I don't see any race conditions in your sample VI. Dataflow looks continuous through the loop and there are no Global or local variables lurking about. The switch is set to Switch Until Released so it will stay on as long as the mouse is clicked on it. Completing the Since each iteration of the loop creates a new waveform and writes it to the hardware regardless of the switch, the clue to your problem lies in how the gaps get into in the data. I think there are two contributors to this. If the gap is a constant 0.1 seconds in length, then try removing the Wait(ms).vi since you have to wait for the AO to complete anyway.
If the gap remains or is greater in length, it could be because the two signals you produce are not the same length. If you run the DC only case, the sample has a duration of 0.1 seconds and contains 100 points. The appended waveform 's length varies depending on which waveform you select. For example, when Waveform 2 is selected with G=0 , the mixed case puts out a sample of 860 points spanning 0.860 of a second. This is the sum of the 100 samples generated in your original DC signal appended to 250 samples from Simulate Signal 2 and for each pass thru your Half Wave Blocker vi which add 102 more samples each. Try generating another 760 DC samples and appending to the original signal in the DC (False) case. Now your samples will be the same length.
If gaps remain, data production from your AO might run smoother if you move the AO Channel creation, timing and clearing outside the while loop. Take a look at the example Cont Gen Voltage Wfm-Int Clk- Non-Regeneration.vi for some clues as to how to do this. If you don't need to create a different reference array on each iteration, move that outside the loop as Yamaeda suggested.
JohnCS
