There are several options, but I would recommend using the already existing file as your data buffer. This will replace the internal data structures (queue or functional global) for the main data, not the decimated preview. I would do it as follows:
- Replace the number of points front panel control with a path control called Filename. This will allow you to specify your file.
- Modify the command enum so Change NPoints is Change Filename (only needed for us humans).
- In the Value Change event (top loop) for Filename (create one if it is not already there, delete the nPoints one if it is still there), stuff the command queue with Change Filename.
- Create a local shift register in the bottom task handler loop (similar to the one in the traditional memory store and browse) to hold your file reference.
- In the Change Filename frame of the task handler, close the file reference (if it is valid) and open a new file reference using the TDMS functions (use the file name terminal to read the file name).
- Replace all instances of reading or generating the main data with the corresponding TDMS function using the new file reference. For example, in the Change Filename case, replace the Sine Wave.vi with a TDMS Read.vi which reads the corresponding subset. Make your chunk size 65,000 points for best speed.
I have never done this with TDMS (only because I stopped working with file I/O before TDMS was invented), but have done it with raw binary, NI-HWS, and HDF5. All worked well, so you should have no problems.
