Originally Posted by amirm
Arny, Arny, Arny, the noise is the noise. It makes no difference if it is the silence at the end or beginning. Here is the "lead out" to make you happy
This time I circled the more critical region around 3 KHz. We see all the same points I made in my previous post. We are still at -108 dB in the 3K area. And far higher in low frequencies which would fool a meter showing a "dumb" SPL value devoid of psychoacoustics.
Amir, don't confuse the fact that I now know that I have to take you though audio 101 a step at a time with the individual steps.
So the first step was to get you looking at the same part of the same audio track that I was (ironically, based on your choice of tracks which is poor when it comes to proving your point).
Now, I have to teach you how to analyze noise levels. The rookie mistake we are concentrating on today is the same one I just corrected JA for when we were talking about amplifier SNR. Note that he had the good sense to stop arguing that point with me. You didn't. ;-)
The mistake I corrected JA for is using a FFT to establish what a noise floor is. This fallacy is based on the fact that when it comes to random signals, a FFT will give a lower noise floor the more points you put into the analysis.
The correct way to measure noise in a frequency range is to filter the noise so that you are measuring the noise in just the desired frequency range, and then measure the output of the filter. This is for example how we use things like "A weighting": First we filter the noise with a pre-defined filter and then we measure the output of the filter.
When working with Audition we can easily measure noise correctly because Audition has a incredible set of filters that we can use to synthesize just about any filter we want.
Looking at the Fletcher Munson curve:
We see that human hearing reaches its peak sensitivity in the range from 2.5-4.5 KHz. If I apply what Audition calls a "scientific filter" with Butterworth characteristic and 4th order roll-offs with these points, and then measure the results I get the following numbers:
Min Sample Value: -21.24 -14.81
Max Sample Value: 22.78 17.95
Peak Amplitude: -63.16 dB -65.23 dB
Possibly Clipped: 0 0
DC Offset: 0 0
Minimum RMS Power: -94 dB -90.38 dB
Maximum RMS Power: -81.81 dB -81.72 dB
Average RMS Power: -90.15 dB -87.82 dB
Total RMS Power: -89.92 dB -87.73 dB
Actual Bit Depth: 32 Bits 32 Bits
Using RMS Window of 20 ms
Bottom line - Still nothing that can't be handled with properly noise shaped 16 bit quantization.