At last year's Hannover Fair in Germany, Interest Group Sercos (IGS) announced an upgrade to the next generation of Sercos interface. The biggest change for anyone used to working with the interface is the move away from optical fiber to an Ethernet physical layer.
The biggest reason for the move to Ethernet is cost. The Ethernet hardware layer lowers the cost/node. Until now, Sercos used a standard chip designed by the Sercos group. The new standard can use field-programmable gate arrays (FPGAs) or communication controllers that cost half as much. In addition, the cost of cabling and particularly connectors will also be less than half that of optical connectors. And doing away with fiber-optic couplings also helps contain costs.
The other motivation for physical Ethernet is the availability of the IP layer for programming. So instead of plugging a specific Sercos interface card into a PC, users can now use a laptop with an Ethernet port to set up drives.
Sercos III will be backward compatible with older versions of Sercos in topology, profiles, synchronization, and message structures. It also lets slave devices communicate. Plus, it adds fault tolerance in case of a break in the network ring. Currently, the Sercos protocol has a service channel that can be used for transferring data for communication as well as for parameters or diagnostics. Keeping the service channel makes Sercos III downward compatible. Plus, an optional IP channel transfers real-time or nonreal-time data using standard Ethernet frames.
Sercos has traditionally been known for solid real-time performance. Instead of implementing Sercos on top of Ethernet's TCP layer, Sercos is the baseline with an IP layer on top of it. Although its hardware protocol is Ethernet, everything from the data-link layer on up is written specifically for Sercos. This preserves the high performance on the twisted-pair protocol.
In a typical TCP/IP stack, there is significant overhead in the form of error handling, data transmission and reception, and establishing connections. Sercos has stripped out quite a bit of this overhead and so doesn't waste bandwidth. The raw throughput of Sercos III will be the ordinary 100-Mbits/sec Ethernet rate. The increase from 16Mbits/sec, the rate for second-generation Sercos, boosts the available bandwidth and increases the number of hardware nodes.
The new Sercos halves the minimum cycle time of the current spec from 62.5 to 31.25 µsec. The increased data rate of Ethernet means more bandwidth, so it's still possible to connect more slave devices despite the shorter cycle time. It's now possible to close current loops in the drive using Sercos, as well as velocity, position, and interpolation loops. Despite the ability to close current loops with Sercos III, higher forms of loop closure allow higher performance when using intelligent digital drives, as the processing expanded with every drive axis added to the system. With current loop control, the machine performance is limited to the control's processing capability, whereas in higher forms of loop closure, this power is shared between the control and drive processors.
An additional feature of Sercos III is the capability of redundant, self-healing connections. Traditionally, Sercos had a ring topology, which required a transmit and receive for every device, with the last device required to be closed back to the controller. Now, this last cable is optional; a line protocol starts off at the master device and can emulate a ring in a single cable with a separate transmit and receive pair. For instance, in a packaging line several hundred meters long, the Sercos III bus can run down the length of the whole machine without having to return to the control. Closing the ring not only provides a redundant connection path to the control, but it's also self-healing; if one cable is cut, the protocol automatically switches over to line configuration without interrupting communications. Another advantage to the Sercos III protocol when compared to other Ethernet-based solutions is that it doesn't require costly switches to distribute the protocol or guarantee performance. The only hardware required is already factored into the estimated 12 Euro/node cost.
One concern with the move away from optical fiber is noise immunity. Joe Biando of Bosch Rexroth, Hoffman Estates, Ill., thinks noise won't be a problem. "Twisted-pair cable is generally extremely noise immune because it uses a differential driver and receiver," says Biando. So any noise that shows up on the one half of the cable shows up in the reverse polarity on the other side of the cable, effectively canceling out. Theoretically, there will be more noise in the twisted pair than in the optical cable simply because of the copper conductors. But in practice, the difference shouldn't be great.
The first prototypes of Sercos III hardware should be available this year, with products set to roll in 2005.
The origins of Sercos
In the 1980s, the German Electrical and Electronic Manufacturer's Association, or ZVEI, and the Society of German Economic Advisers initiated a consortium to specify an open digital interface for digital drives. Sercos was the result of this effort. The first generation of Sercos had transfer rates of 2 and 4 Mbits/sec and mainly handled advanced machine-tool applications. The interface expanded worldwide and became an IEC standard in 1995.
The second generation followed in 1999. Transfer rates increased to 8 and 16 Mbits/sec and the service channel for the transfer of nonsynchronous data was expanded. This technology has been available since 2001 in the Sercos816 ASIC with downward compatibility to the first generation.
Sercos features collision-free data transmission based on a time-slot mechanism together with an efficient communications protocol to ensure determinism. Up to 40 axes can be synchronized over Sercos with a cycle time of 1 msec and jitter less than 1 msec. Other classic features include hardware-based synchronization, efficient protocols, and transfer of real-time and nonreal-time data. Plus, Sercos communication mechanisms don't need a great deal of computing power, so the application processor can fulfill these tasks.
Sercos III performance parameters
|8 bytes||31.25 µsec||8||Torque command, actual position|
|12 bytes||250 µsec||70||Speed command and actual value,
position command and actual value
|32 bytes||1 msec||150||Numerous command and actual values|
|16 bytes||1 msec||240 (max. #
|Numerous command and actual values|
|For 8-byte data, Sercos III cuts the cycle time to 31.25 msec, half that of the previous generation of Sercos.|