EtherREGEN: The long development thread. [Some Gen2 dev. pics and update starting on page 92.]

[JohnSwenson]

12 hours ago, diecaster said:

 

So all the RJ45 ports and the optical port will be equal and all will clean the clock phase noise along with all the other noise from the Ethernet signals?

As far as the clock noise attenuation goes there are two domains: one has the 4 RJ45 jacks and the SFP cage. The other side has the 5th RJ45 jack.

 

So going from any of the 4 jacks or SFP cage to the 5th jack blocks the clock noise. Going from the 5th jack to any of the 4 jacks or the SFP cage blocks the clock noise. Going from a 4 jack to another 4 jack does NOT block the clock noise and going from a 4 jack to the SFP cage does NOT block the clock noise. So as long as the 5th jack is involved in the path then clock noise is blocked.

 

John S.

[diecaster]

12 minutes ago, JohnSwenson said:

As far as the clock noise attenuation goes there are two domains: one has the 4 RJ45 jacks and the SFP cage. The other side has the 5th RJ45 jack.

 

So going from any of the 4 jacks or SFP cage to the 5th jack blocks the clock noise. Going from the 5th jack to any of the 4 jacks or the SFP cage blocks the clock noise. Going from a 4 jack to another 4 jack does NOT block the clock noise and going from a 4 jack to the SFP cage does NOT block the clock noise. So as long as the 5th jack is involved in the path then clock noise is blocked.

 

John S.

 

So is Superdad wrong in correcting what I wrote a few posts above? In other words, I was correct in what I said?

[JohnSwenson]

1 minute ago, diecaster said:

 

So is Superdad wrong in correcting what I wrote a few posts above? In other words, I was correct in what I said?

Your statement said that ALL the RJ45 jacks and the SFP cage are the same, this is not true.

 

Going from J1 to J2 does not block clock noise, but going from J1 to J5 does. To me at least this does not agree with "All the RJ45 jacks are the same".

 

Now going from J1 to J5 and going from J2 to J5 is the same , but going from J1 to J2 is different.

 

Think of it this way, there are 4 RJ45 jacks and the SFP cage all connected to a very high quality switch, another port on the switch is connected to the special isolation circuitry which blocks leakage and clock noise, the other side of that circuitry connects to another RJ45. This isolation circuitry (both leakage and clock) is effective in both directions.

 

That's it. If that doesn't make sense I don't know any other way to explain it.

 

John S.

 

 

[diecaster]

2 minutes ago, JohnSwenson said:

Your statement said that ALL the RJ45 jacks and the SFP cage are the same, this is not true.

 

I made no such statement! I said:

 

Quote

 

John stated very clearly that there would be one super clean port that was an RJ45. That port must be used to get the clock phase noise removed from the Ethernet signal. So both sides are NOT the same.

 

John intimated to me that the "cleaning" of that special RJ45 port goes either way. So, to get the best of both worlds, you want to have a setup where the EtherREGEN is fed by an RJ45 through the special RJ45 port and the opticalRendu is fed by optical from the optical out of the EtherREGEN.

 

 

Then "thyname" said:

 

Quote

That was a while back. Things have changed since then.

 

Then Alex said:

Quote

 

@thyname was more correct.  The design has changed a lot since John posted the information you based the above on.  The EtherREGEN is now extremely symmetrical.  I don't have time to detail it all now (my assistant keeps walking in asking for the tested LPS=1.2 boards).  

Lots of good detailed info will come in due time.

Thanks,

--Alex C.

 

 

Then I asked:

 

Quote

So all the RJ45 ports and the optical port will be equal and all will clean the clock phase noise along with all the other noise from the Ethernet signals?

 

So, what did I say that was wrong?

[JohnSwenson]

You said ALL the RJ45 ports and the SFP port will be equal. What I am saying is they are NOT ALL equal. How can they ALL be equal if sending a signal from RJ45 jack 1 to RJ45 jack 2 is different from sending a signal from RJ45 jack 1 to RJ45 jack 5? IF the behavior fo the two paths is different how can they ALL be the same?

 

Now maybe you are thinking what you are saying is that all the jacks in the 4plex RJ45 jack are the same, THAT is true. But saying "ALL the RJ45 jacks" includes those 4 plus the 5th jack on the other side of the isolation. It IS different than the other 4. Thus saying ALL the RJ45 jacks are the same is not true.

 

I think there is some weird misunderstanding what the term ALL means here. So I'll just stop here, I've said what I can say about this.

 

John S.

[diecaster]

2 minutes ago, JohnSwenson said:

You said ALL the RJ45 ports and the SFP port will be equal. What I am saying is they are NOT ALL equal. How can they ALL be equal if sending a signal from RJ45 jack 1 to RJ45 jack 2 is different from sending a signal from RJ45 jack 1 to RJ45 jack 5? IF the behavior fo the two paths is different how can they ALL be the same?

 

Now maybe you are thinking what you are saying is that all the jacks in the 4plex RJ45 jack are the same, THAT is true. But saying "ALL the RJ45 jacks" includes those 4 plus the 5th jack on the other side of the isolation. It IS different than the other 4. Thus saying ALL the RJ45 jacks are the same is not true.

 

I think there is some weird misunderstanding what the term ALL means here. So I'll just stop here, I've said what I can say about this.

 

John S.

 

Where did I say they were equal? I said they were NOT equal!!!

[diecaster]

@JohnSwenson

 

I said this:

 

"John intimated to me that the "cleaning" of that special RJ45 port goes either way. So, to get the best of both worlds, you want to have a setup where the EtherREGEN is fed by an RJ45 through the special RJ45 port and the opticalRendu is fed by optical from the optical out of the EtherREGEN."

 

Then "thyname" and Alex said I was not correct.....

[LowMidHigh]

Guys, just drop it.

The rabbit hole is deep enough.

[BlueDL]

Well, it’s all clear now. To me, the three stand-out takeaways from this are (to capitalise on the clock noise blocking):

 

1. You could run a wired Ethernet feed into the clean(est) port and output optical Ethernet from the SFP to an OpticalRendu.

 

2. You could run a wired Ethernet feed to a standard port and output wired Ethernet from the clean(est) port.

 

3. You could run an optical Ethernet feed to the SFP port and output wired Ethernet from the clean(est) port.

 

Initially, I will likely opt for No. 3 until I can afford an OpticalRendu.

[Superdad]

20 minutes ago, BlueDL said:

Well, it’s all clear now. To me, the two stand-out takeaways from this are (to capitalise on the clock noise blocking):

 

1. You could run a wired Ethernet feed into the clean(est) port and output optical Ethernet from the SFP to an OpticalRendu.

 

2. You could run a wired Ethernet feed to a standard port and output wired Ethernet from the clean(est) port.

 

3. You could run an optical Ethernet feed to the SFP port and output wired Ethernet from the clean(est) port.

 

Initially, I will likely opt for No. 3 until I can afford an OpticalRendu.

 

Yes, you have all that correct except that—because the design now has separate femto clocks, clock synthesizers, and ultra low-jitter differential flip-flops on both sides of the active digital isolator circuit/moat, with top-notch power networks everywhere—there really is not one “cleanest” port any longer.

They are all very clean—with great clocking—and they all block leakage currents (due to other design choices), it is just that as John explained—and you understood—one copper port is entirely in its own domain, so best use of the EtherREGEN will be as you describe, using both sides.

We’ll publish a block diagram at some point that will help folks get a visual feel for it.  Will make more sense to all then.

[BlueDL]

Yeah, I guess that’s why I used the term cleanEST, to acknowledge that they’re all very well implemented but the use of ‘Jack 5’ is required to obtain the clock noise blocking benefit.

 

Maybe I should ask JR on the SystemOptique thread, but can you advise what optical cable to use, e.g. LC-LC Single Mode, Simplex cable?

 

Is Single Mode with one connector (Simplex) the optimum choice? Maybe I’m getting ahead of myself here.

 

Thanks.

[Superdad]

1 minute ago, BlueDL said:

Maybe I’m getting ahead of myself here.

 

Yes. Not just ahead of yourself, but ahead of us!  

Frankly the goal of EtherREGEN is to make none of that matter—to the point where copper in and out (across the isolation/dual-reclocking most of course) is as good as optical-to-copper (in either direction as previously explained).

We’ll have to see.

 

But as far as choice of SFP module, mode, and cable, that’s not on our radar at the moment.  If one is going to use the SFP cage on EtherREGEN, the module (optical or copper) must be Gigabit as John is not including the code to auto-negotiate down to 100Mbps (same with the opticalRendu BTW).

 

Whatever you do though, please do not bring up EtherREGEN in the Sonore SystemOptique thread!  Jesus and I are trying to keep discussion of both forthcoming products in their respective lanes.  There is enough confusion already! 

[BlueDL]

Will you be doing ‘Optical vs Wired’ trials before the product goes on sale?

 

I will need to rip out my 25 feet of SC-SC terminated optical cable anyway, but it’d be good to know what to replace it with come that time, i.e. LC-LC vs UTP vs STP.

[Jiffi32]

1 hour ago, Superdad said:

 

Yes, you have all that correct except that—because the design now has separate femto clocks, clock synthesizers, and ultra low-jitter differential flip-flops on both sides of the active digital isolator circuit/moat, with top-notch power networks everywhere—there really is not one “cleanest” port any longer.

They are all very clean—with great clocking—and they all block leakage currents (due to other design choices), it is just that as John explained—and you understood—one copper port is entirely in its own domain, so best use of the EtherREGEN will be as you describe, using both sides.

We’ll publish a block diagram at some point that will help folks get a visual feel for it.  Will make more sense to all then.

 

Thank you for the new information on the EtherRegen, @Superdad @JohnSwenson now we have a clean side and a not clean side as it were ( active digital isolator circuit/moat) is it possible to add another port, be it copper or Optical on the same side as port 5?

[Solstice380]

I’m thinking that UpTone is trying to bring the eGEN to production ASAP and that all of our requests aren’t going to change the design at this point.   If you prefer a different switch’s feature set, go for that one!

[Jiffi32]

1 hour ago, Solstice380 said:

I’m thinking that UpTone is trying to bring the eGEN to production ASAP and that all of our requests aren’t going to change the design at this point.   If you prefer a different switch’s feature set, go for that one!

The EtherREGEN does have the features i'm after, I was just asking if it was possible to add a port, expecting a 'no' answer but there is never any harm in asking if something would be possible is there? 🙂

 

[Superdad]

4 hours ago, Jiffi32 said:

 

Thank you for the new information on the EtherRegen, @Superdad @JohnSwenson now we have a clean side and a not clean side as it were...

 

Still not quite. I guess for now we’ll give up on trying to explain that BOTH SIDES ARE EQUALLY CLEAN, but that in order for two connections to have their timing domains separated (such that phase noise in either does not cause power-plane pulses on the opposite side), one must use the lone port 5 that is by itself on one side.

 

4 hours ago, Jiffi32 said:

is it possible to add another port, be it copper or Optical on the same side as port 5?

 

While that is a seemingly reasonable request, it is not technically feasible. To explain why would require us to reveal some key secrets about the design.  That’s not going to happen. 

[Superdad]

5 hours ago, BlueDL said:

I will need to rip out my 25 feet of SC-SC terminated optical cable anyway, but it’d be good to know what to replace it with come that time, i.e. LC-LC vs UTP vs STP.

 

Why rip out your existing optical cable?  While all SFP cages accept only the narrower LC pair connector, you can but for $20 SC-female to LC-male adapters.

[BlueDL]

Really?  I guess there’s no connection loss to consider here, so that might be a good way to go. Thanks.

[R1200CL]

How about add another optical port, and remove one of the RJ45 ?  (J1 to J4)

 

And if possible as a last design change, would adding a SFP cage with RJ45 be a bad idea ?

[Superdad]

12 minutes ago, R1200CL said:

How about add another optical port, and remove one of the RJ45 ?  (J1 to J4)

 

Again, not possible.  Would require addition of another full Ethernet switch chip and we already have two in there.  Not going to explain the real technical reasons about this.  And I don't see any point of having two SFP cages.

 

12 minutes ago, R1200CL said:

And if possible as a last design change, would adding a SFP cage with RJ45 be a bad idea ?

 

One can already stick an RJ45 copper SFP module into the one cage--though again there is no point to do so unless the other 4 copper jacks on that side (and the other 1 across the clocking moat) are not enough.

==========

 

Please folks, we have been working on the EtherREGEN design for almost a year.  There are a mass of technical considerations that went into it which go well beyond the external port compliment--but which do impact what is feasible.  If people need more ports than provided on one unit, they can buy a second one.  So I politely request that suggestions for redesign be held in check.  

 

[Bones13]

My feeling is that they have designed a switch that will do well in the average computeraudiophile equipment stack. Copper or Optical in, a "special" copper to the streamer/server, and a couple of other "good" copper outs for additional equipment (BluRay/NAS/Video Streamer) (my stack has networked audio streamer / video streamer / Oppo BR )  It should work GREAT for that purpose. I assume that some people will have other needs, and wants, and need to consider other options.

 

I currently have a switch on the wall in my mancave that feeds my computer + printer, and a line over to my switch in the A/V stack. I need to look in my closet again, I think my AV switch is the third in line - EdgeLite Router - switch in the closet, mancave switch then the AV switch.

 

I use a NAS in a closet to run RoonCore, and store music. My streamer is an integrated copper ethernet fed streamer/dac (Bricasti M12)

 

I am consider network changes to minimize switch count prior to the streamer, and lower electrical runs in the mancave.

 

1) running copper ethernet from the mancave ethernet wall jack straight to the new etherRegen, and then either copper or fiber back over to a switch under the computer + printer (which is ~15 foot of wall away from each other) adds an optical line, does not reduce copper line, does lower switch count

 

2) re-arrange my closet system to allow a run straight from the router up to the mancave, and reduce the switch count till the streamer (probably should be done anyway, and might have already been done)

 

3) install a new switch in the mancave with an SFP slot to feed to the etherRegen (eliminates one current copper line)

[Superdad]

5 minutes ago, Bones13 said:

My feeling is that they have designed a switch that will do well...

 

All your ideas sound fine.  

But keep in mind that if EtherREGEN achieves its design goal, it really won't matter--to your final, DAC-attached steamer--how many Ethernet switches came before it or the power supplies on those.  (Assuming you have that final music end-point connected by itself to one side of the EtherREGEN's clocking moat.)  Of course the proof will be "in the pudding" as they say. 

[R1200CL]

Will there be an EtherPCB ?

[Bones13]

good point, so just do it as simply as I can. TYVM