Pri-tal was very influential, and his powers of persuasion paid off in making backwards compatibility a prime consideration. We had many long discussions on this topic at Motorola because we had a huge investment in VMEbus to protect.
“On VME, the most important specification enhancement was the evolution to VME64x and eventually to 2eSST. The other major impact was VITA 46 (VPX), which kept the VME form factor but replaced the connectors with high-speed ones and added serial fabrics to move data,” remarks Steve Edwards, CTO for Curtiss-Wright Controls Embedded Computing.
“VME64x, because of the 2 mm P0 connector extension and GEOID (geographic addressing),” says Vincent Chuffart, Product Marketing Manager, Kontron.
Michael Munroe, Product Specialist at Elma Bustronic Corporation, mentioned four enhancements that had a big impact: VME64x, PMC, VPX, and Force Computers’ electronic bus grant daisy chain patent. He chose VME64x for the addition of 3.3 VDC signaling, shielded front panels, and the addition of a P0 connector. PMC has become the de facto mezzanine interface for VME and other 6U form factors. VPX gives VME a breakthrough revolutionary migration path for complex systems of the future with hybrid backplane bridges. Force Computers’ electronic bus grant daisy chain patent simplified a challenging problem for VME system designers: Early-generation products required jumpers on the backplane to handle bus grants. This was difficult and error prone, requiring access to the backplane to make changes and leading to errors with missing or misplaced jumpers. The first question asked by technical support was to query you about how the backplane was jumpered, because, most often, a jumper was missing or out of place.
What was the most unusual or strangest application using VMEbus that you ever saw?
VMEbus has found homes throughout our solar system. The list of design wins could go on forever. Here are few of note:
“Without a doubt, the most unusual VMEbus application was the VME technology deployed on the Mars Rovers, Opportunity, and Spirit,” shares Edwards. The Mars Rover Opportunity is still taking photos, sending back data, and roaming the surface of Mars today. Designed for only 30 days of service, it has logged more than 20.86 miles and more than 2,700 Martian days. VME technology actually made its first trip to Mars in 1998 after IBM and Lockheed Martin Federal Systems developed a radiation-hardened VME system that was selected for the Sagan Lander.
“An area where we have seen significant growth in VME is the railway market. While it may not be a ‘strange’ application, it is definitely an unanticipated migration from the harsh environments of avionics to the mobile world of trains and railway systems. The triple-redundant A602 VME single board computer is especially useful in several areas of railway applications, including signaling systems, wayside control, and automated train operation, since failure may result in loss of human life (trains derailing or colliding). In response to this increasing use of VME, MEN Mikro is now preparing a railway certification package for the A602 for safety levels SIL 3 and SIL 4, anticipated to be available at the end of the first quarter in 2012,” explains Barbara Schmitz, Chief Marketing Officer at MEN.
“GDCA manufactured the VME-based Heurikon V4F boards for a company [that] used them in two different and interesting applications. One of them was a screening device that was used by airport security to detect explosives in passenger baggage, and the second one was a precision wood saw mill. It was a good thing they never accidently switched the applications, although some airline luggage looks like they might have done so,” quips Martin Plotkin, founder of GDCA, Inc.
My two favorite applications for VME were as control systems for the “Tower of Terror” at Disney World and in the launch control systems for the Space Shuttle program. It often made me wonder if the two organizations shared design ideas! Other interesting applications were the VME-based voting system in the parliament of China, the Chyron text overlay system for television, and many telecom digital switches.
VME was first used in many industrial applications long before it gained such a stronghold in military applications. Recently, I was driving through New Mexico past the Very Large Array astronomical radio observatory where I saw several of my Motorola VME boards in a dark and dusty lab processing data for the array. Many of these applications are event driven, whereas VMEbus excels over today’s switch fabric architectures that favor data-driven applications.
What was your most difficult design challenge with VME products?
“Because VME is limited in terms of power in a rugged environment, the most difficult challenges involve designing 70-90 W VME processor cards. VPX is more forgiving in this regard since it allows 1" pitch and will allow cards to draw more power from the backplane,” specifies Edwards. Power has always been a challenge, and it was not unusual to see two- and three-slot boards just to get enough power for a single slot.
“Parallel PCI bus extension through P0,” answers Chuffart. Bridging PCI bus has challenged engineers for years as PCI bus became much more than a local processor bus.
“GDCA provides the long-term support for the MVME162 board from Motorola for their end customers. In 2008, we realized that one of the key ASICs on that board was obsolete – with no availability in the market. This was a critical issue for customers since the redesign of their applications was going to have a major impact on their costs and lead time to their customers. GDCA replaced that ASIC with a functionally equivalent FPGA to ensure that, except for a layout change to accommodate the new footprint, there would be no impact on the board design,” details Plotkin.
Thank God for human memory. The early days of VMEbus had many challenges that failed to make the list. I’m pulling this from my own fading memories, but who can forget these issues?
The moving target of the first two or three revisions of the specification caused some conflict. From the introduction in October 1981 to sometime in the 1985/1986 timeframe, the specification was a bit dynamic. Thousands of copies of the specification were printed and passed out, only to have to be redistributed months later. I personally carried boxes and boxes in my car, at times giving up to five seminars a day on VMEbus to rooms packed with engineers. Luckily, there were no major changes and board designers were able to move forward with hundreds of new designs.
The lack of bridge chips to simplify the bus interface was a challenge. Early VME boards could easily use 30 percent or more of the board space for a VME interface. Few could afford ASICs, and FPGAs were yet to be practical. A whole battle ensued for several years in the late 1980’s over who would do a chip and how it would be marketed. In 1987, at the invitation of Joe Ramunni (Heurikon), 12 VMEbus manufacturers met to explore the interest in collaborating for the design and development of a comprehensive VMEbus interface chipset. It was years before anything came of this effort. In the meantime, Force Computers, Motorola, PLX, and others developed their own bridge chips. A VITA consortium developed the VMEbus Interface Chip (VIC), which was marketed and sold by VTC. Eventually Tundra Semiconductor emerged with the Universe chip to save the day.
Another major faux paux in the VMEbus specification was the lack of a defined software architecture to move data along the bus. Being direct memory architecture, this was not considered a big issue in the early days, but as real-time operating systems emerged and the number of devices on the bus exploded, writing device drivers kept a lot of software engineers gainfully employed. Architectures today enjoy the Ethernet, PCI Express, and Serial RapidIO protocols supported by serial switch fabrics and PCI bus.
Which time period in the history of VMEbus was most influential?
When asked this question, the responses were all over the map, and rightfully so. Most of it is dependent on the individual’s perspective and history with VMEbus. Pri-tal was particularly fond of the early days. During that time, his name was synonymous with VMEbus, and many consider him the Grandfather of VME, myself included. Thus his answer to the question is: “The period immediately after the formation of VITA and the establishment of its technical committee as an open forum for all who joined to develop the architecture and drive the industry.” Pri-tal really pushed the open standards concept and fought to prevent trademarks and patents that would slow down the market acceptance of VMEbus or show favoritism to any one company. That sentiment is still felt today in the VITA tagline of “Open Standards, Open Markets” and in its industry-leading ex ante policy.
Other influential periods occurred in the history of VME. Companies focused on the military markets were especially impressed with two specific time periods. “There were two critical periods in the history of VME. The first was in the early ’90s, when VME first started to gain use within military systems, which today is the major market for VMEbus and its derivatives. The Commander Perry memo of 1994 introduced the NDI initiative, which later morphed into the COTS initiative encouraging more use of off-the-shelf products. Later, between the years of 2003 and 2005, the VITA community made the decision to move beyond VMEbus and design a follow-on spec, VPX, which would support serial fabrics. We have seen VPX take over in the past several years, to the point where it is the major bus architecture for new deployed military programs,” details Edwards.
Another colorful period for VME was in the mid 1990’s as the PowerPC architecture emerged and found solid homes on VME single board computers. Chuffart mentions “PREP architecture: PowerPC + PCI bus + VME + Linux. Most current deployments of VME still have this architecture.” PREP became the hardware and software definition for anyone developing PowerPC-based hardware.
The Motorola 88K processor was to be the evolutionary path for the 68K processors, but it had few supporters. When Apple, IBM, and Motorola teamed up to develop the PowerPC Architecture, VME suppliers quickly embraced the technology. Ironically, the Motorola Computer Group was one of the very last companies to fully get on the PowerPC bandwagon. This was because of internal political struggles and a solid family of 88K single board computers that ranged from a single processor to eight processors on a VME computing platform. Once Motorola cleared the internal issues, it was full steam ahead on PowerPC until the AIM alliance fell apart. Fortunately, we had a skunk works team working on a PowerPC project that we were able to launch in a few short weeks.
Munroe points out the years from 1985 to 2001 until CompactPCI came along and gave VME a serious run for its money. Using the same 6U form factor but the PCI bus as the interconnect, CompactPCI was able to better leverage this new parallel bus and eliminate extra bus bridges, thus simplifying designs and improving performance. During those years, we saw all types of attempts to improve VME performance: VXI, RaceWay, SCSA, Autobahn, and 2eSST were but a few of the alternatives developed for VMEbus.
What would your company have done differently regarding its strategy in the VME market?
Most companies were very happy with the strategies that they took. I did enjoy Chuffart’s comment: “Keep 68K products alive a lot longer.” Kontron is an amalgamation of many smaller companies, many of which enjoyed years of success as VME suppliers, most with 68K-based products. I certainly understand the strategy of holding on to 68K processor-based products for as long as possible. VMEbus was built around the 68K processor bus, and as a result, worked best with products built around the 68K architecture. Everything else always required a little extra bus logic to work, and then there were always issues. Who can forget the endian byte order issues between the Intel x86 little endian and the Motorola 68K big endian?
The 68K processors had an average of about three years between major releases from 68000 to 68010 to 68020 to 68030 and finally the 68040. That timeframe was long enough to complete a board, get a few solid design wins, and make some serious profit from the design efforts before moving on to the next generation. Many 68K products are still in production today as evidence of the long life cycles of the 68K generation of VMEbus products. Fifteen years or more was not at all unexpected.
Plotkin still remembers the day he organized the first VERSAbus user’s group meeting – prior to the existence of VME. It was during that meeting that some of the key sponsors of the VERSAbus user’s group left the room and came back later to talk about the birth of the VMEbus standard – VERSAbus in Eurocard format. It is interesting that GDCA now supports the legacy VME products that were originally manufactured by Motorola and others. So Plotkin has not only witnessed the birth of VMEbus, but the company he founded has developed the product legacy management expertise that supports these VME products and keeps them alive as long as customers need them.
The Intel and PowerPC architectures rolled out so much faster, and board designers were always playing catch-up to the latest-generation processors. It has slowed a bit in the past years, as the Intel Architecture has become the dominant processor on new products. Five- to seven-year product life cycles are more typical today, causing some applications to have difficult life cycles to manage.
Many companies tried to make VME a system-level strategy. Motorola was the most successful with a large number of personal and server computer families based on VME technology. This gave them a tremendous market advantage because they were one of the few companies that developed a complete system, allowing them to better understand how a full system worked as other boards were integrated into the backplane. Mercury Computer Systems had some system expertise, but in those days VME was a front end to their much-larger processing arrays.
Today, many companies are wishing that they had hung on to VME a bit longer. During these tough economic times, many applications are upgrading those very durable VME systems instead of doing complete replacements.
What was the most influential product over the past 30 years?
Responses to this question tended to favor the home team, as nearly every company claims to have at least one “most influential product.”
“Curtiss-Wright’s CHAMP-AV6 and VPX6-85: They were the very first VPX products to market and helped to validate the choices made in VITA 46,” remarks Edwards. From my perspective at Motorola, several of the 68K-based CHAMP products gave us a real run for the money.
“Our A602 is a 6U FPGA-based, triple-redundant 64-bit VMEbus SBC that employs a lock-step architecture, keeping software development at a minimum. The redundant lock-step system increases system reliability, so the SEU-resistant A602 runs the same set of operations in parallel to ensure that the programming only views the hardware components once,” explains Schmitz.
“The MVME147 … introduced revolutionary levels of integration and completely changed the industry. As an example, no more memory cards, Ethernet controller boards, etc. The only products were processor boards (e.g., GP, DSP, etc.) with their required complement of I/O,” says Pri-tal.
“The ABG connector, first from Augat,” answers Munroe.
In my many discussions with customers and suppliers over the years, one product in particular continues to quickly come to mind: the MVME147 mentioned by Pri-tal. I have a particular closeness to this product, as it was my first major product in my role as VME Product Manager at the Motorola Computer Group. It was a true VME single board computer. Though not the first in the industry, it was the most integrated and most popular. It used an internally developed ASIC for the VMEbus interface, saving a substantial amount of board space that was then dedicated to other functionality. We often promoted it as the equivalent of five other VME boards: a processor, memory, Ethernet controller, SCSI bus controller, and general purpose I/O board. Originally it shipped with 4 MB of DRAM and an option to add 4 MB more on a full-size 6U board. Today you can’t even buy a single DRAM chip that small. More than 1 million copies of that board were shipped, and it experienced five major redesigns over its active production life as components went obsolete, especially DRAMs. The MVME147 also set new price points for VME boards and even shipped with a lifetime warranty. The MVME147 led to the MVME162 and MVME167 that also reached similar shipment milestones. Hooray for 68K!
VME still a leader
VME has had a very colorful and profitable run in the embedded computing industry. It continues to enjoy a sizable market with the introduction of new technologies like VPX, which give VME a great migration path. I expect to see products with their roots in VMEbus for many years to come.
A big “thank you” goes to the following for their contribution to this article:
Steve Edwards, CTO, Curtiss-Wright Controls Embedded Computing
Barbara Schmitz, Chief Marketing Officer, MEN Mikro Elektronik GmbH
Vincent Chuffart, Product Marketing Manager, Kontron
Michael Munroe, Product Specialist,Elma Bustronic Corporation
Shlomo Pri-tal, Chief Technology and Strategy Officer at Emerson Network Power, Embedded Computing
Martin Plotkin, founder of GDCA, Inc.