The EtherREGEN thread for various network, cable, power experiences and experiments

[Superdad]

2 hours ago, JohnSwenson said:

HAVE FUN,

 

Thanks so much for posting that John. You said it all with exactly the facts and tone I would wish to convey. 

I can’t believe how far behind I got in keeping up with posts in our own forum these past few days. Just spent the whole weekend preparing the final 95 March EtherREGENs for shipment this week—plus tracking down and ordering parts for the next 250 unit run (can’t find 1,000 of those SilLabs clock buffer chips in stock anywhere the world yet )

 

I am grateful that there are many enthusiastic and knowledgeable EtherREGEN users pitching in here to assist each other in having fun with @JohnSwenson’s unique creation! 

—Alex C.

[AfterDark.]

15 hours ago, MasterWarzombie said:

it is shielded so it is not a good cable. I'm teasing . My Wireworld cable has the same connectors.

 

same : 

https://www.adark.co/products/afterdark-constellation-cfs-cat7-telegartner-mfp8-edition

 

it is however recommended by @ AfterDark

 

Hi! MasterWarzombie, 

 

The AfterDark. Constellation Network cable is tailor made for use it on the port B of EtherREGEN. The braid outside the network cable was insulated by 3M electrical tape and not touching the metal plug of Telegartner MFP8 plug. This is to avoiding the noise is feed to the port B of EtherREGEN.  The best sounding network cable for UpTone EtherREGEN shall following this approach.

 

The Telegartner can provide the excellent contact point for the lan port, so make the shield cable not touching the Telegartner metal plug is the way to go. 

 

 

Please noted that when we change the plug for some customers, we found that Wireworld and Audioquest Diamond Lan cable is connected to the braid of the cable, so it was not recommended. So, make sure it is modify with the above arrangements. 

 

Hopes it can help to explain the design. 

 

Thanks and have a nice day.

 

Best Regards,

Adrian

AfterDark. 

[Johnnydev]

audioquest has of their RJ / E cables the shielding only connected to the plug on one side, namely the receiving side in their case. So no ground loops that way.

[James Stephens]

20 hours ago, James Stephens said:

 

 

I use ethernet cables from Ghent Audio that are shielded via the JSSG360 method ... the shield is not connected at the ends and so it brings the traditional benefits of shielding ... namely blocking EMI .. but also blocks leakage currents. I use them everywhere because although I only have a single fiber connection to the A-side of my ER, there are other places in my network .. router, modem etc where I desire to block leakage currents.

 

Best,

James

 

Correction from John on Isolation thread Page 3 as I had this wrong (apologies all) 

 

 

====

Just a word of clarification: the shield tied/not tied Ethernet cable ONLY affects true ground loops. High impedance leakage current will go through ANY Ethernet cable (except one that is not plugged in!)

 

Quick refresher:

 

Ground loop: caused by voltage difference (50/60Hz) between safety grounds on different outlets.

 

Leakage loop: current flowing from mains AC (NOT safety ground), through power supply to DC output. Can travel to other boxes and either go through leakage in another power supply or get shunted through a safety ground connection.

 

I've written extensively about this in other posts.

 

John S

=====

 

Best,

James

 

 

[Johnnydev]

I just got the EMPEROR GIESEMANN EVA 75OHM SPECIAL EDITION ordered from Afterdark's friendly and helpful Adrian. The delivery will take a few more weeks, but I am very curious.

[fuzzypoodle]

@johnnydev may I ask why you went for that and not something like Mutec with 8 clocking interfaces ?

[Johnnydev]

53 minutes ago, fuzzypoodle said:

@johnnydev may I ask why you went for that and not something like Mutec with 8 clocking interfaces ?

I think this is better and cheaper.

[JohnSwenson]

21 hours ago, MasterWarzombie said:

 

Merci @JohnSwenson;)

If the equipment is connected to the same Gigawatt power strip for example or the same power conditioner, that does not solve these possible concerns ground loops and leakage loops?

I understand that the quest for the perfect audiophile is fraught with pitfalls.

As I recently posted only the true ground loops go through two safety grounds, thus only these will be eliminated with everything connected to the same power strip. Leakage loops are very different. They form a loop from AC mains (hot, neutral or whatever you have) NOT the safety ground, through the power supply to the DC output, through whatever is powered by that DC, potentially through interconnects to other boxes, through THEIR power supplies and back to the AC main. Having everything on the same power strip does not help this at all since it does not depend on voltage differences between outlets.

 

Back on ground loops, saying "everything is on the same power strip" so you can use shield tied cable only works as long as EVERYTHING with a safety ground connection including computers, servers etc. is on he same strip. Network equipment with only 2 pin connections cannot contribute to ground loops, but CAN very strongly contribute to leakage loops which don't depend on the safety ground.

 

Sorry I don't have any time to go into more on this, I HAVE written extensively about it in posts on this forum.

 

John S.

[One and a half]

With an impending delivery occurring for the EtherRegen, thought to put feeble brain to paper and see how to connect the device to the rest of the system.

 

With all of the cabling shielded including AC power lines, may as well continue with the signal and 9V power lines to the EtherRegen and the ifi Micro USB 3. 

 

To avoid rising voltages caused by shield currents, the 9V lines are short. Shields are connected at the plug end as well bonded to the metal frame of the junction power box. The large 6mm (10AWG) bonding the Lumin, ER (A-Side) and the junction box maintain the same potential  to the earthed 0V at the Acopian power supply.

 

The thought of a separate LPSU for the Micro USB 3 and Ether Regen did receive some thought. Not only the cost, but potential leakage paths across the AC line and bonding the two 0V togther would have created a worse problem. The noise from the A side of the ER is worse than the Micro USB in any case, a pragmatic solution with just the one PSU. In any case the noise form the A side would be far less less than the standard ifi iPower SMPS for the Micro USB3. 

 

The STP (Shielded twisted pair) of the Lumin bonds the B side to the same ground as the Lumin, antenna effects are minimal due to the short cable and being shielded anyway.

 

The Acopian supply (9V, 6.5A)  has two sense terminals, like the JS2,  that compensates for long leads voltage drops, so the desired voltage won't drop (too far).

 

Normally the audio system is powered down with the Music server always on. IP addresses are assigned by the router to the MAC address of the one of the JCAT Net Femto NICs. The other NIC is DHCP for the Lumin, now ER.. so the IP address should be transparent from the JCAT through to the Lumin, apart from a speed change, 1GB/s to 100MB/s.

 

The ER & Acopian will power up with the rest of the sources, the Lumin takes some time to establish a network connection, so the ER should be functional by then.

 

Networked audio EtherRegen _ USB3.pdf

[MartinT]

Going back to the "everything in digital is important" mantra.  I had freed up a 12V Chinese LPSU which had been upgraded by Coherent with low noise regulators and high quality caps.  Putting that in place of the Paul Hynes PR3i to power my 4G router made a significant sound quality improvement.  Noise is everything in power supplies for digital equipment, but it's crazy how I can hear the difference powering the component at the very start of the chain, ending at the DAC.

 

Turning to the end of the chain, Coherent are now upgrading my Mutec MC-3+ USB for better performance (it was already modified for 5V DC operation).  Since this directly feeds my DAC via AES, it's critical for best sound quality.  I have a loan Mutec while waiting for mine to be returned.

 

Relevance to the ER?  It's in the middle of the chain and doing its job very well.

[R1200CL]

On 3/16/2021 at 5:58 AM, One and a half said:

The thought of a separate LPSU for the Micro USB 3 and Ether Regen did receive some thought. Not only the cost, but potential leakage paths across the AC line and bonding the two 0V togther would have created a worse problem. The noise from the A side of the ER is worse than the Micro USB in any case, a pragmatic solution with just the one PSU. In any case the noise form the A side would be far less less than the standard ifi iPower SMPS for the Micro USB3. 

 

I’m quite sure you’re braking the moat by this setup. Mostly by using shared power and probably the creative way to ground. (In addition Vbus power is a challenge). 

In your case I can’t see any reason to use the ground screw on the EtherRegen. Read the user manual.

You could also test with fiber in. Adding fiber interface to your PC. Or even use the opticalModule. 
 

Without starting a discussion about the iFi Micro USB 3.0, I take the chance to suggest to remove it. At least as a test.

USB cables is an issue as well. And good Vbus power. I guess a USPCB won’t fit in your system or ?

 

On 3/16/2021 at 5:58 AM, One and a half said:

The STP (Shielded twisted pair) of the Lumin bonds the B side to the same ground as the Lumin, antenna effects are minimal due to the short cable and being shielded anyway.

This goes against everything we learned about how to do ethernet wiring. Allowing possible loops to travel along the connected shield of the ethernet cable. 

 

On 3/16/2021 at 5:58 AM, One and a half said:

The other NIC is DHCP for the Lumin, now ER.. so the IP address should be transparent from the JCAT through to the Lumin, apart from a speed change, 1GB/s to 100MB/s.

You may have a problem here. Data sheet Lumin says only 1 GB. @wklie can you comment on that ?

 

Further you may benefit a lot to reverse the EtherRegen and use fiber into Lumin. 

Will be interesting to hear how things works out.

 

BTW: You using both 230 VAC as well as 110 VAC ?

An other area of concern or ?

 

 

[wklie]

All Lumin products except S1 / A1 / T1 / D1 have 1000Mbps Gigabit Ethernet.

 

The oldest Lumin S1 / A1 / T1 / D1 only had 100Mbps Fast Ethernet.

[James Stephens]

Hi all,

 

Interesting point of note  ... as I understand it the ER was limited to 100Mbps on the "clean" B-side to reduce jitter since it is less noisy that GbE.

 

Best,

James

 

[R1200CL]

On 3/9/2021 at 3:12 AM, JohnSwenson said:

Yes the SFP+ modules work the same way, they are just electrical/optical converters. No clocking, no format conversion. They use a different laser that can be modulated faster and a different driver chip that can drive that laser very quickly.

 

The SFP+ interface was designed to be compatible with regular SFP so you should be able to use SFP+ modules in an ER. Just make sure you use the same modules on both ends. There are so many different variants of SFP modules that trying to mix and match is a recipe for disaster.

 

As to phase noise, I'm not sure. 10Gb has such a short pulse that jitter in the whole system has to be very small, but I'm not sure how much of that is in the actual module and how much is in the electronics feeding the module. Measuring that in situ is going to be very tough.  It could possibly be done with a simple system with a very low phase noise 10MHz LVDS clock going into the transmit side of an SFP cage, then another SFP cage with the receiver LVDS going into a very low jitter LVDS to single ended 50 ohm output (I know the circuit for this, but it is pretty expensive) and feed that into the phase station. THAT will tell you what the phase noise of the module itself is (well at least the combination of transmitter in one module and receiver in another). Again another piece of custom test equipment that I'm not going to have time to build for a long time.

 

John S.

Hi John

 

Please confirm that EtherRegen is working with 10 Ghz SFP+ modules, so there isn’t any confusion about this. 
(I know there exists 1 GB SFP+, but I’m not asking about those. They will of cause work). 
 

Andreas

[One and a half]

14 hours ago, R1200CL said:

I’m quite sure you’re braking the moat by this setup. Mostly by using shared power and probably the creative way to ground. (In addition Vbus power is a challenge). 

In your case I can’t see any reason to use the ground screw on the EtherRegen. Read the user manual.

 

From recent discussions here, the last few pages, the ground thumbscrew is on the A side only. So any voltage differences on the A side Ethernet shells and the Lumin's case won't be, that's the whole idea of that 6mm wire.

 

14 hours ago, R1200CL said:

You could also test with fiber in. Adding fiber interface to your PC. Or even use the opticalModule.

 

 

Without starting a discussion about the iFi Micro USB 3.0, I take the chance to suggest to remove it. At least as a test.

USB cables is an issue as well. And good Vbus power. I guess a USPCB won’t fit in your system or ?

 

This goes against everything we learned about how to do ethernet wiring. Allowing possible loops to travel along the connected shield of the ethernet cable. 

 

You may have a problem here. Data sheet Lumin says only 1 GB. @wklie can you comment on that ?

 

Further you may benefit a lot to reverse the EtherRegen and use fiber into Lumin. 

Will be interesting to hear how things works out.

 

BTW: You using both 230 VAC as well as 110 VAC ?

An other area of concern or ?

 

 

 

[FIndingit]

Every (computer) audiophile should own a simple multimeter. You can check every possible ground point continuity. No need to ask here every time. 
 

[roman410]

11 hours ago, James Stephens said:

Hi all,

 

Interesting point of note  ... as I understand it the ER was limited to 100Mbps on the "clean" B-side to reduce jitter since it is less noisy that GbE.

 

Best,

James

 

If I remember good, Alex from Uptone noted somewhere, the reason of speed limited to 100Mbps on the "B clean side" ethernet port it is only because the time when the etherRegen was dizained 1 GB ISOLATOR chips was not on the market.

[James Stephens]

7 hours ago, roman410 said:

 

If I remember good, Alex from Uptone noted somewhere, the reason of speed limited to 100Mbps on the "B clean side" ethernet port it is only because the time when the etherRegen was dizained 1 GB ISOLATOR chips was not on the market.

 

Right ... so I imagine 100Mbps was a better choice for the B "clean" side .... back then .... but things change and maybe that's not true now ...

 

The slow one now
Will later be fast
As the present now
Will later be past.

 

Best,

James

 

[Superdad]

13 minutes ago, James Stephens said:

Right ... so I imagine 100Mbps was a better choice for the B "clean" side .... back then .... but things change and maybe that's not true now ...

 

A great many aspects of the EtherREGEN's architecture would have to change to support Gigabit on the 'B' side. Including clock speed for our reclocking flip-flops, the main switch chip itself, and the power networks (much more PS current to put across the moat).

We have designed on paper a much more elaborate EtherREGEN that would include Gigabit and fiber on both sides, more ports, etc. But it would require a lot more expensive parts, an in-board dual domain dedicated power supply, and a much larger case.  You would be looking at likely a $3K switch (which, if our competitors has similar build cost and tech would charge $6K for...).

 

We are sure people would buy this--our dealers in Asia bug us all the time for such--and clearly we are leaving $$$ on the table by not leveraging the huge success of EtherREGEN.  But we have several other groundbreaking (and affordable) products under development (not Ethernet-related) and finishing design, getting to production and launch of those will take at least the balance of this year.  So I advise that no one hold their breath for an EtherREGEN Premier anytime soon.

[JohnSwenson]

16 hours ago, R1200CL said:

Hi John

 

Please confirm that EtherRegen is working with 10 Ghz SFP+ modules, so there isn’t any confusion about this. 
(I know there exists 1 GB SFP+, but I’m not asking about those. They will of cause work). 
 

Andreas

 

OK, here is the rundown.

 

Basic statement:

ALL SFP+ modules will work in SFP cages.

Neither an SFP module nor an SFP+ module by itself, knows anything about speed or protocol of the signal going through them. There is no clocking going on in the module itself. There is no auto-negotiation speeds or anything like that happening at all.

 

All of the above happens in circuitry on the board containing the SFP cage, NOT in the module itself.

 

The SFP+ connector interface is exactly the same as the SFP interface except for one pin I will explain further down. The SFP+ was specifically designed so an SFP+ module WILL work in an SFP cage.  This means that as long as SFP+ modules meet the specs they will ALL work in SFP cages, period.

 

The one thing that will NOT work, is if the boxes are running at different speeds. Since there is no auto-negotiation you cannot have a switch with a 10GbE port, connected to a device with a 1GbE port. Since the speeds of the BOARDS don't match, they will not talk. Plugging SFP+ modules into SFP cages on both sides WILL work for ALL SFP+ modules, as long as you use identical modules on both sides. There are a lot of different types of SFP+ modules, if you use different modules on the two ends you have a good probability of it not working.

 

I have in fact tried this, I have two identical SFP+ modules and appropriate cable and plug one into an EtherRegen and the other into a cheap gigbit FMC and it does in fact work! Same with an oR and oM. As long as the speed of the port on both ends is the same, it will work. Note the speed is determined by the BOARD the SFP+ module is plugged into, NOT the module itself.

 

Now the technical background as to what is going on.

 

There are three prime differences between the SFP+ and SFP modules,

1:  SFP+ modules are designed to run faster (I hope that is obvious). The lasers are different, the optical receivers are different so they CAN handle higer frequency signals

 

2: pre-emphasis

 

3: pre-emphasis control pin

 

And one board difference that does not actually matter to the module.

 

#1, The modules have a maximum speed they will run at (10 gigabit per second), but that does NOT mean they cannot run slower. The actual lower limit is around 1MHz (which is determined by the coupling capacitors inside the module) . Since the signal that goes in is the optical that comes out, and the optical that comes in is the electrical that come out, that means you can run signals from 1 mega bit per second to 10 gigabit per second over the SAME module. What determines the speed is the electronics sending the signal to the module.

 

#2, Pre-emphasis. IF the module if designed to run long distances (those infamous 10km or more modules) the edges of the optical pulses will get degraded enough that at 10Gb there is not much edge left. At 1Gb the "bit window" is wide enough that the signal can be detected. In order to receive a 10 gigabit  signal over such long distances, pre-emphasis is added at the transmitter. This is essentially over-shoot on the optical signal, the laser puts out more optical power for a few picoseconds right at the edges. The means that after the edge degradation while traveling over 10km there IS enough signal to detect at the other end. The pre-emphasis does not hurt anything if the link is short or slow. It DOES take a little more power, and the laser lifetime is shortened by a tiny amount.

 

Note the pre-emphasis is not needed for modules that are designed to only work at shorter distances, such as 850nm modules.

 

Thus modules desinged for short distances do not have pre-emphasis, modules designed for long distances do have pre-amphasis. There is a third type that can have the pre-emphasis turned on or off. Note that the only gain you get from turning pre-emphasis off is saving a little power and maybe a half hour longer lifespan. Having the pre-mphasis on does NOT harm the reception of slower or shorter systems.

 

#3 pre-emphasis control. Since the module knows nothing about the signal going through it, the outside world has to tell the module to turn pre-emphasis on or off. One of the pins on the pins on the SFP connector is always grounded, so the SFP+ uses this pin as the pre-emphasis control, when it is low pre-emphasis is turned off, when high pre-emphasis is turned on. Thus plugging an SFP+ module into an SFP cage will turn off pre-emphasis.

 

Note that SFP+ modules designed for shorter distance will ignore this pin, since they don't have pre-emphasis at all. SFP modules that are designed for long distance and have pre-emphasis always turned on will also ignore this pin (but will still work fine with slower signals). Only modules that have controllable pre-emphasis will look at this pin.

 

Signal difference. The SFP spec uses LVDS format for the signals going in and out, SFP+ uses CML. CML circuits will run much faster than LVDS so they are used for SFP+. They ARE different, BUT all the signals going in and out of both SFP and SFP+ have a capacitor in series, which means that INSIDE the module they are exactly the same. The result is that the signals going in and out of an SFP cage work just fine with an SFP+ module.

 

So in summary,

all SFP+ modules will work fine in SFP cages as long as you don't mix speeds of the devices, both ends must be running at gigabit. If you use a switch with 10GbE ports that can be switched to 1GbE, they will work fine, but If the port ONLY runs at 10GbE, you cannot use it with a link where the other end is SFP.

 

I hope this clears up the confusion.

 

John S.
 

[R1200CL]

29 minutes ago, JohnSwenson said:

I hope this clears up the confusion.

Fantastic explanation.

Thanks a lot John. 
 

30 minutes ago, JohnSwenson said:

Signal difference. The SFP spec uses LVDS format for the signals going in and out, SFP+ uses CML. CML circuits will run much faster than LVDS so they are used for SFP+. They ARE different, BUT all the signals going in and out of both SFP and SFP+ have a capacitor in series, which means that INSIDE the module they are exactly the same. The result is that the signals going in and out of an SFP cage work just fine with an SFP+ module.

Could this be understood as two modules with same laser and transfer technology (singel mode / multi mode), but one being SFP and the other SFP+, should in theory perform equal with respect to the parameters we as audiophile are interested in ?

 

This LVDS vs CML is something determined by circuits in the switch or where the cage is mounted, so if any difference, it will be determined somewhere after cage ?

 

 

[MarkusBarkus]

1 hour ago, JohnSwenson said:

I hope this clears up the confusion.

...for about a half-hour, looks like. You're a prince for providing that level of detail, John. Thanks to you and Alex for your guidance and intelligence on this topic. Cheers...

[JohnSwenson]

2 hours ago, R1200CL said:

Fantastic explanation.

Thanks a lot John. 
 

Could this be understood as two modules with same laser and transfer technology (singel mode / multi mode), but one being SFP and the other SFP+, should in theory perform equal with respect to the parameters we as audiophile are interested in ?

 

This LVDS vs CML is something determined by circuits in the switch or where the cage is mounted, so if any difference, it will be determined somewhere after cage ?

 

 

As far as I know the lasers and optical receivers used in all SFP+ modules ARE different from the ones used in SFP modules so there will always be differences. SFP+ modules use several different types of lasers and optical receivers, this is part of why  you should always use the the same modules on each end of a link. The lasers etc used in SFP+ modules are different than the ones used in SFP modules. One of the primary differences is that the circuits that will run at 10Gb speeds always take more power than the ones designed to just go to 1Gb. Thus all the manufacturers don't use SFP+ components in SFP modules.

 

LVDS and CML are two different high speed differential output stages. When you buy a high speed chip it will come with either LVDS or CML outputs. It turns out that CML can run at considerably higher speeds, but takes more power. So CML is primarily used for VERY fast chips.  CML is only used on "high end" chips that need to run very fast, so they will inherently be made to be very low phase noise. LVDS is used at MANY different speed levels so there are some chips that have pretty poor phase noise and others that are VERY good. The best LVDS and the best CML chips seem to be neck and neck as far as  phase noise is concerned.

 

So IF the board you are building has an SFP cage, you will use chips with the very best LVDS implementations (as we do in the ER), if you are building a board that can go 10 gigabit, you use chips with CML. IF you use gigabit in both, the noise component of the output buffers will be about the same, BUT the CML one will take more power.

 

John S.

 

 

[James Stephens]

All this talk of SFP vs SFP+ modules has got me wondering about potential sound quality differences between modules and also single mode vs multimode and whether it might make a difference SQ-wise.

 

What do people think?

 

I am currently using single mode SFP's for my OM to ER fiber run but I ran both single mode and multimode fiber from my basement too in case I got the urge to experiment lol ....

 

Best,

James

[R1200CL]

7 hours ago, JohnSwenson said:

The lasers etc used in SFP+ modules are different than the ones used in SFP modules.

The attached data sheet of SFP+ 10 GB FTLX1475D3BTL and the SFP 1GB FTLF1421P1BTL is both singel mode 1310nm DFB laser. But I guess it’s still different some how. Maybe not important. 
 

If you know what to look for in the data sheet telling us good or very good, please let us know. 
The jitter numbers for the FTLF1421P1BTL seems very good, and was the lowest number I could find among Finisar modules. 

Thanks again for a great explanation. 

 

finisar_ftlf1421p1xtl_oc-48_ir-1_stm_s-16.1_rohs_compliant_pluggable_sfp_transceiver_product_specification_rev_b1.pdf finisar_ftlx1475d3btl_rohs-6_compliant_10g_10km_1310nm_single_mode_datacom_sfp_transceiver_productspecrevb1_0.pdf