Originally Posted by pbz06
Anyone know what the audible threshold is when running Audyssey (XT32) is without disrupting the chirps? I've never gotten an warnings.
I have an open floor plan and never realized how loud my fridge is. A constant humming/buzzing sound and it's about 35-40dB when I'm testing the MLP. There also seems to be a car driving by every time it gets to my subwoofer, argh. Loud swoosh sound as it drives by. The world is against me!
I have never heard anyone cite a specific threshold, but noise floors in rooms probably range from about 35db to as high as 45 or 50db, so I suspect that the threshold is pretty high. There has always been some disagreement about whether it is better to turn everything in a room off, such as the HVAC system, or a refrigerator, or whatever, or to just calibrate with the room as it would normally be during playback. Personally, I have tried it both ways and felt that the audible differences were too insignificant to notice. So, I typically allow the noise floor in my room, which is already fairly low, to be whatever it is. YMMV, but I suspect that it is more a matter of personal degrees of OCD than it is a matter of tangible audible differences.
The issue of cars driving by at inopportune moments is a little different, because that represents a momentary (or longer) change in the normal noise floor. The loud "swoosh" sound as a car drives by may be at too high a frequency to affect your subwoofers. If a loud truck slowly rumbled by, I think that I would want to stop and repeat the process. My own OCD would probably require at least that much. In practical terms, though, the earlier discussion of clustering and centroids is applicable to this situation as well.
It takes some time to move microphones between each of the calibration positions. So, it's unlikely that a random noise would persist for more than one mic position. And, some random sound during a single mic position, which isn't even loud enough to make Audyssey stop the test, is probably not going to make any difference in the final result. That is because the fuzzy-weighted averaging is going to treat that specific ping as an outlier anyway, if it even changes the overall frequency response to start with, and is not going to give it any weight. Instead, it's going to average based on more uniform clusters. That's not an ironclad guarantee, but I think its a reasonable conclusion.
I think that microphone positions can matter, and once we find a good calibration pattern, it's a good idea to stick with it. But, exactitude to the millimeter is probably a little over the top, as is an excessive focus on the noise floor, and on random short-term noises during calibration. It's certainly a YMMV question, which people can test for themselves either by listening, or by measuring. But, I suspect that the calibration process is actually a little more flexible, and more forgiving, than some of us who are implementing it.