Wavelength Ethernet Spacelator

[tboooe]

Have you guys seen this?  I dont understand all the technical jargon but it sure sounds cool!  @Superdad curious on your thoughts since Uptone is developing a similar product.

 

https://parttimeaudiophile.com/2018/10/10/wavelength-ethernet-spacelator-rmaf-2018/

[stevebythebay]

Seems this one will front end the DAC or say a dCS Network Bridge or similar Roon Ready or other D-D converter.  Looks like Gordon is after the same sorts of filtering and reclocking.  But there's little else to go on at this point.

[firedog]

This isn't a switch. It's a reclocker and filter. It  filters ethernet thru optical ins and outs, as far as I understand. Good power supplies and regulation, etc.

Not exactly the same as the Uptone product, if I'm understanding what both of them are.
 

Anyway, it looks like Uptone is going to be on the market first, and if past efforts mean anything, the Uptone product will cost less. 

 

Both products sound interesting. I'll certainly want to hear what early adopters think of them. 

[Superdad]

5 hours ago, tboooe said:

@Superdad curious on your thoughts since Uptone is developing a similar product.

 

2 hours ago, firedog said:

This isn't a switch. It's a reclocker and filter. It  filters ethernet thru optical ins and outs, as far as I understand. Good power supplies and regulation, etc.

Not exactly the same as the Uptone product, if I'm understanding what both of them are.

 

Hi Gents:

 

Well there is a great deal I could say--regarding the topology and functioning of our in-development EtherREGEN--but doing so right at this moment would be a strategic business mistake since revealing the data paths, isolation methods, and clocking placement would give away too much.

 

Gordon too is purposely and rightfully being vague as well, so there is not much to comment on with regards his design. But I will say this:

Much as with USB, there is no such thing as a "reclocker" for Ethernet.  With USB such is done with a hub chip (or a USB hub core in an FPGA) to "regenerate" the data through a protocol engine.  Likewise, Ethernet must go through some form of MAC processor.  Certainly Gordon could pick from numerous forms of Ethernet MAC processor chips, or simply an Ethernet switch chip and use one port in and one out. We chose to use an Ethernet switch chip and to provide a few ports because it does not add much to the cost.

 

But our choice of Ethernet chips (yeah it looks like we will use two because of some latest thinking), is very specific, based on availability of a certain form of MII ports for internal chip-to-chip transmission (I'm purposely not saying). Yet what matters as much or more is the type of isolator (we prefer GMRs as they are much lower jitter than optical isolators), the lines run to them, and the very high speed, ultra-low-jitter differential flip-flops used after them.  Plus low-jitter clock synthesizers on both sides of the "moat," too many LT3045 regs throughout, and a small pile of other tricks--are all adding up to one rather expensive BoM (bill of materials)--far beyond what we have seen in the recent crop of switches modded with OCXO clocks.

 

Clearly I've already said more than I intended (it's just hard not to because we are excited about this thing!).   

I just want to strongly caution folks about categorizing some of these devices (especially not ours) as switches, filters, reclockers, when technically there can be so much more--or so much less--going on inside.

 

Cheers,

--Alex C.

  

[ssh]

Hi Alex,

 

I have two LPS 1.0 feeding an ISO REGEN and an UltraRendu into a Yggy (mods up to date). The switch is a Netgear FS 108. Will the new switch, or the LPS 1.2 upgrade be the best bang for the dollar?

Thanks.

 

 

 

[audioking87]

Something similar:

https://www.hifistatement.net/tests/item/2362-artistic-fidelity-isolator-und-usb-audio-interface

 

 

[AudioDoctor]

Considering my past experience with Wavelength products, I am really excited to try this in front of my Aurender.

[Superdad]

2 hours ago, audioking87 said:

Something similar:
https://www.hifistatement.net/tests/item/2362-artistic-fidelity-isolator-und-usb-audio-interface

 

Sorry, not similar at all.  That is a $1,400 USB isolation system.  The current topic is about Ethernet isolation.

[audioking87]

Ah ok this has Ethernet as well for isolation. Good luck. 

[Cebolla]

On 10/10/2018 at 9:25 PM, Superdad said:

Well there is a great deal I could say--regarding the topology and functioning of our in-development EtherREGEN--but doing so right at this moment would be a strategic business mistake since revealing the data paths, isolation methods, and clocking placement would give away too much.

Gordon too is purposely and rightfully being vague as well

 

6 hours ago, Superdad said:

The current topic is about Ethernet isolation.

 

Indeed - please don't be tempted to name your isolator a spacelator too!

[Albrecht]

On 10/10/2018 at 1:25 PM, Superdad said:

 

 

Hi Gents:

 

Well there is a great deal I could say--regarding the topology and functioning of our in-development EtherREGEN--but doing so right at this moment would be a strategic business mistake since revealing the data paths, isolation methods, and clocking placement would give away too much.

 

Gordon too is purposely and rightfully being vague as well, so there is not much to comment on with regards his design. But I will say this:

Much as with USB, there is no such thing as a "reclocker" for Ethernet.  With USB such is done with a hub chip (or a USB hub core in an FPGA) to "regenerate" the data through a protocol engine.  Likewise, Ethernet must go through some form of MAC processor.  Certainly Gordon could pick from numerous forms of Ethernet MAC processor chips, or simply an Ethernet switch chip and use one port in and one out. We chose to use an Ethernet switch chip and to provide a few ports because it does not add much to the cost.

 

But our choice of Ethernet chips (yeah it looks like we will use two because of some latest thinking), is very specific, based on availability of a certain form of MII ports for internal chip-to-chip transmission (I'm purposely not saying). Yet what matters as much or more is the type of isolator (we prefer GMRs as they are much lower jitter than optical isolators), the lines run to them, and the very high speed, ultra-low-jitter differential flip-flops used after them.  Plus low-jitter clock synthesizers on both sides of the "moat," too many LT3045 regs throughout, and a small pile of other tricks--are all adding up to one rather expensive BoM (bill of materials)--far beyond what we have seen in the recent crop of switches modded with OCXO clocks.

 

Clearly I've already said more than I intended (it's just hard not to because we are excited about this thing!).   

I just want to strongly caution folks about categorizing some of these devices (especially not ours) as switches, filters, reclockers, when technically there can be so much more--or so much less--going on inside.

 

Cheers,

--Alex C.

  

Hey, hey, Alex....

Now you know that you're getting the kids all wound up about this thing. You just put a while bunch of Godiva Espresso beans out on the table here.....  Please don't get us bouncing off the walls too early, - or there's going to be TONS of clean-up.   ?

 

Needless to say, - can't wait for this thing.....

 

 

[stevebythebay]

Alex, unless I’m not reading the tea leaves properly, it sounds like like you guys are still in what seems a development phase, with various designs still in testing. So, likely still many months before final prototyping, right. Which means next quarter for shipping product, right?

[plissken]

I would love to put some of these isolation devices through a testing rig:

 

Setup a Cisco 2960 Switch:

 

CAAUDIOSWITCH(config)#int r f0/1-2

CAAUDIOSWITCH(config-if-range)#channel-group 1 mode active

CAAUDIOSWITCH(config-if-range)#

Creating a port-channel interface Port-channel 1

 

CAAUDIOSWITCH(config-if-range)#int port

CAAUDIOSWITCH(config-if-range)#int port-channel 1

CAAUDIOSWITCH(config-if)#swi

CAAUDIOSWITCH(config-if)#switchport mode access

CAAUDIOSWITCH(config-if)#switchport non

CAAUDIOSWITCH(config-if)#switchport nonegotiate

 

This will put ports 1/2 on the switch into a LaCP LAG

 

In turn install an Intel two port server nic in teaming mode (also LaCP active).

 

This way with port aggregation in place we can actively unplug one Ethernet cable at a time with out the listener knowing and see if any one can actually hear the difference.

 

[Arpiben]

10 minutes ago, plissken said:

I would love to put some of these isolation devices through a testing rig:

 

Setup a Cisco 2960 Switch:

 

CAAUDIOSWITCH(config)#int r f0/1-2

CAAUDIOSWITCH(config-if-range)#channel-group 1 mode active

CAAUDIOSWITCH(config-if-range)#

Creating a port-channel interface Port-channel 1

 

CAAUDIOSWITCH(config-if-range)#int port

CAAUDIOSWITCH(config-if-range)#int port-channel 1

CAAUDIOSWITCH(config-if)#swi

CAAUDIOSWITCH(config-if)#switchport mode access

CAAUDIOSWITCH(config-if)#switchport non

CAAUDIOSWITCH(config-if)#switchport nonegotiate

 

This will put ports 1/2 on the switch into a LaCP LAG

 

In turn install an Intel two port server nic in teaming mode (also LaCP active).

 

This way with port aggregation in place we can actively unplug one Ethernet cable at a time with out the listener knowing and see if any one can actually hear the difference.

 

A few remarks:

- for implementing LACP protocol you need both devices able to deal with it.

- in principle ETH LAG scatters the traffic among ports in aggregation based on Mac addresses /VLans,etc algorithms. Therefore traffic will be shared but not split.

- You will need to pay attention to buffer sizes as well as switching times.

[One and a half]

Further to the OP, Gordon Rankin discusses the Wavelength Ethernet Spacelator in an interview with Darko starting here at 10:52. Gordon rethought the Ethernet Repeater as a basis for the Spacelator. More on Ethernet Repeaters here. 

 

Isolation is achieved with transformers that isolate the HF noise front and back, while optical isolation provides for LF frequency attenuation. Passive filtering is not possible, just too lossy my guess. Two linear power supplies take care of no SMPS. The Spacelator reduces the number of errors in the Ethernet Transmission, so there's no need to correct or resend, avoiding processing and the extra noise that goes with that. 

From the interview, it appears the clocking is performed internally, with no external clock input mentioned. Delivery is 2019.

 

If the Spacelator was used in series with a spur line from a dual NIC or JCAT Net Card Femto, that signal into a streamer should be quiet!  The Spacelator is a purpose built device to tackle noise, rather than souped up Ethernet switches with tacked on clocks, have to start somewhere.

[Miska]

18 minutes ago, One and a half said:

The Spacelator reduces the number of errors in the Ethernet Transmission, so there's no need to correct or resend, avoiding processing and the extra noise that goes with that. 

From the interview, it appears the clocking is performed internally, with no external clock input mentioned. Delivery is 2019.

 

Errors? What errors? I have several terabytes of traffic gone through with zero errors.

 

Retransmissions happen usually due to higher levels in the OSI stack that Ethernet level devices don't know/understand about and thus has no impact over.

 

And there's nothing to "reclock" on Ethernet. One could try to reduce the noise impact by using spread spectrum clocking though (clock with as high as possible random jitter), although many devices use it already. Unfortunately audiophile gear usually goes totally opposite way on this though... (Ethernet clock has no relation what so ever to audio clocks)

 

[One and a half]

Just now, Miska said:

 

Errors? What errors? I have several terabytes of traffic gone through with zero errors.

 

Retransmissions happen usually due to higher levels in the OSI stack that Ethernet level devices don't know/understand about and thus has no impact over.

 

And there's nothing to "reclock" on Ethernet. One could try to reduce the noise impact by using spread spectrum clocking though (clock with as high as possible random jitter), although many devices use it already. Unfortunately audiophile gear usually goes totally opposite way on this though... (Ethernet clock has no relation what so ever to audio clocks)

 

I'm just passing the message on from the interview. 

[Miska]

2 minutes ago, One and a half said:

I'm just passing the message on from the interview. 

 

Yes, I totally understand that.

 

Just commenting about the information, not shooting the messenger!

[jabbr]

6 hours ago, Miska said:

 

 

And there's nothing to "reclock" on Ethernet. One could try to reduce the noise impact by using spread spectrum clocking though (clock with as high as possible random jitter), although many devices use it already. Unfortunately audiophile gear usually goes totally opposite way on this though... (Ethernet clock has no relation what so ever to audio clocks)

 

 

 

As of Ethernet 10g which is now decade+ old and more so 25g 40g 100g, the end to end jitter is specified to have a tight upper limit. (So spread spectrum wouldn’t meet spec)

 

There are perhaps some some improvements with latency associated with good clocks on NICs for example the Solarflare “Flareon” 7322F if you want something exotic but commercially available ... would be really hard to beat good commercial fiberoptic switched/NICs in isolation, latency or any other measurable parameter  

 

Its not just that there isn’t anything to reclock, it’s that there is no reasonable way to improve on what’s already there unless you are comparing to 20-30 year old offerings. 

[Miska]

20 minutes ago, jabbr said:

As of Ethernet 10g which is now decade+ old and more so 25g 40g 100g, the end to end jitter is specified to have a tight upper limit. (So spread spectrum wouldn’t meet spec)

 

I have not spent time reading 10+G network specs, so I don't have clear picture of their frequency constellation patterns. I'm quite disappointed if they actually transmit any clock (that you could reclock in first place). Someone needs to begin modeling signaling of high speed military radio networks that have advanced randomized frequency hopping. So much that the comms looks like background noise. Although lot of this is also trickling down to high speed mobile/wireless networks (WiFi, LTE, etc).

 

There's for example this doc about these in CDMA:

http://www.wireless.ece.ufl.edu/twong/Notes/CDMA/ch2.pdf

And OFDM variant used in many many different technologies these days. xDSL and similar wiring systems also use these kind of techniques due to lot of interference and low channel separation on the telephone cables.

 

One of the best Ethernet specs so far to improve noise for audio use is 802.3az / Green Ethernet. Newer switches and NICs have better and better capabilities to measure cable length and attenuation and adjust transmit powers accordingly. Would be kind of stupid to blast at full power over short cable at powers needed for 100 meters long cable.

 

32 minutes ago, jabbr said:

There are perhaps some some improvements with latency associated with good clocks on NICs for example the Solarflare “Flareon” 7322F if you want something exotic but commercially available ... would be really hard to beat good commercial fiberoptic switched/NICs in isolation, latency or any other measurable parameter

 

Yes, although latency of local networks is totally non-issue for audio playback. Since this is not Wallstreet trading application...

 

But certainly I'd like to see on R&S or similar network analyzer how the audiophile products fare against HPE/Cisco gear.

 

Anywa, for example NAA can tolerate about 500 milliseconds of network jitter, and would thus work even across the internet from other side of the globe. Also bandwidth needs of audio vs capacity of 1 / 10 / 25 Gbps network links is so low, that there's plenty to spare for other things.

 

[Arpiben]

For anyone interested in Ethernet synchronisation aspects:

- SyncE or synchronous Ethernet for transfer clock purposes (ITU-T G8261/2/4)

- 1588 PTP (Precision Time Protocol)

 

 

 

[mansr]

2 hours ago, Miska said:

I'm quite disappointed if they actually transmit any clock

Everything except the lowly UART has some kind of clock transmitted along with the data.

[jabbr]

2 hours ago, Miska said:

I have not spent time reading 10+G network specs, so I don't have clear picture of their frequency constellation patterns. I'm quite disappointed if they actually transmit any clock (that you could reclock in first place).

 

Basically as of 10GBase-X compliance with eye-pattern for end-to-end jitter in picosecond range rather than specs for individual components.

 

The “audiophile” brands are promoting the idea that reclocking a network switch, for example, will affect jitter at the receiver — and it will if the network switch is crappy — but 10GBase-X already takes this into account by specifying end-to-end jitter  

 

Quote

 

Someone needs to begin modeling signaling of high speed military radio networks that have advanced randomized frequency hopping.

 

A different interesting use of spread spectrum than the military has been doing for at least 25 years.

 

Of course 10G speeds are not necessary for audio — I’m just pointing out that the really old 10mbs and 100mbs networks tolerated/used poor jitter components — 1Gbe less jitter but when 10G was introduced, the jitter specs became vastly better and the 10G switches are already very low jitter. If an audiophile switch is introduced it should be compared — end to end — against 10G equipment not 100M equipment  

[plissken]

12 hours ago, Arpiben said:

A few remarks:

- for implementing LACP protocol you need both devices able to deal with it.

- in principle ETH LAG scatters the traffic among ports in aggregation based on Mac addresses /VLans,etc algorithms. Therefore traffic will be shared but not split.

- You will need to pay attention to buffer sizes as well as switching times.

 

I understand LACP quite well thank you. You are missing the point:

 

You can remove a cable during playback and lose zero connectivity. 

 

Nope, you don't need to pay attention to buffers: You can plug one cable back in, wait a second and remove the other. 

 

I suggested a computer with a LACP teamed NIC. Also I'm pretty sure the port channel group on the switch will work as an access port w/o the end device being configured for LACP.

 

 

[Arpiben]

7 minutes ago, plissken said:

 

I understand LACP quite well thank you. You are missing the point:

 

You can remove a cable during playback and lose zero connectivity. 

 

Nope, you don't need to pay attention to buffers: You can plug one cable back in, wait a second and remove the other. 

 

 

What I wanted to point is that the switch is not hitless (free of bit errors).

By unplugging one cable you trigger a Loss Of Frame/Input that allow to switch faster-> a few ms

By waiting for LACP packets loss the switching is much longer-> less than 1s