Question

Connect no longer bit-perfect?


It looks like the Connect is no longer bit-perfect. Here's my evidence: let's discuss this.

First, I constructed a wav file of pink noise with amplitude ramping up from zero to digital max and back to zero.
I play this through my Connect and record the SPDIF output from the coax output into my PC.
The recording uses a Scarlett 8i6 audio interface set to use the Connect as master clock.
I record into a DAW (Sonar) multiple times - all instances are identical.
However, this recorded signal is not quite the same as the original wav file - it can be up to -21dB different.
See https://www.dropbox.com/s/t8od479xo9hi5el/connect_diff.PNG?dl=1
Note the expanded scale on the difference (third) track.

It looks like the difference gets larger when the signal is larger. To confirm this, I import the
original and difference files into Matlab and plot the raw data (difference vs original). There is clearly audio compression
happening here. See https://www.dropbox.com/s/p1yq6wcqafvnhaj/diff_vs_orig.png?dl=1
The scale is such that digital maximum is 1.

There also appears to be a slight bias when the waveform is negative and the signal is below the
compression threshold. See an expanded version of the previous plot
https://www.dropbox.com/s/9001tl9mkle4wly/diff_vs_orig_zoom.png?dl=1

Happy to answer questions about the method and conclusions.

Cheers, Peter.

p.s. Volume is set to fixed - I haven't tried variable.
In a loopback test (8i6 out from DAW to 8i6 in, no Sonos gear involved), I get bit-perfect cancellation.

This topic has been closed for further comments. You can use the search bar to find a similar topic, or create a new one by clicking Create Topic at the top of the page.

453 replies

We are talking about differences that, if they exist, are extremely subtle and are likely to suffer from a lot of noise and error (statistically speaking).

No, we're not... We're talking chalk and cheese differences in sound quality...

Or, put another way, "extraordinary claims demand extraordinary proof".

The claim is being made that Sonos kit can equal the performance of high-end hi-fi - that is indeed an extraordinary claim, and it certainly does demand extraordinary proof - from the person that made the claim.


The claim is being made that Sonos kit can equal the performance of high-end hi-fi - that is indeed an extraordinary claim, and it certainly does demand extraordinary proof - from the person that made the claim.


Why? This is a purely subjective opinion, Kumar has repeatedly said it is purely subjective and, depending on one's taste, could very well be perceived as the opposite by others. One should not have to "prove" their opinion, especially if they specifically deny they are making some kind of universal truth.


The claim is being made that Sonos kit can equal the performance of high-end hi-fi - that is indeed an extraordinary claim, and it certainly does demand extraordinary proof - from the person that made the claim.


Why? This is a purely subjective opinion, Kumar has repeatedly said it is purely subjective and, depending on one's taste, could very well be perceived as the opposite by others. One should not have to "prove" their opinion, especially if they specifically deny they are making some kind of universal truth.


A number of the statements that Kumar makes are not subjective opinions, they are stated as fact. If they were qualified (e.g. 'in my personal opinion') then I wouldn't be raising this - as you say, everyone is entitled to their own opinion.


A number of the statements that Kumar makes are not subjective opinions, they are stated as fact. If they were qualified (e.g. 'in my personal opinion') then I wouldn't be raising this - as you say, everyone is entitled to their own opinion.


Kumar replied to your accusation of this with the following:

Particularly when I have also often said that speaker selection is a subjective matter of preferences between one good and another just as good, best done by listening to it in the kinds of environments that it is designed for, and when it is set up to deliver all of its designed performance.


Time to drop the cudgel.
One should not have to "prove" their opinion, especially if they specifically deny they are making some kind of universal truth.

How does this apply to those who have other possibly controversial opinions about audio matters? For example, that MQA sounds better to them, or an external DAC sounds better to them than the internal DAC on a Connect, or that 24 bit sounds better to them?

Do we not need a consistent standard here?
I was specifically referring to ABX testing of formats like MQA, rather than Kumar's personal opinion of Sonos speakers against high-end hifi systems.

But even in that case, I would argue that if one were to take a fully scientific view, ABX testing always starts from the principle that the two sources are indistinguishable and tries to prove otherwise. no other approach is valid. In some cases proving otherwise is easy.

Regarding MQA, here is a rather interesting article on the subject: http://archimago.blogspot.co.uk/2016/01/measurements-mqa-master-quality.html

Cheers,

Keith


How does this apply to those who have other possibly controversial opinions about audio matters? For example, that MQA sounds better to them, or an external DAC sounds better to them than the internal DAC on a Connect, or that 24 bit sounds better to them?

Do we not need a consistent standard here?


As stated, opinions are fine. It's once someone starts making definitive claims that they need to start giving proof. But that's not really an apt description of what happens. What really happens is this:

Hi-res sounds better to me, it's night and day!

Well, that could be due to remastering, because biology, physics and math states it isn't possible for the Hi-res to sound better. Have you done an ABX with a downsampled version from the same master? Any time this has been tested in the past, the results pointed to the differences being in the mastering, so if you really can tell the difference, that would be a very big thing.

I don't have to, I know what I hear! Hi-res is better! Who are you to question my ears? Are you telling me my $$$ system is no better than your crap? Are you a trained listener? I have golden ears! People once thought the world was flat! Go away, you old know it all!

Compare this with Kumar, who fully admits that bias, $$$, expectations, shiny dials, and even a dram of tipple can can color one's opinion vs reality. Put another way, Kumar's claims are not extraordinary, it is perfectly possible for one to subjectively like the sound from Sonos over an audiophile system. Even so, Kumar makes no claim that his views are universal, he expressly states they are subjective.

Most Hi-res fans, on the other hand, make extraordinary claims which violate known laws of math, physics, and biology, yet they refuse to accept there may be other factors at work. No bias, no placebo effect, no mastering differences, it's all in the higher resolution.

See the difference?

PS - I deliberately used Hi-res and not MQA in my examples. I fully believe anyone who states MQA sounds superior. I'm also aware that given the MQA specs, it's more than likely this is due to remastering by Meridian than anything in the codec. However, since Meridian conveniently doesn't allow stripping out the raw data for downsampling to CD quality, it is unfair of me to request an ABX test. Meridian is the high-end of snake oil. Kudos to them.
Let me see if I can summarise your post:

If anyone states that, in their opinion, they prefer A over B, you are fine with this.

You would be more comfortable if they also stated that they understood the various subjective factors that could influence their opinion.

You are skeptical if their opinion conflicts with your understanding of the laws of biology, physics and maths.

If anyone makes a statement of fact, they need to back it up with acceptable evidence or it remains as opinion.

No-one should call you old 🙂 (I am with you on this wholeheartedly).
That's about right. Except change "your understanding of the laws" to "the laws", unless you have examples of where my understanding is incorrect? This stuff is not really up for debate, it's established science. By very definition, it would need scientifically rigorous proof to overturn it. Hence why we insist on it.

As for the old part, I couldn't care less. If they stoop to that childish stuff, I know I've won.
Good, I'm glad we essentially agree. I will, however stick with "your understanding of the laws". None of us has perfect knowledge of the laws that apply here. We can only judge someone's opinion according to our own understanding of the biological/physics/maths laws. There might be some real-world physics that complicates an otherwise ideal picture, and there might be gaps in our understanding. I'll give you two examples:
From my limited reading, I understand that DACs don't theoretically need to apply a low-pass filter prior to output. In practice most or all do. I'm yet to fully understand this.
The dithering plot on Monty's xiph web page is for a single frequency, and only goes up to about 9kHz. What happens beyond that, and what happens for multiple frequencies? Stay tuned for more on this.

Thanks for the discussion.
By the way, I'm not using the term "laws" correctly (though I'm not capitalizing it, so there is that). Being pedantic, it should say something like "established science in" not "laws of". I'm also specifically referring to things like being able to hear hypersonics, or saying the higher the sampling rate, the less jaggies in the sine wave. In other words saying things that misrepresent the science or make impossible claims. I and others agree there is a grey area in the bit-depth dithering argument, albeit one which will very rarely be applicable to real world recordings.

And I'll be the first to state I'm lacking in knowledge on DACs, which is why I rarely argue them, except to say that effective, high quality DAC chips are a commodity, so very high priced ones are the equivalent of high end cables. I also think that anything which adds color is either flawed (unlikely given my former point), or engineered that way, and thus questionable as to the motive behind the engineered coloring of the sound.

to subjectively like the sound from Sonos over an audiophile system.

Admirable summary, the quoted post.

I only differ on the part quoted above.

Why can I not claim that Sonos is also an audiophile system at certain price points? I would therefore rephrase to say - "to subjectively like the sound from Sonos over another audiophile system".

I would even drop the redundant word "subjectively". Why redundant?- leads to the second thing I want to say and some here will get the point I am making, about saying "in my opinion" all the time, recalling an episode from The Good Wife - some interesting viewing by the way, the terrible title notwithstanding:

Early in one episode the lawyer in the courtroom says something and the lady judge, a prissy and anal looking character interjects to say - " In your opinion". The lawyer, a smart lady, is taken aback and asks, "Excuse me"? Judge: "You should first say - In your opinion....." Lawyer hesitates and rephrases. This happens a couple of times and by now the jury is also rolling their eyes.

Sometime later, the judge says something and the lawyer stands up and says " Excuse me, your honour". Judge looks irritated and says" What?". The lawyer says "You mean, in your opinion". Jury sniggers. Judge looks like she is sitting on a haemorrhoid, searching in her mind to see if she can hold the lawyer in contempt, but reluctantly realises she cannot. And that is the last time this kind of dialogue happens in that episode.

:D

Back on topic: I checked my zp90 and based on the hardware version I see for it, there has been nothing done to it, so I presume it is bit perfect. But here is what Sonos ought to clarify, and isn't saying very clearly:

In my case, I should be seeing more effective normalisation for all my other kit compared to what I am seeing in the zp90 zone. If that was to be the case, it ought to be quite obvious because this is a lot easier to pick up compared to sound quality differences, particularly in grouped mode with other kit. No such thing though, I notice no difference in how normalisation works, or doesn't, across all my zones when grouped or otherwise.

On the other hand, someone with an affected Connect, should be seeing normalisation now work as effectively in the Connect zone as in all other zones. Has anyone found that to be the case?

And if not, combined with the kind of thing I am seeing with my untouched zp90, begs the question: why not just revert Connect back to bit perfect? No one has claimed any ABX tested way of proving that this matters, but at least Connect can continue to remain dressed in pristine white without feeling guilty about doing so!

Perhaps this subject is now of no commercial relevance to Sonos; they have largely moved on in life from Connect and even Connect Amp?
I'm lacking in knowledge on DACs, which is why I rarely argue them
The thing about DACs though that it is a lot easier to ABX them compared to say, amplifiers. For one, level matching is much easier in the kind of set up I used where it was a non issue; with amp comparisons, instrument based level matching is essential and cannot be avoided. The second thing about DAC comparisons that makes things simpler than with amps is that with DACs there isn't the amp/speaker interaction of the kind that can muddy the waters via things like sensitivity and impedance mismatches, clipping and the like.

By the way, I believe that the same archimago folks found that the DAC in the midget sized and USD 35 Chromecast Audio was as good as any DAC audibly needs to be. Not conclusive on the DAC subject of course, but yet another tested data point. In my hearing I found no differences in sound in casual testing; what I did not care for was the clunky UI of the CCA and absence of hard buttons.

If one can put up with that and the slightly industrial looks of the speakers, I would not be very surprised to see a set up like a CCA supplying music to a pair of JBL LSR 308 active speakers give a 5 pair a credible challenge at a much lower price point - even with CCA thrown in, at just half the price of the 5 pair.

From my limited reading, I understand that DACs don't theoretically need to apply a low-pass filter prior to output. In practice most or all do. I'm yet to fully understand this.

Peter, I am sure that there are many places on the net that has both knowledge bases and discussion fora on the subject of DACs and everything else related to audio. One I have found to be very useful at times is Hydrogen Audio, that you might want to look at.

Note though that it is a tough crowd out there that often expects you to have first read what is in their knowledge base. Compared to there, everyone here is a softie. Also read their terms of service including the (in)famous TOS 8.

If you want a quick lick at the flavour you will find there, see:
https://hydrogenaud.io/index.php/topic,112803.0.html
Kumar, I never stated you couldn't claim that Sonos is also an audiophile system at certain price points. Matter of fact, I explicitly stated you can subjectively claim anything you wish. Even more so, you have expressed a willingness to justify, or otherwise rationalize your claims, which is far more than others do. For this I applaud you. 🙂
I logged in to Hydrogen Audio after a long time today and found this on MQA:
https://hydrogenaud.io/index.php/topic,107666.0.html
The thread starts a couple of years ago, and over its two years nothing new has turned up in it on MQA. Early in the thread there is an interesting statement by a poster, quoted below:
"This format only exists to intentionally deliver low fidelity music to consumers without Meridian hardware, to sell them expensive Meridian decoders."
The thread might be interesting reading for those thinking of pursuing the MQA road, for some robust views on the subject.
I also found some talk in the thread on the Stuart rejection of the archimago assessment of MQA - I did not read that in any depth.
Regarding DACs, I think there is an important distinction to be made between the DAC chipset and the DAC appliance.

The DAC appliance obviously contains a DAC chipset, but also contains many other components including analogue pre-amplifiers which may or may not colour the audio over and above the capabilities of the DAC chipset. A DAC may have a world-class DAC chipset but if, for instance, the analogue output stages are badly designed then the DAC could sound bad. I'm not saying this actually happens in practice (high-quality designs and components are pretty much a commodity these days), but it *could* happen, hence the important distinction.

I'll mention I studied DAC chipsets briefly almost 30 years ago when I did my EE degree. I've not studied them in detail since, and there have been a lot of changes and improvements since then, so I'm certainly not an expert.

I have, however, done some research on DAC chipsets a few years ago to try to understand the capabilities, the market, etc. So I can state the following:

Pretty much every audio DAC chipset made in the last 15 years is capable of 24-bit resolution. The vast majority are also capable of higher sampling rates than 44.1/48k.

Pretty much every audio DAC chipset made in the last 15 years will be "flat" within the normal limits of human hearing and will have very low noise (well below 16 bits). Basically, it can be assumed that the DAC chip does not colour the sound significantly, and that all DAC chips sound the same. If there is colouration in an implementation, it's due to the design and other components of the system (e.g. output preamps) rather than the DAC chip itself having a specific "sonic signature".

A "world class" audio DAC chip (whatever that is) costs a few dollars in bulk. The difference in cost between "high-end" expensive DAC chips and cheap commodity ones is a dollar or two. Usually the more expensive chips are more expensive because they include more capabilities (e.g. buffering, DSP, etc.) or more channels (e.g. for surround sound support), not because they sound better.

I thought it was important to bring these up because there is, IMO, a huge amount of mystique and BS around chipsets.

The same happens in the guitar world where people claim that certain production runs of "vintage" op-amps sound better in distortion pedals than the modern versions, even though those op-amps have been in continuous production with the exact same design and manufacturing process since they were launched.

Cheers,

Keith
A DAC may have a world-class DAC chipset but if, for instance, the analogue output stages are badly designed then the DAC could sound bad.
Pretty much every audio DAC chipset made in the last 15 years will be "flat" within the normal limits of human hearing and will have very low noise (well below 16 bits).
If there is colouration in an implementation, it's due to the design and other components of the system (e.g. output preamps) rather than the DAC chip itself having a specific "sonic signature".

Where does the anti aliasing and other required filtering happen? is that another thing external to the chipset like the output stages that you refer to are? Or is it part of the chipset?

Where does the anti aliasing and other required filtering happen? is that another thing external to the chipset like the output stages that you refer to are? Or is it part of the chipset?


The filter is usually on the chipset, although many chipsets allow you to bypass the onboard chipset and use an external one.

Arguably, the filtering is one area where these chips could have sonic differences. I'm personally not convinced: the digital filters used in modern DAC chips are actually very good and most of them use similar designs. If there is a difference in this respect it's not a massive one, and it's not strongly related to the cost of the chip.

Cheers,

Keith


The filter is usually on the chipset,

Less scope for an integrator to screw up things if this is the case, I would think?
I am almost certain that this would be the case for something like the Chromecast Audio device, seeing its form factor. In fact it might even be an integrated board with amps on it as well. The fact that this little device, that sells for just USD 35, also turns in the kind of transparent performance via its analog outputs that it has been tested and found to do, indicates the state of affairs in general on the DAC subject. In my opinion.:D
It is my understanding that DACs also need an analog filter (called a reconstruction filter) on their output. The conversion of discrete digital information into analog voltage 44100 times per second leaves a jagged signal, and this must be smoothed to provide a continuous voltage with no aliasing or supersonic artifacts. This filter has a tough job - it has to be steep enough to go from zero gain at 20kHz to something very low at 22.05kHz (less than -60dB?) Such a filter will potentially have phase problems. This is where oversampling can be very helpful. I don't know how Sonos implement this filter, or whether it is an integral part of the DAC chip.
I wonder where Sonos implement their brickwall limiter? Older Connects don't have it, but software versions prior to 6.0 don't either. So it would seem to need a mixture of hardware and software. If software can activate it, presumably it comes before the DAC, so it is in the digital domain. In which case, why is additional hardware needed?

I presume that the brickwall limiter applies to all Sonos units, not just the Connect. Does this mean it is just software? Or are older Play units immune to it because they lack the right hardware?
If you are a "high end" audio company, you can just leave the reconstruction filter out, lol.

http://www.stereophile.com/content/audio-note-cd-41x-cd-player-measurements

To his credit, technical reviewer Atkinson trashes it in his summary. The subjective reviewer, of course, gave it a rave review. Typical audiophile nonsense.

To his credit, technical reviewer Atkinson trashes it in his summary.

Does he? I am not so sure, these are his last words:
"does it sound good because of how it measures or despite it?" - not ruling out that it sounds good, is he?
This from a USD 12K bit of kit. Is this a digital version of the preference among some for distorted valve amp sound?
Typical audiophile nonsense.
I came across this today: "It's audiophile only if it's inconvenient". It seems to me one could also say: It's audiophile only if it causes anxiety. And It's High end only if it is expensive.