What is a TAP Integrity test?

This test is created automatically, and in most cases will work without debug. This test verifies that the chain is operable, and that the scan path (shown in bold red in figure 1) is intact.

This is done by verifying the two least significant bits of the Instruction Register of each device which is mandated by the standard to be “01”, and which is captured during the CAPTURE-IR state.

Whenever possible, this test also verifies that the correct devices have been loaded onto the board by checking their IDCODES against a 32-bit value provided in the BSDL file, if provided.

Figure 1 : Boundary-scan TAP Integrity Test checks the TAP of each device

If IDCODES are not provided, the data captured by the Instruction Register during the CAPTURE-IR state is verified, which provides some capability for device identification.

So basically, TAP integrity test looks for the basic integrity of the devices' TAP controllers and the boundary-scan chain. Examples of problems that could be encountered include:

• one of the devices is bad: simply does not work
• a wrong device was loaded
• one of the TDI or TDO connections is bad (for example, an open between u3 and u4)
• TCK is disconnected from one or more devices

Integrity tests are an integral part of every chain test. Integrity test occurs before each powered shorts test, before each interconnect test, before each bus wire test, and before each connect test.

Any failure during TAP Integrity test causes the board to power down and a general failure message is generated. The message provides a diagnosis to the general area (for example, to a device), but it does not say exactly where the failure occurred. A failure could occur in a device upstream from the device that the message cites as the starting point for looking for the cause of the failure. Also, the diagnosis is limited in its ability to pinpoint the cause of the failure.

The type of failures detected during this test would prevent the chain itself from working properly: if the chain is bad, you cannot do meaningful testing. So the chain must be repaired before further testing can be done.