Optical Network Configurations

[jabbr]

Regarding SFPs, I’ve bought a bunch on eBay with great results (brand name), and from fs.com (but these aren’t brand name) 

 

You can also get dual 1g/10g SFP+ modules which usually work in SFP ports and there are even longer distance rated SFPs (single mode) but frankly I can’t personally hear a difference — I might possibly with single mode but I might be fooling myself, then again some people can hear dramatics things after 30sec, that I just can’t hear. I’ve also tested 10, 40 & 100 Gbe modules on the server side for which I can’t get one iota of difference — perhaps I’m  nearly deaf. 

 

I’d say start with LX and if you can hear a difference try BX or even ZX or dual 1g/10g but I’m happy with LX.

[bobfa]

1 hour ago, barrows said:

@jabbr, Are you able to recommend a specific, single mode, 1 Gbe SFP module for use with the Sonore optical products off the top of your head?

I am interested in testing single mode fiber in my set up (Although RMAF preparations and product building is keeping me very busy right now). 

Barrows,

 

I am using the SM BIDI  SFPs from FS.com.  The following picture shows what I purchased from them.  NOTE that you have to have two different modules as they transmit and receive on different frequencies.   These are  a single fiber .

 

NOTE that Others are using similar SFPs from Ubiquity.

 

I am setup with two SFPs in my Ubiquity switch and one in a FMC to my server and one in my Signature Rendu SE Optical.  

 

 

 

 

Here is the switch end with only one cable plugged in...

[jabbr]

5 minutes ago, bobfa said:

I am using the SM BIDI  SFPs from FS.com.  The following picture shows what I purchased from them.  NOTE that you have to have two different modules a

as they transmit and receive on different frequencies.   There is a single fiber in use not two.

 

Thats one of the really cool things you can do with single mode fiber — you are doubling the bandwidth by multiplexing two optical signals. 100Gbe can multiplex 4 x 25G — crazy modules. 

[Ralf11]

On 7/16/2019 at 5:50 PM, jabbr said:

...cheap FMCs and switches, but I’ve personally moved on from that. (they work fine)

 

just a quick back up here - why did you move on?

 

and to what?

[asdf1000]

21 hours ago, vortecjr said:

They are not as simple as you think. To answer your question though we do not have plans to publish them at this time. 

 

Understood. Although a little disappointing that you talk about engineering and measurements but won't let @JohnSwenson eye pattern diagrams...  which I know he would have as part of development, since he has talked about improving USB signal integrity for years...

 

[marce]

16 hours ago, jabbr said:

 

The power supply used to power the opticalRendu is more important than merely for the SFP module and network clock because it’s also supplying the USB and many DACs are bus powered. In all cases the USB touches the DAC.

 

For the FMC I think any reasonably good PSU would suffice, eg linear if the FMC is near the audio equipment. The fiberoptic cable will not transmit leakage current!

 

Professional equipment uses SMPS — I assume well designed because the jitter specs are very tight. But they don’t use really cheap wall warts!

And the finesse of power supplies is not just the main supply, more critical are local decoupling and isolated supply islands with either LDO's or PI filtering for local circuitry... These have to be done correctly when the board is designed, instrumentation and some control stuff has upwards of 20 separate supplies. A basic design with Ethernet, USB, processors should have a separate controlled supply for each section (one 0V plane,  in most instances) for best noise mitigation.

 

[marce]

17 minutes ago, Em2016 said:

 

Understood. Although a little disappointing that you talk about engineering and measurements but won't let @JohnSwenson eye pattern diagrams...  which I know he would have as part of development, since he has talked about improving USB signal integrity for years...

 

What does an eye pattern tell you though, as long as the required opening is clear the signal will get trough for a specific interface... The tighter the lines are the better the signal, but other information is required, such as supply noise, waveforms etc.

[asdf1000]

4 minutes ago, marce said:

The tighter the lines are the better the signal

 

But doesn't lower supply noise etc, result in tighter lines, so better signal integrity?

 

Sure if eye patterns for the below don't show any difference, it may not tell the complete story. But if there is an improvement in eye pattern with each of the below, that's a good thing?

 

1. ultraRendu alone

2. ultraRendu + opticalModule

3. ultraRendu + etherRegen

 

Using LPS-1.2 to power everything.

 

[marce]

Lower power supply noise always helps, but digital is also immune to noise to a certain degree, so the interface could work perfectly, the PSU noise could effect other circuitry.

Signal integrity is the opposite face of a coin, EMC being the other face. Both are critical in the design  of a product, the levels you need to achieve differ for different implementations and uses of circuitry. That said I apply the same diligence and OCD to all PCB layouts and I am a great believer in over engineering to an extent.

The results you seek would be interesting and informative I do agree. My view on the front end supply is as long as it is a decent supply it should not make any difference to the results as the main and critical part of any power delivery system is already on the board as I explained above... One would hope.

[John769]

@barrows Cisco GLC-LH-SM (x 2) confirmed working  OM -> Cisco 2960.  They were cheap s/h on ebay.

 

 

 

 

[Arpiben]

@Em2016

You may have a look at the following paper for a global approach of Jitter in general.

https://www.keysight.com/upload/cmc_upload/All/Clock_Jitter_Analysis_2008.pdf.

The document is not recent but provides a good overview.

From eye patterns you retrieve all relevant information.

Nevertheless, there are several ways to look at or measure noise in time or frequency domain.

Usually involved test equipment are: phase meters, spectrum analysers, real time scopes (Time Interval Error TIE +Digital Signal Processing), clock references (Rubidium,GPS), etc...

 

Unfortunately, when dealing with noise proof of concepts or comparisons (SFPs, crystals, clocks,etc...) there is no other way than providing proper measurements with significant differences.

Obviously above testing equipment are rather expensive (>100 kEuros/ kUSD each) and require expertise.

I am not expecting any audio companies to have the required instruments in their labs.

Keep in mind also that those publishing specifications are doing it:

- in controlled environment ( best possible conditions)

- in comparison with information provided by competitors.

(Certifications are performed by independent labs )

 

Nothing is perfect. C'est la vie...

Rgds.

 

 

[marce]

There is a different requirement for longer distance signal transmission than the few feet we use in our homes...

[jabbr]

40 minutes ago, marce said:

There is a different requirement for longer distance signal transmission than the few feet we use in our homes...

Yes IEEE 802.3-2005 I believe includes specifications for metropolitan/wide area networks. 

 

1000base-lx/lh goes to 10 kM (single mode fiber) as does 10GBase-LR

10GBase-ER goes to 40 kM and the nonstandard 1000base-ZX and 10GBase-ZR go to 70 kM+!

 

So obviously if you need really good transceivers and a really tight signal upfront in order to be able to handle the degradation over distance — which is why if upstream jitter somehow made its way into the system, it wouldn’t work as specified. Upstream noise from servers does not cross modern fiberoptic networks. 

[Arpiben]

What about Short Range 10 Gbps ones -> 400m ( 850nm/ IEEE 802.3-2012 10GBASE-SR/SW3) ?

Let me have a look at IEE 802.3 Jitter requirements by curiosity 😉

[Arpiben]

49 minutes ago, Arpiben said:

What about Short Range 10 Gbps ones -> 400m ( 850nm/ IEEE 802.3-2012 10GBASE-SR/SW3) ?

Let me have a look at IEE 802.3 Jitter requirements by curiosity 😉

 

According to IEEE 802.3-2018 10 GBASE-SW/SR & 10 GBASE - LW/LR do share same transmitter eye mask definitions.

 

  

 

 

[barrows]

5 hours ago, marce said:

My view on the front end supply is as long as it is a decent supply it should not make any difference to the results as the main and critical part of any power delivery system is already on the board as I explained above... One would hope

While I would agree, to a point (and to address marce's point, John Swenson has addressed the local power supply needs with great attention to detail, many separate ultra low noise/impedance linear regs for different sections, with multiple islands, etc on the main board of the opticalRendu and Signature Rendu SEoptical, and great attention to proper decoupling), the main power supply and first pre-regulator supply still needs to be as good as is possible for the best possible performance.  Although I agree with @marce, that it would be a fools errand to pay so much attention to the main supply if the local supplies were not properly engineered first.

 

As to the view of sharing measurements, such as eye pattern, etc.  This is a tricky area to get into for a manufacturer for many reasons.  Most customers would have no idea how to interpret things like this (and for those who do, great, I am not attempting to dumb you down!), and what would one compare these with?  Where are eye pattern diagrams from other products for comparisons?  And, even worse, if the eye pattern tests are not performed under the same conditions, one could be making a big mistake to compare one eye pattern diagram with another.  So, if a manufacturer shared measurements like these, it can do more harm than good.  Instead, I would suggest it is the manufacturers job to make the best product they can, and it is the users and reviewers job to evaluate the product for its performance.  Additionally, in my experience in this industry, I have seen many times when a manufacturer publishes a measurement, and then the public comes back and criticizes the measurement, suggesting that the manufacturer has been disingenuous about how they made a given measurement, creating a "controversy".  It is usually better for independent third parties to evaluate components, rather than for manufacturers to try and "prove" performance.  

I think it would be cool though, if an independent third party (reviewer) tested various Renderer's for both eye pattern on the USB data lines, and noise on the USB power lines under the same test conditions and published the results.  

I have seen HiFi News testing of USB cables with eye pattern results (I wonder how that LUSH USB cable would do on an eye pattern test...).

 

OR:  Or, one could just listen to the products.  The improvement in sound quality from the ultraRendu to the opticalRendu in the same system is not difficult to hear, for example, as those who have listened have reported.

[barrows]

5 hours ago, John769 said:

@barrows Cisco GLC-LH-SM (x 2) confirmed working  OM -> Cisco 2960.  They were cheap s/h on ebay.

 

 

 

 

Thanks, I have these bookmarked and will likely pick up a pair for testing here, but probably not until after RMAF as preparations are taking a lot of time now.  BTW folks, hope to see some of you at the Show!  We will have a really nice set up there (I hope the new venue's rooms are going to be a bit better than the old venue).

[jabbr]

1 hour ago, barrows said:

And, even worse, if the eye pattern tests are not performed under the same conditions, one could be making a big mistake to compare one eye pattern diagram with another.  So, if a manufacturer shared measurements like these, it can do more harm than good.  Instead, I would suggest it is the manufacturers job to make the best product they can, and it is the users and reviewers job to evaluate the product for its performance.

 

The reason the IEEE 802.3 networking standards were upgraded after 2002 was to avoid this line of reasoning. Modern Ethernet standards include compliance testing to that customers can have more confidence that network switches, NICs, SFP modules etc will work together. If we didn’t have network interoperability then we wouldn’t have an Internet. That said your equipment does not claim to be complaint with these upgraded standards so eye pattern testing is not required for your products. 

 

Most customers don’t look at the eye pattern testing, rather the IEEE standard that is claimed to be compliant. Customers assume that companies like Cisco, Brocade, Mellanox and Intel have done the compliance testing required. Companies like Tektronix and Keysight/Agilent/HP make the testing straightforward assuming you have their equipment. Each new generation of specifications eg 800Gbe(!) requires new advances in the measurement tools, and keeps Tek & Keysight in business 😉

[barrows]

8 minutes ago, jabbr said:

The reason the IEEE 802.3 networking standards were upgraded after 2002 was to avoid this line of reasoning

 

I was referring to customers comparing eye pattern diagrams from the USB output from different sources.  I do not think your comments are relevant to this.  My comment was in response to questions asking about the USB output.

 

While the standards which you refer to are great for professionals (or enthusiastic amateurs) well versed in the subject, i still do not believe that the average customer is such, and knows how to properly interpret such tests.  Clearly I am not referring to you here though!

[jabbr]

32 minutes ago, barrows said:

 

I was referring to customers comparing eye pattern diagrams from the USB output from different sources.  I do not think your comments are relevant to this.  My comment was in response to questions asking about the USB output.

 

Again, I don’t want to have an all encompassing discussion here, rather focus on optical Ethernet. My own impression regarding USB, which I suspect is similar to yours, is that it is the upgraded USB Vbus/power supply which makes the most difference for the DAC. 

 

Quote

While the standards which you refer to are great for professionals (or enthusiastic amateurs) well versed in the subject, i still do not believe that the average customer is such, and knows how to properly interpret such tests.  Clearly I am not referring to you here though!

 

My claim is that fiberoptic Ethernet compliant to the modern standards does not transmit noise from upstream servers to the DAC. I am able to back this claim up with the extensive information and testing that has been done on such equipment, in case anyone wishes to question me but no one needs to understand those nitty gritty details in order to understand what I am saying. There may be other ways to isolate noise from the DAC, perhaps equally effective, but I haven’t seen anywhere near the detailed testing nor analysis done on other methods so I’m not claiming anything else. I’m willing to be proven wrong but haven’t seen anything close to the data that would be required to do that.

[R1200CL]

1 hour ago, barrows said:

It is usually better for independent third parties to evaluate components, rather than for manufacturers to try and "prove" performance.  

I think it would be cool though, if an independent third party (reviewer) tested various Renderer's for both eye pattern on the USB data lines, and noise on the USB power lines under the same test conditions and published the results.  

I like this idea.

What would it cost pr. test ?

 

And should we add some DAC’s as well and the In the measurement chain ?

 

I’m asking what would be a fair setup ?

Including power supplies. 

[barrows]

8 minutes ago, jabbr said:

Again, I don’t want to have an all encompassing discussion here, rather focus on optical Ethernet. My own impression regarding USB, which I suspect is similar to yours, is that it is the upgraded USB Vbus/power supply which makes the most difference for the DAC. 

Well, certainly the very clean USB power is an important feature of the opticalRendu.  But John has some interesting ideas RE USB signal integrity, and I would not contradict him on that (these ideas are discussed in depth elsewhere on these forums).

 

8 minutes ago, jabbr said:

My claim is that fiberoptic Ethernet compliant to the modern standards does not transmit noise from upstream servers to the DAC

Agreed, that is why we incorporated it into the systemOptique product line.

 

But we ARE getting a bit off topic here, sorry for that...  Folks, let's try and keep the discussion here to @jabbr's topic.

[jabbr]

18 hours ago, marce said:

 

Signal integrity is the opposite face of a coin, EMC being the other face.

 

Assuming both single and multimode SFPs and fiberoptic cable’s completely isolate EMI from the source eg server, is there reason to believe that the improved electronics in the single mode SFP cause less EMI from the destination?

 

I know that the VSel lasers are considered to be substantially better, but are the opticalelectronic transducers different?

 

18 hours ago, marce said:

 

 One would hope.

 

[Arpiben]

6 hours ago, jabbr said:

 

Assuming both single and multimode SFPs and fiberoptic cable’s completely isolate EMI from the source eg server, is there reason to believe that the improved electronics in the single mode SFP cause less EMI from the destination?

 

I know that the VSel lasers are considered to be substantially better, but are the opticalelectronic transducers different?

 

 

 

Hi @jabbr

 

Sorry but I am not fully understanding. As far as I know:

• MMF SFP transceivers are using VCSEL lasers
• SMF SFP transceivers are using DFB or Fabry-Perot lasers.

Therefore I am expecting Multimode SFP electronics to cause less EMI from the destination.

You are expecting the opposite. Sincerely I have no clue and would appreciate if you can elaborate.

Rgds.

 

MMF: Multi Mode Fibers 

SMF: Single Mode Fibers

[jabbr]

1 hour ago, Arpiben said:

 

Sorry but I am not fully understanding. As far as I know:

• MMF SFP transceivers are using VCSEL lasers
• SMF SFP transceivers are using DFB or Fabry-Perot lasers.

 

I don’t fully understand either. My initial statement came from what seems to be some really really old information back in the days when LEDs were used for multimode. There also are a bunch of different types of single mode and even a bunch of types of newer lasers. 

 

That said, I looked up some Finisar literature and they use VCSEL for MMF and mostly DFB for SMF with SFP+ modules. 

 

Its hard for me to predict which would result in less EMI because SMF itself needs less power because more efficient coupling of laser to fiber yet VCSEL needs less initial excitation. 

 

1 hour ago, Arpiben said:

Therefore I am expecting Multimode SFP electronics to cause less EMI from the destination.

 

That could be the case. I am not sure.