How to choose Bad Packet Ratio limit

[Tim (PassMark)]

Yes the error count as displayed by BurnInTest in the summary is the BurnInTest error count (not an individual packet error count) and only increased when the chosen bad packet ratio percentage has been exceeded.

The percentage value displayed on the test tile while the test is running is the current actual packet error percentage count (since the last BurnInTest error was raised).

[Charlie]

Dave
That would be great to have the percentage in the log, better than those errors. If understand correctly, the number of errors displayed is not the number of bad bytes or bad packets. It is the number of times the bad packet percentage exceeded the limit AND the software decided there were enough samples to statistically say the error was exceeded.
Thanks for looking into this for a future release.

[David (PassMark)]

Wifi errors will depend heavily on the local radio conditions (interference, antennas in use, distance to access point, 2.4Ghz or 5Ghz, etc..).

While the test is running the percentage error is displayed on screen. So you could just collect the information while the test is running.

After the test is complete you can get the number of errors and the number of operations ('operations' in this case being the number of bytes transmitted). From that value you can work out the percentage again.

But it would be better if we just logged the percentage value to the log as part of Level 1 or level 2 logging. So we'll look at doing this in the next patch release.

[Charlie]

We recently released a production test using BurninTest V9.0 Pro that includes a WiFi test. We use the standard network test pointing to the IP address of wireless router. We used 'bad packet ratio generates error' with a limit of 0.1% and timeout 4000mS. Now we are failing tests in the production environment, presumably due to increased interference. We need to set a new, more realistic error limit. I have 2 questions
1 - Is there a way to store / track the error percentage? This number disappears at the end of the test. I would like to run long tests to characterize performance.
2 - Is there some logical way to set this limit, other than running tests and tracking the error percentage with known good UUTs?
Charlie S.