ZooZe said:
Actually I cannot tell much about the project yet. but here goes some short answers:
Our mics should be omni-directional and have low noise (maybe ECM8000) The analysis is made upon the time of arrival of the waves to different mics (about milliseconds/hundreds of microseconds differences). And there are a couple of sources in the environment that produce acoustic waves.
OK, a couple of things about microphones: small diaphragm omnidirectional microphones are universally used for measurement because their small diaphragm size gives them the best transient response and flattest frequency response. However, due to the small size of the diaphrgam, the signal strength is low relative to the self-noise of the capsule. On top of that, various microphone brands have circuit designs that will be quiet or noisy, which can make the situation somewhat worse.
So you will need to define an appropriate signal to noise ratio, which means we need to know your planned signal level, and the nature of the signal. Another thing to understand is that microphone self-noise typically manifests itself as random broad spectrum noise, usually "pink" noise, which is random noise that decreases in intensity by 3dB/octave.
This is important, because while you might see a noise spec that reads, for example, 24dBA (for A-weighted), and you measure your signal strength at a given point as 20dBA, that doesn't mean you will be unable to detect the signal over the noise, especially if your signal is a coherent high frequency wave. This is because the self-noise rating is integrated over a broad frequency spectrum, such that the noise at a given frequency might be more like 10dBA, for example. Thus, you should only have difficulty if you are trying to detect very quiet or low-frequency sounds.
Which brings up another issue, the environmental noise of your test lab. You will need to establish that as a baseline if you are trying to detect quiet sounds, because that could also create a noise floor.
Finally, you say that you are recording two sources; if the content of the sources is identical, then you will be unable to discern which source is which at an given point with a single microphone. If the distances from the two sources to a given microphone are different, of course the signal level would be different, even so, without knowing the nature of your experiment, I would suggest using different frequencies for the two sources such that they are easily distinguishable. Otherwise, you would need to use a pair of directional microphones at each point, which would drive up the cost, and potentially limit your ability to detect very high frequency signals.
So, one of our concerns is that, the time of the arrival of the waves to different mics should not be distorted when moved to the PC. That is the internal synchronizations of the PC (PCI workings and `sound card` itself) shouldn't change these delays. If this is the case, we would have to build custom circuits for this task
.
At a 100usec resolution, that will not be too tough of a task; that corresponds to a 20kHz sample rate (10kHz frequency * 2 for Nyquist sample frequency). In fact it's relatively easy for PCs to manage much higher sample rates.
However, you will need to define your necessary resolution, because if it gets much lower than 100usec, you will have problems with calibration. Given that sound travels 340m/sec at sea level, that gives you 3.4cm in 100msec. So you would need to be able to calibrate microphone placement with that accuracy; not terribly difficult, but if you decide you need say 10usec resolution, now you would need a 200kHz sample rate (192kHz is practical), and 3.4mm placement accuracy. That would also place a strain on the microphones, since there are no inexpensive microphones that are calibrated to 100kHz.
However, with a 48kHz sample rate, you get 42usec resolution with 1.4cm placement accuracy. That is reasonable in terms of cost for the interface and ability of a PC to record simultaneous tracks.
I'm not really a PC/interface guru, so I will leave that to someone else, but I would think you'd would 3 or 4 firewire interfaces with built-in mic preamps, but I really don't know how many of those you could chain, or what spec a PC needs to hit to do that many tracks simultaneously. Another option is a soundcard that can take 4 ADAT inputs (RME makes one), and then get four 8 channel ADAT interfaces (Behringer makes an inexpensive one, there are others by Presonus and several others I don't recall)
You really don't need to worry about distortion in the preamplifier or digital converter stage (the interface), if you are just trying to detect arrivals, I guarantee you we are a LOT more particular about quality than you will be
