Optimising optical: converter, SFP module and cable choices

Following multiple exchanges in which @Anupc has offered us the benefit of 20 (I think) years of professional use of optical circuits, I’m keen to try optical myself. I’ve also been made aware of the dangers of uncertified (or questionably certified) components from certain geographies and/or bought via certain channels, and I want to make sure I get it right.

Getting it right means assembling good quality, standards-compliant components which work together and don’t cost me an absolute fortune. I will happily stick at the budget end if I can get the same performance of course, and won’t be anywhere near spending premium sums to buy premium equipment unless the mid-range experiement proves enlightening.

I’ve had a poke around and tried to learn as much as I can but this has revealed a number of different factors which need to be considered. Some are important to simply assemble a circuit which works (!) but others may be more or less important to audio applications which is clearly our concern here. Yes, we’re talking performance (speed/reliability) and noise.

Here are some considerations I’ve come across. It would be good to know for each:

  • whether you have personal experience of any particular option (+ve or -ve)
  • whether you know of the reliable experiences of others you respect (+ve or -ve)
  • which option you’d recommend and why
  • any price bracket indicators (in any currency!)
  • any stories of having compared options

I’m really not interested in measurements unless they can be correlated with peersonal experiences.

SFP or SFP+
I have no idea why one might be better or worse than the other. Any experiences? Any actual facts?

Multi-mode or single-mode
I’ve read that single mode might have lower reflection and therefore be better for audio but this could be, well you know, some sort of “internet wisdom” which gets passed around so much that it gains more credibility than it should.

Wavelength
Apparently the cable and modules/converters need to be on the same wavelength (who knew?). If so, which and why?

Speed
1G? 10G? Why?

Makes/models
Potentially a minefield where everyone recommends the one they use but I have to ask.

For example I came across a recommendation for a Finisar FTLX 1475 D3 BTL because it uses a laser instead of an LED. Completely irrelevant unless one is noiser than the other?

There may be more. If you want to comment on any of these, it would be really helpful if you could part-quote the above so we know which particular attribute you’re talking about. Of course feel free to expound on the topic in general!

Many thanks. @Anupc and all.

2 Likes

@TheFlash

A quick overview of my setup (I am leaving for Valencia so I don’t have much time)

When I am back I will replace the current fibre cable with (Corning SMF-28 Ultra fibre)

Torben

1 Like

Excellent, thanks Torben.

I can see:

  • acknowledgement of the importance of minimising noise - quiet power supplies, ferrite beads
  • Startech media converters and SFPs
  • well-regarded Cisco Meraki switch
  • Corning optical cable

Did you evaluate other makes than Corning for cable and Startech for media coverter/SFP?

For completeness, how long is your optical cable please? I don’t ask because this will have any impact on its performance; if it is say only 1m long then you would have had the option of a simpler path and might have chosen this topology for sonic reasons but if it say 10m then that presents different constraints.

Great stuff, many thanks.

@TheFlash

I did evaluate different generic SFP’s, FMC’s, multimode (cheap ones) but end up with the current solution.

No I did not test Finisar - not sure it will work with Cisco.

Make sure that you use Grade B connector and bend insensitive fiber optic patch cable. I use a 2 meter version.

Have a nice WE

Torben

PS: Here a picture of the last mile :slight_smile: 50 cm from FMC to streamer/DAC

RJ45 connector on the transmitter side (FMC) is stripped with scotch tape.

Ferrite beads to reduce common mode interference.

For the most part the decision between SFP and SFP+ comes down to speed - how much do you want and what is supported?

SFP maxes out at 1 Gbps, SFP+ at I believe 10 Gbps.

If you’re connecting between a media converter and the optical SFP/SFP+ port on your router, the decision’s been made for you.

If you’re using a pair of media converters for galvanic isolation purposes (or lightning protection, as I do) it’s really up to you. My switch can only do 1 Gbps so there was no need to buy an SFP±capable converter or SFP+ fiber adapters.

Now, multi-mode or single-mode? It kind of depends upon you, again.

Personally I dealt with single-mode throughout the portion of my career I spent at Bell Labs, so single mode it was for me.

Wavelength isn’t really important, except for the fact that the SFP/SFP+ adapters on each end have to agree about what’s being used for what - one has to receive on the transmit wavelength of the other and vice-versa.

Models? Again, ex-Bell Labs, so I prefer lasers. :grin:

My setup has a media converter on one side (TP-Link MC220L), a single mode cable and a matched pair of Ubiquiti SFPs (a kit, UACC-OM-SM-1G-S-2):

This 2-pack kit includes complementary SFP modules with simplex LC connectors. The module with the blue latch features a 1310 nm transmit and 1550 nm receive wavelength while the yellow latched module has a reverse configuration.

as one is plugged directly into my EdgeRouter 4’s SFP port.

2 Likes

My setup:

World → Unifi gateway → 1g ethernet to switches (all Netgear) → NUC10i7/ROCK → Netgear SFP bridge → standard double-fiber cable → ROBOfiber SFP bridge → SoTM Isocat 6 → dCS Rossini Apex Loop port (100mbps). Ethernet cables on Rossini side are non-shielded (Cat5 or 6) and very short.

The ROBOfiber is powered by an Ultracap LPS-1 power supply. The ROBOfiber is also able to be set to run at just 100mbps.

Most audiophile switches use 100mbps on the audio side so that’s why I set it up like that. I did not do listening tests.

5 Likes

Finisair SFP’s appear to be very good with a great reputation, however the issue of compatibility caught me out (I hasten to add, through no fault of Finisair).

I bought three Finisair SFP+'s which the Intel network card in our router recognised, but refused to enable as they weren’t Intel branded. I then bought two Intel SFP+'s, which arrived - branded Finisair with IIRC the same part numbers as before, but with an additional Intel logo!

As a second part of this cautionary tale, I had to replace the third Finisair with a Ubiquiti SFP+ as the UniFi switch couldn’t negotiate the 10Gbps connection on a reboot following firmware upgrades without unplugging the SFP and plugging it back in again.

TLDR It’s worth checking which SFP’s your equipment is certified to work with. While many will happily accept any SFP, it doesn’t mean any SFP will necessarily work without any issues

1 Like

The difference between the two is that single-mode uses a single path (or mode) of light that runs straight down the length of the cable, whereas multi-mode has multiple paths with the light bouncing off the circumference of the fibre.

These light reflections reduce the effective maximum length of the cable (as the light is attenuated more rapidly) and increases the time (and variability) between when the pulse of light first arrives and when the pulse ends as one photon might take a longer path than another. Theoretically, this can lead to jitter within the ethernet signalling, although I think this is easily handled by the hardware.

On the audio switch, I simply chose single mode to reduce the number of variables and SFP’s to try out.

2 Likes

Thanks Jeremy,
This reflects (see what I did there?) what I’d read elsewhere. Single Mode it is then.

1 Like

Excellent, many thanks.

Very thorough, Bill, many thanks.

No SFP slot on router so I’d be using a pair of converters.

I have to say, from what I’ve read, Corning are considered the gold standard in optical fibre. That said, I’m not convinced I would hear a difference between them and another top brand, but life’s too short to evaluate everything…

1 Like

Other than the difference in bit-rates involved, it won’t matter, not to the bits anyway.

Also, contrary to the claims by some Audiophile Switch vendors (e.g. Innuous), 10G/SFP+ have better (lower) noise characteristics, due to the significantly higher transmission rate involved.

Some audiophile make preposterous claims online. It makes absolutely no difference in the home context.

Please don’t forget first principles to which you yourself agreed, the bits are not changing, and noise doesn’t travel over optical fibre cables!!

Laser diodes associated with single-mode transmission are generally more expensive, but has the potential for much further reach (up to as much as 40km) because of the lower dispersion of the wavelength over single-mode fibre. Whereas multiple-mode lasers are much cheaper, but transmissions incurs significant dispersion and thus has a much shorter reach.

None of this has any consequence to the bit error rate, and no consequence in the home environment.

The typical optical SFPs used in audio are what’s generally known as “grey optics”, typically 1310nm and 1550nm. They’re standard wavelengths defined by the ITU and IEEE for 1G/10G optical transmission.

You can also BIDI optics (combines and TX and RX onto a single fibre instead of a pair), as well as coloured optics (CWDM and DWDM optics with a very wide range of frequencies); designed for multiplexing multiple 10G wavelengths onto a single pair of fibre via optical couplers. But none of these have any real use in most home environments.

Finisar (now II-VI) are one of the World leaders in pluggables, and are in fact a key OEM supplier to many vendors. Some audiophiles who claim one SFP is better than another don’t actually realise that many of the SFPs are in fact identical, just branded/labeled differently :wink:

More to the point, please don’t forget first principles to which you yourself agreed, the bits are not changing, and noise doesn’t travel over optical fibre cables!! SFPs and optical cables have no impact to sound quality. Impacting sound quality goes against basic physics and engineering of how these things work. Those who claim to hear otherwise, are IMHO delusional :slight_smile:

5 Likes

Just wanted to mention the fibre should match the wavelength in use too

Perfect, many thanks!

So… Since we are discussing this… What are people’s views on using 100mbps or 1gbps ethernet? The argument for lower speed seems to be that the electromagnetic noise put out is lower (ie lower power and more limited bandwidth) than the faster speed. I don’t know that that makes sense to me but I can see having fewer digital transitions might produce overall less noise.

1 Like

I believe that’s the theory, however I believe the dCS implementation buffers the music. In which case, one might find that getting the music data buffered more quickly means there is less data going over the interface when the music starts playing.

The argument is a little different: the network interface produces electromagnetic noise with each bit transition. Those bit transitions are happening all the time regardless of whether you’re sending audio data. I think the argument is there are 10x more of those transitions per second and thus more electromagnetic noise, at 1gbps than 100mbps. I don’t know if this noise is lower than cosmic radiation or our presence in the room, but I think that’s the argument.

3 Likes

I wonder, if dCS would know that a 100mbps ethernet port would sound better than a 1000mbps one, wouldn’t they have made the former the standard connection, instead of a loop port?

Well let me ask this:

Why is there a loop port at 100mbps, with its own separate controller? I mean, why bother at all?

What is the use of this port especially given the slower speed and the fact that the controller is set up in parallel, ie it is completely equivalent to the 1g connection?

What exactly is a loop usecase? Put a computer on a slow connection?