Optical Network Configurations

[Audio]

1310nm!! YAH!!!  Yellow!  Got it right the first time! 

 

 

(Audio)

[jabbr]

15 hours ago, wklie said:

A few days ago we had a Lumin X1 event at Volent Speaker Hong Kong office. The host played the same music after switching among copper LAN, multi-mode SFP, single-mode 1310nm SFP, and single-mode 1550nm SFP available from fs.com, and also cheap fiber vs Corning glass fiber.  The host and the guests agreed that different SFP modules sounded different.

 

There was no universal consensus on which sounds best, but I believe there are more people who preferred the 1310nm SFP to others.  People also preferred Corning glass fiber (but not easy to purchase in a usable terminated form).

 

Very nice!

 

Is Corning that hard to find? One of the biggest benefits is "ClearCurve" which is greatly flexible e.g. : https://www.ebay.com/itm/2m-LC-LC-Duplex-9-125um-Corning-ClearCurve-Single-Mode-Bend-Insensitive-Fiber/232757309843?hash=item36316ab193:g:9~QAAOSwzFVa7Oco

 

People could tell between 1310nm and 1550nm? Interesting... I've personally settled on 1310nm singlemode ... haven't extensively compared fs.com fiber vs Corning brand name (both work ;)

 

[Lobbster]

I would like to experiment with fiber optic and connect my switches - Netgear GS748t (house network backbone) to a Cisco C2960 8TS-JP (listening room). Would appreciate any advice to confirm I'm looking at the right parts:

https://www.fs.com/products/13282.html

https://www.fs.com/products/11774.html

https://www.fs.com/products/44158.html

I need less than 50ft of cable.

 

Also wondered if fiber optic internet provides any noise/isolation benefits versus cable/wire ?

I was recently connected to Telus's new PureFibre network for free but haven't subscribed. I depise Telus, but if it improves Tidal streaming it might be worth changing.

 

TIA!

[jabbr]

Those look reasonable.

 

Better to err on side of cable being too long than too short.

 

Would doubt that fiberoptic internet to the home would provide significant SQ improvement considering that the networks are isolated. The most important connection is the connection closest to your DAC, and for example if you have a direct fiber input DAC then singlemode electronics (1310 nM) might be advisable. But who knows  

[Lobbster]

Huh!?...

 

I ordered the supplies yesterday at 8:32 am. This morning at 12:49 am (?) I got the shipping notice & UPS Tracking #.  Package arrived at destination at 12:30pm (damn work). Basically the rest is plug and play and I'm now listening to an isolated fiber network connection.

 

First, massive props to Fiber Store, www.fs.com - Pretty sure that's the fastest delivery I've ever experienced.

 

I'm not sure I can offer any experienced comments, but I can play louder and it sounds much better now. Maybe that's all that really matters. Definitely worthy of the $40 investment. Thanks for the help and sharing! 

[Lobbster]

Well maybe I can offer this as a clearly audible improvement of my switch to fiber optic; I've been playing Higher Love by Steve Winwood alot lately (16/44.1). Not because it's particularly outstanding track, yet I've been persistently bothered by the cymbal work crumbling. Again, nothing special about the cymbal work, but it sounded like crud on my system. I've been testing various sheilding combinations with the Lush ^2 USB cable using this track as a reference. Even with other decent USB cables (Cardas & LH) nothing has tamed this problem. However, now the crumbling is gone and I'm listening to rather unspectacular cymbals - but they're whole again. Now I can relegate this track to the 80's dust bin and move on. Not sure if this is consistent with others experiences, but so far so good.

[jabbr]

Off the chart clock on Fiberoptic Ethernet NIC

 

Solarflare SFN7322F ultra low latency, Stratum 3 clock SFP+ (10GbE) NIC

 

Solarflare are primo NICs even better than the Intel x520 ... these are the highest spec clock on any 10G NICs I know of. If precise clocks are audible check these out!

 

https://www.solarflare.com/Media/Default/PDFs/SF-111380-CD-LATEST_Solarflare_SFN7322F_Product_Brief.pdf

[tboooe]

1 hour ago, jabbr said:

Off the chart clock on Fiberoptic Ethernet NIC

 

Solarflare SFN7322F ultra low latency, Stratum 3 clock SFP+ (10GbE) NIC

 

Solarflare are primo NICs even better than the Intel x520 ... these are the highest spec clock on any 10G NICs I know of. If precise clocks are audible check these out!

 

https://www.solarflare.com/Media/Default/PDFs/SF-111380-CD-LATEST_Solarflare_SFN7322F_Product_Brief.pdf

Very cool...only $350 on eBay.  

[jabbr]

19 hours ago, tboooe said:

Very cool...only $350 on eBay.  

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F263945138820

 

$199 while they last  

 

Solarflare is a favorite in the high frequency low latency trading crowd which has undoubtedly moved on to 100 Gbe...

[User471]

On 5/15/2018 at 2:34 AM, jabbr said:

There are many price points — I’ve suggested the ClearFog feeding USB at the low but excellent end — it would be interesting to compare this + ISO Regen vs one of the Rendu family using copper Ethernet. Still at the lower end particularly if you own a USB DAC.

 

Hi @jabbr

 

Unfortunately my ethernet feed comes into my listening room via ethernet-over-mains ?

 

No choice sadly

 

So by the time it gets to my endpoint I'm guessing it's pretty noisy (x2 switches, only one with iFi SMPS, plus the ethernet-over-mains gear and the ring main itself!!)

 

I'm toying with the idea of using FMCs between my nearest (gigabit) switch and the endpoint (<5m), so I have a couple of questions

 

1

In terms of performance for this setup does it matter particularly what spec the FMCs are (multimode vs single mode for example or 10/100 vs 10/100/1000)?

 

2

Does the ClearFog fare any better in terms of SMPS/SM regulator noise than a 'standard' FMC with an appropriate LPS? (I'm guessing not as it has a very wide input DC range) If not are you aware of any other ARM based or similar kit that could be used that doesn't have these issues (I'm happy to set any device up as a router in Linux)?

 

Many thanks for any help/advice

 

Neill

[jabbr]

7 hours ago, User471 said:

 

Hi @jabbr

 

Unfortunately my ethernet feed comes into my listening room via ethernet-over-mains ?

 

No choice sadly

 

So by the time it gets to my endpoint I'm guessing it's pretty noisy (x2 switches, only one with iFi SMPS, plus the ethernet-over-mains gear and the ring main itself!!)

 

I'm toying with the idea of using FMCs between my nearest (gigabit) switch and the endpoint (<5m), so I have a couple of questions

 

1

In terms of performance for this setup does it matter particularly what spec the FMCs are (multimode vs single mode for example or 10/100 vs 10/100/1000)?

 

I haven’t done detailed testing of FMCs and aren’t still using but prefer the one with SFP port that way you can test out single vs multimode for yourself. 

 

7 hours ago, User471 said:

 

2

Does the ClearFog fare any better in terms of SMPS/SM regulator noise than a 'standard' FMC with an appropriate LPS? (I'm guessing not as it has a very wide input DC range) If not are you aware of any other ARM based or similar kit that could be used that doesn't have these issues (I'm happy to set any device up as a router in Linux)?

 

The ClearFog isn’t an FMC at all rather an Ethernet endpoint. It is unique in having an onboard SFP port. I power with an LPS.

 

There are other, more expensive devices with SFP(+) ports such as the MachiattoBin but I haven’t (yet) used.

 

 

[Ralf11]

How about WiFi?

 

I do suspect that ethernet over mains is a huge noise source; but see some other posts saying the freq. is easily filtered out...

[User471]

9 hours ago, jabbr said:

The ClearFog isn’t an FMC at all rather an Ethernet endpoint. It is unique in having an onboard SFP port. I power with an LPS.

 

 

I appreciate that, but my plan was to run linux and use it as an FMC in effect (much as on our blog you describe using it as an endpoint, Ehernet in/USB out, I was planning on ethernet in/fiber out)

 

I already have an endpoint (ethernet in) I want to keep

 

Looking about it seems highly like that all the single board PCs/FMCs out there have some sort of SM regulation on the power circuits of the boards, I was enquiring if anybody knew of a board that hadn't

[jabbr]

10 hours ago, Ralf11 said:

How about WiFi?

You can add a WiFi card to the ClearFog Pro and use it as a router. or firewall.

[jabbr]

1 hour ago, User471 said:

Looking about it seems highly like that all the single board PCs/FMCs out there have some sort of SM regulation on the power circuits of the boards, I was enquiring if anybody knew of a board that hadn't

 

Unclear that onboard switchers are a problem where modern well designed power distribution networks mix and match. More unclear that dc-dc switchers are a problem in an FMC. 

 

The UpTone EtherRegen might have what you are looking for. @JohnSwenson would know if there are onboard DC-DC switchers.

[User471]

5 hours ago, jabbr said:

You can add a WiFi card to the ClearFog Pro and use it as a router. or firewall.

 

I might well do that but I will PM @JohnSwenson to see if he can help

[JohnSwenson]

First off some clarifications about the differences between SMPS and DC/DC switching regulators. An SMPS plugs into the mains, it chops up the mains voltage into high frequency pulses, feeds that through a transformer, on the other side rectifies the high frequency pulses coming from the transformer, filters it into fairly clean DC, measures the DC and sends a feedback signal through another transformer or optical coupler to the front end to modify the pulse width so the DC output voltage is correct.

 

These have low level high frequency noise on the output AND input, but also have both high impedance and low impedance leakage current from the AC line to the DC output. This is what causes a lot of the issues with SMPS, NOT the high frequency noise on the output.

 

A DC/DC converter does NOT connect to the AC mains, its input is DC from some other power supply. Their purpose is to convert one DC voltage to another (either up or down). Their big advantage over a linear regulator is that the POWER is constant, not the current. With a linear regulator the current going in is the same as coming out. With a switching DC/DC converter the current going in depends on the power on the output.

 

Say for example you have 10V coming in and you want 1 volt coming out. If your load takes 1 amp, a linear regulator will need 10V at one amp on the input, that means 9 watts of power are dissipated in the regulator (ie a BIG heat sink). With a DC/DC converter the power on the output is 1 watt (1V x 1A = 1W). So on the input it takes 1W / 10V = 0.1A. Now in reality the converters are not 100% efficient, maybe say 85%, so the power needed by the input will be a little higher, in this case 1.15W, so the input current is 0.15A. This means the DC/DC converter is dissipating 0.15W instead of 9W, 60 times lower!

 

The DC/DC converter DOES have high frequency output noise, but it is usually at a much higher frequency than an SMPS. SMPS usually run at 40KHz to 70KH, DC/DC converters usually run at 500KHz to 1MHz. This is a big difference. At 60KHz a filter is fairly large and costs some money. The same amount of attenuation at 1MHz takes very small very inexpensive components. The result is that IF you use a filter on the output, you can get a huge attenuation for very small board space and cost when using a DC/DC converter. Of course you DO have to spring for the filter, very cost constrained devices usually don't.

 

The DC/DC converter does NOT create any extra leakage since it is not connected to the mains. Whatever leakage exists from the power supply is still there, the DC/DC converter does not attenuate it, but it doesn't increase it either.

 

With a DC/DC converter you can always add a linear regulator after the converter, good ones can decrease the high frequency noise from the converter by 80-100dB. So a combination of the filter AND a linear regulator reduces the high frequency output to much less than the inherent noise of the regulator, it is essentially gone. So if properly implemented a DC/DC converter does not have any down side and it drastically reduces the amount of heat that has to be dissipated.

 

All that said, very low cost devices are not going to go to any of those measures, they are going to use a cheap DC/DC converter and and feed the output directly to the chips.

 

On the etherRegen, there ARE a few DC/DC converters in the design. But they have the good filters and very good linear regulators after the converter. I have tried using some very sensitive test equipment and have not seen ANY sign of the DC/DC converter on the final DC.

 

One other issue with DC/DC converters is EMI, the converter can radiate high frequency signals that can be picked up by other traces on the board. I am using some very special converters that have extremely low EMI, I can JUST barely measure it, where I can measure strong emission from other converters. Even though the EMI is so low I am laying out the board so that pickup by other signals is quite small. Again I have not been able to see any hint of this in prototypes.

 

The whole reason for using the DC/DC converters is to drastically cut down on the amount of heat the box has to dissipate, AND the high cost of getting the heat from the regulators TO the heatsinks.  I'm quite confidant that this is done in such a way that there will be NO impact of this on the operation of the device.

 

John S.

 

 

[Superdad]

Of course that was a great post by @JohnSwenson.  The only thing I feel inclined to add is that while the EtherREGEN board is already up to eleven(!) LT3042/45 ultra-ultra-low-noise regulators (at around $3.50 each in quantity), we may have almost as many of the fancy, low-emmisons DC-DC buck regulators he mentioned.  And those too are about $3.50 each.  All that does not even include the $15 isolating regulator module that we are using to get voltage across the "moat."  

 

It's crazy but the bill-of-materials for the state-of-the-art power regulation in the EtherREGEN could come close to equalling the total cost for the Ethernet switches, magnetics, PHYs, clock synthesizers, XO, digital isolators, and high-speed low-jitter differential flip-flops!

 

And some of the above parts are not always well stocked.  I am sitting here today searching stock and placing advance orders for some of the more difficult items.

I sure hope this all turns out to be worth it!  

 

[R1200CL]

12 hours ago, Superdad said:

It's crazy but the bill-of-materials for the state-of-the-art power regulation in the EtherREGEN could come close to equalling the total cost for the Ethernet switches, magnetics, PHYs, clock synthesizers, XO, digital isolators, and high-speed low-jitter differential flip-flops!

 

I suspect this power design will benefit from other upcomming products from John ?

 

Maybe even be used in a possible upgrade of existing products.

 

If you like to go even more crazy, you can cover the radiating parts with mu-metal ?

That part will easily add between $100 and $200 itself to the BOM, not to mention the possible need for a larger case, with additinal cost of pakking and freight. 

 

Alex, do you see that some of your products in the future will move away from SMPS, and iclude AC as power input ?

 

Wouldn’t such an approach be the best technical solution?

But maybe the cost will be to high or ?

[mourip]

I have been using a pair of TP-Link MC200CM's for a couple of years to isolate the clean side of my system. These sound better to me than a straight ethernet connection to my switch.

 

They have one problem however. If I power cycle my endpoint for any reason the FMC's lose their link and do not automatically reestablish it when the endpoint boots up. I then have to re-plug the ethernet cable, sometimes many, many times, before the FMC's link up again. Once linked, they remain so until the endpoint reboots again.

 

Has anyone else seen this or know what the issue might be?

[wklie]

@mourip take note of the current dip switches of MC200CM and experiment with those, especially try not using AUTO.

[mourip]

7 hours ago, wklie said:

@mourip take note of the current dip switches of MC200CM and experiment with those, especially try not using AUTO.

 

Will do. I had forgotten about those switches. It does seem like they get stuck in a negotiation loop...

 

Thanks.

[pl_svn]

afair if you are using two MCs those switches must be set to *Force*

[mourip]

Thanks for the suggestions. Unfortunately it did not fix the issue. The good news is that after a reboot if I re-plug the ethernet cable enough time it eventually syncs.

[the_doc735]

Hi,

 

can someone make informed comments on this please:

https://audiophilestyle.com/forums/topic/55562-electrically-isolate-your-networked-audiomicrorendu-problem/

 

many thanks!