It's been said that oversampling captures very high resolution but as long as the system is downsampled back it can still mirror higher frequencies below Nyquist Frequency which causes aliasing anyway.

I'll put it another way:
If we record any signal with a microphone, higher frequencies may be captured and mirror into our digital limited (Nyquist Fequency) system.
Oversampling (PWM for instance) should solve this issue by capturing higher frequencies.

I have been wondering how ADC converter with oversampling techniques leaves out mirror frequencies (aliasing) if it continues with PCM techniques that can't capture higher frequencies and reflects them back into the system spectrum (upto Nyquist Frequencies).

If the PCM converter reflects higher frequencies back (aliasing) what's the difference between putting previous Oversampling converter before an Aliasing filter which followed by PCM converter and leaving the PCM converter as the first link with preliminary LP Aliasing Filter ?

why does it matter where higher frequencies come from to the PCM converter ?
in both cases, PCM converter has to deal with frequencies above 22.050Hz that come in: straight from a microphone or from another converter (such as PWM delta-sigma) right after the microphone.

thanks..

Oversampling convertors generally use simple filters in front to prevent aliasing at the very high sample rate, then apply digital filtering to the digital domain signal to bandpass for audio use.

JR

Simply put, under conditions of no oversampling, the aliasing frequencies will fold down to be near the audio frequencies being sampled, thereby causing enharmonic distortion much like a ring modulator does by producing the sum and differences of two frequencies beating against each other. At a simple sampling rate of 44.1K (lower Nyquist freq of 22.05Khz), the difference frequencies will fall within the human audio range and therefore will appear as audible distortion.

The better solution is to over sample because that will raise the Nyquist frequency and will cause more of the enharmonic sum and difference frequencies to fall above the range of human hearing, thereby making the aliasing filter more effective.

Think about it in terms of sum and difference frequencies. At 44.1Khz sampling without oversampling, much of the enharmonic distortion from the beating of two frequencies will clearly fall in the audio band. But increase the sample frequency to 5.6Mhz by using 128X oversampling instead, suddenly the enharmonic sum and difference frequencies fall way above 20,000Hz, therefore causing the distortion created to be inaudible.

The key is that enharmonic distortion is produced rather than harmonic distortion. If only harmonic distortion of the Nyquist frequency was present, we would not have much of a problem because harmonics are always higher in frequency. With enharmonic distortion however, the frequency images produced can be lower in frequency as well which is where the problem lies.

Yup. And by oversampling, another advantage is that you can push the cutoff point of the analog filter very high, much higher than 20Khz, up to the point where its inherent colorations will be too far beyond the audio range to affect the sound in any way. Following that, the digital filter is applied when bringing it down to 44.1Khz (or whichever rate), and that filter can be (or at least, 'should be') designed to have little to no effect on the audio signal, unlike analog filters which all exhibit some unavoidable coloration.

Really ingenious engineering when you think about it.

Yup, those Yamaha engineers were pretty smart.



-tINY