KMG wrote:unfortunately in russian, use google translate
Unfortunately Google isn't helping much. If you or another Russian speaker can find the time, maybe you could give us a summary.
Here's what I did to compare the samples:
1. Used only one (the right) channel, to make things a bit easier and clearer
2.
Aligned the samples as best I could. I found that they were
not perfectly aligned to begin with, so be aware if you're doing this kind of test.
3. Compared the three by subtracting one from the other (A-B, A-C and B-C), by reversing the phase.
It's just a quick test, I didn't use long averaging for smoother and more precise results, but I did use the same part of the song in each case.
Here's the baseline, peaking around -1 dBFS with -11 dB RMS:
http://xonedesign.ro/publik/captest/baseline.png This is the right channel of the SBE715 sample, and the other ones look just about the same.
Here are the results of the subtractions (I'm hoping the file names are self-explanatory):
http://xonedesign.ro/publik/captest/SBE715vsSOZO.png
http://xonedesign.ro/publik/captest/SBE715vsWIMA.png
http://xonedesign.ro/publik/captest/SOZOvsWIMA.png
What one can conclude:
1. The difference is lower than -23 dB RMS down. In other words, less than one part in 14 if the flu hasn't affected my math.
2. The difference is mostly broadband, that is mostly a matter of
loudness (signal level) and not of
frequency respose.
I have to say, this resource is a lot better than most
silly "experiments" on YouTube. Wish I had more time to look into it, but the points I've made above provide me with plenty of evidence to infer that any difference in cap type is less significant than systematic errors, even with these fairly controlled conditions.
Ignorance more frequently begets confidence than does knowledge. (Charles Darwin)