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

[lpost]

51 minutes ago, Jud said:

Are shield-linked metal connectors required to achieve Cat 7 or Cat 8 specifications? Or to ask more specifically, does anyone know of any Cat 7 or Cat 8 cables with plastic connectors, or don't these exist? If they do exist, I might be interested in trying one between the gateway and the ER.

 

Yes, CAT 7 and 8 are both shielded (STP) designs and require the metal shielded connectors, as does 6A STP (as opposed to 6A UTP).  

 

There are other variants of shielding:

 

Individual shield (U/FTP)

Individual shielding with aluminum foil for each twisted pair or quad. Common names: pair in metal foil, shielded twisted pair, screened twisted pair. This type of shielding protects cable from external EMI entering or exiting the cable and also protects neighboring pairs from crosstalk.

Overall shield (F/UTP, S/UTP, and SF/UTP)

Overall foil, braided shield or braiding with foil across all of the pairs within the 100 ohm twisted pair cable. Common names: foiled twisted pair, shielded twisted pair, screened twisted pair. This type of shielding helps prevent EMI from entering or exiting the cable.

Individual and overall shield (F/FTP, S/FTP, and SF/FTP)

Individual shielding using foil between the twisted pair sets, and also an outer foil or braided shielding. Common names: fully shielded twisted pair, screened foiled twisted pair, shielded foiled twisted pair, screened shielded twisted pair, shielded screened twisted pair. This type of shielding helps prevent EMI from entering or exiting the cable and also protects neighboring pairs from crosstalk.

 

Shielded Cat 5e, Cat 6/6A, and Cat 8/8.1 cables typically have F/UTP construction, while shielded Cat 7/7A and Cat 8.2 cables use S/FTP construction

 

U = unshielded

F = foil shielding

S = braided shielding (outer layer only)

TP = twisted pair

TQ = twisted pair, individual shielding in quads

[lpost]

There should be no need for attenuators unless you're running LR - Long Range transceivers, in which case don't do this, as they are designed to shoot many kilometers.  SX - mulitmode and LX - singlemode should work just fine with a 1M fiber patch.  The lasers won't over drive the receiver.

 

If the receiver is being over driven, you likely won't get link/signal and if you do it will drop and recover over and over until the receiver is burned out.

[lpost]

Interesting.  I wonder if he found a difference or tried all copper Ethernet.  It's difficult for me to justify how the optical wavelength used could impact the sound and flatten or crush the sound stage; perhaps unless the 850nm SFP were faulty but again, it's a situation of either signal or no signal.  It can be too low or too hot but this will only cause retries or dropouts of the data.

 

It's cheap to experiment but use caution to not to lose the forest for the trees (or music).

 

fyi and fwiw, I have an ER in my system and very much like what it's brought to the system.  I've got copper from a Cisco 2960G feeding it and 850nm multi-mode out to my server with PCIe board then I2S to my DAC.

[lpost]

20 hours ago, GryphonGuy said:

 

 

Perhaps 1550nm is worth trying...

[lpost]

On 1/27/2020 at 6:12 PM, GryphonGuy said:

 

Hi lpost,

 

Yes I did try all copper Ethernet and quite a few different combinations of same brand before and after the ER and different brands before and after the ER. Cables in Cat 6A, Cat7 and Cat8 were used and they varied with the amount of shielding on the cables from none (UTP) to unshielded cable but foil shielded pairs (U/FTP) and fully shielded (S/FTP). Any form of shielding was better than unshielded before the ER. The most pleasing copper connection was Cable Matters Cat8 before the ER and Tera Grand Cat7 after the ER BUT the OM5 fibre cable's transparency (with the 1310nm transceivers) and airiness trounced the copper connection in my system. Of course YMMV. 

 

Unfortunately I do not have an OM4 fibre optic cable long enough to try it to see if it is something special about the OM5 cable but the reason the OM5 cable was introduced, according to sales literature, was to get greater distance using 850nm transceivers.

 

It was a surprise to me too that the wavelength of the transceivers mattered especially since the cable is supposedly optimised for the 850nm wavelength signals. It was not that subtle that the 1310nm transceivers opened the air around the instruments in the mix. The transceivers are from the Fiberstore custom made for my ubiquity networking gear.

 

Regards

GG

 

I've got a long-since retired HP 10Ge dual SFP+ nic (found it in a cabinet in our storage room at the office) that I first need to be sure Archlinux has drivers for...if it does, I'll try some of the 1310nm and perhaps 1550nm transceivers and patch.  It appears that OM3-OM5 support 850-1300nm.  I may also try 100Mb SFP 1310nm as I've got fiber to my server thus the ER is B-A and fed at 100Mb.  Lots of combos to try, awesome that fs.com has transceivers for so cheap.

 

I'm thinking these low power, only 2KM, 100Mb could be the ticket.  Shouldn't need attenuators but who knows how many ma they draw, could be less efficient that newer but could be less as it operates so 'slowly'.

 

https://www.fs.com/products/23583.html

[lpost]

Minimum 2M cable.  Power specs listed.

 

https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/fast-ethernet-sfp-modules/product_data_sheet0900aecd801f931c.html

[lpost]

Well no luck.  The card is not seen by the kernel and it get's very hot very quickly.  It's way old.

 

Anyone know of a reasonably priced/cheap PCIe card with a single open SFP that Archlinux has a driver included?  A model that is a few years old will do fine.  

[lpost]

Have you successfully run Intel X520-DA1 with other branded SFP?

 

FS.Com only has a couple 'Intel' SFP to choose from...

 

Looks like I can simply force it 'unsupported' via:

 

In /etc/modprobe.d/ixgbe.conf
options ixgbe allow_unsupported_sfp=1,1  <-BOTH ports

[lpost]

On 2/11/2020 at 9:13 PM, kennyb123 said:

The attenuators tended to reveal a bit more to me after swapping one in for another that had been in place for a few days at least.  While I have had the Startech SFPs for a while, my system got quite a better over the weekend after shifting a few things around.  I’ll have to spend more time with each of these to be able to reach any firm conclusions.  I guess I’m actually somewhat relieved that the differences were pretty subtle.  

 

How about some lesser dB attenuators on LX?  ZX is WAY powerful (40KM) and if it sounds better with 10dB of attenuation, it follows that LX (10KM) would be affected by say 2-5dB attenuation.  At $3 each, I'm going to pick up an assortment to try but I'm only going with 1310nm LX (10KM) and a 10m SMF patch.

[lpost]

I'll have the parts I ordered by the end of next week.  Perhaps the primary difference is LED vs laser.  SX transceivers get away with LED for the light source, while LX and ZX et al use lasers.

 

I purchased 4dB and 12dB attenuators just for fun to try.  These were the only ones that were in US warehouse for quick ship, otherwise they come from Asia and that could take weeks with the current virus situation.

 

So, I'll be comparing SX (850nm) LED to LX (1310nm) 10k lasers with and without attenuation over a 5m single mode patch.

 

My path is Cisco 2960G (AV switch) feeds ER B-side > A-side SFP out to Pink Faun 2.16 PCIe fiber NIC.

 

It's sounding astoundingly good and I'll be impressed if the LX improves things further but for $75 I was willing to run the experiment.

 

I'll follow up in a couple weeks

[lpost]

On 2/15/2020 at 7:23 AM, lpost said:

I'll be comparing SX (850nm) LED to LX (1310nm) 10k lasers with and without attenuation over a 5m single mode patch.

 

My path is Cisco 2960G (AV switch) feeds ER B-side > A-side SFP out to Pink Faun 2.16 PCIe fiber NIC.

 

It's only been a couple hours but I must say I'm impressed with the LX 1310nm transceivers with 12dB attenuators.  It worked without attenuators, of course, but there is an 'ease' with them in vs. without.  I listened for 30 mins. or so before pulling the server to clean it out and add...EMI absorption.

 

I was needing to pull my server off the rack and clean it as it's been nearly 2 years since installed.  While grounded I used a static-free brush and a blower to clean it all out.  It wasn't as dirty as I thought, it's fan-less, but still I feel better with it clean.

 

I must add I broke the cardinal rule of change only one thing at a time but the effort to move a 70lbs server is significant for me.  I added a good number of 3M AB5020 Adhesive EMI Absorber to many components within the server.  I've read not to over do it, but I don't understand how more radiated interference within the box could be a good thing vs. less.  Again, conserving effort, I put a layer on every cap head, IC with or without a smooth aluminum heatsink and probably most importantly, numerous layers between and either side of the memory modules.  I just set the strips on each side as opposed to sticking them on the memory as I'll be trying some APacer modules soon.

 

I can recommend both tweaks as worthwhile and both are quite inexpensive by any standard let alone audiophilla standards.