1) The "sturm sequence failure" does not mean there is anything wrong with your job, rather that there is at least one more mode close to where you stopped extracting frequencies.
Take as an example a system where you set the input up to extract all frequencies below 95 hz. The extraction routine will stop as soon as a frequency is found that exceeds 95 - for this discussion, assume this is 100hz. How do you know there isn't another frequency very close to this 100hz that may affect your results? You don't unless you rerun, changing the frequency cut-off. However, a better way is to implement the Sturm Sequence Check, which determines how many frequencies exist between 0 Hz and some other value (not what they are, just how many). So we take your last frequency (100 hz in this case) and add 5%, which in this case yields 105 hz. We then use the Sturm Sequence Check to determine how many frequencies exist between 0 and 105 hz. If the count is the same as the number of frequencies just extracted (between 0 and 100hz), then we report that the check passed. If we get a higher count, we report that the check failed.
There are some models that will always fail. For example, consider a cantilever running in the "X" direction. For every mode in the "Y", there is a corresponding identical mode in the "Z". Regardless of what your cut-off frequency is, the extraction will stop at the first mode of any given pair. The Sturm Sequence Check, by adding 5%, will always find (at least) the 1st mode of the pair, and therefore report a failure.
2) You would want to use the higher frequency, because the results then include the response from more modes. However, you don't want to get ridiculous here either - the more modes you include, the longer the job runs.
3) In general, you don't know how many modes you need to include in the analysis. For one job 10 may be sufficient, for another you may need 50. This is unknown, so a better way is to say "give me all the modes below 'xxxx' hz." You need to include enough modes so that your results include at least 90% of the system response to the applied load. There are various "rules of thumb" to guide you indetermining this value, such as the discussion you referenced in the Applications Guide.
_________________________
Regards,
Richard Ay - Consultant