Linear Power Supplies, Optical Connections or do nothing?

Following general forum advice, I was looking to power at least my English Electric 8 switch with a linear power supply, but also possibly my router and other switches as potential improvements in my digital chain. A multi output LPS like the HDPlex 300W would accomplish all of this or I could try various well documented other options from Silent Angel, MCRU, SBooster etc.

On approaching a HDPlex supplier however it was suggested I could achieve more by introducing an optical element to my ethernet chain, which would remove virtually 100% of any noise and this aproach would even avoid the need for LPSs?

I’ve since read a good deal of information on other forums (and some on this one) on the benefit of introducing an optical element to ethernet chains, some claiming optical isolation for them (even with reasonably high end streamers / DACs) was the equivalent of a black box upgrade! What I’ve also learnt is introducing optical into the ethernet chain could be done quite simply via two Fibre Media Converters and a Fibre Cable (cost £80 est.), then add an LPS on the ‘clean’ side to improve more, or if you want to really go for it there are more esoteric solutions involving multiple optical switches or offerings from audio suppliers such as Sonore (Optical Module Deluxe) etc.

I have tried some limited testing within my current system (i.e. not spending anything) to try and evaluate any benefits of removing any noise (if there indeed is any!) from my digital chain. I can stream Roon to my Sony TV which connects via optical to my NDX 2 but this has limitations in bandwidth (48khz max) the TV can handle via Google Cast and obviously loops in the TVs processing. That said, with even a 16khz stream the ‘pseudo optical’ route I employed didn’t appear to add anything. I’ve also tried streaming to my NDX 2 via MConnect and UPNP which will stream hi-res, which should avoid any cable/switch noise but I feel my cabled route including long runs, a Cisco 2960 and EE8 actually offers more and not just in ease of use. So my question to myself there would be, do I actually have any digital noise to remove or am I chasing my tail?

I also read somewhere on this forum that the newer Naim streamers are optimised to remove some nasties in the digital chain, and I’ve spent some time getting my ethernet chain right via switches and quality cables, so maybe that’s why I’m not feeling any major change streaming via UPNP or via my ‘pseudo optical’ route?

I’d appreciate any advice from anyone who has actually implemented an optical link and got any major uplift for their efforts? If so, it would be great to hear what was implemented too? I can go for LPS/s or an upmarket optical run, but would prefer not to do both…

My digital chain currently runs like this:

Router → Netgear GS108 (with TV, Sky, DVD attached) → in wall CAT 5E → Cisco 2960 → Roon NUC → back to GS108 → DeLock Ethernet Isolator → EE8 → NDX 2

Thanks for any constructive advice!

@james_n uses the ADOT system and may be able to advise.

1 Like

The whole ‘optical removes noise’ meme starts to become annoying. It only removes all noise if, and only if, the receiving device has an optical in, which has been carefully implemented for audio purposes.

Any other solution introduces noise through the power supply of the optical-to-ethernet translation device, none of which are engineered for audio purposes and deploys electrical isolation of the ethernet connection (which connects to your streamer) because that’s kind of irrelevant for data networking use if the ports and cables are within spec. Audio use needs a far better isolation than network use from electric noise creeping in over the electric connections.

Furthermore, using those optical-to-ethernet boxes has a rather distinct and profound impact on the stereo imaging, something you might like or dislike. Beware of that.

Given your current setup, I’d say you already have far too many stuff in the ethernet chain already. Remember that those Netgears already have quite good port isolation. You can actually overdo it.

If you’d apply some logical thinking, I’d move the Cisco 2960 to the other end and run an ethernet cable from there into the NDX2. Connect that one from one Netgear. Isolation enough and it probably will lead to a far more musical sounding system, instead of the rather analytical sound that will invariably be the result of stacking so much isolators in your ethernet chain. You might like the character of the EE8 better, so you can try and use that one instead of the Cisco.

Just try to take out everything first it is my advice, you wouldn’t be the first one to be surprised that it is an improvement after tweaking so far down the path. The EE8 is a rather poor performing switch when it comes down to isolation of noise of the power supply getting into the ports. Having a Netgear in front of it (the first one) with a long run cable through the wall is isolation enough though.

If you consider adding an LPS, the last switch before the NDX2 is the one that needs it.


Thank you, all useful advice.
It is actually only one Netgear after the router, feeding the wall CAT 5E and various other devices, plus the EE8. The Roon feed comes back to the same switch. I implemented that before the EE8 but kept it to try removing traffic/noise from the EE8 so I’ll try all those connections straight to the EE8.
I’ll also try the other suggestion, swapping the Netgear and Cisco. The Cisco in my office currently feeds the NUC, the in wall CAT 5E, a PC and printer however is a bit of a brute to have in a lounge due to size and all the data connections front and bulky IEC at the back. Worth a try though.
I found moving the NUC into my office away from the audio definitely brought a good change, even though all the SMPSs are on a different power feed than the Naim gear.
I also understand the point about optical devices needing to be geared to audio, it was suggested I use a Sonore Optical Module with a LPS as an end point before the NDX 2, but as you say, I may already have enough complexity going on. I’m still curious though if that would remove noise or actually do nothing of note.

With only a lay-man’s knowledge, it seems to me this whole area is a mix of engineering and voodoo.

Excellent logical questioning summary by @jmtennapel above.

For me, I go back to 1st principles and try to have as little done to the signal as possible, which is what we always aimed for before streaming arrived. I have serious misgivings about devices that regenerate or translate from ethernet to optical and back again – my scepticism dictates that if anything is gained in these processes, there is a likelihood that something will also be lost.



I was at the High End Munich, speaking to a lot of engineers. It is becoming less voodoo and more engineering each year.

Most of the voodoo is induced by forums like these. No engineer is stacking all these devices in the chain, they are focussing on isolating their devices better, so these become immune to all the Ethernet tweaking. I have seen and heard (from the engineers) lots of convergent solutions, which they independently arrive at.

My takeaway from these conversations: optical works, but only if it is used in a re-engineerd way, meaning the device has an optical network input as a way to isolate from the external network. But those optical connections are not always TCP / IP based, but propriety solutions solving some other issues. In the end, some of the ‘pollution’ that causes specific forms of jitter in the D to A conversion stage is already in the signal, and then it doesn’t matter anymore if that signal is electrical or light based. The pollution is just passed on.

Furthermore, at some stage the light has to be translated back into signals and that introduces new sources of unwanted electrical noise.

In short: optical works to isolate from network noise, but not all electrical noise.


FWIW - I have a Qobuz streaming setup that works rather well (good SQ)
A. NUC7i7NHE1 (fanless in AKASA Plato X7D case, 16GB RAM, Win10, Intel Optane 800P NVMe SSD) powered with an LPS (Paul Hynes SR7T 19VDC/10A). I happen to have a spare rail on this LPS, otherwise it is overkill!
B. RJ45 to USB converter : TP-Link UE300
C. FMC (2) : TP-Link MC-110MCS powered by one LPS with 2x 9VDC/2A rails (China via ebay)
D. Switch : (an older model metal case TP-Link TL-SG1008D) with a 7VDC/2A LPS (DIY with an old 1st gen Ted-P board)
30 ft flat patch cable: Tera Grand CAT7
5 ft patch cable: Startech CAT6
2 m patch cable: Canachoice CAT8
1 m fiber patch cable: Tripp Lite N506-01M SC/SC 50/125

Router—CAT 7—D—CAT6—C1 —Fiber—C2—CAT8—B—A

1 Like

I will start by saying that i have not invested in anything beyond standard domestic grade ethernet. I also have lot of experience on the commercial side of optical networking. I certainly understand the theory of isolation with an optical link. However my main concern is that any investment on the optical side could be easily undone if the PSU driving the streamer side optical connection is not up to scratch as that could just reintroduce noise that the optical link is looking to remove. You also have the issue of multiple signal conversions potentially adding jitter into the signal.

Also I know from my commercial work with optical that cleanliness is critical when making optical connections. Any amount of dust or dirt can impact signal strength and error rates.


Thank you everyone on here for the good advice.

I’ve had a good look at optical solutions from ADOT, ENO, Sonore, Cisco switch optical connections and the cheaper 2 x FMC options. There are of course Melco’s and Innuos devices etc. but the price begins to escalate and then a black box upgrade comes into reckoning.

I also got some helpful advice from a Linn DSM owner who has FMCs and LPSs on every network node, his NUC and elsewhere!

I will do some testing by rationalising my LAN first, see what that does, if anything.

I also read with interest that the Naim NDS/NDX White Paper specifies 2 levels of optical isolation within the streamer, I’d assume that was carried forward to the newer streamers (if not more) so I’m not sure what adding more optical isolation will achieve…

I’m also unclear on is what LPSs actually bring to the party. There are a few members on here who have say they’ve gained using an LPS with the EE8 switch, also on Cisco switches modified to use an LPS and various Sonore/Uptone/FMC devices with LPSs added.
Can anyone clarify for me, does an LPS on a switch or similar node reduce only the noise transmitted as EMI/RFI down a streaming cable, or actually reduce noise transmitted with the signal (or possibly both)? If it’ improves the actual signal, it seems to me (with only basic knowledge) that adding an optical link will not necessarily help without cleaning up noisy network nodes first?

If I’m going to invest in something I’d like to be reasonably clear where to start (after rationalising my network) and how far I then need (or want) to go!

Apologies if that’s more questions than anyone feels like answering!

A switch isolates the electrical noise on the ethernet connection by the standard properties of having port isolation. However, that port isolation is not 100%. Furthermore, the power connection of the switch itself introduces new electrical noise into the signal and this is travelled down into the streamer. Switches are not designed with audio use in mind, and the amount of electrical noise the power supply adds is has no impact for data transport applications.

Changing the power supply for one which introduce as little interference as possible has a positive effect, especially for switches where the port-to-port noise (not so good port isolation) is high.

It doesn’t have to be a linear power supply though, a clean power supply is more important. A slow responding LPS has a negative effect on timing in the D to A conversion stage, so that is another thing to consider. The thing has to be fast responding for use with a switch.

It is not just about RFI/EMI, but every form of electrical interference. EMI/RFI is an isolation problem in the cable itself. Using a shielded ethernet cable (Cat 6 or higher are shielded by specification) will remedy this. Things like common mode noise play a part as well, so having both the( last) switch before the streamer and the streamer sharing the same ground is important as well. Even more important than the power supply used.

ps: the Netgear GS108 has both good port isolation and comes with a pretty clean power supply …


Thank you @jmtennapel , that makes it much clearer.

All my ethernet patch leads are Cat 6A with floating shields, except a Chord Epic on the last leg. Also I use IFI power supplies, standard on GS108/Sky Router and X on EE8 so there is some cleaning already in place, but I’ll study available LPSs some more, after I’ve tried swapping the NG and Cisco switches in case there are any gains there.

That’s not a good idea, you then turn the cat cables into antenna’s for RFI/EMI. Shielding properly connected is needed to have proper shielding.


I had the Netgear gs105 some years ago. Was astonished at how the sound improved when I put a good linear ps on it. It was an Hdplex at that time.

1 Like

I thought the floating shield means the shield is disconnected from the plugs, so RFI/EMI is not passed through the plugs but the cable length is still shielded. May be my misunderstanding but other forum members reported positive outcomes from floating shield cables?

They appreciate the noise induced by this. :man_shrugging:

1 Like