I've discovered that my integrated sound on my motherboard is a huge POS using the digital aliasing test website.    It seems to alias fairly severely.

It is an ADI 1896 chip. Seems like little else could be much cheaper!

Most ADCs are some version of the sigma-delta. Sampling rates of 64X and 128X are the norm, the output of which is then decimated to the desired sample rate (1/64, 1/128). Only mild anti-aliasing filters are normally employed, nothing else being deemed necessary.

Several years ago I tested a number of soundcards (all I could get my hands on) and published results in a couple or three audio forums. A few people tried the experiments, some had much more expensive soundcards to test.

No soundcards were found that did not evidence very easily observed aliasing above 20kHz. Increasing resolution on the spectral display shows there is always some aliasing signal down to almost 20Hz. A search in this forum, and on the Audiomasters forum
www.audiomastersforum.net
should find several threads on the topic. They will describe the methodology and some might still have screen displays of the results attached. I suspect these results are as good as any soundcards get in this respect.

I made clear my view that the aliasing, although easily detected with test tones, does not matter with music. There was general widespread agreement. First, little music has much energy above 20kHz. Second, most music is recorded with microphones, most of which have little response at those higher frequencies. There is thus little high frequency signal that might alias to begin with. Third, human hearing can not detect much at the higher frequencies (above 20kHz) where the strongest aliasing distortion will occur (assuming anything to produce it).

I don’t know if there is any method to determine how much aliasing distortion is present in recorded music, but the probability is that it is not audible -- i.e. the aliasing does not matter in any practical sense.. If you made one recording at 44.1kHz and another simultaneously at 96kHz, of the same source, noone would be able to hear which is which, (the strong part of the aliasing occurs in the first few kHz below the Nyquist limit, therefore for 96kHz sampling it is in the 45kHz to 48kHz range).

By observation, results are much worse for multi-media/gaming soundcard that aromatically resample everything to 48kHz, if 44.1kHz output is being used. Results are much cleaner if recording at 48kHz, probably due to the fairly poor quality hardware resampling many of those card do. Whether or not the aliasing distortion is a significant component of those soundcards that actually sound poor is an open question, but it is unlikely that the aliasing matters under any other practical situation.

For any professional or semi-pro card, and perhaps any of the better gamming cards, the aliasing, although always operating (given anything in the input to creating it) is probably never a factor in how the music sounds.

Interesting.  My sound card aliases unbelievably at 44.1k.  Using foobar's resampler to upsample to 48k cuts down drastically on aliasing, but I can still detect it on test tones with my ears, not by plotting a frequency spectrum of the output or something.  My Sansa and (just tested) my laptop don't produce audible aliasing at either 44.1 or 48 khz sample rates on the same test tone.

No soundcards were found that did not evidence very easily observed aliasing above 20kHz. Increasing resolution on the spectral display shows there is always some aliasing signal down to almost 20Hz.

In that case its probably just simple distortion not aliasing.  Those test tones are within a couple dB of full scale.  Blast enough energy in one single tone into an otherwise silent DAC and you'll hear all sorts of nonlinear effects.  They're probably 60 dB down or more, but you'll still hear them since theres nothing to mask the sound (unlike in real audio).

Quote from: Arnold B. Krueger on 2010-03-28 22:50:22

It is an ADI 1896 chip. Seems like little else could be much cheaper!

If that is supposed to mean AD1896, at a price where little else could be much cheaper, I'd like to order one...  Sounds great!

I've discovered that my integrated sound on my motherboard is a huge POS using the digital aliasing test website. It seems to alias fairly severely. Resampling to 48,000 or 96,000 hz in foobar mitigates, but doesn't completely solve, the problem....
...Most of my audio is FLACs made from plain old audio CDs, so supporting super high sampling rates, anything more than two channels, or resolutions >16 bits are useless to me.

-strictly speaking doesn't 'aliasing' apply only to ADC, whereas the complementary phenomenon for DAC is 'imaging'?  Or are we still talking about 'aliasing' if  resampling of a digital signal is involved?

Since aliasing happens during sampling or downsampling (where the "input" frequency is above half the sample rate), I don't see how you can get aliasing with a file ripped from a CD!

Since aliasing happens during sampling or downsampling (where the "input" frequency is above half the sample rate), I don't see how you can get aliasing with a file ripped from a CD!  (Unless the the original CD already has aliasing on it, or if you're downsampling to to something under 44.1kHz.)