Innovating Technology

A serious shortcoming in off-the-shelf PHYs – from the data monitoring point-of view – is the “intrusive” nature of regular PHY silicon. Commercial PHYs are architected to provide both receive (RX) and transmit (TX) functionality since regular bus devices want to receive as well as send data. Therefore, it is a requirement that these devices fully participate in the tree identification, bus enumeration and become real participants on the bus. This, however, is not necessarily the case for pure “data monitoring” devices. If the goal is to provide only data monitoring (listening) functionality, it is actually desired that the “monitor” is not part of the bus as it influences and changes the system under test. The dynamic nodeID enumeration in IEEE1394 would stay unaffected when “connecting” a non-intrusive monitor.

The figure below demonstrates the typical problem. A number of IEEE1394b-2002 devices are connected via 1394b connections. The addition of an additional device is shown in the middle/right graphic. All devices’ nodeIDs get re-enumerated as a new tree-identification process reconfigures the bus. In this example the bus monitor ends up at the bottom of the tree with the lowest nodeID. When added to the bus via a different connection location there is fair chance that the newly added device wins the arbitration and becomes the root node (device with the highest node ID). Only in this case all other nodeIDs stay the same. However, it must be pointed out that for most testing applications this is a highly undesirable configuration. In IEEE1394 the root node has a very specific status with some important bus supervision functions. Having the monitor be the root and having it take over these responsibilities is a severe change of the system under test and bears very little resemblance to the original configuration.

Another issue is the input/output buffer delay when using regular PHY silicon. PHYs work as signal re-conditioners and re-transmitters. Signals or data received on one port are actually reconditioned and retransmitted with a delay on all other connected ports. From a pure data analysis / network optimization point-of-view, this is an undesired feature for an “analyzer” and should be avoided.