Direct Attach Cable Thread

audiobomber · June 19

There's a lot of interest in passive DAC cables, in various threads. This thread could provide a central knowledge base. My questions:

1. Are DAC cables brand-sensitive, like SFPs?

2. Can I connect a DAC cable from a 10G NIC to a 1G FMC?

It appears that the cable that Taiko recommends is available in the US:

https://www.fs.com/products/36699.html?attribute=8774&id=2370594

https://taikoaudio.com/taiko-2020/product/extra-dac-cable/

audiobomber · June 19

@Superdad explained why a DAC cable is a good choice, where length allows:

https://audiophilestyle.com/forums/topic/38968-etherregen-the-long-development-thread-some-gen2-dev-pics-and-update-starting-on-page-92/?do=findComment&comment=1282600

He said that a passiveDAC cable is the correct choice for audiophile use;

https://audiophilestyle.com/forums/topic/38968-etherregen-the-long-development-thread-some-gen2-dev-pics-and-update-starting-on-page-92/?do=findComment&comment=1282680

Passive vs. Active DAC cable clarification:

https://community.fs.com/blog/understanding-the-key-variations-active-vs-passive-sfp-cables.html

audiobomber · June 19

6 hours ago, audiobomber said:

He said that a passiveDAC cable is the correct choice for audiophile use;

https://audiophilestyle.com/forums/topic/38968-etherregen-the-long-development-thread-some-gen2-dev-pics-and-update-starting-on-page-92/?do=findComment&comment=1282680

Note that this link may glitch, goes to the post below, so you may need to scroll up one post. Chris is aware.

di-fi · June 19

9 hours ago, Superdad said:

Despite the construction of these cables being highly standardized (100-Ohm pairs mandatory, copper wire, PTFE dielectric, and shielding), that has not prevented one company from recently introducing a ridiculously priced $1,000 “audiophile” version with silver-plated wire.

That's exactly my point. Copper cables have undergone numerous iterations in the audio world, available at various price points and a range of SQ. Could it be the same for DAC cables? For example, a silver-plated DAC cable can also be found as a budget version on Amazon for $39.95.(Silver-coated copper wire Passive Direct Attach Copper Twinax Cable).

The two examples I mentioned in my earlier post might be exceptions, but when you take a look at both it seems that these cables are quite different and not standardized. This makes comparisons worthwhile if it leads to noticeable audible differences. Consistency between different lengths of the same brand, yes. But I’m afraid we find ourselves in the same situation again with different brands. I would indeed recommend to make a purchase only when you know what the DAC cable is made of (is it a standardised type?) or if you trust the ears of the person recommending it . Enjoy!

PS. feel free to transfer this post to the new thread by @audiobomber

Mops911 · June 19

1 hour ago, di-fi said:

Amazon for $39.95.(Silver-coated copper wire Passive Direct Attach Copper Twinax Cable).

This is QSFP, not compatible afaik...

did somebody find a solid-core TwinAx cable for DAC?

Technomad · June 19

I’d be surprised if you find one - solid core tends to be for permanent installs, not for drop cables.

Superdad · June 20

2 hours ago, di-fi said:

That's exactly my point. Copper cables have undergone numerous iterations in the audio world, available at various price points and a range of SQ. Could it be the same for DAC cables? For example, a silver-plated DAC cable can also be found as a budget version on Amazon for $39.95.(Silver-coated copper wire Passive Direct Attach Copper Twinax Cable).

While we truly doubt that either silver-plating or wire gauge are going to make a difference in SFP DAC cable performance, the dielectric used to insulate each wire in the two pairs very definitely matters. Look for Teflon, PTFE, or foamed-Teflon.

2 hours ago, di-fi said:

PS. feel free to transfer this post to the new thread by @audiobomber

I'd like to move your post and the two that follow it, but have moderator privileges/power only in threads here in the UpTone sponsored forum area. Can't move posts to public forum threads.

Superdad · June 20

11 hours ago, audiobomber said:

My questions:

1. Are DAC cables brand-sensitive, like SFPs?

Other possible compatibility issues aside, the issue of some brands of network switches (and NICs?) such as Cisco--requiring whatever is plugged into their SFP/SFP+ cage to have a hard-coded EPROM for identification--may indeed apply to usage of SFP/SFP+ Direct Attach Cables.

In fact, you can see here that the big reseller FS.com allows you to specify the equipment brand at each end for gear that requires such customization for compatibility. https://www.fs.com/products/156993.html

11 hours ago, audiobomber said:

2. Can I connect a DAC cable from a 10G NIC to a 1G FMC?

So here is the scoop--and this is important:

There is never ANY speed "negotiation" between gear connected via SFP/SFP+ ports!

If a 10G-capable box with cage sees an "SFP+" thing (fiber transceiver or DAC) plugged in, then it is going to send at 10G (exception only for equipment that allows you to reprogram to force 1G).

When shopping for Direct Attach Cables in the SFP/SFP+ form factor (as opposed to QSFP, SFP28, etc.), you will notice that for the same price ($11 at FS.com) you can buy either an SFP or SFP+ passive DAC--and they look identical. What is the difference? NONE, except a single pin tied to ground (I think in the SFP+ version since 10G came after 1G) which tells a 10G-capable device/switch/FMC that the "module" (optical transceiver or DAC) is willing to accept data at 10G.

[And as with fiber optic transceivers, you can use SFP+ modules between two 1G SFP ports, but not between 1G and 10G boxes.]

So to your question:

As with fiber transceivers, if you want to use a DAC to connect a 1G device to a 10G-capable device, you must use only SFP ends, not SFP+ ends. That will keep the 10G-capable box talking at just 1G.

It is that simple, and unlike with fiber optic transceivers, there is zero performance difference between SFP and SFP+ Direct Attach Cables.

TRHH · June 20

If you want to reduce potential Electromagnetic interference (EMI) noise in your setup, than DAC would not be the best solution.

Torben

i.deklein · June 20

8 hours ago, di-fi said:

That's exactly my point. Copper cables have undergone numerous iterations in the audio world, available at various price points and a range of SQ. Could it be the same for DAC cables? For example, a silver-plated DAC cable can also be found as a budget version on Amazon for $39.95.(Silver-coated copper wire Passive Direct Attach Copper Twinax Cable).

The two examples I mentioned in my earlier post might be exceptions, but when you take a look at both it seems that these cables are quite different and not standardized. This makes comparisons worthwhile if it leads to noticeable audible differences. Consistency between different lengths of the same brand, yes. But I’m afraid we find ourselves in the same situation again with different brands. I would indeed recommend to make a purchase only when you know what the DAC cable is made of (is it a standardised type?) or if you trust the ears of the person recommending it . Enjoy!

PS. feel free to transfer this post to the new thread by @audiobomber

@Superdad Please explane why passive DAC cables for streaming is better, compered to isolation via optic fiber. All regards.

Superdad · June 20

13 minutes ago, i.deklein said:

@Superdad Please explane why passive DAC cables for streaming is better, compered to isolation via optic fiber. All regards.

Hi:

I did explain in detail—a couple of pages back—how, in the usage context of connection between the highly isolated (clock/data/power) domain of one side of an EtherREGEN and an endpoint with SFP cage, a Direct Attach Cable can perform better, with less induced ground-plane noise and less added jitter.

But for other use cases—including the network feed to an EtherREGEN—optical transceivers and the isolation from leakage currents they provide will still be better than a DAC cable connection.

Please see here:

Superdad · June 20

1 hour ago, TRHH said:

If you want to reduce potential Electromagnetic interference (EMI) noise in your setup, than DAC would not be the best solution.

Usage contact is important. Elsewhere I wrote:

In the usage context of connection between the highly isolated (clock/data/power) domain of one side of an EtherREGEN and an endpoint with SFP cage, a Direct Attach Cable can perform better, with less induced ground-plane noise and less added jitter.

But for other use cases—including the network feed to an EtherREGEN—optical transceivers and the isolation from leakage currents they provide will still be better than a DAC cable connection.

I went a bit more in-depth beginning with some posts here:

By the way, I’m not very fond of using generic generalized terms such as EMI. I prefer to discuss the actual unwanted electrical sources, the ways they propagate, and how they actually impact the digital audio endpoint (i.e. the D/A converter’s master clock pin via ground-plane noise).

TRHH · June 20

27 minutes ago, Superdad said:

I prefer to discuss the actual unwanted electrical sources, the ways they propagate, and how they actually impact the digital audio endpoint (i.e. the D/A converter’s master clock pin via ground-plane noise).

Than there is a lot to cover: Mode conversion, common mode noise and EMI.

Torben

di-fi · June 20

9 hours ago, Superdad said:

I'd like to move your post and the two that follow it, but have moderator privileges/power only in threads here in the UpTone sponsored forum area. Can't move posts to public forum threads.

@The Computer Audiophile?

alecm · June 20

@i.deklein

Direct Attach Cables (DAC) do not use a five-layer transmission model (PAM-5) (2,1,0,-1,-2v), as Gigabit Ethernet does for RJ45. Instead, they use LVDS to provide high-speed and reliable data transfer between devices with SFP/SFP+ ports. This significantly minimizes noise and jitter

di-fi · June 20

Quote

A Direct Attach Copper cable or a “DAC cable” is a twinax copper cable terminated with SFP connectors that connects directly the SFP ports (or line cards) within active equipment, such as switches, routers, servers or data storage devices, in a data network.

For audio a DAC cable is typically used as the last connection to the endpoint. The length should not exceed 7m. and does not have a significant impact on SQ, however there other variables in the composition of a DAC cable that could impact sound.

Copper cables have undergone numerous iterations in the audio world, available at various price points and a range of SQ. Could it be the same for DAC cables? For example, a silver-plated DAC cable can also be found as a budget version on Amazon for $39.95.(Silver-coated copper wire Passive Direct Attach Copper Twinax Cable).

The two examples I mentioned in my earlier post might be exceptions, but when you take a look at both it seems that these cables are quite different and not standardized. This makes comparisons worthwhile if it leads to noticeable audible differences.

Consistency between different lengths of the same brand, yes. But I’m afraid we find ourselves in the same situation again with different brands. I would indeed recommend to make a purchase only when you know what the DAC cable is made of (is it a standardised type?) or if you trust the ears of the person recommending it . Enjoy!

The Computer Audiophile · June 20

16 hours ago, di-fi said:

PS. feel free to transfer this post to the new thread by @audiobomber

13 hours ago, Superdad said:

I'd like to move your post and the two that follow it, but have moderator privileges/power only in threads here in the UpTone sponsored forum area. Can't move posts to public forum threads.

Posts moved (correctly I think?)

audiobomber · June 20

7 hours ago, Superdad said:

Usage contact is important. Elsewhere I wrote:

In the usage context of connection between the highly isolated (clock/data/power) domain of one side of an EtherREGEN and an endpoint with SFP cage, a Direct Attach Cable can perform better, with less induced ground-plane noise and less added jitter.

I expect that even without the isolation provided by an ER, a DAC cable would be acceptable in a system where the two end devices are grounded to the same point, such that no ground current would flow between them through the DAC cable.

SQFIRST · June 20

From what I have gathered (please correct me), the use of a passive Direct Attach cable makes sense when everything else (EMI, RFI, Common mode, Power etc.) have already been optimized/reduced. This will most likely be at the closest, if not the final, connection to the audio endpoint. The advantage over other types of digital connection is that it uses low power and minimal electronics thereby providing least negative impact. And it is only possible when the connection is between two SFP ports.

R1200CL · June 20

1 hour ago, alecm said:

@i.deklein

Direct Attach Cables (DAC) do not use a five-layer transmission model (PAM-5) (2,1,0,-1,-2v), as Gigabit Ethernet does for RJ45. Instead, they use LVDS to provide high-speed and reliable data transfer between devices with SFP/SFP+ ports. This significantly minimizes noise and jitter

This is very interesting and good to know.

Then @Superdad or @JohnSwenson

Is there any difference in chips you decide to use in conjunction with the SFP cage ?

Does 10Gb has better specs that’s in our interest to have ?

And is there a difference between the opticalRendu and ER2 in this design that may affect jitter etc when using a DAC ?

With using proper PS, can we forget about ground loops between at least those two devices?

Is it possible one way or another to measure optical connections vs fiber between those two’s, or make some predictions maybe?

(whatever we want to measure and is realistic to do).

SQFIRST · June 20

1 hour ago, di-fi said:

This makes comparisons worthwhile if it leads to noticeable audible differences.

Without any doubt this will be going on! :)

Will be interesting to hear back what comes up.

di-fi · June 20

39 minutes ago, SQFIRST said:

Without any doubt this will be going on! :)

Will be interesting to hear back what comes up.

@MartinT no pressure, but you are invited ;-), @Mike Rubin?

di-fi · June 20

1 hour ago, SQFIRST said:

Without any doubt this will be going on! :)

Will be interesting to hear back what comes up.

I completely agree. In my opinion, one DAC cable can vary significantly from another, phisically I mean. I'm unsure if any audible comparisons have been done since Taiko started using the FS branded cable (maybe because few in that user group seem inclined to experiment).

As long as users are aware of the potential differences and possible sound quality variations between DAC cables, we can't be certain of the outcomes, can we? For instance, what impact might different shielding have? What about the connectors? I think it's important to know specifically because using this cable as the final connection should have the most influence on sound quality.

Mike Rubin · June 20

43 minutes ago, di-fi said:

@MartinT no pressure, but you are invited ;-), @Mike Rubin?

I will be interested in the conclusions. I no longer offer detailed opinions about what hear in my own system, though, as I recently have been fitted with hearing aids and because my room setup precludes my system ever from sounding its best.

To experiment, I bought this cheap cable:

https://www.amazon.com/dp/B097DH9MK6?ref=ppx_pop_mob_ap_share

The cable runs from an Optical Module Deluxe (not EtherRegen) to a Sonore Signature Rendu SE Optical. It was plug and play and worked immediately.

I just had my Sonore Signature SE upgraded and still am getting used to its radically different sound signature, but I already had reverted to Finisar SFP's from the Cisco AOC that replaced them in the SE before the modification. The new cable has been in the system just a couple of days and it clearly sounds quite different to my defective ears. Better? I think so, but I need more time to be sure. Particularly, there seems to be more mid-bass presence, which makes the presentation "weightier" and a bit more dynamic, but I have too many new variables in play and am just getting used to hearing aids, so take that all with a grain of salt.

Superdad · June 20

2 hours ago, audiobomber said:

I expect that even without the isolation provided by an ER, a DAC cable would be acceptable in a system where the two end devices are grounded to the same point, such that no ground current would flow between them through the DAC cable.

Maybe, maybe not.

So much depends on what else is going on in the box with the upstream SFP. It is not just about "grounds."

Speaking of which, when you say "two end devices are grounded to the same point, such that no ground current would flow between them through the DAC cable", it brings up the question of WHICH "grounds" you are speaking of: On the AC side of the power supplies in use? Or of the -Ve/zero-Volt "ground" domains of the power supplies for the devices?

And don't be too sure than ground and especially leakage currents would not still pass between the units. Likely via the shell/shields, not so much via the LVDS pairs.

But really, there are other "noise" elements, including induced phase-noise/jitter that could come from some average upstream SFP port.

What I am trying to say is:

Do not expect to be get great performance from a Direct Attach Cable between for example a $20 FMC and an EtherREGEN--even if you use linear power supplies on both.

3 hours ago, alecm said:

@i.deklein

Direct Attach Cables (DAC) do not use a five-layer transmission model (PAM-5) (2,1,0,-1,-2v), as Gigabit Ethernet does for RJ45. Instead, they use LVDS to provide high-speed and reliable data transfer between devices with SFP/SFP+ ports. This significantly minimizes noise and jitter

Correct. That's what I was explaining over in the EtherREGEN thread a week ago. All RJ45 ports have transformer magnetics and connect to PHY chip (even if integrated into the switch chip) that handles the multi-level encoding (3 levels for 100Mbps and 5 for Gigabit).

[The LVDS pairs at the back of every SFP/SFP+ cage are signals--from switch chip or controller or transceiver chip--in what is called Fiber Mode/1000BASE-X. It is pretty much the same as SGMII (used between Ethernet chips on a PCB) but with some differences including some negotiation if supported.]

But the topic of this thread, Direct Attach Cables, is really about the potential performance difference between usage of a DAC versus SFP fiber-optic transceivers. The latter put the LVDS pairs though complex circuitry, with a laser driver, at very high speeds, and received at the other end by pin-diodes. It is all pulses at very high speed and high current--so there is some generated ground-plane noise.

1 hour ago, R1200CL said:

This is very interesting and good to know.

Then @Superdad or @JohnSwenson

Is there any difference in chips you decide to use in conjunction with the SFP cage ?

For both the 'A' side port of the original EtherREGEN and the 'B' side SFP of EtherREGEN Gen2, there is not any transceiver chip between the switch chip and the LVDS signals (SGMII in 1000BASE-X "fiber mode") connected to the back of the cages.

But in the original EtherREGEN, there is not an instance of our special ultra-low-jitter (just 200 femtoseconds!) reclocking flip-flop between the main switch chip and the SFP cage (there is one on the 'A' side, but it is reclocking the data coming back from the 'B' side--so that all 'A' side ports are quite good).

In EtherREGEN Gen2, the data goes from the main switch chip (a newer, lower power part of a different brand) over the high-speed differential digital isolators sitting across the PCB "moat"/gap, and then into our reclocking flip-flop.

If the RJ45 is the port being used for the 'B' side, then the main switch chip is sending data over the moat in SGMII mode--through the flip-flop and then into an SGMII>UTP transceiver (PHY) chip to feed the RJ45.

But when the SFP cage is to be used (it is one or the other only: if ER Gen2 sees an active SFP connection at power on it turns off the RJ45 PHY and sets the main switch chip to send data in 1000BASE-X "fiber mode") the beautiful differential pairs (perhaps the lowest jitter 1G signals ever!) coming off the flip-flop go directly to the 'B' side SFP cage.

So from the above you can see that while it is neat for people to experiment with Direct Attach Cables from the original EtherREGEN's 'A' side SFP port (still fully isolated as long as you don't load up all the other 'A' side RJ45s with funky stuff), what we are most excited for--at least for users with SFP-cage endpoints--is the use of a DAC from EtherREGEN Gen2's SFP cage.

Apologies to @audiobomber for briefly draging your Direct Attach Cable thread back to EtherREGEN. There is just so much overlap.

(And you know how much I like to attempt to bring clarity to a subject--before it wanders off into random conjecture. This is an Audiophile Style thread after all--we are not at WBF or Head-Fi! )

Direct Attach Cable Thread

Recommended Posts

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Create an account or sign in to comment

Create an account

Sign in