These graphs are from MQA claim, not from the manufacturers hardware reviewed by Stereophile. We have to wait for actual hardware with the so called ‘better improved encoders’ to do the final testing. There are plenty of opportunities in the future to see this. I think MQA is going to have a tough time convincing people as more test data start to review the inherent limitations; namely ‘ultrasonic noise’ and ‘aliasing’ due to use of ‘leaky filters’ This is inherent design core which I believe are not going away.
However, you got the best recordings from DXD and DSD that has proven to be superior and one can really hear on your existing components. High Resolution has been here all along! So enjoy the music!
I think MQA will never convince people who don’t want to be convinced anyway. If we go on to the average listener who buys the latest MQA device to enjoy music as music, They are going to just love it and get a lot more from the music than they do today.
In TV, picture quality is going up and in Audio the same thing will come for the masses.
There’s a lot of unconvinced people out there, so I guessed it is tough time for them to keep doing all the ‘claims’. Besides we already have Hi-Res audio, Full HD and 4K TVs available for years, there’s nothing revolutionary we are taking here. It is just another format looking to make $$$. I’m fine with that, as long I don’t have pay for it.
Qobuz has been streaming Hi-Res for quite sometimes now and it sounds much better than Tidal Masters. The streaming is absolutely lossless and don’t require special equipment to do the unfold; this is best you can get directly from the recoding studios - peace of mind!
Since Qobuz is apparently coming for Roon (competition, not acquisition), it’s time for Roon to take on Qobuz with it’s own uber quality streaming service.
I’d be happy to pay the same subscription cost, on top of my Roon subscription of course. I probably should mention this on a more relevant thread.
This is what I don’t understand. If you don’t like it don’t listen to it. I mean it really is that easy. MQA sounds absolutely phenomenal in my system. I do not care about the science and the graphs. I am interested only in what I hear and MQA is just about all I listen to now.
MQA is an end-to-end system; this cannot be ignored. MQA’s inventors have forgotten more about digital filters than most of the forum “experts” will ever know.
Indeed–and you, surely, due to your exposure, know more than most folks, too. But to be clear, for the 2L file, I recorded the output of an MQA decoder, so what I recorded reflects end-to-end processing.
It is true, however–as others have written–that 1. this is just one of countless encoding methods; 2. that it’s being “fixed” if it hasn’t already; and 3. I couldn’t hear it–at all–anyway.
MQA/Stuart frequently emphasize the importance of eliminating pre-ringing. Ordinarily, that would mean NOT using linear-phase filters. What we normally call “minimum-phase” filters delay the higher frequencies so that the pre-ringing follows the main peak. (You know this of course.) But here, “minimum-phase” isn’t really the point; the point is the word that’s usually left off: “causal”–which simply means that nothing precedes the actual event (which is taken to be the main peak). In one of its patent applications, MQA mentions such a technique while emphasizing that it is not “minimum-phase”, which only means that the phase delay is the least that’s required to achieve the objective of ‘causality.’ The fact that this appears in an MQA patent application, though, merely means that the company owns the rights to that invention (or has applied for it)–not that the invention has actually been implemented in MQA.
I would, though, refer to a couple of points from the “Part 1” article referenced above by Dr. Tone. First, note that the “upsampling renderer” used by MQA has a “causal” character. (Is it minimum-phase? I don’t know.) Notice also that when the decoder is turned off (so the “upsampling renderer” is no longer in the circuit) and a linear-phase reconstruction filter is chosen, you get a sort of hybrid output–part linear-phase, part causal (Fig. 4 in the article). It’s a superposition of the linear-phase reconstruction filter and the encoder’s characteristics–so the latter appear to be causal (if not necessarily minimum-phase).
There is one other point though, which I’ll just hint at, since it hasn’t been published yet. I do not fully understand this, but the usual pre-ringing seen with a linear-phase filter is very likely related to the use of a sinc(x) function as sampling kernel:
What would happen if you used a different sampling kernel, with a different shape? Is it possible then that you could get rid of pre-ringing without delaying high frequencies?
I do not yet know if MQA is doing this, but there are strong hints.
Em, no. It’s not that easy. It’s a significant pain in the ass in Roon to identify which of multiple Tidal versions of an album is Redbook vs. MQA. If Roon would make this obvious, then yes: don’t like, don’t listen. But the current interface is pray and play, then check to see what version you ended up with. Yuck.
Ordinarily it would. Bob Stuart and Peter Craven have submitted a patent application for an up-sampling filter which is essentially linear phase, but which applies selective group delay to the ringing frequencies and/or those which trigger ringing. This pushes the ringing into the main response peak (or just after) which masks it. Apparently, the filter is idempotent, so it actually prevents ringing caused by filters later on in the chain.
I have no idea if MQA uses such a filter, but it seems likely. Bob talks about MQA files being “pre-apodized” and this seems like a likely candidate.
Pretty sure I skimmed that application; probably the one that indicates that what they’re proposing is NOT minimum-phase. But I don’t recall them saying that it was LINEAR phase. Doesn’t mean they didn’t; I easily could have missed it. My assumption at the time was that it had MORE phase shift than a minimum-phase filter would.
I think there’ s a contradiction there; any filter with selective group delay would necessarily NOT be linear phase–right? I’m not an expert.
When I read the paper, I inferred that it was a new type of filter and was, as I said, essentially linear phase, with targeted group delay. Clearly they feel that it is innovative enough to warrant a patent.
My interpretation is that minimum phase filters, with their inherent progressive group delay, are the antithesis of MQA. In essence, minimum phase filters cause temporal blur.
“Linear phase” means all frequencies are delayed by the same amount.
Presumed phase nonlinearity is one of the frequent criticisms of MQA, in my conversations with designers who don’t support the technology. Have a look at John Siau’s “manufacturer’s response” to my review of the Benchmark DAC3. (Oddly, as I’ve just noticed, it isn’t online, so I’ll type in the relevant part):
“The similarity between the sound of the DAC3 an MQA is no accident. Like all Benchmark products, the DAC3 has an extended 0.1Hz low-frequency response in the analog section. This extension eliminates delays in the low frequencies. This keeps the lows in the correct time alignment with the high frequencies. The result is that reverb tails are not obscured by the late rendering of the bass. This correct time alignment provides solid bass and a noticeable improvement in the audibility of high-frequency reverberant details (a greater sense of 3D space). MQA arrives at a similar result by delaying the high frequencies using a minimum-phase filter. This approach tends to expose the reverberant details, but it also distorts the time-domain response across the entire audible spectrum.”
I don’t know if he’s right. He’s an expert in digital design, but has no inside information. But it does show how, with a single trip (no opportunity for delays to accrue), especially when both ends of the process are controlled, such delays could have the desired acoustic effect.
Jim, I am unsure if your question is more complex than this. But the simple answer is that a linear phase, brick wall digital filter has a perfect, infinite sinc impulse response. And, transformationally, the inverse holds true. To change the shape of the impulse response would mean to use a different digital filter, one that could have different frequency and/or phase responses.
and to eat it too 
