Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: sampling (Read 2379 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

sampling

When oversampling at (for example) 192kHz, the samples are usually quantized to only a few bits resolution.
How is it possible to still have a "signal to error" ratio comparible with a 16 bit recording (97.8 dB)?
Thanks for the input!

sampling

Reply #1 –
The noise is distributed over a wider frequency band (and is shifted upwards with noise shaping). That gets you equivalent SNR at 20-20k.

sampling

Reply #2 –
DSD provides an extreme example that shows how SNR is not determined by bit depth alone:

DSD is a 1 bit system with a sample rate of 2.8224 MHz.  When measured over its entire bandwidth (1.4112 MHz) its SNR is less than 6 dB.  However, DSD achieves an excellent SNR (up to about 120 dB) when measured from 20 Hz to 20 kHz.  DSD uses very aggressive noise shaping to move the quantization noise out of band and into high frequencies.

The in-band SNR limits of a digital system are determined by the following:

1) Bit depth (about 6 dB per bit)
2) Sample rate (about 3 dB for every doubling of the sampling frequency if noise-shaping is not used)
3) Noise Shaping (substantial improvement - especially at high sample rates)

Noise shaping is only effective when there is spectrum space available to "hide" the noise.  The noise shaping that is often used on 44.1 kHz CDs moves much of the quantization noise the the band between 20 kHz and 22 kHz.  Packing the noise into this 2 kHz band can substantially improve the perceived SNR as well as the measured 20 Hz to 20 kHz SNR.  96 kHz has a much wider band available for noise shaping - (20 kHz to 48 kHz).  96 kHz has a 26 kHz band available for noise shaping while 44.1 kHz has only 2 kHz available.  DSD has 1391 kHz available for "hiding" quantization noise.

Noise shaping must be re-applied whenever a signal is modified (gain change, EQ, etc.).  Cascaded noise shaping processes can be very damaging due to rapid build-up of out-of-band noise.  For this reason DSD in not a good format for use in recording and editing.

Noise shaping is rarely used at 96 kHz because most 96 kHz systems support 24-bit data.  For this reason 96 kHz 24-bit systems have gained widespread acceptance in pro-audio applications.  96 kHz 24-bit audio can be edited easily without needing to apply potentially-damaging noise shaping.

John Siau
Vice President
Benchmark Media Systems, Inc.