|ARM Technical Support Knowledge Articles|
During AMBA test output signals other than those used for test purposes are not guaranteed to remain in a safe state.
During AMBA test the output signals of the various processors will transition depending on the test currently being run. Signals other than the AMBA test signals may transition and enter states which may alter the systems behaviour.
Signals of particular note are the TAP controller output signals such as TAPSM, SCREG and the decoded boundary scan signals for scan chain 3. It is possible for the TAP controller to be tested and for the TAP controller output signals to enable the system or boundary scan chains.
If boundary scan is implemented on the device under test and the boundary scan cells are controlled by the ARM TAP controller then it is possible for the boundary scan chain to be selected by the test controller. This would result in the boundary scan cells switching from functional mode to scan mode, thus causing the AMBA tests to fail.
Signals which will affect system behaviour can be suppressed by combining them with the AMBA grant signal for the test interface controller. If the ARM output signal is required to remain high then they can be logically anded with the grant signal, otherwise the grant signal can be inverted and logically anded with the required signal.
An alternative approach is to use a D-type flip flop with its D input tied to a logic 1, the reset connected to the JTAG reset signal nTRST and the clock input connected to the JTAG clock TCK. The output of the flip flop can then be used to disable the ARM TAP controller outputs when low, and enable them when high. Thus at reset all TAP controller outputs are disabled and if AMBA test is used then they remain disabled as TCK will be static. However, in the event of any JTAG activity the flip flop will be clocked and all TAP outputs will become enabled.
Article last edited on: 2008-09-09 15:47:38
Did you find this article helpful? Yes No
How can we improve this article?