SMPS and grounding

[Ryelands]

8 hours ago, rickca said:

Does grounding the SMPS mean using an LPS is no longer as compelling an idea?

 

As what follows is a bit off-topic wrt Sonore thread, I'm chipping in here. I hope that's OK but move/delete as appropriate.

 

I've been powering the downstream end of a music-system-only LAN with an LPS-1 for several months now. I have neither the wit nor the kit to replicate John's fascinating measurements but I'd be surprised if an SMPS, however well configured, were to better the LPS-1. As I experimented over months to get where I am now, I'll describe my setup. It might be of wider interest.

 

A TP-Link FMC feeds music data from a remote server (OK, an old XP box in the back room) and passes it through an EMO EN-70 LAN isolator to a Zytec ES105A switch. That switch connects to a second switch which connects to a third which connects via another EN-70 to the elderly Fit-PC2 I now proudly call my Network Audio Adapter.

 

The LAN devices connect to the LPS-1 via home-made, 24awg quad-core cables. The Fit-PC2 has its own LPS-1 and connects to an Intona USB isolator also powered by its own LPS-1. I think I can honestly describe my digital "front end" as off-grid.

 

The LAN cables are made from decent (Excel) CAT5 stock and Telegartner connectors. The system runs at 10MB/sec, allowing cable pairs 3 & 4 to be omitted. That tweak definitely matters. The cables are screened using self-adhesive Cu sheet, each with a JSSG-style drain wire. (My wife says that all that shiny copper makes the rack look like a tart's dressing table. I can't say.)

 

The lesson for me was that if I change ANY part of the above almost absurdly elaborate chain I lose sound quality (SQ). If I swap in even a decent LAN cable, remove one of the isolators, by-pass even one of the Zyxel switches, I lose SQ. The Meanwell SMPSs that run the LPS-1s are powered via an isolating transformer. If remove the transformers, SQ falls. If I leave all four pairs in the LAN cables, SQ falls. If I run the LAN at 100MB/sec, SQ falls sharply.

 

Worse, though I already knew that Zyxel switches typically sounded better than other makes and that all the 10/100 switches I tried sounded better than a Netgear Gigabit switch, when by chance I swapped one of the Zyxel v3 switches for a Zyxel v1, SQ improved sharply. (Different controller chips.) It also improved when I made the cases "sort-of Faraday shields" by filing away paint so the metal base made proper contact with the lid.

 

I could have tried one of those "audio grade" switches but it seems they all use Gigabit-speed devices and are not cheap.

 

All that said, SQ has very much improved. I'm not talking "more inner micro-detail in the singer's mid-range" or the like: what I got was a Wham-Bam in your face improvement. The system came alive even though in essence all I'd changed was the network cable between a server and a NAA. It suggests to me that the signal quality is markedly better than what I've been used to for several years now or, put a different way, LAN links are almost as pollutant as KVMs.

 

BTW, when I replaced the decent linear PS that drove the server end of the link (3 more Zyxel v1 switches and FMC but no isolators) with a USB charger thingie, SQ again improved. I plan to try an LPS-1 there when funds permit.

 

I'm not suggesting the chain could not be simplified, only that I don't at present know where to begin despite John saying "you will have to use something like the LPS-1 which will get rid of it all". The LPS-1 is superb at powering LAN devices but IME it's only a start.

 

[Ryelands]

15 minutes ago, Theobetley said:

I don't think anyone else has come up with eliminating wires within an ethernet cable and running 10/100 in the ethernet driver.

 

The 10/100 thing came from a passing remark in a post by John S a while back about Gigabit drivers being much more complex and busy, which makes sense. I did think up the wire-snipping bit all by myself but have since been told (rather snootily) that it's been done before.

 

BTW, I found a big difference between running at 10 and 100 MB/sec. Easy enough to try in many systems.

[Ryelands]

1 hour ago, mansr said:

Of course, I don't believe any of this actually matters. I'm merely pointing out that forcing a lower link speed isn't going to do much at the software level, so if you think this is important, you're going about it the wrong way.

I don't have the advantage of belief but I have experimented at length with LAN-related variables in an audio context. My only regret is that I didn't do it years ago.

 

I didn't advance an hypothesis as to why link speed might affect SQ nor am I clear what "the software level" is.

 

In light of John's passing remark, I did a five-minute trial, heard a worthwhile improvement and so continued to experiment at length. Mostly successfully.

 

I've repeated the trials on different systems with all the usual precautions. Others kind enough to repeat them on their systems generally confirmed what I reported.

 

I accept what you say about older NICs. Both systems I use as described link to the audio adapter via 10/100 PCI Realtek NICs at least 10 years old, probably more.

[Ryelands]

44 minutes ago, mansr said:

Sorry if I misunderstood you.

No problem but thanks anyway. As I quoted John from memory, I may not have got his point quite right. My point OTOH is that he's to blame for all this.

[Ryelands]

33 minutes ago, jaaptina said:

Has anyone actually experienced a boost in SQ after applying John's SMPS grounding hack?

Not quite a straight answer to your question but I have three LPS-1s each driven by a stock SMPS (Meanwell ?). If I connect the MWs straight to the mains, the SQ is not as good as it is if I connect them via Jon Risch-style isolating transformers. The improvment is marked and readily repeatable. Whether it's down to a quirk in my system, I can't say but it's a damn nuisance because the transformers are crying out for a case on both aesthetic and safety grounds which I really, really don't want to make. And they take up too much space.

 

So, last night, I tried the tweak (the JSGT?) on the MW that powers the DAC's clock and thus (or so I guessed) the one most likely to show up an SQ loss. Several hours in and so far, so good - no detectable SQ loss compared to the isolators. I'll be trying the second MW shortly and will report back.

 

Should anyone feel the temptation to ask what the effect of using the isolators and the JSGT might be, resist it . . .

[Ryelands]

On 31/03/2018 at 12:16 AM, wwc said:

Could you be more specific to the model # of the Zyxel "v1 switch?"  Thanks!

Humble apologies but I've only spotted this query. It's probably an age thing.

 

Should it still be on any interest, I use ZyXel ES105As and find that the now-hard-to-get V1 sounds better than the more common V3s. But, to get the best from either and all-but eliminate the gap between them, replace the 3.3v regulator at the input with an LT3045-based device. See pic of a v3 box though it's easier to do a neater job with shorter leads with the v1.

 

Neater still, wait until Uptone Audio releases its upcoming audio-grade switch which, I note with interest, is 10/100MB/sec only.

 

HTH

 

Dave

[Ryelands]

24 minutes ago, oneguy said:

UpTone switch?