CD oversampling
- Thread starter zener
- Start date
To put it in the simplest terms, oversampling is a form of digital feedback.
On a multibit CD player (e.g. tda1541a), it's removal can lead to a far more direct and immediate sound. However, one then needs to deal with the aliasing noise.
Oversampling is not a sales gimmic, but has been a defacto technique of achieving better "technical performance" (i.e. greater signal to noise) for any given bit rate.
On a multibit CD player (e.g. tda1541a), it's removal can lead to a far more direct and immediate sound. However, one then needs to deal with the aliasing noise.
Oversampling is not a sales gimmic, but has been a defacto technique of achieving better "technical performance" (i.e. greater signal to noise) for any given bit rate.
Werner
pfm Member
This is entirely wrong.
Sorry.
The correct replay of sampled audio requires a reconstruction low-pass filter after the conversion to analogue.
The sampling theorem, in all its beauty, prescribes a particular type of filter: Sinc(x).
The job of the Sinc filter is to remove the raw staircase output of the DAC, turning it into a fluid and continuous signal exactly identical to the band-limited signal that was sampled during recording.
This filter cannot be realised in the analogue domain.
However, oversampling the original data allows to implement this filter (largely) in the digital domain, where it is feasible to arbitrary accuracy (this is simply an engineering problem). After this the signal is converted to analogue and a mild analogue filter completes the process.
Turning this oversampling with digital reconstruction filtering off allows the signal's ultrasonic images ('images', emphatically not 'aliases') which correspond to the staircase steps to pass on to the rest of the system. It is then up to your speakers and ultimately your ears to provide the reconstruction filtering and reject the images. Which they will, being hard-limited to 12-20kHz depending on your age. All of this may sound different, it is also demonstrably less accurate. But then a turntable also isn't very accurate, and many people like its sound (perhaps for very valid reasons).
The terms oversampling/upsampling also are used in the context of ADCs and of delta-sigma modulation, where they mean somewhat different but conceptually related things.
PigletsDad
My intelligence test came back negative.
When people do the No-Over-Sampling (NOS) mod on a player or DAC, the most obvious effect is a treble cut; this is presumably why people like it, especially in systems that sound bright or thin.
I like it when I find stuff that I don't understand and is, in fact, counter-intuitive. Two things:
1. The oversampling process occurs in a low-pass filter, which has the effect of removing HF gunk.
2. Disabling the oversampling (and presumably the low-pass filter) actually reduces the perceived high frequencies.
Totally counter-intuitive! Can anyone shed light on this?
1. The oversampling process occurs in a low-pass filter, which has the effect of removing HF gunk.
2. Disabling the oversampling (and presumably the low-pass filter) actually reduces the perceived high frequencies.
Totally counter-intuitive! Can anyone shed light on this?
PigletsDad
My intelligence test came back negative.
The sampling theorem applies for infinitely short samples, which are reconstructed as a sequence of infinitely short pulses, then low pass filtered. Real DACs don't do this; you get a succession of constant values, rather than short pulses that return to zero. This results in a loss of HF, of about 4dB by the time you get to the Nyquist frequency.
As NOS converters produce lots of HF garbage just outside the audio band, they need complex analog filter stages to clean it up; a few designs include some ad-hoc HF boost to counteract the droop in frequency response.
Over-sampling designs use a digital interpolation process, to construct a succession of smooth signal values at a higher sample rate. In many cases, there is a digital feedback process, which keeps track of the running errors in the conversion and/or interpolation process. The converter now runs at a higher sample rate. The aperture effect, due to the rectangular rather than pulse output waveform, occurs at the higher frequency, and has negligible effect on the response. The junk is now well separated from the audio band, and can be smoothed out with a simple filter.
Bitstream and related converter types (as used in say the Sabre chip in the Buffalo DACs) take the idea further; they use a very high oversampling ratio, and a converter with a small number of bits; in some designs just 1. The digital processing shapes the noise, so that the quantisation error lies outside the audio band.
Read http://en.wikipedia.org/wiki/Sampling_(signal_processing) especially the section "Practical implications"
As NOS converters produce lots of HF garbage just outside the audio band, they need complex analog filter stages to clean it up; a few designs include some ad-hoc HF boost to counteract the droop in frequency response.
Over-sampling designs use a digital interpolation process, to construct a succession of smooth signal values at a higher sample rate. In many cases, there is a digital feedback process, which keeps track of the running errors in the conversion and/or interpolation process. The converter now runs at a higher sample rate. The aperture effect, due to the rectangular rather than pulse output waveform, occurs at the higher frequency, and has negligible effect on the response. The junk is now well separated from the audio band, and can be smoothed out with a simple filter.
Bitstream and related converter types (as used in say the Sabre chip in the Buffalo DACs) take the idea further; they use a very high oversampling ratio, and a converter with a small number of bits; in some designs just 1. The digital processing shapes the noise, so that the quantisation error lies outside the audio band.
Read http://en.wikipedia.org/wiki/Sampling_(signal_processing) especially the section "Practical implications"
Werner
pfm Member
Not intuitive at all, but exact nevertheless, is this:
Take a single flat-topped impulse with duration 1/44100 second, i.e. a single sample as if emitted by a non-oversampling DAC.
The Fourier transform of that impulse is a function of frequency that looks
like sin(f)/f = Sinc(f) . Yep, Sinc again, this is (no) coincidence!
Evaluating that function at 20kHz gives that -4dB. Really.
If you take the ideal sample representation, i.e. an impulse of infinitely-small
duration, then its Fourier transform is a flat line at 0dB. Evaluating that at 20kHz
obviously gives 0dB, or no treble loss.
Take a single flat-topped impulse with duration 1/44100 second, i.e. a single sample as if emitted by a non-oversampling DAC.
The Fourier transform of that impulse is a function of frequency that looks
like sin(f)/f = Sinc(f) . Yep, Sinc again, this is (no) coincidence!
Evaluating that function at 20kHz gives that -4dB. Really.
If you take the ideal sample representation, i.e. an impulse of infinitely-small
duration, then its Fourier transform is a flat line at 0dB. Evaluating that at 20kHz
obviously gives 0dB, or no treble loss.