Problems with Powered Opens on ICT

If you’re using Powered Opens on In-Circuit Test, you may
not be getting the test coverage you expect.

Known variously as Cover-Extend, Powered Framescan, and
ToggleScan, Powered Opens (hereafter referred to as “PO”) is used whenever test
points cannot be placed on nets to provide the stimulus required for capacitive
sensing to detect opens. Instead, boundary scan is used for the stimulus.

As test access on ICT is continually eroding based on
successive generations of denser, more complex board designs, PO seemed like a
good way of making up for lost test coverage. However, limitations of this
technology have rendered it largely ineffective.

The issues can be summarized in a handful of bullet points:

  • Long scan chains create a very low toggle
    frequency at the target cell, making capacitive measurements inconclusive (for
    frequency domain-based measurements).
  • The boundary scan cell toggle frequency will
    drift, due to non-deterministic scanning/handshaking between the PC running the
    application and the hardware itself (again this applies primarily to PO that
    uses frequency measurements).
  • Two opens on a differential pair will escape
  • Shorts on the connector side will escape
  • Target boundary scan devices need to be
    populated and often have their clocks running for PO to work, which creates
    cooling, noise and higher fixture cost issues.
  • Often, differential receiver cells (like clocks
    and RX pins) are input-only and/or not self-monitoring, so shorts coverage can
    be limited or non-existent.

There are other issues, but this captures the majority.
Other solutions intended to address some of the above, such as EXTEST_TOGGLE
and the recently ratified IEEE 1149.8.1 initiative, either introduce new
problems or await widespread adoption inside chips, which may be years away. In
the meantime, consider carefully before investing in PO – let the buyer beware.
There are better alternatives.

Alan Sguigna