Spotify is a pretty decent interface. It is not Roon. But it includes Spotify Connect, which is a game changer for sound quality. Apple Music is an ok interface, but here you have the incredible ability to stream to every Apple device, including streaming directly to my Apple Watch over cellular (which I use at the gym all the time). So even though the interfaces are not as good as Roon, they are good enough (certainly as good as Tidal and Qobuz), and they are available EVERYWHERE, not just at home.
(Btw, Spotify can also stream directly to the Apple Watch, but the handshaking of credentials is flaky and oftentimes you need your phone to revalidate credentials - and if you left your phone at home you’re screwed)
Actually, it is the A I have the biggest problem with. MQA encoding is not done on the spot, it is sent some place and it gets encoded either by automated blanket algos (99% of the material) or by actual MQA people working on the encoding.
But I have heard from trustworthy sources that the main reason recording engineers and producers do not like MQA is because it changes the sound. They produced the masters LISTENING to the files they then send to MQA, only to receive an encoded file that has a slight but noticeable “Eq” applied to it, which is NOT what they settled on. So the A is not just not true, it is the opposite of true.
Listening to it, I came to the conclusion that the MQA version (unfolded and rendered by my DAC, a dCS Rossini) sounded better. A frequency graph aggregated to the entire track is a very specific measure of a tracks content, it says very little about distortion, transients, phase response, etc, so I would not really regard it as a measure of quality, especially when you’re comparing two very similar sounding tracks.
Of course. I was only commenting on your own comment about dynamic range.
I didn’t comment on anything else.
I like reading about others technical analyses and doing my own micky mouse analyses And so far I see no value in M , Q or A in the way it’s been implemented so far, in general.
This isn’t a life or death thing for me (unlike some others).
Just friendly observations and discussion here.
The crowd seems more focused on Apple Music news at the moment , so should be safe from a heated flame throwing MQA debate
Search is much faster and works much better than v3.5
I think they made a good step forward.
New Tidal and Qobuz releases take a day or two to show up in Roon every week, so Audirvana has been great for me for years every Friday to Sunday, playing direct to HQPlayer for me.
And I can add to my Tidal playlists (for away from home) direct in Audirvana.
Not sure I understand… I can add to favorites and playlists in both Tidal and Qobuz apps from anywhere, and those will show up in Roon. How is this different with Audirvana Studio?
I don’t like the subscription scheme for something I barely use so I am not going to try Audirvana Studio. I already own two licensees for A3.5 (I needed three machines: mac pro, mac mini (Roon core machine), and macbook).
I don’t know what the two plots are that you are showing but they are not sufficiently alike in any of the detail structure to be comparable measurements. Look at the detail in the slope over the first 20 kHz. Plainly they don’t represent the same thing. There have been enough careful comparisons of MQA and PCM to know that the envelope shape matches exactly over that region. This is definitely a problem.
At what point in the processing chain are the Audirvana measurements taken from? The plot doesn’t extend beyond 48 kHz (the Nyquist of 96k) so is it supposed to be a comparison of the PCM file immediately after FLAC decoding versus the first-unfold (to 96kHz) of the MQA? Is the plot averaged over the whole file? Is it a peak or mean dB value. Are the two plots sufficiently time-aligned to allow a close comparison?
Interesting question… The first unfold gives the same result regardless of the software and hardware. The second step, rendering, is simply a signaling of what filter/upsampling to use, and it will depend on the hardware down the chain. I don’t think Audirvana (or any other software for that matter) can (is allowed to) do rendering (ie upsampling in software based on the MQA filter choice), so this has to be the frequency response after the first unfold.
Yes, the plot is averaged over the whole file, this is standard practice. Time alignment does not really matter here as it is aggregated over the length of the track. I believe each frequency point is the max over the whole file, but not 100% sure in this case.
Add 5 seconds of silence at the beginning of the track. The resulting plot will be the same. You are correct that the analysis algorithm is taking a few samples to determine the frequency, but the discrepancies from that time alignment are very small in my experience. Each tick mark here is 25dB.
More importantly, an aggregated plot like that is telling me almost nothing about the sound of the file.
I would say less than 10% of the hi-res I have listened to sounds better than a high quality redbook (ie 16/44) recording. Really. A great DAC playing a high quality redbook file will sound AMAZING.
And so far I see no value in M , Q or A in the way it’s been implemented so far, in general.
