You have it all wrong, and don’t understand how this works. The original encoded file from recording company is 352,8. The first unfolding is 88.2. The rest of unfolding must take place in a DAC. The MQA decoder does no up or downsampling at all. It decoding the ogami. If an ogami exist. MQA exist also without ogami. Any encoded file displayed as MQA 44.1 or 48kHz has no ogami. Roon and probably other decoders will by an error claim it has.
The transfer rate in this case is FlAC 44.1 and bit depth is 24.
No, I'm sorry but you are the one who has it all wrong. A 352.8k original is downsampled to 88.2k, which is then folded into a 44.1k container file. Undecoded, the MQA file will play back at 44.1k. Unfolded ("core") it will play back as 88.2k. Fully rendered - "the rest of the unfolding in the DAC" as you say - it will play back with the DAC and/or playback software
showing 352.8k - but that is because the "rest of the unfolding" is what MQA calls the final render, where in Bob Stuart's sly words they "put the sample rate back to what it was." The 352.8k sample rate is
generated -
not decoded or restored - via a quadrupling of the native 88.2k sample rate of the MQA file - it's not the actual original samples from the original 325.8k PCM file that the record company file started out with.
So once again, here is the chain:
- Record company 352.8k PCM original (assuming in this example that they have an original at that resolution)
- Record company runs that original through MQA encoder
- MQA encoder downsamples 352.8k to 88.2k - throws out 3/4 of the samples (I know you think this can't be true, but it is)
- MQA encoder then "folds" the the 88.2k into a 44.1k MQA container, via lossy compression of half the samples, encoding that lossy-compressed data in the lower bits of the files
- MQA "first unfold"/software decode on the playback end decodes the 44.1k container file back into 88.2k
- Hardware DAC "final render" upsamples the 88.2k to 352.8k and applies allegedly "custom" MQA reconstruction filter
Crucially - and the bit you don't get - Step #6 does
not "unfold" 352.8k sample rate information that is encoded in the MQA file - all that's encoded in the MQA file is 88.2k. The additional samples of the original 352.8k PCM file the record company started with are gone forever, and the MQA final render simply makes 3 additional copies of each sample, thereby quadrupling the number of samples, in order to bring the file back up to 352.8k.
Now, it is not surprising that there is confusion about this - MQA has been misleading in their communication about it from the beginning. By calling it "audio origami" and referring to a "first unfold," they have strongly implied that there are multiple "folds" and that the full 352.8k sample rate info can somehow be folded up/encoded into a file that ends up as a 44.1k container. That's impossible and is not what happens.
Personally I don't care if the file is 88.2k, 176.4k, or 352.8k - there's no sonic difference. But MQA is marketing them as 176.4k and 352.8k and Bob Stuart is claiming (incorrectly) that "new neuroscience research" shows that the ultrasonic information from those higher sample rates is meaningful. So he and MQA are being deeply misleading, allowing the impression to circulate that something beyond 88.2k/96k is contained in an MQA file when it most certainly is not.
If you take a 352.8k PCM file and downsample it to 96k yourself in Audacity or any other audio editor, and then run that 96k file through the MQA encoder, when you play it back on an MQA-capable system and DAC, it will register as 96k. If you take the same file and do not downsample it to 96k yourself and run it through the MQA encoder, the MQA encoder will downsample it to 96k before doing any other processing. But because the file you gave the MQA encoder was 352.8k, when you play back that file on an MQA-capable system and DAC, it will register as 352.8k. In both cases the file has been downsampled to 96k - it's the same file. But in one case the MQA DAC will show it as 96k while in the other it will upsample it and show it as 352.8k.