Ethernet, USB, Optical or Coax to my DAC??

[4est]

If I may add to this, the ground is likely still connected through the power supplies/AC lines. As has been noted elsewhere, ground currents can create noise effecting digital triggering. Also, if the signal can go through something, so can noise and balanced signals are not a universal panacea. They tend to only reject noise that is on both lines simultaneously. That is great for 60hz hum, but much less so for RF.

Just to expand a little more with Ethernet connections, there is an isolation transformer built into every NIC, yes so a ground connection is not required and would cause more pain. The system is a true differential like an XLR if you like.

 

Now that is not available with USB, so the DAC needs to make its own "isolation", by other techniques, buffering, and locally re-clocking, then the issue is gone anyway.

 

Usually a shielded Ethernet cable has a limit of 70m whereas unshielded is 100m. The problem with Ethernet is getting the protocol conversion from TCP/IP to USB2, PS Audio failed with the Bridge, Sotm got it right with the 100-s (check the front page of CA, review is there).

[mac_and_dac]

If I may add to this, the ground is likely still connected through the power supplies/AC lines. As has been noted elsewhere, ground currents can create noise effecting digital triggering. Also, if the signal can go through something, so can noise and balanced signals are not a universal panacea. They tend to only reject noise that is on both lines simultaneously. That is great for 60hz hum, but much less so for RF.

 

Given the above, I am surprised that Toslink is not more popular as a device interconnection medium.

 

Quite a few negative things have been written about S/PDIF and Toslink as transmission media for digital music information, but galvanic isolation is surely a pretty large tick in the 'Positive' column?

 

I actually chose a glass interconnect between my transport and DAC based on listening tests alone, and only later came to understand the reasons why it can compete with supposedly superior electrical interconnects.

[barrows]

Is this true for the ESS Sabre DAC? My understanding from the WP on this chip is that it dumps the S/PDIF interleaved clock data completely and uses a separate downstream mechanism for timing the L/R channel data.

 

(See: http://www.esstech.com/PDF/sabrewp.pdf)

 

I use Toslink as the connection between my transport and my EE (Sabre) DAC - primarily for galvanic isolation - and the results are really, really good ... possibly due to this jitter correction technique.

 

Ref. the Ethernet comments elsewhere: Ethernet is a 'reliable' protocol, like TCP/IP, so occurrences of 'bad' data being transmitted should be pretty rare. And even if some PCM data were sent 'bad' or 'corrupted', it would surely exhibit itself in the form of clicks or pops.

 

ESS Sabre DACs use a built in asynchronous sample rate converter to re-clock the SPDIF input, John S. indeed addresses this approach in his post. In my experience, the ESS Sabre produces better results if one feeds it a direct I2S data stream (bypassing the onboard SPDIF receiver) with the ASRC disabled. Of course this approach means one best take care of the I2S feed and make sure it is very low in jitter to begin with, as the ESS DAC will not "fix" the jitter with the ASRC disabled. Before anyone tries to suggest that the ESS' ASRC cannot be disabled, understand that if one supplies a synchronous masterclock, both the ESS' DPLL and ASRC are essentially disabled; that is, those features do nothing when the bit and master clocks are synchronous.

[JohnSwenson]

The ESS chips have a builtin ASRC which converts the input sample rate to whatever local clock the DAC is using so it doesn't need to sync the local clock to the external data rate. BUT it does this with some pretty hefty DSP on samples, I personally do not like the sound of this, I think the "cure" is much worse than the "disease" in this case. The ESS chips can be run with the ASRC turned off, I think they sound much better this way, but then the DAC designer has to actually work hard to get clocking right. A few DACs with ESS chips do this but they are a fairly small minority.

 

TOSLINK connections DO block low frequency ground loop issues, but the receivers have a lot of jitter, they are not great. A well done glass TOSLINK can significantly lower the jitter compared to a plastic one (I don't have time to go into WHY in this post) but you still have to live with the jitter of the optical receiver. BTW very good optical receivers DO exist, but none of them are for TOSLINK (ie work at right wavelength and have the appropriate connector). It's certainly possible for somebody to make a REALLY good TOSLINK receiver, but so far nobody has done it. I guess the people that have the technology don't consider the market big enough. I do know a company that could easily do it, but they don't touch anything unless there is a $100million profit potential! Somehow I don't think the Audiophile market is big enough.

 

John S.

 

Is this true for the ESS Sabre DAC? My understanding from the WP on this chip is that it dumps the S/PDIF interleaved clock data completely and uses a separate downstream mechanism for timing the L/R channel data.

 

(See: http://www.esstech.com/PDF/sabrewp.pdf)

 

I use Toslink as the connection between my transport and my EE (Sabre) DAC - primarily for galvanic isolation - and the results are really, really good ... possibly due to this jitter correction technique.

 

Ref. the Ethernet comments elsewhere: Ethernet is a 'reliable' protocol, like TCP/IP, so occurrences of 'bad' data being transmitted should be pretty rare. And even if some PCM data were sent 'bad' or 'corrupted', it would surely exhibit itself in the form of clicks or pops.

[mac_and_dac]

Thanks guys - very informative stuff.

 

I actually looked a few months ago to see if I could by an uprated Toslink receiver to mod into my DAC, and found nothing. The OEM bits look so cheap. I work with telco kit: maybe I'll 'borrow' an SFP and see if I can tweak it to work as an Rx i/f for my DAC. ;-)

[tranz]

This is true for USB, but not for LAN, because data goes through line transformators in both sides- basically the lines are galvanically isolated.

Ground connection is possible only with grounded/shielded cables, mainly used in outdoor application (there grounded shield is needed because danger of lightning).

Ordinary LAN cables for indoor use don't have any shield - thats why ground connection is not possible.

 

Good point about the transformers, which is another benefit over USB.

 

But...a few more thoughts regarding cleaning up the ethernet.

 

- Not all transformers are built equally hence the existence of medical grade isolators

 

- Noise can enter at any connected point in the path. Perhaps a noisy switching supply on a PC or router connected to the network. Or even a fan connected to the same power grid.

 

When I turn on or off my ceiling fan motor, the spike impact is so large that packets of data are dropped and internet radio drops a second.

 

- Noise can enter through the cables being too close to e.g. electricity lines, motors, fluorescent lights/ballasts.

 

- Importance of cable quality in rejecting noise or be able to benefit from the differential structure of ethernet. This link is good reading:

 

Is Your Cat 6 Cable a Dog? -- Blue Jeans Cable

 

Cheers

[mrvco]

For all the debate over Twisted Pair Ethernet Cabling, has anyone integrated an SFP port into a media server or renderer so Multi-mode fiber could be utilized? I've had good luck with MMF links in high-RF environments for work and it is relatively inexpensive (especially in "audiophile" terms).

[jabbr]

For all the debate over Twisted Pair Ethernet Cabling, has anyone integrated an SFP port into a media server or renderer so Multi-mode fiber could be utilized? I've had good luck with MMF links in high-RF environments for work and it is relatively inexpensive (especially in "audiophile" terms).

 

Yeah. I think the problem is that audiophile companies wouldn't easily be able to sell fancy pure diamond pre quantum entangled platinum doped cables. And someone is sure to complain about sub-femptosecond jitter issues in the SFP transceiver.

[jabbr]

Good point about the transformers, which is another benefit over USB.

 

But...a few more thoughts regarding cleaning up the ethernet.

 

- Not all transformers are built equally hence the existence of medical grade isolators

 

- Noise can enter at any connected point in the path. Perhaps a noisy switching supply on a PC or router connected to the network. Or even a fan connected to the same power grid.

 

When I turn on or off my ceiling fan motor, the spike impact is so large that packets of data are dropped and internet radio drops a second.

 

- Noise can enter through the cables being too close to e.g. electricity lines, motors, fluorescent lights/ballasts.

 

- Importance of cable quality in rejecting noise or be able to benefit from the differential structure of ethernet. This link is good reading:

 

Is Your Cat 6 Cable a Dog? -- Blue Jeans Cable

 

Cheers

 

Feel free to use either shielded Cat 6a or Cat 7 cables, interconnects etc. Ferrite cores at the ends. There's no question that noise will trash a 10 Gbe signal. You can isolate a noisy server room with an optical SFP out to a clean audio room.

[jabbr]

For example: http://www.mellanox.com/related-docs/prod_adapter_cards/ConnectX3_EN_Card.pdf

 

Can be had in the $250 range if you look ... now tell me: who would consider a $1K ethernet cable over this?

[Habanero Monk]

Feel free to use either shielded Cat 6a or Cat 7 cables, interconnects etc. Ferrite cores at the ends. There's no question that noise will trash a 10 Gbe signal. You can isolate a noisy server room with an optical SFP out to a clean audio room.

 

Even better and certainly 100% obtainable in homes: Wire less.

 

Nothing to worry about at that point.

 

The only thing that can remotely introduce sound is induced, external noise on the wire. If you have zero of that then a cable is a cable is a cable at that point.

 

I had a September 27th date with a Polkie all setup but he declined at the two weeks prior. This was to see him compare his Belkin, another uber high end Ethernet cable of his, and my BJC certified Ethernet cables.

 

I'll consider the 'Audiophile Ethernet Cable' foolishness put to rest as I can't get anyone to agree to the testing I took a decent amount of time and materials to setup.