Your post unfortunately misses some of the fundamentals of sampling theory, both A to D and D to A. It would take too long to go into all the detail, but let's just take a few basics to get the ball rolling.
ADC and DAC have many things in common. First, higher sample rate only buys you higher frequency reproduction. It IN NO WAY makes the reproduction within Nyquist any better (filter issues aside - more on that later). More bits buys you greater dynamic range and lower noise floor. It in no way makes the reproduction of the higher bits any better. Since the very best (and very expensive) analog electronics today have a practical limit of the equivalent of about 22 1/2 bits, anything below bit 22 (and change) is conveying the self noise of the electronics and contains nothing of your desired signal.
The key is that you must have a band limited signal. Nyquist and Shannon showed that you can perfectly recreate any frequency information below one half the sample rate. Fourier showed that all complex waves can be broken down into their individual sine components. Between these two concepts, anything below Nyquist can be captured, and reproduced again.
One problem you run into is how to band limit the signal. If you don't filter the Nyquist frequency and above, you are left with aliases on the ADC side, and image frequencies on the DAC. Since there is no theoretically perfect filter where all frequencies below Nyquist are perfectly passed, and all frequencies above are perfectly rejected, you run into many issues.
The three parts of a filter's response are the passband (frequencies you want to keep), the stopband (frequencies you want to get rid of) and the transition band (where the rolloff happens between pass and stop bands). You also get filter ripple to either side of the rolloff, which is deviations in the response at to either side of the rolloff region. Depending on whether you are using analog, digital, IIR, or FIR filtering, you get a host of other problems. Most is done with digital FIR filters these days.
IIR stands for Infinite Impulse response and is the typical EQ we see in music, which equates to minimum phase analog designs, and most digital plug in EQs as well. FIR is Finte Impulse Response and usually (but not necessarily) exhibits a linear phase response. These will always be digital. IIR filters have problems of phase distortion in the form of group delay, or certain frequencies arriving out of time with others. FIR filters delay all frequencies by the same amout, thereby maintaing a linear phase response, but they exhibit problematic pre-ringing. This is a bit of signal that arrives ahead of the main lobe of the filter and in practice can smear transients and skew imaging and such. The steeper the filters, the worse the problems.
Filtering is a big deal. Higher sample rates will allow you to move filter issues such as rolloff and ripple above the range of human hearing, and also allow you to use much gentler filters which will reduce ringing problems. So, analog electronics that perform well in both ADC and DAC, and filtering for both ADC and DAC are a huge part of the sound of converters. There is a bunch more that goes into why a converter sounds good or bad, but I think I've rambled long enough for now.
Quote:
Originally Posted by jamsmith  Someone enlightten me ont enlighten me a little on the asthetics vs the science on AD conversion. Now certainly DA conversion has so many factor as you are taking a finite set of snapshots of an analog signal and trying to recreated a infinitely variable signal while handling slew and distortion. And the real challenges of that are largley in the analog circuitry of the DAC,
But DA conversion is a matter of getting the instantenous voltage level of a signal. Other than the quality of the sample hold circuit, that is very simple task. What is it you are looking for in these higher quality ADCs. Are they some how skewing the sampling in a way the translates better in DA process?
I can forever see the challenges or perfecting DA conversion, but after all these years, I just can see how AD conversion should anymore than a matter of more bits and higher sampling rates. |