Well, since I have been forced to stay at home because outside 
, I have had time to ask myself a lot of questions about improving SQ.
Besides, this is a question that we are exploring here in FR on a nearby Naim forum (switch cascades, various switches like here, RJ …).
The results obtained and test reports are more or less the same across the Channel …
Strange coincidences or established facts? But it is a remarkable fact.
Some irony in your post? Or I didn’t catched it…
Thank you FR, to what I read here, and as usual, you know very well summarize things 
…
It helps a lot (in reading) long threads … :sourire:
No irony at all. why do you say that ?
Adding any device to your network can introduce noise. Adding a second switch to my network lowered the noise floor.
Thanks for clarifying. When you said that results are the same on both sides of the Channel, I was wondering if it was some irony.
As a french chicken, I don’t always understand all the hidden meanings.
The lowering of noise has two main dimensions - time and amplitude. In fact all noise has both a temporal and amplitude property - a noise spectrum when looked at in an averaging way.
For switches it may be more the timing noise that propagates through box to box if not re-clocked in a good way, so cascaded switches may just be tightening the timing jitter temporal spread.
It is not that any noise on the data packets matters in of themselves - they don’t matter at all to data integrity, but the associated coupling of especially very high frequency noise is difficult to manage-away. Once it is present on so-called earth and other voltage rails it temporally will do what it is electrically capable of doing no matter what people have decided otherwise it can’t do - it will do; meaning it will interact with all the tuned circuits formed in supply decoupling and recharge circuits and ripple through the lot, even to the clock feeding the final DtoA converter chips - and that is possibly where we ‘hear’ what happens.
Every level of pre-filtering by cascading re-clocking to narrow the available range of frequency spread should help.
I just noticed it sounded better - wondered why - don’t know for sure why, but have many reasons as to why it could be happening.
In meantime people can try implement if they like the result or if they prefer argue about it - whatever suits them best. 
DB.
DB, you said that noise doesn’t matter that much on data integrity.
At the beginning, I was using a cheap Ethernet switch and entry level Ethernet cables. Then I had a linear ps on that switch : the sound was nicer, bigger, more fluent and textured. For me it was the consequence of the drop of the noise in the Ethernet chain, with the help of the linear ps ( vs the stock smps).
If the drop in noise give better sound, with more textures and openness, better defined instruments, is it not a relationship between data integrity and noise , in the Ethernet chain ?
If you look at integrity of data - this has a specific meaning - the data values as such is not impacted at all, or nothing we use would work.
But associated with any data transfer is the fact it is facilitated by a medium and not just an abstraction of mathematics - this is where people seem to diverge on how the effect is manifesting.
The medium be it electrical or optical has a frequency component due to transition of high to low (1 to 0) of the bits and eventually it gets converted into an electrical signal - I’m suggesting that this could be where the absolute noise levels of the environment where that happens could influence the final jitter - not as measured on a time-average way but over shorter intervals that will get transposed into noise and coloration in the DtoA process.
We hear the result in an analogue signal from the DAC/Streamer - it is not data-errors - so it must be ‘something else’ - I just give my opinion of the sort of thing it may be.
Personally I’ve worked with high-frequency design and getting the noise floor low enough for the dynamic range required for audio is not simple - perhaps we are hearing it was never achieved well - until recently. 
In any case it is just experiment and discussion - I don’t ‘know’ but I know that I don’t know so that can help not rule things out as impossible.
DB.
What a neat and refreshing post with very interesting thoughts.
I know it may not be the best time to talk about our selfish endeavour but I put a SOTM clock on the EtherRegens and a really significant improvement was heard if anyone is interested.
So you feed a reference clock from this other device into the ER and it improved perceived performance - I’m not surprised as that is why they left the capability on the switch to be able to do that.
Can you give any detail of what you felt improved - and any more info about the clock source and how you connected it?
…yes I know it is dread times right now, but I’m still interested in such things, as I may have more time at home to experiment…
DB.
Hi DB.
Good to hear from you again. Yes, as I said, it’s difficult to get to excited in these difficult times but i know most on here will get comfort from the forum, especially if they have time on their hands.
I was lucky enough to pick up a used sotm clock at a price I had to give a try. It has 4 clock BBC outputs. So I moved one ER back to the supplied power supply and put the Sean Jacob’s supply onto the Clock. I only had standard cheap bnc clock leads to connect it, I have ordered Chord Shawline to try but understandably will be a few weeks delivery.
I tried the clock with one ER and from the very beginning it was clear the bass improved, more depth and definition while the mid and treble opened up and sounded more like vinyl and less digital. I, then compared each ER powered differently, one standard and one sean Jacob’s both with the clock and it was very close, very small difference now. I then cascaded the switches and it improved although not by much.
Would I pay the full price on the clock, having heard and got used to the improvements it would be hard to go back.
Thanks for the info - useful to ponder ahead.
So you figure the external clock did more than a PS upgrade - or just different?
It is nice to have the option to try in future on the ER right now.
Mine is still working in its basic config with as-supplied PS and I’m pretty happy with it, but nice to consider other options to try ahead.
It is not that I’m avoiding the doom stuff, it is just I can multi-task and find it personally healthier for me to be doing so!
DB.
Yes, bigger improvement than ps upgrade but also the ps upgrade was less when the clock was independent.
Not read yet, but very soon. The review compares 5 different Ethernet cables, from Blue Jeans cat 6a to expensive ones .
Oh dear! they used the ‘b’ word!
Really ? ‘respect’ that you actually read it all thru, I nodded off sometime in the first paragraph.
What is b world?
If I can summarize the review, the reviewer found the Blue Jeans cat 6 very good for its price, and even approaching, on some aspects, the most costly cables of the test.
The more expensive, but not the most in the test, the totaldac one, gave more colors and textures, meat on the bones, vs the blue jeans.
The least expensive, the Star Trek, was the worst.
I had similar results with the blue jeans cat 6a vs my Audioquest diamond. The BJ was very good for its price, not bright, opened, with good prat. The Forrest and cinnamon, more expensive, were more bright and less musical.
However the diamond was on another league definitely, specially for a more richer sound, with more nuances, textures and colors.
Boutique
" … I had amassed a small collection of ethernet cables ranging in price from $5 to well beyond $1,000 USD. Some grabbed in a pinch at Best Buy, others sent to me from boutique, high-end companies that specialize in bespoke cables."
He alos says “There are no measurements, double-blind tests or specifications to refer to” … just like the Naim ethernet Maniac thread.