TIDAL can only be streamed up to 16-bit/44.1kHz FLAC on Sonos. Sonos doesn’t support MQA.

As stated above, I am not using Sonos Connect to decode or convert the PCM stream. I am bypassing the internal DAC.

Sonos Connect passes through MQA to my external DAC, which indicates that it is natively unfolding MQA (blue MQA LED on DacMagic). The issue is that the sample rate remains 44.1kHz.

Therefore, my question, once again: What is the maximum sample rate that Sonos Connect passes through its TOSlink and Digital Coax outputs?

Most MQA content on Tidal  is actually 44.1 inside a FLAC file, saying that fairly certain that the digital output gets converted to 44.1kHz, 24 bit.

may be of use to the OP

https://www.audiosciencereview.com/forum/index.php?threads/review-and-measurements-of-sonos-connect-streamer.8038/

How relevant is the above URL, Bell M? It is dated 2019 and Sonos have updated the software many times since then. My Sonos devices are all up to date with latest versions.

Does current (as of 24 Oct 2023) Sonos Connect S2 pass through 24 bit/44.1kHz over digital out without truncating to 16bit?

It’s relevant as it still current / valid.

There is no current Sonos Connect S2, insofar as it was replaced by Port on September 12, 2019. The ASR article is therefore just as valid now as it was then.

Connect (and ZPx0 before it) has always had a 24-bit/44.1kHz digital out. In Fixed Volume it was found to be bit-perfect, which I guess is why the cherished blue LED is lighting on the DAC.

I have to ask though: if sound quality is the objective why on earth is MQA being favoured? Thankfully it’s now going the way of the dodo, and Tidal is finally replacing it with genuine, unbutchered HiRes FLAC.

I am afraid, you have it wrong, Ratty: My Connect uses Sonos OS S2, according to the Sonos controller app on my smartphone (which app is also the S2 version).

Version: 15.9 (build 75146030)

The Connect has been updated at every Sonos OS update roll-out, so it is CURRENT in as far as the latest software. In this context, I repeat my question: How relevant is the above article, dated 2019? Sonos could have updated the software not to truncate 24 bit to 16 bit.

Does anyone has authoritative, as opposed to speculative guesses, at the correct answer?

As for MQA - I am not particularly hooked on it, but rather the question is much wider than MQA: will the Sonos Connect pass through 24bit at 48 or higher sample rates of ordinary (non MQA) FLAC files to the DagMagic over digital out, or will it truncate and downsample before passing the PCM stream to the DacMagic?

I don’t know precisely how the ASR 24-bit test signal was delivered but, sure, it could have been truncated. Sonos currently only offers 24-bit support from certain sources, such as local files, Amazon and Qobuz. For Tidal Sonos only claims 16-bit support, so perhaps your little blue LED was being illuminated by (horror of horrors) MQA-CD files? 

My point about the Connect/ZPx0 digital out however stands. The S/PDIF has always been 24-bits wide, to allow for the volume adjustment of 16-bit samples down to -48dB without loss of information. 

On the format support question, Sonos doesn’t support anything over 48kHz sampling rate. Its native sampling rate is 44.1kHz, and I have a suspicion that it may convert 48kHz to 44.1kHz before sending it out the S/PDIF. I haven’t looked in a long while. To be frank, if you’re wanting a bit-perfect streamer at higher sampling rates you’d need to consider alternatives.

At last - I have come across an official Sonos article about Tidal on Sonos. They support up to 16 bit/44.1kHz. No luck with better sample rates or bit depth

From their guide document: “ Stream CD quality audio (up to 16-bit, 44.1 kHz) ”

https://support.sonos.com/en-us/services/tidal?language=en_US

Which leads into the argument over how much resolution is sufficient…

Don't begin with “human ears can't hear beyond 20kHz”. That argument ignores the principle of superposition. Even though one cannot hear frequencies above 20kHz, if the signal amplitudes of such ultrasound frequencies are not at zero crossing at corresponding sample points in time of frequencies that we can hear, these ultrasound signals will either add to or subtract from the total signal amplitude - in other words, still contribute to the overall sound colour and detail that we perceive. But it is a game of diminishing returns. For 24 bit (and 16bit), beyond 44.1kHz the improvement decreases non-linearly. At 48 kHz you should still hear an improvement. At 96 kHz less so, and at 192 kHz, my guess is that one will be hard-pressed or need exquisite sound equipment to hear a clear improvement. So, perhaps, I should settle for 16bit/44.1kHz and downgrade my Tidal HiFi Plus to HiFi. Or buy a network streamer that does pass through 24bit 48kHz to my DAC.

Seems like Sonos, which has never sold itself as a ‘audiophile’ company, may not be the right choice for you. 

Internally the players have had 24bit data paths since inception in 2005, however, there was not enough internal processing power or network bandwidth to deal with 24bit high sample rate content. While there are advantages when using 32 bit high sample rate during post processing in the studio, there is a complete shortage of peer reviewed studies proving that anything beyond 44.1/16 is necessary for distribution. By the way some studio people feel that 192/24 is a lame, outdated, low sample rate format.

Rather than chasing word length and bit rate, you are better off chasing quality mastering and absence of compression. I’ve had UK, Japanese, and US LP pressings, CD’s, and Mobile Fidelity CD and LP of the same master tape to compare. They sounded so different that it was hard for me to satisfy myself that they all originated from the same studio master tape. The US LP’s and CD’s were clearly the worst. Carefully comparing the Mobile Fidelity CD and LP, when using a very high quality CD player and turntable ($200 units need not apply), the CD and LP were almost indistinguishable until the LP issued a ‘tic”. After enough listening it became clear that the CD had a lower noise floor, distortion, and a wider dynamic range. Casual audiophiles typically walked out mad after they misidentified CD or LP a few times, claiming that I somehow tricked them or that my equipment was not good enough.

Do you have a cite for the bold above?  Preferably from an accredited scientific journal?  

I am shocked at the horrible mixing and mastering out there. Some songs of the 80's, 90's and early 2000's are terribly bright. Unbearably so (reason - mass music consumption on poor quality portable music players needed boosted treble content to sound OK).

But further to the point of sample rate: A good bit-stream DAC can reconstruct the sampled points with very low error, but what about in between sampled points? Those in between points are approximated by some interpolation based upon assumptions. Nyquist-Shannon theory may state you need to sample only twice the highest interesting frequency (40kHz for 20kHz) - which holds true for a pure sine wave -  but in practice you need to oversample most real-world signals at least four times for reasonable reconstruction. So, 96 kHz sample rate means that you guess about half as many times as with 44.1kHz. Remember, what you have not measured (sampled), you do not know for a fact. You can only approximate it from what you do know (do have sampled). So, higher sample rates do bring more accurate reconstruction. But again, diminishing returns at rates increase.

Yet another benefit of higher sample rate is that it eases the stress on filter design. The higher the sample frequency, the slower the digital filter can be without risking aliasing folding into the audible spectrum (up to 20kHz). As you likely know, a slower filter (more gradual cut-off) has a more benign behaviour in that it has less pre and post ringing artefacts than a faster filter (faster cut-off). Reproduction tends to sound more like old analogue on LP and less bright, yet detailed - this one is subjective.

I agree with Buzz: Good recordings, good mixing and good mastering at 24bit/48kHz, released as 16bit/44.1kHz should sound quite brilliant on a good sound system. I have some such recordings and a medium grade sound system (Sonos Connect, QED Performance Graphite TOSlink, DagMagic 200M, Cambridge Audio CXA60 amp, Van den Hul Clear Water speaker cables, B&W DM1600 speakers (ages old). For CDs, I use an old Rotel RCD965BX CD player with digital out to DagMagic 200M (QED Performance digital coaxial interlink).

At an AES convention Tomlinson Holman presented a paper describing his preamp design criteria. He is a limited bandwidth believer. After the presentation he was challenged by a person who claimed to have observed 400KHz on a spectrum analyzer connected to his turntable -- therefore the 20KHz limit on the Holman preamp was limiting the music. Behind me there was a phono cartridge designer and a disk cutting head designer mumbling that “we cannot play these frequencies” and “we cannot cut those frequencies”, implying that the 400KHz was a spurious response possibly the result of a cartridge resonance. Tomlinson’s argument was that limited bandwidth results in less intermodulation and much cleaner output. Mixing 32KHz and 31KHz signals can result in a 1KHz, audible intermodulation product plus harmonics. In other words mixing of inaudible ultrasonic signals can result in a blizzard of audible trash.

The wide and narrow bandwidth camps will forever be at odds with each other.

A narrow bandwidth person has no need for sample rates beyond 44.1, nor do the ears of adults.

In the studio there is plenty of room for mathematical mischief during digital processing. High sample rates and wide words minimize the risks. 

https://www.britannica.com/science/principle-of-superposition-wave-motion

My request for a cite is for the part after "in other words".  Can you cite a study which proves superposition actually "colours the sound and detail that we perceive"?   Better yet, one which finds that it  improves what we perceive.  Because I know of several where hypersonics actually added intermodulation distortion in the playback stream, which is never beneficial to the sound.

Cite for the  bolded?

Not just should sound so, but also actually do. And in my assessment of what is a good sound system, after having used many others in the days of my audiophile pursuits now abandoned, I find that Sonos can be as good a sound system as those I have used, for the sound quality it can deliver.

Why then tie oneself into knots by succumbing to all the digital snake oil being hawked by marketers today?

To the quoted words I will only add: placed appropriately in a decent domestic acoustic environment.

Even Spotify streams are, with some rare exceptions, just as good to listen to as their CD equivalents, except perhaps where high quality headphones are being used for listening.

Other than Sonos Connect, all familiar names from a past life; what is medium grade about these bits of kit?!

And Sonos Connect belongs in this set for sound quality it delivers. Presumably, the Rotel CDP has analog line out as well, so the DAC magic interposed between that and the amp is another familiar audiophile fancy! I once had another magic box called a valve based output stage buffer which essentially was a impedance matcher, that had another separate box that was optional - a “better” power supply for the magic box…Boxes and more boxes

Kumar, I have seen ridiculously expensive kit displayed in hifi shops around Cape Town. My stuff are decidedly medium grade compared to those things.

Concerning the Rotel 965, I have run that one for years with analogue out. Great CDP. But the DagMagic has brought improved detail, texture, stage acoustics, and better composure when the music gets complicated and loud (when the 965’s built-in DAC would get a bit rough). It was most apparent when I briefly switched back to the Rotel analogue out after running with the DagMagic for a bit. The drop in treble clarity & detail was quite audible.

So, now the Rotel serves as a CD transport. More boxes, as you say.

Home audio of that type is based on the false notion that the more ridiculous the cost, the better it must sound. This taps into cognitive biases that afflict all humans. I would submit that your kit may be medium grade only in what it cost you compared to the kit you refer to, not in the sound it delivers in your listening spaces.

As to the DAC magic - you hear what you do but make sure that what you conclude is not driven by similar bias, and by sound level differences that need to be as little at 0.1dB to return similar improvements that you hear. 

The only way to be sure that the difference is real is if it survives a precision sound level matched blind listening test in a manner that is statistically reliable. 

Having installed the DAC and liking what it does, I am not suggesting you run such a test. But it might be something to bear in mind before making the next hardware uprade.

Sampling (signal processing) - Wikipedia

At each sample step all frequencies present superimpose and result in the total amplitude sampled at that sample step. The issue of precision comes with reconstruction in the DAC.

Note the remark in the above reference that the original signal can be reconstructed up to the Nyquist limit (half sample rate). Each sample is reconstructed with great accuracy by a bit stream (delta-Sigma) DAC. However, the detail between reconstructed samples are approximated by interpolation. The higher the sample rate, the fewer the in-between sample step approximations, hence more accurate reproduction.

Also, the issue of reconstruction and sample rates affects predominantly the upper audible range - above 10kHz, where much of the detail of acoustics and much of the tonal colour are represented for the human ear. Below 10kHz signal, the 44.1kHz sample rate is above 4 times over-sampled. E.g. at 1kHz, a 44.1kHz sample rate is 44.1 x over-sampled - ample, very minute interpolation errors during reconstruction. But at 20kHz, a 44.1 kHz sample rate is only just more than the Nyquist minimum of twice the recorded frequency - more scope for interpolation errors vs the case of a 1kHz signal. So, increasingly between 10kHz and 20kHz is where the issue lies with CD reproduction often sounding unnatural and bright. Here higher oversampling (48-192kHz) have the greatest benefit, if diminishing with raising sample rate.

Many players have a lot of leftover trash in their output -- due to leakage on the PC board, not because of any difficulties with the math. This trash results in intermodulation distortion. External DAC’s may minimize this trash. The external DAC could smooth out some clock jitter too.

Oversampling at playback cannot insert information that was never present, but it reduces potential mischief caused by the reconstruction filter. Too many people bolted out of the Nyquist-Shannon lecture flapping their arms, shouting “jaggies”, “jaggies”, “jaggies” and never stayed for the second part about the reconstruction filter.

I give up.  

The reconstruction filter is exactly where the interpolation happens that, in my view, is the root cause of the perceived tonal colourations of various DAC implementations. See the references below:

chttps://en.wikipedia.org/wiki/Reconstruction_filter]

ihttps://academic-accelerator.com/encyclopedia/reconstruction-filter]

A few DAC makers and implementers get it "right” to avoid bright treble, yet have full, true detail up to 20kHz. The rest never get it spot on. But given any reconstruction filter, I maintain my point that, above 10kHz, we run into increasing challenges for interpolation, because we are going below the practical over-sampling factor of 4 as we progress towards 20 kHz, increasing the demand on the approximation made between sampling points by the interpolation algorithm - the goal being minimal error between approximation and what the true amplitude would have been at this time step, were it sampled in the first place. The simple solution is higher sample rate, at the cost of larger files and increased demand on the electronics. Higher sample rate is not only to ease the design of anti-aliasing filters.