I am myself a software developer, doing TCP/multicast network streaming for one of big guys in the SF Bay Area (aka Silicon Valley!).
Thanks. That was very interesting and explained clearly. Provides another complication in why some may or may not find differences between their FLAC and WAV rips, too. As far as this track goes, the differences in sound are very subtle, chiefly perceptible in how Fraser’s voice floats above the hash of guitar and underlying drum machine. Do you know the track? I often use it in demos because of this. It’s caused many a retailer to blench…. 
My history with this was wondering whether the Qobuz file would sound better than my domestic rip, partly because this was the only track I recalled sounding better on the CD555 than the ND555. I was surprised that the Qobuz file sounded very slightly worse, and frustrated with verifying which was which (had to display the filename and search for the underbars or lack thereof). That was when I noticed the difference in length. I have no idea of the why. Even the processing argument doesn’t convince me greatly, since I generally can’t tell the difference between FLAC and WAV of the same resolution…
Thanks, glad it helped!
My head is done in now, had a bad day at work because the customer IT departments who deploy our software (or try to) consist mostly of freakin’ imbeciles, which is why I am so giddyed-up that just kept going now in the evening 
So whatever I hear now is worthless. But I have your files and will try one of these days if I get around to it.
(I can try with UPnP without transcoding, meaning that the NDX2 actually sees the different flac files, and I can also try with transcoding to WAV on the UPnP server, so that the NDX2 sees the WAV that is the same from both, plus I can try with Roon which goes even further as its RAAT protocol sends essentially the raw PCM.)
OK this explains some things 
1 Like
From memory (which is to say that what follows could be untrue) I first noticed the difference in sound on Roon, using the Nucleus.
Good to know, thanks. Will try.
1 Like
That’s certainly so, both for FLAC and MQA (which is much smaller). And FLAC is still huge for hi-res, in particular 192/24, it has to be said.
Half the file size means that the streaming provider needs half he disk space and half the network data, and everything that goes along with other infrastructure costs, which means a lot of cost savings. Netflix, for instance, is investing a lot of money into improvements to movie file compression for this reason. (And I believe to have read that when a new movie codec was developed, they went and re-encoded their whole gigantic library with the new codec, which tells you something about how much it means in savings).
Probably also explains why Spotify is not too keen on duplicating their OGG library in 44.1/16 or even hi-res files.
1 Like
When I first dipped my toes into streaming, I thought so, too. Not with the default SBC codec, which really is poor, but with AptX I thought that this was pretty OK. It took me a while to realize that while nothing was obviously wrong with it, the music left me cold and I listened less and less.
1 Like
The only difference I can think of is that one of the files only needs to be transcoded, whereas the other also needs to be decompressed. But then I live in Lalaland….
You are correct about the PhoenixNet.
Oh, what a tease! I take you’ve never tried one, then… 
This is also my understanding.
There should not be a difference between the sound quality of something streamed remotely compared to a LAN. Buffer underruns can affect the stream but this will cause drop outs as opposed to subtle differences in sound quality. Once the bits are in, they’re in.
It is more likely that the music being compared is not exactly like for like. Who knows what a subscription streaming service, or the supplying label/distributor might do to the material to make it sound “better”? There’s just no way of knowing.
1 Like
… we might be at cross purposes … my post was specific to DSD (bitstream) processing vs LPCM when streamed from a NAS, not from a web service.
I was referring back to the question in the OP Mike, after responding to your point about how DSD is handled.
1 Like
Not as good as hi res, but completely listenable for none critical listening. Otherwise i am totally a 24bit snob!!
That’s what I meant, completely listenable but it still didn’t get me excited on a subliminal level, so I drifted out of listening. Once I switched to a proper streamer, I couldn’t get enough.
Edit: And this may well be because an NDX2/555 is more exciting than an AptX receiver 
but it’s how I experienced it, and without rigorous testing there’s no way to tell really. Besides, other things changed …
3 Likes
This has been covered previously by @Simon-in-Suffolk who is an expert in these matters. Differences were found when streaming Tidal compared to locally and this could be seen on his analysis kit. The bits are the same but inter-frame* timing was different and 1st generation streamers were more affected than the 2nd generation ones.
Bubble proxy servers suddenly became popular as these made Tidal to effectively be a local stream and the differences were no longer. 
- As I remember it was something like this.
Hi @Guinnless - @Harry says there should be no difference if exactly the same media (master) is being sent from the cloud and home NAS - although that is a big ‘if’, and I agree with that
However in the real world streamers and related equipment are not perfect devices and there can be interactions or interference with themselves caused by the process of the device itself operating - and certainly with the first gen Naim streamers I did measure different network frame timing patterns and how certain patterns created a certain ‘sonic signature’
I shared with Naim whilst they were developing the newer streamers - I have noticed the newer streamers don’t exhibit this in the same way - and to be fair the new streamers work differently in this regard ie how they spool the streamed data feed from the server.
2 Likes
Hi James, I just took a look at my settings, and it appears that I don’t have all the options you do. Particularly the “enable native DSD processing”. Not sure if I have to click on something else to activate that window?
You need to select the correct option from the “sample rate conversion” dropdown first.
