Well you will need to run test scripts for every single application that your software takes care of, across all the major interfaces it will encounter.
Do you have a counter for example that cannot go abour 65536 or below 0? If so, you need a test to test both sides of the limits as well as random values within to see if your software conforms. If it goes out of spec is there an error ahndler that will give the user a clearly stated idea of the problem (or a imiter that wont allow the entry?). This needs to be tested.
Basically for every single function that your software addresses, you need to try fail testing it both high and low, as well as within tolerance.
Record all results meticulously. Write the testing procedure in such a way that the testing team can easily use Yes/No checkboxes for tests, as well as commetns sections after every set of related tests, and a place where you can sign and date after each applicable set of testing.
In the beginning of your testing protocol, you need to list the software, build, etc, and also outline the reasons for your testing and how you will go about it. This is also the place to put in serial numbers, listing test equipment used as well as their serial numbers, and calibration dates if required.
If you use Windows XP, 98, Linux, or whatever, they all need to be listed and signed off as working correctly by the SME prior to your software testing.
Finally, when you have a coherent document, have a place up front for those in charge to sign off against their name and title that this testing is preapproved for implementation.
At the end, create a certification page for post approval sign offs after all testign has been completed and passed. You may wantto attachan appendix in the end outlining any deficiencies you had that went against testing (both critical and non critical). Any critical failures in the software should be addressed before release. If time, prioritize the non critical deficiencies and take care of the highest priorities elft until you run out of time before release. Non critical defiicencies can still be present in this testing procedure that is signed off as complete, as long as a solution to address these problems in a later build is outlined.
I know that it sounds like a lot, and I know I have left out a few things, but that should get you startedd anyway.
Good luck!