Maybe that JIS 1.4% reading from MI is a peak unweighted, whereas the track is RMS/Average weighted. That would account for the difference (/sqrt2).
You are absolutely right.
I was led here accidentally after typing "NAB wow and flutter" in Google Search. I searched this today because some MI users requested the support of weighted and unweighted RMS W & F defined by NAB/JIS. Currently MI only supports weighted and unweighted Peak W & F defined in AES6-2008 (r2012), which is also compatible with IEC 60386, IEEE Std-193, CCIR 409-2, and DIN 45507.
Strictly speaking, this "peak" value is not exactly the maximum value. It is a "peak" value determined through a so-called “2-Sigma” statistical method. In this method, the cumulative time for which the instantaneous frequency (speed) deviation exceeds the "peak" W & F is equal to 5% of the total time. Assuming the modulating signal is a sine wave, then "peak" W & F is sin(0.95*90 degree)= 99.7% of the real maximum value. Anyway, 99.7% is almost the same as 100%
, but this is a more robust method statistically if the modulating signal is not sinusoidal and contains noises.
Thanks for the standard WAV files. I have tested 3KHz NAB-JIS 0.1.wav as follows. The measured weighted peak flutter is 0.1407%, which is very close to the theoretical value: 0.1%*1.414*0.997=0.1410% of this calibration file. One trivial issue with this calibration file is that it contains a zero portion at the beginning and the end. These zeros need to be removed for a better measurement accuracy if the WAV file will be directly used for analysis, or they should not be sampled in the acquired data.
The unweighted value is the same as the weighted value as the modulating frequency is 4 Hz. In the following figure, Ch.A (Blue) shows the original FM signal while Ch.B (red) shows the FM-demodulated signal (which is 4 Hz).
BTW, the FM function of the signal generator in MI can be used to generate the calibration file as well.