Jump to content
IGNORED

Optical Network Configurations


Recommended Posts

My current: Cisco 2960>MM fiber (Cisco OEM SFP) to opticalModule>Ghent JSSG Catsnake ethernet> opticalModule> SM fiber (Finisair LX SFP)> opticalRendu. 

 

I wanted to stick to ethernet out from the switch to the oM and then fiber to the oR so I could sell one of the oMs, but having the double fiber bridge just sounds that much better. 

SERVER CLOSET (in office directly below living room stereo):NUC 7i5BNH with Roon ROCK (ZeroZone 12V on the NUC)>Cisco 2690L-16PS switch>Sonore opticalModule (Uptone LPS 1.2)>

LIVING ROOM: Sonore opticalRendu Roon version (Sonore Power Supply)> Shunyata Venom USB>Naim DAC V1>Witchhat DIN>Naim NAP 160 Bolt Down>Chord Rumor 2>Audio Physic Compact Classics. OFFICE: opticalModule> Sonore microRendu 1.4> Matrix Mini-i Pro 3> Naim NAP 110>NACA5>KEF Ls50's. BJC 6a and Ghent Catsnake 6a JSSG ethernet; AC cables: Shunyata Venom NR V-10; Audience Forte F3; Ice Age copper/copper; Sean Jacobs CHC PowerBlack, Moon Audio DIN>RCA, USB A>C. Isolation: Herbie's Audio Lab. 

Link to comment
6 hours ago, jabbr said:

SFP/SFP+/SFP28/QSFP28 modules


Leave out QSFP28. That’s another type of modules. 
 

SFP+ and SFP28 can go in same slot

SPF+ should not be inserted into SFP

SPF can be inserted into SFP+

 

I would chose any of these SFP
FTLF1424P2xCR
FTLF1424P2xCD 

FTLF1424P2xTR
FTLF1424P2xTD 

FTLF1424P2xCL
FTLF1424P2xCV


I’m not sure what last 2 letters indicates. 
 

I’m interested in finding out if any of Finisar modules is using CDR. As that may reduce jitter. 


image.jpeg.11791827372eafc9efd92b8e61a2b62a.jpeg

image.jpeg.84d3555e5fe24dd2de11d8e22c1719fe.jpeg

Link to comment
10 hours ago, R1200CL said:

I’m interested in finding out if any of Finisar modules is using CDR. As that may reduce jitter. 

 

Jitter performance is part of the spec. The light source really doesn't matter.

 

Jitter only happens on data transfer. With 10GBe and a theoretical 1250MB/s a 50MB 16/44.1 track will transfer in 4/100ths of a second.

 

I'm now convinced we have a comprehension problem and a flat out refusal to face the realities of high speed networks and cached playback systems.

 

 

 

 

Link to comment

And I’m convinced we have a dichotomy between how something should sound according to the theory and how it actually sounds according to practice. 

 

Theory tells us that bit perfect files should sound the same no matter which devices they pass through but practice tells us they don’t.  

 

The problem is that the practice is based on an illusion created in our brain, so the question is, what changes do our brains perceive that our measuring instruments don’t? 

 

The argument I don’t buy for one minute is that the changes are due to my imagination. There’s no way my imagination can add additional venue ambience, richer timbre and harmonics, addition complexity to strings, bows, voices,  increasingly detailed microdynamics and all the rest, consistently, reversible, reproducibly.  My brain may be be smart but its not that smart that it can generate the information that allows me to differentiate a Guarneri from a Stradivarius 

Link to comment
46 minutes ago, Blackmorec said:

And I’m convinced we have a dichotomy between how something should sound according to the theory and how it actually sounds according to practice. 

I have no dichotomy, as my network is entirely silent. It has no sound!

 

I have stated my design decisions, my research, product selections and practices. I have experience with many switches and modules.

 

You may have your own experiences. I can’t say what you hear nor you I. All I can say is what I do and why, and what I hear: nothing but music (and kids in the background these days of Covid)

 

Moreover, this thread is about what works in practice.

Custom room treatments for headphone users.

Link to comment
2 hours ago, plissken said:

With 10GBe and a theoretical 1250MB/s a 50MB 16/44.1 track will transfer in 4/100ths of a second.

Perhaps a good topic for another thread. I don’t think we are anywhere near this in practice. It would be good to know how each app transfers data. Some say it’s transferred at wire speed then communication stops, but I’ve had many tracks stop shortly after pulling the plug. 
 

anyway, I don’t want to details this thread. 

Founder of Audiophile Style | My Audio Systems AudiophileStyleStickerWhite2.0.png AudiophileStyleStickerWhite7.1.4.png

Link to comment

Let me expand: if there are specific measurements comparing jitter or not, or patterns of power usage among various SFP modules etc, there is a thread in the Object-Fi section appropriate to that.

 

The original intention of this thread is to discuss exacty what works with what ... which SFP modules and SFP+ modules work with each switch and endpoint. It is possible that different SFP modules, particularly those attached to the endpoint might have a different SQ. Does single vs multimode have different SQ? In those cases where specific modules are used with specific endpoints and specific SQ differentes are noted, then it is appropriate to discuss.

 

Most of the posts here have been on topic, for example very recent posts by @R1200CL and @charlesphoto

Custom room treatments for headphone users.

Link to comment
14 hours ago, R1200CL said:

I’m interested in finding out if any of Finisar modules is using CDR. As that may reduce jitter. 

My understanding is that with SFP(+) the clock data recovery is after the module. In the XFP module you showed, its on the module and you can see how much larger the modules are.

Custom room treatments for headphone users.

Link to comment

IMO SFP rolling is somewhat chasing the dragon. I honestly can't tell you if there was a SQ difference when I replaced the final oM to oR leg of fiber with single mode vs multimode, and unfortunately due to my set up I have no way of AB'ing without pulling cable and stressing everything out. So what is done is done, and it sounds perfectly fine to me (I got Finisar FTLF1324P2BTV-B1). I've had much bigger SQ changes lately adding a bevy of AC iPurifiers and some new AC power cables to the system, and some Herbie's Super Stabilizers and feet to my PSU and oR. I wouldn't get too hung up on this end f things and just use what works and is affordable, and work on overall system synergy. Clean power is probably the biggest game changer. 

SERVER CLOSET (in office directly below living room stereo):NUC 7i5BNH with Roon ROCK (ZeroZone 12V on the NUC)>Cisco 2690L-16PS switch>Sonore opticalModule (Uptone LPS 1.2)>

LIVING ROOM: Sonore opticalRendu Roon version (Sonore Power Supply)> Shunyata Venom USB>Naim DAC V1>Witchhat DIN>Naim NAP 160 Bolt Down>Chord Rumor 2>Audio Physic Compact Classics. OFFICE: opticalModule> Sonore microRendu 1.4> Matrix Mini-i Pro 3> Naim NAP 110>NACA5>KEF Ls50's. BJC 6a and Ghent Catsnake 6a JSSG ethernet; AC cables: Shunyata Venom NR V-10; Audience Forte F3; Ice Age copper/copper; Sean Jacobs CHC PowerBlack, Moon Audio DIN>RCA, USB A>C. Isolation: Herbie's Audio Lab. 

Link to comment
15 minutes ago, FIndingit said:

What are your fmcs sitting on, i.e. have you taken care of vibration control? I’ve heard meaningful differences using different feet under several brands of converters. 

 

I'm using cut pieces of Herbie's dbNeutralizer as feet. But mine are also outside the hifi area nd in a server closet below.

SERVER CLOSET (in office directly below living room stereo):NUC 7i5BNH with Roon ROCK (ZeroZone 12V on the NUC)>Cisco 2690L-16PS switch>Sonore opticalModule (Uptone LPS 1.2)>

LIVING ROOM: Sonore opticalRendu Roon version (Sonore Power Supply)> Shunyata Venom USB>Naim DAC V1>Witchhat DIN>Naim NAP 160 Bolt Down>Chord Rumor 2>Audio Physic Compact Classics. OFFICE: opticalModule> Sonore microRendu 1.4> Matrix Mini-i Pro 3> Naim NAP 110>NACA5>KEF Ls50's. BJC 6a and Ghent Catsnake 6a JSSG ethernet; AC cables: Shunyata Venom NR V-10; Audience Forte F3; Ice Age copper/copper; Sean Jacobs CHC PowerBlack, Moon Audio DIN>RCA, USB A>C. Isolation: Herbie's Audio Lab. 

Link to comment
10 hours ago, R1200CL said:

My understanding is the opposite. 

 

That was my 'understanding' as well, and why I switched to single mode. I'm just saying that YMMV, and in real life terms there's other simple things that one can/should do with their set up that will make a bigger difference than worrying about which SFP to use. It's just that I couldn't say SM was definitively better than MM, though it certainly didn't make the sound worse (I'm pretty sure it was a bit less digitally sounding). I'm not trying to argue this, I just want to put that out there for anybody reading this thread. I probably did it backwards - if I had taken care of some power issues first (which really eliminated the 'digital glare' I kept chasing) then I may have noticed the difference more between MM and SM fiber. 

 

What I can say is the opticalModules are night and day from cheap FMC's no matter which mode you run, and better than the SFP out on my 2017 Cisco 2960PS. And if I try and take out one of the oM's from my 'double bridge' I hear it. 

SERVER CLOSET (in office directly below living room stereo):NUC 7i5BNH with Roon ROCK (ZeroZone 12V on the NUC)>Cisco 2690L-16PS switch>Sonore opticalModule (Uptone LPS 1.2)>

LIVING ROOM: Sonore opticalRendu Roon version (Sonore Power Supply)> Shunyata Venom USB>Naim DAC V1>Witchhat DIN>Naim NAP 160 Bolt Down>Chord Rumor 2>Audio Physic Compact Classics. OFFICE: opticalModule> Sonore microRendu 1.4> Matrix Mini-i Pro 3> Naim NAP 110>NACA5>KEF Ls50's. BJC 6a and Ghent Catsnake 6a JSSG ethernet; AC cables: Shunyata Venom NR V-10; Audience Forte F3; Ice Age copper/copper; Sean Jacobs CHC PowerBlack, Moon Audio DIN>RCA, USB A>C. Isolation: Herbie's Audio Lab. 

Link to comment

For those looking for very good and reasonable SFP’s. 

 

https://www.ebay.co.uk/itm/Finisar-FTLF1421P1BCL-OC-48-STM16-GigaBit-Ethernet-1x-2x-FC-IR-1310nm-SFP/114327996056?hash=item1a9e7ac698:g:PMQAAOSwNNxfIUh7


You will have to email him for shipping quote. The modules is new. 
 

I have studied the various available SPF’s from Finisar.

FTLF1421P1xTL (FTLF1421P1BTL) and the FTLF1421P1xCL (FTLF1421P1BCL)is ones that is listed as available today. Data sheet from 2015. 
 

The only difference I have spotted between those two version is a slightly better operation temperature specification on TL version and some different number is a parameter called LOS De-Assert / LOS Assert. I can’t think it matters for our use. 

 

The 1424 options has data sheets from 2005 and EOL, and hence not listed in Finisar website.



  

 

Link to comment
On 10/6/2020 at 11:57 AM, R1200CL said:

Some info about CDR. 

 

Beginners Guide To Clock Data Recovery
 

http://thenetworksherpa.com/hardware-clock-data-recovery/

 

I don’t know if CDR implementation in SPF’s is the answer, but it seems to serve a purpose that probably should benefit us. 

 

Its a complicated topic! Much more so than at first glance, and I encourage your research ;) CDR needs to be done, the question is where? But honestly the great thing about the specification for 10Gbe+ is that as long as its compliant the manufacturer is asserting that it has done the stressed eye pattern test (not just a simple eye pattern) and that the equipment does not pass along jitter from upstream. So really you can just get it and plug it in. I trust Finisar has done its testing, as well as Cisco, Broadcom, Intel, Mellanox etc

Custom room treatments for headphone users.

Link to comment

@jabbr

 

Have you heard about a White Rabbit Switch?

I wanted to start a tread about if part of that project could be used in design of audio transfers using optical fiber. But I’m not sure how interesting it is. Maybe to technical and advanced?
 

We definitely shouldn’t need a central atomic clock, but maybe some technology that the White Rabbit Project is implementing in switches can be adopted. I guess the only person that could answer that is John Swenson. 
 

There is a lot of information searching the net about this. 
They was also the ones recommended 1310nm and single mode fiber. They also liked APC instead of UPC. Actually they liked to use different wavelength on TR and RX. I leave up to people to search information. If we end up in to much rabbit talk😉, we should probably start a new tread. 
 

1 hour ago, jabbr said:

But honestly the great thing about the specification for 10Gbe+ is that as long as its compliant the manufacturer is asserting that it has done the stressed eye pattern test (not just a simple eye pattern) and that the equipment does not pass along jitter from upstream.


Well, I’m not 100 % sure it’s that simple. Why did John then design the etherRegen and opticalRendu without implementation of SFP+/SFP28. He must have good reasons not to do I hope.  And why is CERN developing almost jitter free switches? (Why aren’t they endorsing 400GB ?). 

Did Sonore and Uptone more or less hide your simple solution from us ?

And why can’t I find any information that buffers will remove jitter outside switches. If it really was that simple, why isn’t there a huge (switching) buffer in the opticalRendu as an example. Not sure how well that would work with RAAT or NAA. 
 

I’ve studied some Finisar data sheets lately. (Only SFP). They have various jitter information. And some even refer to a standard. I have emailed Finisar asking what SFP have the lowest jitter. Probably won’t get an answer. 
I’m also wondering why there is produced fiber that’s have very low transfer rates ? Less power ?


The ones I have suggested (1421) seems to have very low jitter. Still there will be new jitter other places down in the signal processing. 
 

If you are able to explain what we should be looking for in SFP data sheets, that may be helpful.

 

BTW, is it any point producing stressed eye pattern if we’re in 1GB or below. Has anyone done, or why isn’t normal to do ? 

 

We can hope for audio manufacturers to adopt 10GB requirements. I think it may take some time. Unless Uptone and Sonore starts, and everyone keeps on coping them again. 

Maybe someone with god relation to SOtM can ask them to move to 10GB optical interface? If they’re reluctant, why ....

 

Link to comment

https://www.opnt.nl/white-rabbit-switch/

 

White Rabbit Switch
(WRS-3/18)

OPNT is a provider of the 18 Port White Rabbit Switch (WRS-3/18), a key component of many White Rabbit networks. White Rabbit combines optical Gigabit Ethernet connectivity with accurate, reliable, and deterministic time and frequency synchronization, achieving time transfer with sub-nanosecond accuracy over distances of more than 100 km of optical fiber. In combination with OPNT’s unique range of optical amplifiers and optical multiplexers, the range of White Rabbit can be further extended to many hundreds of kilometers, and White Rabbit wavelengths can be made to run in parallel with existing signals in fiber-optic networks.

 

Features

Sub-nanosecond time synchronization accuracy

Optional Low-Jitter Daughterboard factory installed

Truly phase-coherent replicas of 1 PPS and 10 MHz signals at each node

Optical Gigabit Ethernet connectivity and switching capacity

Autonomous and reliable operation

18 SFP interfaces supporting multiple and configurable network time protocols

Housed in a standard 1U 19’’ case

Easy to configure through remote interface

Founder of Audiophile Style | My Audio Systems AudiophileStyleStickerWhite2.0.png AudiophileStyleStickerWhite7.1.4.png

Link to comment

Let's debate the theory of ethernet switches over at the "Objecti-Fi" thread ... my main switch at home is the Mellanox SN2700 100Gbe switch, which I picked up because they consider this ancient history, they are up to 400 Gbe https://www.mellanox.com/news/press_release/mellanox-introduces-spectrum-2-worlds-most-scalable-200-and-400-gigabit-open-ethernet-switch and the 800 gbe spec has been released. Ummm consider the jitter requirements on those puppies. Not theory or research, you can buy this stuff. The jitter is essentially zilch, as specified and as measured.

 

I am *not* saying that 10 Gbe is necessary for audio, and I expect that the opticalRendu is a terrific endpoint, and the EtherRegen a perfectly excellent 100mb/1000Gbe switch. 

 

Why only SFP not SFP(+)? The design and testing requirements are vastly different. The cost of a low volume bespoke design would be cost prohibitive. I looked at a Xilinx Zynq based board as a hobby using a development platform. I can't see making enough of a profit on it to take me away from other projects (It was really really fun to learn though :) 

 

For the record, I can't "hear" a difference between my 100Gbe switch and any of the 10Gbe switches/NICs I own. Nor can I hear the difference between Intel, Mellanox and Solarflare NICs and any speed over 10Gbe (I don't own a 1Gbe NIC). The network is all perfectly silent.

 

 

Custom room treatments for headphone users.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...