[radiowave]

The trade-off is that sample rate conversion introduced its own artefacts, though we have much better algorithms (and of course more CPU power) for it today.

I'm not sure that aliasing was in itself a big problem, you're perhaps refering more to the audio artefacts introduced by the anti-aliasing filter. But in any case this wasn't even the biggest problem with converters back then. IME, quantization distortion was a much bigger issue. You had to watch the signal levels really carefully - too loud and you'd get absolutely brutal distortion (sometimes including wrap-around, where the value reported by the A/D converter goes instantly from maximum value to minimum value), too quiet and everything turns into a grainy-sounding mush.

The world changed in the early 90s when delta-sigma (aka "oversampling") converters came along, pretty much solved all these issues, and so allowed the price of decent converters to steadily fall, as normal semiconductor manufacturing economics kick in.

[marcan_42]

These days the benefit of 48kHz is that you get twice the frequency headroom vs 44.1k (from 20kHz to Nyquist), which means that any processing steps using oversampling and antialiasing filters (which you should be using) will either sound better, use less CPU, or both, since the filter has to be only half as steep on 48k.

But that's for intermediate processing, once you have the master you can convert to 44.1 one final time and that's that. One final sample rate conversion with modern algorithms will not introduce any audible differences.