Toslink, Coax, AES/EBU, USB 2.0 audio, HDMI??


What connection is best for Hi-Fi sound and why? I cant seem to find a Definitive answer. Or does this just come down to application?


Matt M
128x128mattmiller
Post removed 
The toslink glass fiber may have been stated to be better than toslink/coax, however toslink is still a compromised transfer mechanism.  If anybody has googled around, the engineer from Lampizator did some scope tests and showed that the square waveforms from toslink weren't that square at all.  The problem with toslink is not the glass medium, but it is that the LED light-emitting elements just cannot light up fast enough to produce that perfect square wave pulse that the Digital COAX cables can.
---
USB vs COAX is a crapshoot - in some instances COAX is better, but there are some target devices which, apparently, perform better with USB.  Obviously, COAX cannot transfer DSD.
----
COAX s/pdif vs. AES/EBU.  In theory, AES/EBU should be better because it's a balanced digital signal just like balanced XLR audio.  However, I have read some discussion where  COAX will perform better because of proper 75 ohm termination and cable shielding.
---
For the most part, HDMI is the absolutely worst way to transfer digital audio because of ::  1. PCM audio data "words" are broken up and split between HDMI data transmission words, therefore completely breaking the origination "clock" pulses.  The HDMI target device has to receive multiple "HDMI words" at HDMI clock timing and attempt to re-assemble them back into the proper original sampling rate (i.e. 44.1khz, 96khz, etc.).  .... and 2. compressed Dolby Digital and DTS is further compressed because of sharing bandwidth with video data.  Old school compressed DD/DTS is still much better over digital COAX.   The only source that is best for HDMI is the high resolution bluray formats: Dolby TrueHD and DTS-HD MA.
---
The exception for using an HDMI cable is when it is used for i2s digital audio in devices such as PS Audio and Wyred 4 Sound DACS/transports.  However, in this case, it is no longer "truly HDMI audio".  The HDMI cable is just used as the mechanism to transfer i2s data, which is the best way to transfer PCM data.  Even s/pdif COAX is an encoded version of the pure i2s source data.
Sarjan over at 6moons just did a comparo on this very subject:
http://6moons.com/audioreviews2/i2s/1.html

He did his best to keep it equal due to the different cabling and what's available and came away preferring i2s.

All the best,
Nonoise

Given the same quality of devices:

I2S (HDMI connector):

I2S is best because it usually connects directly to the D/A, bypassing any substandard internal reclocking in most DACs. This means that you can drive it with really low jitter devices like this and get superb results:

https://www.audiocircle.com/index.php?topic=156409.0

A high-quality I2S cable is necessary.  Usually at least $800

The downside is that it is difficult to do galvanic isolation for I2S, although there is one add-on HAT board for the Raspberry Pi that does this.

AES/EBU:

AES/EBU has the potential to be second best but the problem is finding a really good low-loss cable with matched impedance throughout, including the connectors. The connectors and wiring inside the DAC are usually not 110 ohms either. The positive is that it has some level of differential common-mode noise rejection. If there is an internal reclocker in the DAC on this input, this will probably limit the performance you can get.

S/PDIF coax:

S/PDIF coax can be outstanding provided that you use a good BNC-BNC cable (RCA adapters as needed) and use a DAC that does NOT have a substandard reclocker internally on the coax input. Such DACs are the Metrum line, Audio Note and others. With these DACs, one can add an external reclocker in the coax cable like the Sychro-Mesh to achieve truly low jitter and therefore excellent imaging, dynamics and liveness:

https://www.audiocircle.com/index.php?topic=157348.0

USB:

USB is the most problematic of the bunch. You must jump through a lot of hoops to get really stellar performance. It is possible though, using the right playback software. A critical piece is the USB cable (I like the Wireworld Platinum) and an external USB cable regenerator like this:

https://sotm-usa.com/collections/sotm-ultra/products/copy-of-tx-usbultra-regenerator-1

Two USB cables are needed with this device.

A well-designed low-jitter XMOS USB interface is critical and galvanic isolation is desirable. Most built-in USB interfaces in DACs are not great IME.

Toslink:

Forget it. Not good enough.

One that you did not mention is actually the best of all of these:

Ethernet:

You can get external Ethernet renderers or Ethernet built-into a DAC. Some support Roon and others support DLNA/UPnP. IME DLNA SQ is better. The computer playback system does not matter, but the playback software does. I personally use freeware; Linn Kinsky/Minimserver/BubbleUPnP. Runs on Linux, Mac and PC. The one requirement is that the router or switch be AQVOX type or similar and that the Ethernet cable from the switch to the Ethernet interface be high quality. I use Wireworld Platinum myself with superb results. Ethernet beats everything else I have heard at shows and in my own products. Here is an external renderer, the Interchange that achieves 10psec of jitter at the end of a 4-foot coax:

https://www.audiocircle.com/index.php?topic=156409.0

You can stream or play local files with Ethernet renderers. Some are built-into the DAC, like my Overdrive SX.

Steve N.

Empirical Audio

What about using DAC build App (such as Oppo Sonica, Oppo105)?  No external wire involved. Directly XLR line-level out to an integrated amp?
Srajan over at 6moons just did a comparo on this very subject:
http://6moons.com/audioreviews2/i2s/1.html

I would agree for most products.  One question is whether the adaptive FIFO reclocking is on all inputs or just some.  Cannot find anything on this from Denafrips.

Steve N.

Empirical Audio

What about using DAC build App (such as Oppo Sonica, Oppo105)? No external wire involved. Directly XLR line-level out to an integrated amp?

Depends on the implementation/design.  With the right design, separates and cables can sound every bit as good, even better than all-in-one.  It's because of the separate power supplies and galvanic isolation you can get with separates and cables.  These things lower jitter and noise.

Steve N.

Empirical Audio