Affordable ethernet filters

I was reading a review of the first version of @tedwoods SoTM filter done 10 years ago and whoever wrote it summed up exactly how I feel about listening to these things. They clearly work, it's unmistakeable, but it's difficult to put your finger on it. Bear in mind in this review was based on a stream from a tower PC that probably generated more noise than data.

Clearly there was a difference. But it fell outside the primitive treble/mid/bass and soundstage parameters reviewers love to discuss. From today's device you'd neither expect any changes there nor will you get any. Your focus has to be on something else. And you have to trust your ears. Then you might hone in on the contrast between sounds and silence particularly during the pianissimo interludes and lazy fades. You might call the direct stream somehow dirtier, sharper or more pixelly; the one with the 'black' box smoother, rounder, less nervous and more settled. In Martina's flageolet exploits, you might detect more grit without the SOtM. Bass transients could seem wirier and leaner. Are you imagining it to justify a $350 expense that doesn't put food into the refrigerator? After all, the direct stream doesn't sound noisier per se. At all. This experiment is about whether additional noise somehow interacts with (intermodulates) the audio signal and perhaps even affects jitter. With emphasis on somehow, the SOtM isolator works in that domain and hearing notices it.
6moons audioreviews: SOtM iSO-CAT6
This was indeed the review of my filter's precursor, the iSO-CAT6.
I've had that as well when I had an all SOtM front end (the so called-"trifecta") but have since replaced it with the latest version, the iSO-CAT7.
The main difference between them was that the former had only one transformer (like so many of the "typical" filters one can still find in both mainstream and "audio" versions), while the latter, as you have pointed out has four (plus related circuitry).
The corresponding 6moons review for that is the following: https://6moons.com/audioreview_articles/sotm-iso-cat7/
The iSO-CAT6 was indeed mild at best at what it did, something I also had personally pointed out in this thread: https://www.whatsbestforum.com/thre...e-sotm-iso-cat7-and-lampizator-amber-4.36104/
Cheers,
Ted
 
This was indeed the review of my filter's precursor, the iSO-CAT6.
I've had that as well when I had an all SOtM front end (the so called-"trifecta") but have since replaced it with the latest version, the iSO-CAT7.
The main difference between them was that the former had only one transformer (like so many of the "typical" filters one can still find in both mainstream and "audio" versions), while the latter, as you have pointed out has four (plus related circuitry).
The corresponding 6moons review for that is the following: https://6moons.com/audioreview_articles/sotm-iso-cat7/
The iSO-CAT6 was indeed mild at best at what it did, something I also had personally pointed out in this thread: https://www.whatsbestforum.com/thre...e-sotm-iso-cat7-and-lampizator-amber-4.36104/
Cheers,
Ted
The iSO-CAT7 does look like an excellent choice.

There was a long thread on a site called Alpha Audio about noise filtering. They were extolling the virtues of using ferrite cores for noise reduction, which is what the Network Acoustics ENO is. You can do it DIY if you can be bothered and they put up some pictures. It's known technology (and very basic tech) that works.

My suspicion is that the Network Acoustics Tempus uses the board from the Buffalo BG-GS2016 switch. It would seem a good choice. It's the board used in the Melco S100 and S100/2 switches (Melco and Buffalo are the same company) and After Dark also do upgraded versions of that switch. There are other people who use the board to make S100 clones.

Having read a load of stuff about filtering, reclocking and the good and bad of switches, there seem to be two things to consider first:
- How much copper ethernet cabling exists before your hifi and how much power supply noise it is likely to carry into your streamer/DAC
- The reclocking going on in your streamer and DAC.

The assumption seems to be that most people have copper ethernet that carries AC and DC noise that needs to be filtered. My network does not, which is why filtering (beyond the filter in my $30 TPLink FMC) has marginal benefit to me. As I use multimode fibre with SC connectors, I don't even need transceivers, as SC connectors plug straight into some TPLink fibre converters.

Alpha report that using a switch with an upgraded clock can be pointless at best and damaging at worst, depending on the streamer/DAC. That was my experience. An upgraded clock made no difference in my system.

So, as you suggested, it makes perfect sense that you could probably replace the PhoenixNet with a $30 TPLink switch and the iSO-CAT7 will clean all the residual AC and DC noise (a $30 TPLink switch has a separate LAN isolator on each port) from the cable going into your streamer.
 
  • Like
Reactions: NaimYourTubes
My suspicion is that the Network Acoustics Tempus uses the board from the Buffalo BG-GS2016 switch. It would seem a good choice.

In talking with them its my understanding they don't - though I may be mistaken

Also I've heard Melco, I find it to be average and nothing special.

All my comments are based on actual real world experience with the gear in various set ups - not theoretical mental exercises based on specs someone else posted 10 years ago in a post that show and prove absolutely nothing.

So while like everything else that is read on a forum my posts should be taken with very little value attached to them, I only talk about real world personal experience.
 
So, as you suggested, it makes perfect sense that you could probably replace the PhoenixNet with a $30 TPLink switch
I can't betray my audiophile leanings that easily...:)
But for full disclosure's sake I should perhaps mention that, just before Xmas, I borrowed a modest Gustard N18 switch from a friend to try out: the standard version, which is what I have been trying out, has only two ports, fiber in, built in linear power supply, all alu construction (no vent holes and such to let EMI/RFI in). This particular one also seems to have an upgraded fuse in it. In conjunction with the SOtM filter after it, I don't miss the effect of the PhoenixNET at all, plus it seems dynamics have improved a bit. I still have a lot of listening to do, but so far so good...
 
Having read a load of stuff about filtering, reclocking and the good and bad of switches, there seem to be two things to consider first:
- How much copper ethernet cabling exists before your hifi and how much power supply noise it is likely to carry into your streamer/DAC
- The reclocking going on in your streamer and DAC.
For general clarity, as I suspect you know this from your tech background, reclocking is a non-issue in the asynchronous ethernet domain as the streamer does the clocking when it transforms the packets of data it receives into a continuous bitstream. "Jitter" in ethernet refers to variability in the arrival time of data packets, which may well be a concern in an overloaded corporate network but is not the same jitter of which we speak in the streamer-and-after world where timing is critical and high quality clocking vital.
The assumption seems to be that most people have copper ethernet that carries AC and DC noise that needs to be filtered. My network does not, which is why filtering (beyond the filter in my $30 TPLink FMC) has marginal benefit to me. As I use multimode fibre with SC connectors, I don't even need transceivers, as SC connectors plug straight into some TPLink fibre converters.
This would be an incorrect assumption. AC and DC noise might well be concerns for some but the primary role of a switch installed just before the streamer is to clean up the RFI noise coming into it on the network cable and to forward a lower noise signal to the streamer.
This is why the proximity to the streamer is an important consideration (obviously in purely digital terms, physical proximity is irrelevant). I've lost count of the number of folk I've heard say they tried an expensive switch and it made little or no difference but on closer examination they've installed it next to the router, where switches were originally designed to provide additional ports and traffic management, instead of where it will make an audible difference which is just before the streamer.

You may have missed my post this morning where I asked you about where any (audiophile or other) switches you had tried were installed and what they were connected to your streamer by. Whether your present system uses FMC's and/or a switch (ignoring the 100 device beast you use), it would be interesting and/or helpful to understand what your streamer is, what is connected directly upstream to it, and what type and length that connection is. Thanks.
 
Last edited:
I surveyed filters in switches, in some cases switches apparently without filters. I provided links to the isolators in the switches.

I looked at IT reviews of some switches. They mainly obsess about speed and ignore power supplies, which audiophiles obsess about.
As it happens, the QNAP switch is regarded in the IT world as a top product, has good isolation, so it makes sense that Zayin and After Dark use it. It occurred to me if you wanted a good copper/fibre switch with good LAN isolation, you could do a lot worse than the QNAP QSW-2104-2S 2-port with a 12v LPS for about $400 all in.


Many of these inline filters are near identical, loads of people make them, and they use a range of transformer modules. I agree stamina would be required given they all do much the same job in terms of DC noise reduction.

The SoTM one you use looks a lot more complex an a cut above the competition. The cheapest one around, the Delock (below) that costs about £10. The transformer module costs about £1.

Yours appears to have four separate isolators, rather than all being in one module as below, plus added regulation. A serious looking product.
View attachment 142544

The Innuos switch looks like a thing of beauty in their minimalist style. Four shielded ports, individual LAN isolators, the switch chip and an OCXO clock. Everything else in the box is the power supply. I understand it operates at 100mbps only. They say that gives the lowest jitter. Seems to me a simple and well thought out exercise in noise reduction.

The PhoenixNet is more about reclocking the signal and Innuos told me it is likely more effective on earlier models like the Zen Mk3 and Zenith Mk3. Instead of getting a PhoenixNet, I upgraded to the Pulsar, which as an OCXO clock on the usb out. My EMO filter sits before the streamer.

View attachment 142546
It's my understanding that the primary driver for using 100mbps over Gigabit is that crosstalk is much lower compared to Gigabit.
 
  • Like
Reactions: ssfas
In talking with them its my understanding they don't - though I may be mistaken

Also I've heard Melco, I find it to be average and nothing special.

All my comments are based on actual real world experience with the gear in various set ups - not theoretical mental exercises based on specs someone else posted 10 years ago in a post that show and prove absolutely nothing.

So while like everything else that is read on a forum my posts should be taken with very little value attached to them, I only talk about real world personal experience.
Would you put diesel into a petrol car to get real world personal experience that it doesn't work? Hopefully you'd research what fuel is compatible before making an expensive mistake.

Likewise, some of these products lack basic compatibility. The Taiko is SFP-only and the Innuos is copper-only. I could use neither.

The NA clues are the port layout, the 12v power, the fact the BS-PS2016 is managed so the ports can be programmed to 2x 100mbs and 6x 1gps, that no one supplies a 2x 100mbs and 6x 1gps board and its popularity.

I've tried all sorts of audio network configurations over 15 years
- Direct copper wired (into Linn Akurate)
- Wireless (into an Auralic Aries and Devialet Expert)
- Copper cabling with a Netgear GS108 switch with linear power supply
- Fibre cabling with Cisco 2960 and Netgear GS110TP switches - unmanaged
- Fibre cabling with a managed Ubiquiti USW 24 PoE switch
- Linear, switch mode and battery power on TPLink FMC
- Adding a Cisco 1110 with upgraded clock and linear power supply after the FMC (a clocking experiment)
- Two inline filters - the Delock and EMO EN-70e
- A VLAN isolating the music data path from the rest of the network traffic, with copper
- A VLAN isolating the music data path from the rest of the network traffic, with fibre
- Native data streaming or converted to DSD256 using HQ Player

I have fibre optic incoming to my house, a managed Ubiquiti USW 24 PoE switch feeding a 25m multimode fibre optic cable to a battery-powered FMC with SC connector so no transceiver needed. The switch is set up with a VLAN isolating the music stream from the rest of the network. Two Belden BJ CAT6a cables joined by a EMO EN-70e LAN isolator feed my streamer.

That's today's real world experience. It has 100% AC isolation and three stages of DC filtering. There is no noise, the soundstage is wide and holographic, and the music is all there.
 
Last edited:
It's my understanding that the primary driver for using 100mbps over Gigabit is that crosstalk is much lower compared to Gigabit.
I've read that as well and it's featured in the Melco and Innuos devices (and I assume others).

Melco recommend that if you're using Roon to use a 1gbps input for bandwidth and a 100mbps output for sound quality.

You can set the port speed in managed switches (see below from my Ubiquiti switch). The Buffalo/Melco is I think the only managed switch made by a company with audiophile credentials, with two ports pre-set at 100mbps.

Screenshot 2024-12-31 at 15.04.46.png
 
  • Like
Reactions: NaimYourTubes
Would you put diesel into a petrol car to get real world personal experience that it doesn't work? Hopefully you'd research what fuel is compatible before making an expensive mistake.

Sorry, I didn't read past this first sentence - it's all sounds like Charlie Brown phone talk to me.

I'm going to keep trying for myself hifi set ups, listening to music and enjoying it and making up my own mind based on real world while others spend hours wasting energy trying to out-Amir each other.

Happy New Year.
 
The Buffalo/Melco is I think the only managed switch made by a company with audiophile credentials, with two ports pre-set at 100mbps.
Alpha Audio claim that managed is not useful for audio.

There are these unmanaged switches from Matrix Audio with 100Mbps outputs:

The B-Side of the EtherRegen outputs 100Mbps.

SOtM renderers can be configured to output 100Mbps. I use an sMS-200 Neo in a secondary system. It does sound better in 100Mbps mode.

The NA Muon and original ENO output 100Mbps,

The Audio Sensibility Signature ethernet cable is two-pair, which drops output to 100Mbps. It sounds excellent.

Not saying that 100Mbps is best, it depends on implementation and system synergy.
 
Last edited:
For general clarity, as I suspect you know this from your tech background, reclocking is a non-issue in the asynchronous ethernet domain as the streamer does the clocking when it transforms the packets of data it receives into a continuous bitstream. "Jitter" in ethernet refers to variability in the arrival time of data packets, which may well be a concern in an overloaded corporate network but is not the same jitter of which we speak in the streamer-and-after world where timing is critical and high quality clocking vital.

This would be an incorrect assumption. AC and DC noise might well be concerns for some but the primary role of a switch installed just before the streamer is to clean up the RFI noise coming into it on the network cable and to forward a lower noise signal to the streamer.
This is why the proximity to the streamer is an important consideration (obviously in purely digital terms, physical proximity is irrelevant). I've lost count of the number of folk I've heard say they tried an expensive switch and it made little or no difference but on closer examination they've installed it next to the router, where switches were originally designed to provide additional ports and traffic management, instead of where it will make an audible difference which is just before the streamer.

You may have missed my post this morning where I asked you about where any (audiophile or other) switches you had tried were installed and what they were connected to your streamer by. Whether your present system uses FMC's and/or a switch (ignoring the 100 device beast you use), it would be interesting and/or helpful to understand what your streamer is, what is connected directly upstream to it, and what type and length that connection is. Thanks.
I agree 100%. I don't have a technical background, and I learned the hard way ($$$) that reclocking in the switch is irrelevant. It made no sonic difference at all.

I appreciate that RFI is an issue on copper ethernet. I have 25m copper and fibre installed. I use the fibre for the hifi (so no RFI issue) and the copper for a wired access point (kept a long way from the audio).

My 25m copper ethernet is AQ Pearl CAT7. I chose it because it's shielded, very robust, can be bought off the reel and is easy to self-terminate with crimpers.

The basic issue on this whole thread is that whilst ethernet is meant to be galvanically isolated, AC and DC leakage is a reality. It does not affect the signal data, just adds electrical noise into your streamer.

I agree on switch location. The exception would be using a switch with fibre output to your streamer. But then who would spend $5,000 on a shiny Taiko switch and put it in a network room or cupboard?

When I tried the reclocking switch I put it between the FMC and streamer, i.e. as close as possible on a 0.75m cable, per your instructions!

My beast of a network provides me with hifi quality sound throughout the house, mostly over wifi, in 12 Roon zones, using 5 AP's (4 using PoE), hence the Ubiquiti USW-24 PoE managed switch. For £290/$350 it's probably the best value product I own. I also have the matching $100 network controller needed to set up VLANs.

The streamer is the Innuos Pulsar. It's designed for ultra-low noise operation. It replaced an Innuos Zen Mk3. For a time I had both machines and used it in their optimal configuration, which is to use the Innuos Sense app on the Zen sending to the Pulsar configured in Sense Endpoint mode. I did this over fibre. I compared it to using the Pulsar in Sense standalone mode, there was no difference so I sold the Zen. Sense is a very lightweight app, and as a controller I prefer it to Roon.

That 2 x Innuos config is also intended for people wanting to use one Innuos mainly as a server and/or Roon Core and another purely as a transport. I delegated Roon Server and library storage to other networked devices, including a Mac Mini loaded with HQ Player.

The other big feature of the Pulsar is the "lite" USB reclocker output lifted from the Innuos Statement. That connects to a Holo May L2 DAC using a 1m UP-OCC silver NeoTech cable. The Holo May is also optimised for usb with their "USB Gen 2.1" module.

It is well known that HQ Player DSD256 or DSD512 using Holo May's native DSD is a brilliant "giant killer" combination with a very analogue sound. For technical reasons I cannot implement that on a VLAN, so all network traffic is hitting the audio system. I switched over yesterday using a VLAN to completely isolate the hifi, using Innuos Sense. Even though I was streaming in native PCM without oversampling, the result is glorious. I do wonder if there is a real audible benefit from using a VLAN to separate from a busy network, which is not possible on audiophile unmanaged switches.
 
Last edited:
Alpha Audio claim that managed is not useful for audio.
I have multiple wireless audio devices that work in zones, up to 6 in each zone, and then the zones can be grouped in Roon or Amazon HD. When guests arrive shortly I will be playing ambient jazz over about 16 speakers in 4 zones on the ground floor of our house, at up to 24/192 PCM, wirelessly. You need a managed switch to ensure that each group is connected to the same access point.
Screenshot 2024-12-31 at 16.36.57.png
As noted above, using a VLAN to isolate the hifi does seem to have audible benefit, and it requires a managed switch and network controller. You would use a managed switch as @NigelB suggests, as a primary switch with a secondary audio switch next to the hifi.
 
Last edited:
I have multiple wireless audio devices that work in zones, up to 6 in each zone, and then the zones can be grouped in Roon or Amazon HD. When guests arrive shortly I will be playing ambient jazz over about 16 speakers in 4 zones on the ground floor of our house, at up to 24/192 PCM, wirelessly. You need a managed switch to ensure that each group is connected to the same access point.
View attachment 142641
As noted above, using a VLAN to isolate the hifi does seem to have audible benefit, and it requires a managed switch and network controller. You would use a managed switch as @NigelB suggests, as a primary switch with a secondary audio switch next to the hifi.
@NigelB didn't actually suggest using a managed switch anywhere as managed vs unmanaged is not an audio concern. You might of course need one for non-audio-quality-improvement purposes.

I am struggling to see what kind of "isolation" using a VLAN offers.

For audio optimisation purposes, what you describe as a secondary switch is actually the only switch which matters.

EDIT: sorry, just spotted "Two Belden BJ CAT6a cables joined by a EMO EN-70e LAN isolator feed my streamer" above and that your streamer is an Innuos Pulsar. I'm a little confused as to where the "25m copper ethernet AQ Pearl CAT7." is in the chain. Please could you clarify the chain leading to your Pulsar, including what hat length the two Cat 6A's are? Thank you.
 
Last edited:
  • Like
Reactions: Republicoftexas69
@NigelB didn't actually suggest using a managed switch anywhere as managed vs unmanaged is not an audio concern. You might of course need one for non-audio-quality-improvement purposes.

I am struggling to see what kind of "isolation" using a VLAN offers.

For audio optimisation purposes, what you describe as a secondary switch is actually the only switch which matters.


Please let me ask again (I can appreciate how busy you must be): what is your streamer, what is connected directly to it and by what type and length of connection? Or do you use only wifi? Thanks

EDIT: sorry, just spotted "Two Belden BJ CAT6a cables joined by a EMO EN-70e LAN isolator feed my streamer" above. What is your streamer please, and what length are the two Cat 6A's. Thank you.
From what I understand, data flows all around a network to all points and is only accepted by the relevant device identified by the IP address. So you can have masses of data flowing to your hifi, but only 1% is your music. A VLAN limits data to designated ports on your switch. So by using a VLAN the only data going to the hifi is music from the net or my music library.

The more active devices are the more EMF they produce, so the less traffic the better - so I would assume.

Plus no one else has access to my VLAN and can't mess with my music.

p.s. Chilling to the Adderley brothers.
 
1735677799623.jpeg
 
From what I understand, data flows all around a network to all points and is only accepted by the relevant device identified by the IP address. So you can have masses of data flowing to your hifi, but only 1% is your music. A VLAN limits data to designated ports on your switch. So by using a VLAN the only data going to the hifi is music from the net or my music library.
Not all around a network, all around a switch/router. If you install a switch just before your streamer - with one ethernet connection in and another out to your hifi, only the 1% will reach your streamer. No VLAN required.

Of course if you don't do this, the "masses of data" won't be audible anyway. It's basically a non-issue as far as sound quality is concerned.
The more active devices are the more EMF they produce, so the less traffic the better - so I would assume.
This is the real point. That's why installing a one-in-one-out switch just before your streamer which can reap huge sonic benefits (it's not the data, it's the noise). When I say "just"... if you stick with no-name Cat-compliant cables, keep it short (1m max) and unshielded; yep, a Cat 6a/7/8 shielded network cable is great anywhere else but the noise the switch stops simply travels down the shield to the streamer. It's an unusual use case but, for the last metre/yard from a dedicated switch to a streamer, an unshielded cable will perform better than a shielded one.
Plus no one else has access to my VLAN and can't mess with my music.
A non-real-world-issue, I'd suggest.
p.s. Chilling to the Adderley brothers.
Nice!
 
Last edited:
  • Like
Reactions: NaimYourTubes
From what I understand, data flows all around a network to all points and is only accepted by the relevant device identified by the IP address. So you can have masses of data flowing to your hifi, but only 1% is your music
That is how a hub or wi-fi works, but a switch separates out the IP addresses and sends only the correct data stream to each device.
 
Last edited:

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu