In practice, I think many DACs suffer from differential non-linearity and aliasing issues and often upsampling can ameliorate things by randomizing non-differential effects and eliminating aliasing artifacts. In theory it should not be necessary but poor quality, poor design and build of DACs means YMMV. I have a DAC that doesn’t care if you upsample or not - it always sounds identical provided the source file is the same. That might be because it fundamentally uses upsampling as the basis for conversion.
In fact 16/44.1kHz CD can reproduce all the audible information up to 20kHz which is sufficient enough (Though some high order harmonics may be missing). The issue is the cut off frequency is around 22.05kHz and any over-sampling filter (typical 8x) need to attenuate this down by at least -90dB or more to prevent aliasing. This close proximity to the audio range, meant filter has to equipped with a sharp cut-off characteristics. This introduces ‘ringing’ in both pre and post typical standard filter design.
MQA partially solves this by using MQA filter (a variant of minimum phase filter) it’s cut off frequency is now at 44.1/48kHz instead of 22.05/24kHz bandwidth of a typical standard filter. This explains the fact MQA core will always output 88.2/96kHz regardless whether the original sample is 44.1/48kHz or higher. MQA may partially solve this ringing issue but because the filter cut off is 2x higher than typical standard filter, there’s a very high chance of aliasing (leaky). MQA filter belong to a group of non-linear filter; meaning distortion will increase exponentially to higher frequencies. Combined with this distortion and aliasing, I wonder this abnormalities will influence the typical sound ‘signature’ of MQA.
The way to fully solve all these issue is to sample higher, in fact the sweet spot is 176.4/192kHz, this is 4x away from the audio range and any filters designed to attenuate at 176.4/192kHz is a slow roll-off; such filter introduces virtually little to no ‘ringings’. Best of all, you don’t need any oversampling though most DAC will oversample to 352.8/384kHz.
I don’t find ‘MQA sounds terrific’ it just another typical sound signature that may suit certain type of listeners; probably they fall into those who appreciate a more ‘analog sounding’.
MQA is lossy in data form but when conveying musical details, it is close to be subjectively transparent. The distortion you are taking about is all about the MQA filters used.
This kind of ringing isn’t audible. This is MQA marketing ■■. If a DAC sounds different or better because of the upsampling then the upsampling is simply improving DAC performance. Examples would be reduction in IMD distortion, aliasing, poor leaky filters or differential non-linearity.
The video covers how it works but neglects to mention that MQA methodology is not true high fidelity because it distorts the original signal by time and amplitude smearing.
The claims that MQA fixes audible issues is false.
Han’s is much like Bob Stuart. They appear to be using the same language as digital audio and software, signal processing, and the like as everyone else. However, scratch the surface and you realize they either don’t know what they are talking about at all, or they are using and abusing the plain meaning of words for their own $ends$…
I watched it a couple of minutes in. That’s all it took for him to make two large factual errors:
That Qobuz won’t be offering real Hi Res
That MQA is Hi Res, up to 6x sample rates (by which he means 6x 44.1/48). MQA is (sort of) regular PCM to 44.1/48, a lossy compression facsimile between 48 and 96, and after that…nothing, notta, zip.