Conventional installation of CAN networks allows the line as only possible wiring topology. Stub lines, if they are used, are very restricted in length. Tree or star structures are in general not possible. This fact results in unnecessary long cables and hence wasted money. In addition, later changes can hardly be realized without drawbacks regarding the maximum speed. While line structured systems are still easier to install than ring structures, a tree or star topology would in many cases better fit the application requirements.

For this reason it is advantageous to free CAN installations from the restrictions of line topology. An elegant solution can be realized by the use of protocol transparent CAN repeaters. These conserve the benfits of the CAN protocol while breaking the restrictions of high frequency line design.

An example: In a production line individual stations are connected via CAN. The backbone of the network is installed in a cabling channel running besides the production line. The length of the cable from the backbone to the outermost CAN node is assumed to be 15 m, a length that is easily reached, if distributed nodes are attached. Because of the structural requirements in a conventional net the cable is pulled back to the cabling channel. With this design each loop adds a cable length of 30 m.

Lets assume that seven of these stations with a length of 20 m between two branch points on the backbone are installed. The resulting cable length can be calculated: 2 x 15 (outer stations to backbone) plus 6 x 20 m (between backbone branches plus 5 x 30 m (loop at inner stations). This totals to a length of 300 m with an approximate cable propogation delay of 1650 ns.

On the other hand consider the same application, but with use of one CAN repeater at each branch point on the backbone. The backbone is now directly connected from the first of the last node in the network. The cable consists of 2 x 15 m (outer stations to backbone) and 6 x 20 m (between backbone branches) which results in a length of 150 m. Added are five branches with one repeater and 15 m of cable each. The total installed cable length is now 225 m. For timing considerations the maximum propagation delay between any two nodes has to be calculated. Using fast repeaters built with todays standard transceivers the internal delay can be assumed to be equal of that of 25 m cable. In this case the longest delay in wiring is between the outer nodes on the outer branches connected with repeaters (critical path). The equivalent length is 2 x 15 m (outer stations of backbone) plus 2 x 25 m (equivalent repeater delay) plus 4 x 20 m (backbone) totalling to 160 m with a delay of 880 ns.

The installed cable length is 300 m in the standard CAN network compared to 225 m in the repeater solution. Assuming costs of 10 Euro/m for installed cable (the reader may insert his values, which should consider cable and installation cost) the savings in cabling costs of 750 Euro. From this savings the cost of 5 repeaters (about 80 Euro including power supply) has to be subtracted: the resulting savings are 350 Euro.

The propagation delay of the wiring is approximately 1650 ns and using repeaters about 880 ns. Due to this fact the maximum possible data rate is almost doubled.

But don't forget the restrictions. Repeaters used in simple line topology decrease the maximum bitrate due to their internal propagation delay. Therefore benefits, both in cost and speed, can only be obtained from a modified wiring structure. As a general result it may be said that the profit of repeater use is high in large networks with geometrical layout that deviates from a line. Besides large production or transportation systems this structure can be found in storage systems or in building automation systems.