pass/fail and a game of snake-in-a-tunnel

Werner Almesberger werner at
Wed Apr 13 22:56:47 EDT 2011

Since the signal is affected by distance, reflections, obstacles,
etc., it can change quite a bit when moving the sender or receiver

We can used this to have a bit of fun. My example setup is the same
as described in the previous mail: two Bens, called "ben" and
"jlime", both running atrf-proxy and attached to a PC.

First, run atrf-path in GUI mode and move one of the Bens (or both)
around until the waveform looks interesting:

atrf-path -g -T -0.5 -p 0 net:ben net:jlime 10

Then hit "Q" to quit. Next, while keeping the Bens in the same
position, do a measurement without the GUI and dump the result to a

atrf-path -T -0.5 -p 0 net:ben net:jlime 10 >out

Check the result with the "plot" script:

./plot out

Use the genpathprof script to generate a profile. E.g., leave a
clearance of 5 dB above and below the reference spectrum:

./genpathprof +5 +5 out >tunnel

Last but not least, run atrf-path with the profile we just

atrf-path -g -T -0.5 -p 0 -P tunnel net:ben net:jlime 10 

The challenge is now to position the Bens again such that the
spectrum ends up in the boundaries. Here's what this can look

The yellow triangle indicates that the spectrum exceeds the
maximum at least at one point. A downward red triangle would be
shown whenn the minimum is crossed. Finally, a green disc would
mean that the spectrum is within the boundaries.

When the green disc appears, hit "P" for "Pass". atrf-path will
then exit, printing "#PASS".

The pass/fail logic works as follows:

- if the measurements are all within the limits (green disc)
  and "P" is pressed, atrf-path prints "#PASS" and exits. "P"
  is ignored in any other situation.

- if at least one measurement is above the maximum, this means
  that the result is "too good to be true" and something is
  wrong with the test setup. Maybe someone is walking past the
  table, a device is not in the right position, there is some
  interference, etc.

  atrf-path shows the yellow upward triangle and ignores all
  keys but "Q" until the disturbance has been resolved.

- "Q" always quits, but prints neither "#PASS" nor "#FAIL".

- finally, if at least one measurement is below the minimum
  and no measurement is above the maximum, the red downward
  triangle appears and "F" can be used to indicate failure.

The assumption here is that the test is made in an imperfect
environment and the operator has to decide when the measurement
is stable and free from external effects.

In a more controlled environment, the non-interactive mode of
atrf-path can be used. There, all results must agree with the
profile for the test to pass. The test fails if all sweeps have
at least one value below the minimum and no value is ever above
the maximum. All other cases indicate some inconsistency and
atrf-path makes no pass/fail decision.

The purpose of this test is to detect manufacturing problems,
affecting the RF side, e.g., components that are missing, not
soldered, shorted, etc. Devices with such problems would show
clearly reduced performance.

- Werner

More information about the discussion mailing list