[carlseibert]

Hi,

I have stumbled on the realization that my Logitech/SlimDevices Squeezebox 3 sounds better when FLAC files are streamed in WAV format rather than in their own format. In other words, conversion on the server-side seems to work better than conversion on the client.

The symptom - or improvement if you prefer - is better high frequency performance. Treble content is better resolved and transients are better defined. The effect is quite subtle but repeatable. And is consistent in character with digital distortions - or the removal of them - that I have heard in the past.

My Squeezebox feeds my DAC via S/PIDF out. Nothing in the systems is changed except the where decoding takes place.

I'm fairly well convinced there is nothing magically or mysteriously wrong with the FLAC format, which some have suggested. I stumbled on this when I played an Apple Lossless file that I had made for my iPod and it sounded better than a FLAC of the same music. Given that the Apple Lossless was transcoded from that very same FLAC, it didn't seem likely at all that there was anything wrong with the data in the FLAC. But the .m4a is decoded on the server side and by default FLAC is decoded on the client. Thus, connecting the dots....

The question is: Has any other SB3 user run into this phenomenon? Has anybody done any experimenting with, say, a stiffer power supply that would shed some light.

If anybody else can reproduce this, I'll post on the SlimDevices boards to see if somebody really involved with the hardware has an idea what might be happening. Sometimes over there a question like this becomes a flame war over the audibility of a tree falling the forest and the engineers who might really know what's what flee like deer from a wildfire. Better to know going in if you have a reproducible issue.

There isn't any downside that I can see to streaming in WAV, except a pretty good hit on network traffic. That said, if performance can be improved in any mode, that would seem to be a good thing generally.

-Carl

 

[m1abrams]

Well before posting this on the SB forums, search their forums first. I know their has been many long debates on this very subject. You are not the only one to have these claims.

First thing you should do is make sure that you are sending FLAC to the client. As you probably know the server has many settings and one of them is to transcode FLAC to mp3 for low bandwidth networks. I believe that is the default for clients connected via wifi. Chances are that is not your issue, just saying make sure.

Also do you have the SB do volume control or is it fixed volume? I recommend setting it to fixed volume for your tests. Replay gain? Is that being used, if so it may be that the server is not applying it but the client is.

Second have you done an ABX? If not do not bother trying to argue the point.

 

[iriverdude]

You should get someone to change options without telling you what you're listening to. I've got a couple of SB3 into DAC. I have not changed file type settings at all.

 

[carlseibert]

M1 - Yes indeed, I was sending FLAC as FLAC. It seemed to be the logical thing to do - lossless, not that bad on network load. And frankly, it seemed to me that decoding on the client would be the best thing to do. It obviously wasn't horrible. In my main system, the Squeezebox sounded a tiny tad better than my CDP used as a transport. (For whatever reason. There is more than one variable there). I didn't have any major dissatisfaction. The "my file type sounds better than your file type" debate didn't seem like a profitable avenue to pursue compared to just listening to music.

And yes, I have the volume controls bypassed. (There are threads on the SlimDevices fora where some folks make a convincing-sounding mathematical argument that the volume control is safe to use, but again, I've never bothered to invest the time to experiment.)

I have wondered if the decoders for various file formats really output the same level from the same data. A decibel or so one way or the other could make a difference in the perception of musical subtleties. But I've always dismissed that as idle musing. I mean, why would they? It doesn't seem logical. But then that's an invitation to those stupid religious audio arguments. We all tend to make assumptions based on our limited knowledge but the systems in play aren't required to behave in accordance with my decades-old memory of high school physics, or the industry's inability to measure something.

-Carl

 

[rederanged]

It is exceedingly unlikely but not entirely impossible that the FLAC decoding algorithm on the Squeezebox is broken in such a way that it is spitting out something other than the unadulterated PCM data that was encoded. No offense, but it is far more likely to be your imagination.

You should really have some ABX data demonstrating that there is an audible difference if you expect the engineers to pay attention. They DO have the ability to compare their decoded data to the original, and I'm pretty sure they would have done this before they released the SB into the wild.

If it helps you sleep better, you can just decode to PCM or WAV on the server and be done with it. It does nearly double the amount of data going onto your network versus FLAC, but even uncompressed PCM is going to take only a few percent of your available bandwidth, assuming you have a decent wireless signal or wired Ethernet.

 

[m1abrams]

ok first I am not saying the volume control has anything to do with sound quality, just for when you are testing to eliminate an issue set it to limit. With the SB volume control is done in the 24bit domain so with 16bit files you have not have loss, and even with 24bit source it is very minimal.

Back to your thing, the decoders in the SB work correctly. Not saying you may not hear anything different. Other things may effect the sound, such as the extra load causing jitter, or taxing the PSU creating more EMI. However I am skepitcal that you are in fact getting a noticable sound difference.

Have you done an ABX?

Volume levels certainly play a role on how well the music sounds to humans. So if you are not sure your volume levels are matched between tests that right there is your issue.
Also note that the industry can measure every aspect of a signal within the audible range. This has been true for sometime now, there is nothing magical about audio signals, nothing unknown.

Now I am not saying you do not hear a difference something could effect the decode. However on the other forum people have tested this pretty well and found that their is no difference. Many people found they had setting flipped that they did not know.

Are you using replay gain, ie is your music tagged for replay gain. If it is and the client is config to use it then your difference is that. I do not think the server side decode uses replay gain.

 

[rederanged]

Quote:

Originally Posted by m1abrams /img/forum/go_quote.gif Are you using replay gain, ie is your music tagged for replay gain. If it is and the client is config to use it then your difference is that. I do not think the server side decode uses replay gain.

x2 -- that is definitely something to check.

 

[carlseibert]

Quote:

Originally Posted by m1abrams /img/forum/go_quote.gif Other things may effect the sound, such as the extra load causing jitter, or taxing the PSU creating more EMI. However I am skepitcal that you are in fact getting a noticable sound difference.

Exactly. And thus the original question, which was "is anybody else hearing this with similar equipment?"

No, I haven't done ABX, or any other kind of blind test. Frankly, I don't think that would have any value here. The question isn't to quantify what I can hear with my ears, my gear and my (hopefully) objective frame of mind. If any or all of those things is broken, we're just out of luck. We have no quantitative idea of the resolution or margin or error of blind subjective testing in audio. I've seen blind tests come back positive for subtle things that aren't supposed to exist, and we've all seen the marketing snake oil where blind testing "proves" there is no difference between preposterous combinations of things.

So, back to the plot. Levels are matched as far as the hardware is concerned - you can do this by switching the server config without touching any controls. My musing that the decoders might have a level offset was pretty much idle musing. I can't imagine that would get caught in open source applications pretty darn quick.

Anybody want to try switching back and forth on their gear and reporting back here?

Of course, if there's anybody out there who can measure known digital issues like jitter or maybe look at a squarewave through both paths, AND has a SqueezeBox and a nice DAC at hand, can we pique your curiosity?

-Carl

 

[carlseibert]

Quote:

Originally Posted by rederanged /img/forum/go_quote.gif They DO have the ability to compare their decoded data to the original, and I'm pretty sure they would have done this before they released the SB into the wild.

I doubt like all get out that the data is different.

I would suspect something altogether fuzzier, like a sagging voltage causing some unpleasantness in the output stream or RF finding its way to the DAC's audio circuitry. And the effect could be something local, like the receiver in my DAC doesn't cope as well as the next guy's.

So, no I wouldn't hold it against the designer if they didn't see this, even assuming that the phenomenon is happening on their end. Frankly, I would hold it against them if it comes down to a price point trade off. They do have a more expensive product on the market, after all.

-Carl

 

[rederanged]

Quote:

Originally Posted by carlseibert /img/forum/go_quote.gif I doubt like all get out that the data is different. I would suspect something altogether fuzzier, like a sagging voltage causing some unpleasantness in the output stream or RF finding its way to the DAC's audio circuitry. And the effect could be something local, like the receiver in my DAC doesn't cope as well as the next guy's. -Carl

OK, I get what you're saying now. I suppose that it is possible that something like that is going on. Do you know if you have to restart Squeezecenter to get settings changes to take effect? I'll try it and see if I can hear any difference in my system. (SB3 -> Parasound D/AC-800 -> Singlepower PPX3 SLAM -> HD650).

 

[rederanged]

I just spent some time switching back and forth with Dave Matthews' #41 (cleanly recorded, fairly dynamic and has detailed hi hats that can expose digital artifacts.) I discovered by looking at my task manager to see when flac.exe runs that I have to restart playback in order to get settings changes to take effect, so I can't switch arbitrarily in the middle of a track. This makes A/B testing a lot harder.

I can't say that I hear any difference between the settings. My gear, auditory memory, and ears are probably not that advanced, though. If you can tell us specifically in which tracks you find the most obvious differences, and what you hear, maybe someone will have the music available and can try to reproduce your findings.

 

[m1abrams]

Quote:

Originally Posted by carlseibert /img/forum/go_quote.gif No, I haven't done ABX, or any other kind of blind test. Frankly, I don't think that would have any value here.

Ok so that ends my contribution to this thread. Without at least trying to eliminate the biggest variable in the equation you have proven you do not understand how proper testing should be handled.

 

[carlseibert]

Hi Rederanged,

Since my last post, I've only been able to steal a half-hour listen, just before work, when I could stop and start the stream from my server. So, in that less-than-ideal framework - amp warming up and glancing at my watch all the time - I tried to jot down some effects you could listen for. I listened to one cut all the way through one way, and then again the other. I deliberately listened to the WAV stream first each time, because my own prejudice is to hear "more" the second time I listen to a song. Normally, I would listen for a longer time, say an hour or two each way, with varied or random music and then decide if I had a preference.

All that said....

I revisited the record on which I first noticed this, Rebbecca Pidgeon's The Raven. It's originally a 24/96 FLAC download from HDTracks, transcoded to 16/44.1 FLAC. On 'A Rose in Spanish Harlem', listen to the bass at the beginning. There's a "prrr" in the strings that's well resolved one way and not as much the other. Reverberation through the whole track is different/better, after the maracas or shaker comes in about half way through, especially. The effect isn't all that subtle on that recording. (I'm not that thrilled with this album, BTW. It's growing on me, but there are only a couple cuts I really like. Pidgeon was cool in "The Heist", though.)

'Because the Night' from 10,000 Maniacs MTV unplugged came up and I listened through to the middle of the next cut, the name of which I didn't write down. My notes say transients,,particularly bass transients, were a little better. Maybe. Effect on this cut was subtle to inaudible.

'A Simple Twist of Fate' from Dylan's 'Blood on the Tracks' produced a bit more detail - listen for fingers (presumably his own) on Dylan's guitar. And the whole recording sounded a tad bit less recessed.

On the Cowboy Junkies' 'Trinity Revisited' (definitely not random play. I love this album) I listened to 'Blue Moon'. Room sounds are different. There is a bit better sense of pace in Margo's vocals. Listen for tiny little details like amplifier sounds after Vic Chesnut comes in, and the harmonica at the very end sounds more "real".

I'm sure there are four-ton reviewer-approved phrases like "the envelope of air" for what I think comes down to a pinch less noise and better resolved, less time-smeared highs. I'm pretty sure that if your system is doing this, you'll hear it after a while. It's by no means the same magnitude as the difference you hear when you listen to a new and better amp or DAC and four bars in you think "Wow! I'm upgrading pronto!", but you'll be able to tell if it's happening.

Everything you hear or don't hear in audio seems to be tremendously dependent on what music is being played. I sometimes wonder when I get that four-bar-in rush if maybe I'm listening to the only thing that particular component can play well. Doing this I noticed I could hear and articulate differences much easier on the super-duper audiophile recording. I was able to point to specific sounds easily during very simple passages and not as much when the music was more complicated, even though my gut tells me a small improvement will do more "good" on complicated music. I really think when reviewers say that you have to listen to a component for months to make a valid judgment about it, they're probably right. Which means that working class slobs like me are generally taking a flyer every time we buy or build something and reviewers get to listen to borrowed gear they can't afford for long stretches of time. Such is life.

Back to the plot: I have a friend who has a Benchmark DAC1. By reputation, the Benchmark is very forgiving of jitter at its input. My friend is out of the country at the moment. When he returns, maybe we can have a listen to my SqueezeBox on his Benchmark vs my Musical Fidelity XDAC v8. It'd be pretty interesting to see if we hear a difference on one, the other or both.

-Carl

 

[rederanged]

The differences you describe are too subtle for my auditory memory to validate. I'm starting to think m1abrams is right, honestly: you really ought have some blind test results before you ask us to try to reproduce your results. Maybe that would be something you can do with your audio buddy with the DAC1?

 

[m1abrams]

The test you describe can be used very well to prove every single persons point as to why you must do a double blind test. I am not discounted your claim that transcoding or not changes the sound of the music. However not doing a blind test is a very careless way to test your claims.