Weakest digital link: DAC - Streamer - Server?


I am curious WHICH of the three is MOST important in digital playback? DAC, Streamer, or Server? I assume DAC is #1, but what about server/streamer?

 

Specifically, I have a Synology NAS which I can host ALL my music on, then using a good streamer play off it via Roon/Upnp, PLUS my streaming services.  Does it make sense to use a SEPARATE server JUST for my music? 

Like is there going to be MUCH difference between these setups?!

1. NAS (music server) --> Good Streamer (ethernet, Roon) --> DAC

2. Dedicated Music Server --> Same streamer (ethernet, Roon) --> DAC

3. Dedicated Music Server with streamer (digital out) --> DAC

 

I understand for SURE, if one connects a NAS USB out to streamer/DAC it will sound WORSE than dedicated music server with low noise components, BUT if I am sending the music over ethernet to a good streamer, WHY would a music server make ANY difference? The streamer would essentially take the same files off my NAS, as if would from Tidal and play in the same fashion... no?! 

WHAT am I missing?!

 

PS. I can also upgrade the NAS with linear PSU to lower noise, but once again, sending audio through ethernet shouldn't matter.

alexb76

I agree with @electroslacker, except about buying the best cables. Especially a digital cable. As long as the cable passes bits it doesn’t matter. Goodness, my $8 HDMI video cable passes a whole lot more information flawlessly than my USB audio cable does.

In the past, some unscrupulous manufacturers have manipulated the standardized analog conversion formulas for their DACs. Basically adding internal EQ to them to make them stand out. That’s not what a DAC is supposed to do. I used to be a programmer and have written other types of conversion formulas. There’s no big mystery to them. Once they’re figured out, they’re figured out ... and standardized. Theoretically, all DACs should output the same, with noise, including jitter, being below audible thresholds. And I think ASR’s measurements continually prove that. I worry about the speakers and the room, not the digital components.

Especially a digital cable. As long as the cable passes bits it doesn’t matter...Theoretically, all DACs should output the same, with noise, including jitter, being below audible thresholds. And I think ASR’s measurements continually prove that. 

No.  Just...no.  If you can’t hear the difference between digital cables I truly feel sorry for you.  Jitter being below audible thresholds?  I guess timing then just doesn’t matter.  People who can’t hear well or don’t trust their own ears fall back on measurements to support the fact that they can’t hear differences.  How awful that must be as an “audiophile,” and I weep for you. 

Jitter being below audible thresholds?  I guess timing then just doesn’t matter. 

Today, for most digital USB in audio, i.e. USB and Ethernet, no, timing does not matter, not even a little. For SPDIF, most modern DAC chips and SPDIF receivers will remove every last bit of normal jitter. The rare time, very rare, make your SPDIF cable longer. Jitter stopped being an issue a long time ago in most setups.  I am sure there are some low volume DAC's that don't have the necessary circuitry to eliminate jitter.

Insulting someone's hearing is not going to make your point when you are unable to prove that your's is any better.

I worry about the speakers and the room, not the digital components.

Those are the two that will affect what your hear more than digital cables and such. As long as we are talking about competent design I think what a lot don’t seem to grasp is DAC’s, cables and to a great extent amplifiers are commoditites and they do very little if nothing to affect what you hear in your room. Get the features you like at the price you're comfortable with. 

A lot of people use amplifiers, tube, many low feedback designs, etc. to alter the tone of their system. Those will definitely sound different from amplifiers designed to be a straight wire with gain. Some speaker cables may be tone controls. I guess an interconnect could be too. Some MIT seemed to be.