Which digital interface sounds best to you?

[KeithL]

I'm a little confused here guys....

According to the description "SINAD" is ONLY a variation on the standard S/N ratio..... specifically it's a variation on the standard "THD + N" measurement.

It is NOT even claimed to be some overarching indicator of audio quality - unless you consider THD and noise to be the ONLY important parameters.

Therefore, it's just a single specification, albeit a potentially useful one.

Oct 31, 2018 11:25:53 GMT -5 garbulky said:

You have to also consider the room/equipment. Something as esoteric as jitter is going to be hard to be heard if there are bass humps, room reflections etc - if it can be heard at all. So everyone should ask - is my setup good enough to even hear it if it's there?

As for Audioscience review keep in mind that currently their BEST performing DAC tested is the Oppo 205 their measurements shows is nearly perfect in measured performance. It tested 20 spots higher than Benchmark DAC1. Now reflect on that and your actual experience with that DAC and ask me which finding is more accurate for your experience?



So here you'll see that Oppo 205 according to their measurements is number 1.  The Benchmark Dac 1 is NUMBER 20 which is EIGHT spots below an LG smartphone. You'll also find the Schiit Ygdrassil  near the bottom of the list a few spots below an HP LAPTOP onboard sound card. So what I'm saying is ....take that site with a grain of salt. 

[Gary Cook]

I voted for USB because my BigEgo sounds great. It also uses a TI/Burr Brown DAC, so maybe it’s not the USB connection that I like the sound of, it’s the DAC implementation

Cheers
Gary

[KeithL]

I wouldn't go so far as to say that it's useless.... but they are rather overstepping.

It looks to me as if they've attempted to create a single "sound quality" metric.... sort of like a "Consumer Reports product rating".... only based on a few technical measurements.

As far as I can tell, they seem to be claiming that, by combining THD+N with all other non-harmonic distortion, in a single metric, they are basically showing "a ratio of what should be there to what shouldn't be there".

The premise being that "whatever has the best ratio of what should be there to what shouldn't be there must be the best overall".

(Obviously they're choosing to ignore everything they didn't bother to measure, and the fact that not all flaws are equally important, or equally audible.)

But, hey, it's "really sciency", and everybody knows that you gotta believe anything that's presented in a clear, colorful, easy to read graph, right? 

Oct 31, 2018 13:01:28 GMT -5 DYohn said:

Oct 31, 2018 12:39:54 GMT -5 garbulky said:

Supposed to be signal to noise and distortion ratio

Something they made up? Why not just list S/N and distortion? Actually, the "noise" part of S/N is by definition part of 'distortion" so all you really need is the THD+N spec. But again, this chart is absolute rubbish and tells me nothing.

[KeithL]

I tend to put most of these little gadgets in the same class as power filters.

Under some circumstances, and with certain other equipment, it's quite possible that some of them actually solve problems or improve performance.

However, under most circumstances, most of them range from unnecessary to useless.

A power line noise filter is a very useful device.... but only if  you have both noisy power lines and equipment that is sensitive to power line noise.

However, many manufacturers of various audio tweaks go out of the way to convince potential customers that their particular widget will improve the sound of ANY system - including yours.

If you have an actual problem - then their gadget will help...

And, if you don't think you have a problem, then you must have a problem which, even though you can't hear it, is limiting the sound quality of your gear in subtle and sinister ways.

(It works out especially well if they can convince you that their gadget has fixed an imaginary problem that you didn't have to begin with; after all, it's really gone, right?)

And, yes, some of them are such pure snake oil that the ONLY problems they can solve are imaginary ones.

Oct 31, 2018 10:03:55 GMT -5 Boomzilla said:

Oct 31, 2018 8:50:54 GMT -5 KeithL said:

I think you need to be much more specific...

And after reading your comments, Keith, I think that such specificity just won't be possible. As you point out, the number of variables is far greater than I suspected, which may well be why the Audioquest Jitterbug and the Schiit Eiter that I've tried didn't do too much for me - My DACs were good enough that I didn't much need what the conditioners were doing, and so I heard little to no difference. All three of the DACs I use (Emotiva Stealth DC-1, Mytek Liberty, and the built-in DAC of the Arcam AVR550) are asynchronous and well-buffered.

[405x5]

Oct 31, 2018 8:50:54 GMT -5 KeithL said:

I think you need to be much more specific... because the interfaces on DACs themselves differ a great deal.

It's not at all as if "all USB inputs perform and sound the same" and "most S/PDIF inputs sound and perform the same".

There is actually a lot of variation between different examples of both.

For example, virtually ALL modern DACS with USB inputs SHOULD have an asynchronous USB interface - which has "an accessory to remove jitter" built in (the data is clocked by the input).

Assuming that's working properly, it should "wipe away" most jitter that is present at the input (but may or may not be especially resistant to noise, and will have some low "floor level" of jitter of its own).

Older DACs, and a very few boutique modern DACs, have the old-style interface, where the clock is regenerated using PLLs... and which is far inferior.

Likewise, many modern DACs, like our DC-1, and the new DC-2, have an ASRC built in - which is also "an accessory to remove jitter" (with Sabre DACs a similar mechanism is built into the DAC chip itself).

So, with all of those, you are NOT "comparing interfaces without any accessories to remove jitter".

Rather you are comparing them including the jitter reduction mechanisms that are already present inside the particular DAC...

(Some can be disabled, and you will notice a difference with them enabled or disabled, but some cannot be switched out.)

And, like everything else, there are better and worse examples of each different type of input.

In fact, on some of their models, Audio G*D used to offer multiple different options for each type of input... 

They offered both different USB input chip-sets and different options for S/PDIF receiver chips.

And Schiit Audio, as I recall, is on their fifth version of their USB interface for their upgradable DACs.

So..... by asking it like that you're sort of asking: "Which tastes better: fruit juice or soda?"

So, really, you should request answers including the particular model and brand of DAC.

I also find your results interesting....

WITHOUT ADDITIONAL INTERNAL JITTER REDUCTION MECHANISMS, and ALL ELSE BEING EQUAL..... (both unlikely conditions).....

- an asynchronous USB input should have the lowest jitter and be most immune to jitter from any upstream components

- S/PDIF should have very low inherent jitter, but is somewhat sensitive to jitter reaching it from upstream

- HDMI connections are relatively bad in terms of jitter

- and old-style NON-asynch USB should be one of the worst (the connection itself has a lot of inherent jitter; and the input mechanisms those DACs use - mostly PLLs - vary widely, but none are especially good at it).

- ALL audio sent over an Ethernet connection MUST be extracted and re-clocked at the destination... so ti depends entirely on how well the receiving device does its job.

Ground noise is another issue entirely.

Devices with actual galvanic isolation on some inputs are very tolerant of noise... but many devices without isolated inputs are also tolerant for other reasons.

While USB-powered DACs tend to be especially sensitive to it (especially if you're connecting them to a separate system which has its own ground, rather than to headphones, which are inherently floated).

I should also point out that jitter affects different DAC architectures differently.

For example....

- Sabre DACs have something more or less equivalent to an ASRC built-in

- Many multi- bit DAcs are claimed to be less sensitive to jitter (for a given amount of jitter present at the chip they produce fewer and lower sidebands than an equivalent D-S model; this makes them "very forgiving" of certain signal flaws).

- From a high level, as a data format, DSD is inherently far more sensitive to jitter than PCM (because it's easier to "look at the data and clock separately" with PCM).

To answer the question.... for myself personally.....
I find a well designed asynch USB input to sound flawless.
(There's no real way to perform a direct comparison... since you're always dealing with different upstream components... but I haven't found anything that I consider to sound better than asynch USB in general.)
And I find S/PDIF inputs to have the ability to be very good, and just as good as USB with a good source, but to be sensitive to what source you use (and that includes things like re-clockers connected between them and the source).

Oct 31, 2018 8:11:50 GMT -5 Boomzilla said:

WITHOUT any accessories to remove noise or jitter, which interface do you find to sound best?

I vote for HDMI (High-Definition Multimedia Interface) with coaxial digital SPDIF (Sony/Philips Digital Interconnect Format) coming in a close second.

I find USB to sound the worst.

What he said

[qdtjni]

Oct 31, 2018 8:11:50 GMT -5 Boomzilla said:

WITHOUT any accessories to remove noise or jitter, which interface do you find to sound best?

I vote for HDMI (High-Definition Multimedia Interface) with coaxial digital SPDIF (Sony/Philips Digital Interconnect Format) coming in a close second.

I find USB to sound the worst.

USB is worse than S/PDIF which is worse than HDMI from the same device or are differences from different devices?

[donh50]

AES (or S/PDIF, the consumer version) has limited error correction and no resend capability. They are also more limited in the amount of data they can handle compared to HDMI or USB. Asynchronous USB with adequate power/ground isolation should work great, though many devices do not have adequate power/ground isolation. I'm with Keith; there are too many variables to just say one is best or not. HDMI targets video and has by far the highest (intrinsic) jitter for audio and like all the others power/ground isolation depends upon implementation.

As for SINAD, that is probably the most widely-used metric outside the audio world (which seems focused on THD, how limiting ) and is defined by things like the IEEE Standards for ADC and DAC testing (including audio -- I helped write some of them). THD+N is more limited and mostly done wrongly in my experience; most test equipment lumps other spurs in with the noise rather than doing any kind of separation so what they actually report is SINAD anyway. The ASR site presents a lot of measurements on the devices Amir tests so pulling out just one is a bit misleading, and a little reading will reveal that most DACs provide way more performance that what we can audibly distinguish. (Disclaimer: I post there and have a series of technical articles posted, and my day job has me doing things like noise, SINAD and jitter measurements from DC through 10+ GHz.)

As for TOSLINK, here ya' go: Toshiba Link en.wikipedia.org/wiki/TOSLINK

IME when USB sounds worse it is due to a few things that aren't really the fault of the USB protocol:
(1) Poor power/ground isolation, an implementation issue. Good DACs do not have that problem.
(2) If the DAC is not asynchronous, jitter on the incoming data is directly translated to the audio output. Few DACs have that issue these days, though still may have clock problems of their own.
(3) OS problems leading to dropouts and latency in the link. Asynchronous DACs with buffers generally handle that problem.

Most any of the interfaces can be great or lousy, it really depends upon the implementation. They should all deliver the same bits to the DAC in the end if they work properly. The latter is always the question...

FWIWFM - Don

[Casey Leedom]

Hhmmm, with regard to "Error Correction", are you talking about things like CRCs + Retransmission? Or are you talking about Forward Error Correction which can correct some errors with redundant information and detect others? If it's the former, I'd be surprised if there would be enough time in a Streaming Audio application to request a new copy of a damaged packet. And if you're hoping for the latter, I don't believe that USB has any standards for FEC.

Casey

[donh50]

USB ain't my day job, nor is AES for that matter. I know more about PCIe and SAS/SATA error correction, but frankly I'm an analog guy so don;t claim expertise for those either. AFAIK, AES does (and thus S/PDIF and TOSLINK do) not have packet retransmission, and most USB DACs don't either. I am sure there are experts here who can 'splain it. I was not talking about CRCs or parity checks built into the packets, and I'm with you in that I don't think any of these have any sort of FEC. CDs have CIRC included, I don't remember what AES has. In any event I do not think data transmission and reception is usually an issue for any of these interfaces as far as sound for consumers is concerned; I think it is more about implementation of power/ground isolation and maybe CDR (clock and data recovery, though jitter would have to be fairly high to be audible). After that, choice of filter implementation (digital and analog) and the output circuitry probably has much more impact than how the bits hit the DAC (the actual DAC device inside the box).

After skimming some of the earlier posts, this is not the sort of thread for an engineer...