What MQA does is not that hard. It is known, but some will argue, that as you increase sampling frequency from 48k to 96k to 192k to 384k etc., the recording sounds better. I have tried it and think it does, but as I say, opinions do vary. A theorem says the maximum frequency that engineers can record is half the sampling frequency. But since the most we can hear is 20khz, 48k should be more than good enough. So why does it sound better? MQA conjecture it is what is called time smear:
Tutorial: Temporal Errors In Audio In our Hierarchical paper [2] section 2.3 Temporal Limits, we explain:
www.stereophile.com
The question then is how do we convey the audible 20kz region but have the lower time smear of high sample rates? Here is how MQA does it. Let us imagine a high sample rate say 4XDSD. Recorders are readily available to do that. Now let us use a triangle function to downscale that to 96k as per the following:
Even at high sample rates, standard PCM audio ‘smears’ important timing information. A new digital format, MQA, promises vastly improved time-domain accuracy — without the huge file sizes.
www.soundonsound.com
This has the effect of introducing a very slow roll-off above 20kz - which being above 20kz is inaudible and chucking samples above 96k away. The result is as in the following:
The issue is these aliasing components that are in the audio band. But here we are in luck. Analysis of recordings shows with the slow roll-off, they are below the notice floor. Secondly, the research also indicates that recordings have nothing but noise above 50kz, so 96 kHz is not a concern. We then transmit the 96 kHz data (MQA uses a tricky way of doing this you can read about by chopping off the bottom bits, recording at 48k, but replacing those chopped off bits with the 24-48khz information - but for simplicity, we assume it is transmitted at 96k). When received, we want to get back the data chucked away. Simple - we linear interpolate - it's all noise anyway, so it will not make any difference if it is about right.
Not that hard - and we now have a very high sampling rate recording with much less time smear to feed into our DAC.
In practice, MQA uses something more sophisticated than triangles and linear interpolation called Bessel functions. In effect, the resampling and reconstruction functions are filters, and they use downsampling-up sampling filters matched to each other. Now, does doing that improve the sound over simple triangles and linear interpolation? Beats me. MQA think it does.
Thanks
Bill