I am setting up an online convolution workstation where you put in your voice or any other sound, select a space from a number of options and the server will convolve them and return the results. In making the Impulse responses for the spaces, I am limited by time and resources to the transient method, using balloon pops. I have recorded them. My understanding is that I need to remove the impulse (the pop) from the recording. 1) is that the case? and 2) How can I tell where the impulse ends and the reflections begin?
Theoretically, the "bang" is suppose to produce a perfectly instantaneous amplitude peak. If you convolve a signal with your full impulse response, you will get the original signal mixed with its reverb in the recorded space, which is what is expected usually.
But you can experiment with fading a very small part of the beginning of your impulse response and check out the result. This should be near to a reverb-only sound.