Often times it is recommended to combine all devices on a board to one Boundary Scan chain. However, there are applications where it is beneficial, if not even required, to split up the devices into two or more chains.
For example, certain debugging or emulation tools may require a target device (the device to be used in debug or emulation mode) to be the only device in a chain. If that is the case, one should separate that device from the others and provide an independent Boundary Scan chain for it. Or, as an alternative, one could design the Boundary Scan chain to be reconfigurable, so that this particular device could be separated from the others through additional scan chain control circuitry (dedicated devices such as Scan Router devices, or ad-hoc circuitry, such as multiplexers). Another example, where multiple scan chains are useful is a Boundary Scan application that requires many shift cycles and the IEEE 1149.1 compliant devices on the Unit under Test support a wide range of maximum TCK frequencies. FLASH programming applications based on Boundary Scan access, for instance, can require millions of shift cycles through the scan chain. The programming time is directly related to the maximum TCK frequency applicable to the scan chain. If the FLASH device is controlled by one of the faster devices in the scan chain, then this device (and possibly other devices rated for the same or a higher maximum TCK frequency) should be kept in a one scan chain, while the other, slower devices are combined in a second, independent scan chain. This can help to dramatically reduce FLASH programming times.
