In this experiment, the signal-to-noise (S/N) ratio and the total harmonic distortion (THD) are measured by converting a 1kHz sine wave. A 1kHz sine wave in 24bit 44.1kHz format is converted into 24bit 48kHz using SSRC without dithering. 123.15dB of S/N ratio and less than 0.00001% of THD are observed.
A 1kHz sine wave with 24bit 48kHz format is converted into 24bit 44.1kHz using SSRC without dithering. 128.58dB of S/N ratio and less than 0.00001% of THD are observed.
Frequency characteristics are observed by converting impulse responses. An impulse in 24bit 44.1kHz format is converted into 24bit 48kHz with SSRC without dithering.
An impulse in 24bit 48kHz format is converted to 24bit 44.1kHz with SSRC without dithering.
In order to test the dithering function, SQAM track 35 is converted into 8bit, 44.1kHz wav file with and without dithering. The spectrum of dithered sample at 468000 samples from the beginning is shown below. We can see that the quantization noise is shaped to fit the absolute threshold of hearing (ATH) curve.
Here is the spectrum of the non-dithered output at the same point. We can see some harmonic distortions.
As of version 1.32, new noise shapers are available. Users can now choose from noise shapers with 7 different strengths made from 2 different ATH curves, for 44.1kHz or 48kHz output fs. New noise shapers are also available for other frequencies, i.e. 8k, 11k, 22k, 88k, 96k, 192k, although not all strength/curve pairs are available for all frequencies.
The following graph shows the frequency characteristics of the new shapers made from the first ATH curve, which is used for SSRC for long years.
The following graph shows the frequency characteristics of the new shapers made from another ATH curve.
The following graph compares the new shapers and the old shapers. ID 90, 91 and 92 are low, mid and high settings in the previous versions, respectively. As you see in the graph, the frequency characteritics for the new noise shapers are much smoother than the previous shapers.