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

Hi Kumar

Sonos DOES respond to volume normalization information contained in the metadata of individual songs. However, this CANNOT be turned off, so songs that follow on from each other on an album can have a volume jump.

In your audiophile days, if an amplifier was advertised as modifying all inputs by applying a 1dB compression, would you buy this in preference to one that was advertised as being a "straight wire with gain"? The Connect applies such a compression without the option to disable it.

Cheers, Peter.
Peter, in my old audiophile days, I would not touch even wireless with a barge pole, let alone what you refer to. Straight enough to not be heard as bent is good enough now.

I am wiser today and I don't care now about anything that I can't hear. Even more so, if it means added convenience and music access.

More importantly, if as you say Sonos responds to normalization information, why do I get volume changes from one album to another, both ripped or bought in iTunes? Or from ripped CDs to Apple Music streams?
"Straight enough to not be heard as bent..." - what a great phrase and philosophy - I love it!

The background to my concerns is this: I bought a new amp recently - a Peachtree Nova 125SE to replace a Cambridge Audio Azur 640a. I believe amps are all much the same, but the Cambridge was cutting out randomly, and the Peachtree has digital inputs that allow me to ditch an external DAC and clear up some cables. Surprisingly, the Peachtree sounds much better through my speakers (an old pair of Duntech Marquis). Two friends commented on this independently - it is a very clear difference. This was using the analog outputs of the Connect. When I tried the digital outputs of the Connect into the Peachtree DAC input, the bass on my favourite test song was substantially worse. It had lost a lot of punch. I tried a number of other songs, and the same difference was clear. So I decided to test the difference between the two setups. The first step in this test was to ensure the Connect passed the digital signal through unchanged. It doesn't. So now I'm left wondering whether the Connect is the cause of the problem.

The Connect should be "a straight wire with gain" so it cannot be suspected of being a source of audio problems.

To your other point. Coincidentally, my partner just complained that there was a substantial difference in volume between tracks she was listening to (not a single album - a Deezer "best of"). I think volume normalization can help somewhat, but it's not perfect. It is an amplitude correction that cannot fully correct for RMS differences caused by compression. Different albums with different mastering engineers use different degrees of compression. Almost all modern songs peak at very close to 0dB, but can have very different loudness as judged by RMS. No amount of amplitude correction can fully compensate.

Cheers, Peter.
the Peachtree sounds much better through my speakers (an old pair of Duntech Marquis). Two friends commented on this independently - it is a very clear difference. This was using the analog outputs of the Connect. When I tried the digital outputs of the Connect into the Peachtree DAC input, the bass on my favourite test song was substantially worse. It had lost a lot of punch.

I think volume normalization can help somewhat, but it's not perfect. It is an amplitude correction that cannot fully correct for RMS differences caused by compression


To the first part; I am sure you know that if you are comparing two bits of kit/sources that have one ending up sounding as little as 0.2dB lower than the other, the one with the lower level will yield sound that is perceived by humans to be thin and lacking body - what you describe as lacking punch may well be that. Now I am not saying that what you heard must be on account of the same reason and there was no other reason for lack of sound quality present, but to say the opposite needs sound levels to be matched to within 0.2dB, something that can only be achieved by instruments. After this is done, if you can still reliably pick out the two sources in a blind test, only then can you conclude that there is something different between the two sources and that the sound from the Connect is bent enough to be heard as such. Note that I am not saying whether it is, or it isn't; I am just saying what needs to be in place first to say so in a way that will be replicable and conclusive of something that is both measurably AND audibly out of whack.

Obtaining the same, measured to be within 0.2dB, sound levels from the Peachtree with digital outputs of the Connect may need the Peachtree volume control to be at a higher setting than when using the analog outputs of the Connect. If thereafter the sound cannot be audibly distinguished, the Connect is still audibly enough of a straight wire, even if it doesn't measure to be so on an instrument. If it can audibly be distinguished after the precise level matching, it isn't. And to avoid expectation bias, you should not know what outputs are in use. Admittedly, not an easy test to set up, but the only reliable and replicable one found thus far, AFAIK. As Hydrogen Audio also continues to say.

As to the part about volume normalisation, if I was to generalise, I'd say with the HighNote label for jazz, I find that I always have to intervene to reduce the volume a little in a mixed playlist. And for Apple music, I have to have the volume control at a higher setting than for ripped/downloaded from iTunes music from a NAS, so a mixed source shuffled playlist from NAS+Apple delivers constantly varying levels of sound and is therefore inconvenient. Should Sonos not be able to fix something like this? I have asked the questions elsewhere, but I am not optimistic of a response.
I agree that comparison of minor differences requires careful technique. The bass punch I'm referring to is not minor. You are right that the digital and analog cases require different volume settings. I have tried subjective matching as I switch. Sometimes I get it too loud, sometimes too soft. The conclusion is always the same - analog output sounds punchier. So, I think this is a case where the difference is sufficiently large that small volume mismatches will not confound the comparison. But as a scientist I remained somewhat skeptical. I was hoping to record the digital and analog signals into my recording software to do a more rigorous comparison. This was when I discovered that the digital signal was not bit-perfect. So I am stuck on this issue for the moment.

If you play the same track ripped via itunes, and streamed from Apple, do you notice a difference?

I just recorded part of a song while auditioning within itunes and compared to my ripped CD version. All this was in the digital domain on a PC. If I left soundcheck on in itunes preferences, the peak was about 3dB quieter than the CD version but the RMS was the same. If I turned soundcheck off, the peak was about the same but the RMS was about 3dB larger. What this means is that itunes is different to the CD version - I suspect itunes applies compression. The itunes version is also out of phase with my CD version - this isn't the case when I use Sonos rather than itunes.

As you can see, there are some interesting issues with streaming music. It's why I prefer ripped CDs from a NAS. If only the Connect would have an option to play them back as intended!

One final point - this is not reducing my enjoyment of music at all. I have listened to so much good music lately and enjoyed it all thoroughly. Some of it comes from podcasts at 128kbps, so I can't be all that fussy! If you get a moment, listen to a track called All Day All Night by the group River Whyless from their latest album We All The Light.

Cheers, Peter.
The bass punch I'm referring to is not minor. You are right that the digital and analog cases require different volume settings. I have tried subjective matching as I switch. Sometimes I get it too loud, sometimes too soft. The conclusion is always the same - analog output sounds punchier. So, I think this is a case where the difference is sufficiently large that small volume mismatches will not confound the comparison. But as a scientist I remained somewhat skeptical.

If you play the same track ripped via itunes, and streamed from Apple, do you notice a difference?

To the last easy part - yes I notice the difference, I think Apple Music is streamed to deliver a lower sound level in general.. This obviously isn't a problem when I am playing any one source - it just takes a one time setting of the volume control on the controller for the desired sound level to be consistently delivered - except for some CDs that are recorded hot. Now I haven't had the time to see if the same music, played via Apple also ends up louder to the same degree compared to other music from Apple, I need to get around to do this some day.

As to the first part - subjective matching doesn't work if test results are to be valid. A 0.2dB difference in sound level has been proved to yield punchier sounding results, so I am not sure this can be dismissed by saying that the punchiness is so large that it could not have been caused by sound level differences; seems like a cart before the horse thing to me, a non scientist. But more importantly, why are you stuck as you put it? Just use analog. And if you want to use digital for reasons of other conveniences, leave the volume control on the amp at a higher setting. No?
On the track volume issue, if you rip to wav format the songs will be louder because there is no volume normalization tag. When itunes rips to a format that allows metadata (I use Apple lossless - ALAC), itunes computes and stores the volume normalization in the track metadata and Sonos then responds to this. Turning off Sound Check in itunes only affects playback within itunes (the tags are always added to formats that allow them.) I wish Sonos had a similar switch! By the way, volume normalization almost always reduces the volume of a track. It couldn't raise the volume because most tracks have peaks close to the maximum, and a volume increase would lead to clipping.

Back to the listening topic. There is no way to use volume to make up for the bass difference I am hearing - I have tried. The attack of the bass cutting through the other instruments has a very specific frequency signature. Increasing the volume alters every frequency. I would like to use the digital input of the Peachtree to free up the only analog input for other purposes. Unfortunately I hear the sound as bent even though I don't want to.

This paragraph may be considered "too much information". My goal has been to record the analog and digital signals so I can demonstrate the difference. Recording the digital output is easy, and it can be lined up perfectly in my audio software with the original wav version. This is where I discovered the 1dB compression. Recording the analog output is also easy, but it requires A2D conversion. I use a fairly high quality audio interface with reputable A2D converters. I also sync the interface SPDIF clock to the Connect. However, there is a problem lining up the waveform with the original - the timing drifts. I didn't expect this, and I don't understand why it happens. It makes it much harder to compare waveforms objectively. I did try to reproduce my original plots and they had a lot of scatter, but I did see some evidence that the analog output also performs compression. When I get some time I will return to these issues.
When itunes rips to a format that allows metadata (I use Apple lossless - ALAC), itunes computes and stores the volume normalization in the track metadata and Sonos then responds to this.

This is where I discovered the 1dB compression.


My CD rips are all in ALAC, while iTunes purchases are in AAC, so the volume normalisation information is available to Sonos, I should think. I just don't think that Sonos is using this effectively in playback.

I hear what you are saying about bass frequency effects that you are seeing and how volume increases don't/won't work selectively. And if using Music Eq is what is needed to restore things - even if that is possible here - then the Connect has stopped being the straight wire, I agree.

A question though: do you think that the measured compression of 1dB that you are seeing will cause this selective frequency effect? Or is there something more that is happening here?
I also notice that Sonos still hasn't replied to my normalisation question posed elsewhere for a second time.

And is noticeably silent here. Either their engineers don't know what is happening or they are choosing silence.
Userlevel 7
Badge +21
I also notice that Sonos still hasn't replied to my normalisation question posed elsewhere for a second time.

And is noticeably silent here. Either their engineers don't know what is happening or they are choosing silence.


I get the feeling that they're choosing silence. I still don't know why they would remove the ability of a component with a digital output to provide the bit-perfect digital output that it used to be capable of, and was well lauded for. They should at least provide some kind of setting to allow bit-perfect digital output again, whether it's tied to the already existing "Fixed Volume" setting or a separate setting that can be enabled only when Fixed Volume is enabled.

Do I personally care about it? Not anymore... my Connect now serves as an input-only. But obviously there are others that do.
Now that's an interesting thought. I didn't realise that you have both ALAC and AAC versions of the same track on a local hard-drive. I thought you were comparing streaming to a local rip. If you have access to both versions locally, and they have different volumes, you can use mp3tag to look at the volume normalization tags. And you could also calculate the peak values and RMS. Better still, I could do that if you can find a song that exhibits this problem that I also have the CD for. Then I could purchase the itunes version and make a comparison. Best chance is a Miles Davis song - I have a lot of his stuff. I also have a fair bit of ECM jazz from some years ago.

I think it unlikely that the 1dB compression is causing this issue, but I can't rule it out either because I can't turn it off! However, look-ahead compression could theoretically act to reduce bass transients (which often have the most energy and hence largest peaks). My suspicion all along has been that the Connect DAC is better than the Peachtree, although most audiophiles would laugh at this statement. A slight confession here: I have recently bought an external DAC that comes highly recommended and is not too expensive (Schiit Bifrost). The idea was to settle the DAC issue once and for all (and this is all good fun as a retirement hobby). However, because the Connect doesn't pass the signal through unchanged, it makes this comparison much more difficult.

Cheers, Peter.
Now that's an interesting thought. I didn't realise that you have both ALAC and AAC versions of the same track on a local hard-drive. I thought you were comparing streaming to a local rip.

My suspicion all along has been that the Connect DAC is better than the Peachtree, although most audiophiles would laugh at this statement..


Peter, I am sorry if I did not explain correctly - I don't have the same track in both, I have bought AAC versions only of music I did not have in my now static CD collection. Now if you are referring to Miles, I have a lot of his music in both formats, but not the same in both. Something Else for instance is a ripped ALAC CD, while the other classic 50's album of his, Round Midnight, is a AAC iTunes purchase, and when both play from my NAS, there is no noticeable sound difference. But I recently bought a new HighNote album of Ron Carter and Houston Person, Chemistry, in AAC and that plays very noticeably louder. As do all other iTunes purchases of the same duo, a favourite. And as do other HighNote albums bought on iTunes - Person, as well as from other artistes on that label. Given this, if there is any purpose to be served in you looking at any of these files, let me know, I will be happy to provide them.

As to the Connect DAC - before I boxed my legacy audiophile kit after buying Sonos, I did a lot of testing, admittedly imperfect, but good enough for my decision making. I compared the sound quality from analog output of the Connect with that from my SACD player, both into high end amplification and Kind of Blue was one of the test albums. I found no difference in quality and decided that I did not need the SACD player. Nor have I heard of anyone that has been able to pick out the Connect DAC in a well constructed level matched listening test against any other DAC - not even ones with five digit price tags. I therefore believe that the Connect DAC is audibly straight enough. As are many others - indeed I strongly suspect that a modern DAC is no longer a factor, having become a reliable commodity at a low price point.

Ironically, going by your findings, it would appear that those that don't think so, and want to use the Connect into another fancier looking external DAC may be ill served by Sonos while doing this, with audibly poorer sound quality than they would obtain via the analog outputs on the Connect!

Do I personally care about it? Not anymore... my Connect now serves as an input-only. But obviously there are others that do.

I don't care about it either, for the same reason and in just one zone where I use it to drive the analog inputs of a legacy amp that is still working well.

But the inadequate way volume normalisation works just now affects all units and all zones, including those served by play units. That's the one I am looking for some response from Sonos.
When I read the thread over I found nothing definitive about why Sonos "degraded" the Connect. It can't have been for volume normalisation because that, as far as I can see, still doesn't work.

So why degrade the digital output at all?

And does this have any effect on sound quality where Connect is the first player in a group that is wirelessly broadcasting a digital stream to other players in the group? Is that stream not bit perfect any more? If so, this is a bigger problem than of being one only for such people that are using the Connect wired to the digital inputs of an external DAC? Based on my experience of the Connect used in that mode, I'd say that I haven't noticed any sound quality loss, but I don't have a reference baseline anymore, so I can't be certain!
Userlevel 7
Badge +26
I also notice that Sonos still hasn't replied to my normalisation question posed elsewhere for a second time.

And is noticeably silent here. Either their engineers don't know what is happening or they are choosing silence.


I just replied to that thread. We just hadn't read it yet.

When I read the thread over I found nothing definitive about why Sonos "degraded" the Connect. It can't have been for volume normalisation because that, as far as I can see, still doesn't work.

Normalization for tracks does work, but we'd be happy to help take a look with you to make sure it's working right with your system.

So why degrade the digital output at all?

We are talking about a minor change to the audio when applying normalization.

And does this have any effect on sound quality where Connect is the first player in a group that is wirelessly broadcasting a digital stream to other players in the group? Is that stream not bit perfect any more? If so, this is a bigger problem than of being one only for such people that are using the Connect wired to the digital inputs of an external DAC? Based on my experience of the Connect used in that mode, I'd say that I haven't noticed any sound quality loss, but I don't have a reference baseline anymore, so I can't be certain!

Short answer, no. When players are grouped together the audio that's sent across the system is the original digital file of the track, retrieved from the source by the group coordinator and then distributed untouched to the others. Each player uses its own DAC and settings on that original file once they receive it.

We are talking about a minor change to the audio when applying normalization.


Normalization is ALWAYS applied, therefore the minor change (compression) is also ALWAYS applied. I can defeat normalization by either removing the tags or using wav files. I CANNOT defeat the compression. It applies to everything I play through a Connect, and presumably every other component of Sonos.

Surely any hi-fi gear that forcibly altered the signal in this way would be reviewed poorly and eventually fail in the hi-fi marketplace. The best audio gear tries to reproduce the signal without adding its own coloration. Surely the Connect should be capable of passing through the digital signal faithfully and unaltered.

Please consider adding a switch so we can choose to disable the coloration, or at least return to the old setting where fixed volume output bypassed it.

Cheers, Peter.

Normalization for tracks does work, but we'd be happy to help take a look with you to make sure it's working right with your system.

When players are grouped together the audio that's sent across the system is the original digital file of the track, retrieved from the source by the group coordinator and then distributed untouched to the others. Each player uses its own DAC and settings on that original file once they receive it.

Ryan, very good, thank you. How do we proceed with seeing if things are ok with my system? It will be easy enough for me to set up a playlist with just two ripped tracks, one of which plays louder than the other. Do you need to know which tracks I will be using in advance?

Peter, to the second paragraph in the quote, this should mean that in grouped mode, your weakened bass problem ought to disappear when the Connect to digital input on Peachtree set up is in use and reappear when the Connect is not in grouped mode? Although Ryan confuses me by saying each player uses its own DAC. In your set up, the Connect obviously will not. Or does it mean that you will still see the bass issue via the Connect fed zone, but the other zones won't see it even if the Connect is the group coordinator? And in a later post you refer to this being a minor change - going by your description of how it affects bass quality, it is a MAJOR issue even if the compression may be minor?
Userlevel 7
Badge +26

Normalization for tracks does work, but we'd be happy to help take a look with you to make sure it's working right with your system.

When players are grouped together the audio that's sent across the system is the original digital file of the track, retrieved from the source by the group coordinator and then distributed untouched to the others. Each player uses its own DAC and settings on that original file once they receive it.

Ryan, very good, thank you. How do we proceed with seeing if things are ok with my system? It will be easy enough for me to set up a playlist with just two ripped tracks, one of which plays louder than the other. Do you need to know which tracks I will be using in advance?


I just opened up an incident for you with the support team to take a closer look into this. They'll want to see a couple tracks where you've noticed trouble like you described above and will likely want to remote in to take a look at the metadata of the tracks themselves. The reference number is: 160913-001584. I'll see if a technician can reach out to you directly, but feel free to give us a call on our support line with that number too.

Peter, to the second paragraph in the quote, this should mean that in grouped mode, your weakened bass problem ought to disappear when the Connect to digital input on Peachtree set up is in use and reappear when the Connect is not in grouped mode? Although Ryan confuses me by saying each player uses its own DAC. In your set up, the Connect obviously will not. Or does it mean that you will still see the bass issue via the Connect fed zone, but the other zones won't see it even if the Connect is the group coordinator? And in a later post you refer to this being a minor change - going by your description of how it affects bass quality, it is a MAJOR issue even if the compression may be minor?
My comment in regards to the DAC was if you were grouping the CONNECT with a PLAY:5 for example, the PLAY:5 would do it's own conversion. In short, when you group players they each get the same starting digital stream from the original source without any alteration done. Once the player gets it they will take it from there as if you'd played the audio right to the unit. tl;dr Grouping doesn't affect audio quality of a stream.
Hey Ryan - Thanks for the clarification. Please note that I think it unlikely the Connect DSP is responsible for the bass issues I hear, but I can't rule it out because I can't turn the DSP off.

How do we make progress on this issue? Should I open a support ticket?

Cheers, Peter.
Userlevel 7
Badge +26
Hey Ryan - Thanks for the clarification. Please note that I think it unlikely the Connect DSP is responsible for the bass issues I hear, but I can't rule it out because I can't turn the DSP off.

How do we make progress on this issue? Should I open a support ticket?

Cheers, Peter.


We're looking into it on our end here right now. We may want to get a ticket started up for you as well, but I'll let you know on that.
That's great news. Thanks Ryan.

I just opened up an incident for you with the support team to take a closer look into this. They'll want to see a couple tracks where you've noticed trouble like you described above and will likely want to remote in to take a look at the metadata of the tracks themselves.

Thank you, Robbie is in touch with me and I will post outcome details, once there is one.
Userlevel 1
Badge
It's been plenty long enough to fix this. First proven in July of last year. I have made up my mind to sell my Connect and get something else if I don't see this corrected (or at least an announcement) by the end of this month, January 2017. Along with it will go all recommendations of Sonos as a serious music system.

There are plenty of good solutions to play my music library and one of the lossless streaming services (I use Tidal) for serious listening. For the sources that Sonos is devoting all of their resources towards (Spotify, Amazon Alexa, etc.), I can play with my PlayBar in the great room as background music. Or on a Bluetooth speaker wherever.

Just guessing, but I think the Connect was just an afterthought anyway, likely sharing a subset of the software in the Connect Amp, which is why we have the output stream being mucked with. Sonos has not demonstrated they are serious about audiophile systems.
Just guessing, but I think the Connect was just an afterthought anyway,.
You are guessing. ZP80 (from which CONNECT is descended) was the second ever 'ZonePlayer' after the ZP100. Specifically designed to link to existing amplification, with or without an outboard DAC. It's been around since 2006. Hardly an afterthought.
Userlevel 1
Badge
Just guessing, but I think the Connect was just an afterthought anyway,.
You are guessing. ZP80 (from which CONNECT is descended) was the second ever 'ZonePlayer' after the ZP100. Specifically designed to link to existing amplification, with or without an outboard DAC. It's been around since 2006. Hardly an afterthought.


Okay, bad guess, but the issue remains.