EtherREGEN: Installation, Usage, Difficulty, Questions thread

[Superdad]

16 hours ago, Schafheide said:

I tried your suggestion, but I was limited to PCM88.2 max. I assume due to network load, because with NAS connected as per my proposal, I can go all the way to DXD.  Does this make sense?

(The NAS has an LPS rather than the supplied SMPS.)

 

It’s not making sense to me.  We have folks doing DSD512 through the EtherREGEN (endpoint on ‘B’ port) and you yourself said you were doing DSD512 to the Holo Red on ‘B’ port.  So I’m puzzled as to what throughput issues arise when you “turn around” the EtherREGEN and put the NAS upstream of it.  

But guess you just need to go with whatever works for your system. 

[GryphonGuy]

My Dante system (clock signals syncing that VLAN) have 8 channels (4 stereo pairs) places a network load of 58Mbits/s at sample rate of 192kHz. Non-clocked audio would be similar or even less demanding. So EtherRegen gen 1 handles this load easily and with great aplomb.

 

GG

 

EDIT: Oh I just thought. Perhaps you are trying to send 32bit PCM and something in your chain is limited to 24bit audio? Dante is 24bit limited so everything in my chain is always 24bit audio.

[Schafheide]

10 hours ago, Superdad said:

 

It’s not making sense to me.  We have folks doing DSD512 through the EtherREGEN (endpoint on ‘B’ port) and you yourself said you were doing DSD512 to the Holo Red on ‘B’ port.  So I’m puzzled as to what throughput issues arise when you “turn around” the EtherREGEN and put the NAS upstream of it.  

But guess you just need to go with whatever works for your system. 

No way do I think that the EtherRegen is the problem.

Perhaps it is my decision to connect the NAS via Cat6A to the Router, then Router via Cat6A to EtherRegen B port.  I an guessing that this placed a lot of load on the Router and it's connection to B port?

[Superdad]

15 hours ago, Schafheide said:

No way do I think that the EtherRegen is the problem.

Perhaps it is my decision to connect the NAS via Cat6A to the Router, then Router via Cat6A to EtherRegen B port.  I an guessing that this placed a lot of load on the Router and it's connection to B port?

 

Okay, but you may recall that my suggestion was to put a cheap, unmanaged switch just upstream of the EtherREGEN ('B' side in your "turned around" B>A set up) and to connect the NAS and the EtherREGEN feed to that. 

While it might conceptually seem the same to use extra ports on your main router for such, that may be the cause of why you found it not to function as well at DSD1024 (as having everything on the 'A' ports of your EtherREGEN).  

Please continue to experiment.

[Schafheide]

I have just ordered a switch.

[mourip]

On 5/1/2023 at 8:00 PM, Superdad said:

 

Actual EtherREGEN failure is extremely rare (about 6 units out of 3,200 so far). Have you contacted me about it?  Most likely something else besides the EtherREGEN.

 

As for the EtherREGEN Gen2 release:  Just last week we received the months overdue samples of the new PHY chip we are using, so at last we can proceed with the next stage test board (integrating these PHYs with the new main switch chip we already have working).  Still, we have some work ahead.  Once this next board is running we can then proceed towards the unified design--with isolation moat, differential flip-flops, the new clock synthesizers, etc.  

Thus while I am optimistic that the path ahead should be smooth, we are not ready to predict a date of final production. Still 2-3 months I would say.

 

I have used an EtherREGEN in my system for about three years. I power it from an HDPlex LPS at 12v. Yesterday I could not connect to my Roon server. I found that the ER had no lights on, either for power or for the data ports. I tried rebooting it and also connecting it to another power source. I also tried dropping the voltage to 9v. The power light now showed but the data lights still did not. The bad news is that once it was powered up I noticed the dreaded electronic smokey smell. I quickly disconnected it. It seems like an internal component has failed. For now I reverted to using a regular switch.

 

I sent you a message via your website to find out where I might stand in terms of repair. Fingers crossed!

 

Best,

Paul

[mourip]

52 minutes ago, mourip said:

 

I have used an EtherREGEN in my system for about three years. I power it from an HDPlex LPS at 12v. Yesterday I could not connect to my Roon server. I found that the ER had no lights on, either for power or for the data ports. I tried rebooting it and also connecting it to another power source. I also tried dropping the voltage to 9v. The power light now showed but the data lights still did not. The bad news is that once it was powered up I noticed the dreaded electronic smokey smell. I quickly disconnected it. It seems like an internal component has failed. For now I reverted to using a regular switch.

 

I sent you a message via your website to find out where I might stand in terms of repair. Fingers crossed!

 

Best,

Paul

More info after reading a couple more previous troubleshooting posts...

 

I moved the ER over to my desktop system and used the stock 7.5v supply. I plugged one cable into my switch and one into my laptop.

 

The ER power light does come on. I tried connecting the two cables as: A to B and B to A and A to A.

 

None of these combinations produced connectivity as evidenced by accessing the Internet. The port LED on the B side does light up and blink. None of the A ports show a light or indicate activity in any situation.

 

Ideas?

 

[Superdad]

3 hours ago, mourip said:

None of these combinations produced connectivity as evidenced by accessing the Internet. The port LED on the B side does light up and blink. None of the A ports show a light or indicate activity in any situation.

 

Ideas?

 

Hi Paul:

Are you certain that the little black slide switch for the Clock is set to INT (prior to applying power to the EtherREGEN)?

 

When I read your first post (wherein you said you smelled something and had been powering from an HDPLEX) I immediately thought that your HDPLEX might have thrown an over-voltage spike and fried Q1, the MOSFET that sits right at the DC input of the EtherREGEN.  We have seen that before. But then it would be unlikely for the main power LED to light.

I saw your e-mail inquiry from last night--and missed your telephone call while eating breakfast. I'll contact you directly today before noon.

--Alex C.

[mourip]

11 minutes ago, Superdad said:

 

Hi Paul:

Are you certain that the little black slide switch for the Clock is set to INT (prior to applying power to the EtherREGEN)?

 

When I read your first post (wherein you said you smelled something and had been powering from an HDPLEX) I immediately thought that your HDPLEX might have thrown an over-voltage spike and fried Q1, the MOSFET that sits right at the DC input of the EtherREGEN.  We have seen that before. But then it would be unlikely for the main power LED to light.

I saw your e-mail inquiry from last night--and missed your telephone call while eating breakfast. I'll contact you directly today before noon.

--Alex C.

Thanks for the reply Alex. I did look at the external clock switch when I was trouble shooting and it was already set to INT

 

I look forward to speaking with you. 

[mitch751]

12 hours ago, mourip said:

More info after reading a couple more previous troubleshooting posts...

 

I moved the ER over to my desktop system and used the stock 7.5v supply. I plugged one cable into my switch and one into my laptop.

 

The ER power light does come on. I tried connecting the two cables as: A to B and B to A and A to A.

 

None of these combinations produced connectivity as evidenced by accessing the Internet. The port LED on the B side does light up and blink. None of the A ports show a light or indicate activity in any situation.

 

Ideas?

 

Hi mourip,

 

I am sorry to hear about it, badly I experienced the same issue last month.

 

I used to charge my ER with JS2, one day I tried to make a comparison of different LPS, I changed  with another LPS made in China (similar to yours), the LPS was 12v so I thought it was safe, and I did not see the ER's LED light up instead of dreaded electronic smokey smell inside the ER. I opened up the case of the LPS and realised the fuse was blown, so I replaced it with another fuse, the LPS managed to turn on but not my ER .... I think the LPS might have thrown an over-voltage spike and fried th ER.

 

I have learned a lesson, LPS not only the sound issue but also the safety one.

[Meridimac]

My music has annoying Pops & Ticks using ER “B” port which disappear with a Netgear GS108 switch instead. This odd behavior only happens with my preferred transport, Allo USBridge Signature running DietPi, not other Roon endpoints (including other DietPi devices) into a USB Regen and the USB port of a Singxer SU-6 DDC.

 

The same Allo Signature works perfectly well from an “A” port into a USB Regen or miniDSP USB-to-SPDIF converter. Does Fast Ethernet vs. Gigabit figure into this?

 

Any ideas?

[Superdad]

22 hours ago, Meridimac said:

My music has annoying Pops & Ticks using ER “B” port which disappear with a Netgear GS108 switch instead. This odd behavior only happens with my preferred transport, Allo USBridge Signature running DietPi, not other Roon endpoints (including other DietPi devices) into a USB Regen and the USB port of a Singxer SU-6 DDC.

 

The same Allo Signature works perfectly well from an “A” port into a USB Regen or miniDSP USB-to-SPDIF converter. Does Fast Ethernet vs. Gigabit figure into this?

 

Any ideas?

 

Hello:

To comment and advise we would want to see a diagram of your system, including what power supplies, what kind of Ethernet cables are being used, what sample rates you are sending, and where your Roon Core Server is connected.

 

One question that immediately pops to mind is if you are using a shield-tied Ethernet cable to your Allo USBridge. ("Shield-tied" being defined as a cable with metal end shells where the outer shield of the cable is connected at both ends; Easy to check that with a multi-meter in "beep"/continuity mode.)

[mourip]

On 9/14/2023 at 10:01 PM, mitch751 said:

Hi mourip,

 

I am sorry to hear about it, badly I experienced the same issue last month.

 

I used to charge my ER with JS2, one day I tried to make a comparison of different LPS, I changed  with another LPS made in China (similar to yours), the LPS was 12v so I thought it was safe, and I did not see the ER's LED light up instead of dreaded electronic smokey smell inside the ER. I opened up the case of the LPS and realised the fuse was blown, so I replaced it with another fuse, the LPS managed to turn on but not my ER .... I think the LPS might have thrown an over-voltage spike and fried th ER.

 

I have learned a lesson, LPS not only the sound issue but also the safety one.

 

This pretty much mirrors my experience. I ran my ER from the same LPS for several years. It also powers my Intel NUC Roon Endpoint. When I came upon the ER it had no lights on at all. In the process of troubleshooting I rebooted the LPS. Once I rebooted the LPS I did get power to the input side of the ER but the A side is toast. I did not add it all up but I think that the LPS had a spike which caused the overvoltage protection to turn off that leg. It probably did not happen fast enough to protect the ER.

 

I had a helpful and enjoyable phone conversation with Alex and we agreed on this cause of death. I am now debating whether to send it in for a post mortem or just purchase a new board. My unit is a few months out of warranty. I could also just wait for Gen2 or pick up a used ER...

[Meridimac]

17 hours ago, Superdad said:

One question that immediately pops to mind is if you are using a shield-tied Ethernet cable to your Allo USBridge. ("Shield-tied" being defined as a cable with metal end shells where the outer shield of the cable is connected at both ends; Easy to check that with a multi-meter in "beep"/continuity mode.)

Hi Alex. I am using Audioquest Carbon cable between Allo and USB REGEN. Complete path is: singlemode fiber into ER powered by 12v Sengterbelle lps and grounded to Shunyata conditioner, ER “B” to long run of CAT5e through walls coupled to ENO Ag filter into Allo Signature which is powered by 5v Allo Shanti lps (both grounded to Shunyata), AQ Carbon into USB REGEN with Uptone USPCB A>B Adapter out and powered by 7.5v LHY Audio lps which also powers the “dirty” side of a Singxer SU-6 DDC. “Clean” Singxer side is powered by 9v Farad Super3 with AES output to Meridian active speakers. Brad

[Superdad]

9 hours ago, Meridimac said:

...ER “B” to long run of CAT5e through walls coupled to ENO Ag filter into Allo Signature which is powered by 5v Allo Shanti lps (both grounded to Shunyata)...

 

Suggestions:

a) Never put ANYTHING in the Ethernet chain after the EtherREGEN--between it and your endpoint (Allo USBridge in your case). You are messing up the signal integrity of the EtherREGEN's lovely 'B' port by putting that big hardwired transformers after it. This is the inside of the ENO Ag "filter":

 

b) Avoid running long cables AFTER the EtherREGEN. A little-known fact is that the signal degradation from Ethernet cables longer than about 8 feet end up activating additional DSP circuits in the PHY/receiver chips at both ends. (This is true for all wired Ethernet.) And that runs counter to the goals of the EtherREGEN.

 

So position the EtherREGEN close to your Allo USBridge and run a less than 2 meter cable between them.

 

You still did not tell us where your Roon Core Server is located/connected in your system.  That matters and can be also be the cause of some issues.

 

[sandston]

This 8ft ethernet cable activating additional DSP circuits in the PHY/receiver chips is new info. I am in no position to refute it but curious how the phy chip knows the cable is 8'? Resistance? If so wouldn't a cat8 have decidedly different resistance than a cat5? Also is this only the phy chip in the ER that does this or is this ethernet protocol throoughtout our systems?

[Superdad]

22 minutes ago, sandston said:

This 8ft ethernet cable activating additional DSP circuits in the PHY/receiver chips is new info. I am in no position to refute it but curious how the phy chip knows the cable is 8'? Resistance? If so wouldn't a cat8 have decidedly different resistance than a cat5?

 

There is much more to signal integrity than just wire resistance.

 

@JohnSwenson would need to address the details of all this. John was a senior engineer for LSI Logic>Avago>Broadcom for 31 years--his specialty being the power networks of custom large scale chips (rocket-science stuff as far as I am concerned), but he also worked on the chip designs of a bunch of PHY transceivers.  (Sometimes he looks at Cisco switches and points to main parts that he was on the design team for.). John is the one who told me about degradation in SI activating the AGC circuits in PHYs.

 

22 minutes ago, sandston said:

Also is this only the phy chip in the ER that does this or is this ethernet protocol throughtout our systems?

 

The issue is not at all exclusive to the EtherREGEN.  All Ethernet PHY chips for UTP--be they on a NIC, in a switch, embedded in the switch chip or as a discrete PHY--have such circuits to deal with degradation of SI with long cables.

[Meridimac]

10 hours ago, Superdad said:

You still did not tell us where your Roon Core Server is located/connected in your system.  That matters and can be also be the cause of some issues.

Thanks for the new info about PHY chips and signal integrity.

 

My Intel NUC running ROCK is connected by Blue Jeans CAT6 to a Cisco 2960 Catalyst switch which is fed via fiber from the router in my home theater where FTTH comes in. Also connected to the Cisco is an FMC with Finisar transceivers into the ER, which has three other endpoints connected to “A” in addition to the long run from “B” to my living room.

[Markus8]

10 hours ago, Superdad said:

Never put ANYTHING in the Ethernet chain after the EtherREGEN--between it and your endpoint (Allo USBridge in your case). You are messing up the signal integrity of the EtherREGEN's lovely 'B' port by putting that big hardwired transformers after it. This is the inside of the ENO Ag "filter":

Well I have been using the latest version of the ENO Muon Pro and it definitely makes a noticeable difference / improvement. 
 

[Meridimac]

34 minutes ago, richard_crl032 said:

 

That's it ? 🤔🤔😏😏

Yes. Network Acoustics is reportedly two guys in the South of England making this stuff by hand. Sounds so good to me that I also bought their AES/EBU filters for my Meridian speakers.

 

It’s a crazy hobby we’ve gotten ourselves into.

 

-And now that Alex tells me I’ve done it all wrong, I can re-jigger everything to get even BETTER performance!

[Superdad]

5 hours ago, Meridimac said:

My Intel NUC running ROCK is connected by Blue Jeans CAT6 to a Cisco 2960 Catalyst switch which is fed via fiber from the router in my home theater where FTTH comes in. Also connected to the Cisco is an FMC with Finisar transceivers into the ER, which has three other endpoints connected to “A” in addition to the long run from “B” to my living room.

 

Since Roon and the Roon RAAT protocol are both known to be a bit fussy--especially with regards extra network latency (delays) as well as with 100Mbps connections, we suggest  that you try attaching your Roon Core Server directly to an EtherREGEN 'A' side port, rather than to your upstream Cisco switch.  See if that eliminates the pops/clicks you hear.

 

4 hours ago, Markus8 said:

Well I have been using the latest version of the ENO Muon Pro and it definitely makes a noticeable difference / improvement. 

 

That's great. My comment was not about the potential sonic efficacy of adding those large transformers to the Ethernet signal. I meant no offense and glad that you enjoy those devices.

But the EtherREGEN's RJ45s--including its 'B' port--already use 12-core-per-port magnetics. And. we have received a number of issue reports from people putting those ENO boxes between EtherREGEN and endpoint. Hence our suggestion that he try placing that box elsewhere (keeping in mind that the version he is using is 100Mbps-only).

 

3 hours ago, Meridimac said:

It’s a crazy hobby we’ve gotten ourselves into.

 

Sure is! 

 

3 hours ago, Meridimac said:

-And now that Alex tells me I’ve done it all wrong, I can re-jigger everything to get even BETTER performance!

 

Well my intention was not to criticize--rather to advise on the most ideal application to obtain the best performance from the EtherREGEN, based on the principals of the design and the facts of Ethernet signal integrity. 

As always, our suggestions are to experiment to see/hear what works best.  But our suggestions are also based on having supported a few thousand users.

 

We see your system has quite a few elements to it and we recognize that it might take some rearrangements and experiments to be able to get the EtherREGEN close to your Allo USBrige--and to directly attach your Roon NUC server to an 'A' port (long cable is fine on that side). It would seem the easiest would be to connect your other endpoint to the Cisco switch, move the EtherREGEN, and remove the ENO "filter" from the 'B' side.

But if you are happy with the current SQ, then perhaps first try just attaching to the Roon Core to 'A' side while leaving the EtherREGEN, long 'B' side cable, and ENO box in place. See if just that takes care of the ticks/pops you are hearing. 

BTW, do you get those with all sample rates?

 

[Meridimac]

2 hours ago, Superdad said:

BTW, do you get those with all sample rates?

Yes, 44 and up.

[JohnSwenson]

On 9/18/2023 at 6:52 PM, sandston said:

This 8ft ethernet cable activating additional DSP circuits in the PHY/receiver chips is new info. I am in no position to refute it but curious how the phy chip knows the cable is 8'? Resistance? If so wouldn't a cat8 have decidedly different resistance than a cat5? Also is this only the phy chip in the ER that does this or is this ethernet protocol throoughtout our systems?

This gets a little deep but I'll try and condense it. A gigbit Ethernet signal is NOT binary, it actually have 5 possible voltage levels on each pair, PLUS the fact that it is bidirectional on each pair (the 5 voltages from each end are superimposed on each other), this makes signal integrity on the pairs very important, it doesn't take much degradation to make the signal hard to read for the receiver. At the 100M maximum length the signal degrades a lot. In order to read this reliably the receiver applies a complex array of methods to reconstruct the signal.

 

All modern PHY chips run the input signal through a very fast ADC. If the cable is short there is little degradation and nothing needs to be done it is easy to determine the different levels. As the cable gets longer and the degradation increases the data from the ADC is fed through DSP (Digital Signal Processing) steps which reverse the degradation of the cable. The chip looks at the signal and makes a guess as to what degradation is happening (different cable materials, construction "layup" and connectors all play a part in the degradation) and applies what it thinks is the best DSP steps to use. If it still isn't good enough it changes the DSP and tries again.

 

The issue here is that more degraded the signal is the more DSP steps are needed to "clean it up". Each of the DSP steps are clocked at high frequency and produce current spikes on the chip internal power networks, this generates voltage noise on those power nets which increases the jitter of the output signals. The more DSP is needed to clean up the signal to more jitter on the output signals.

 

Note that you do not need short cables to "get the bits", that is what all the DSP is all about, it takes care of the degradation from long cables so the chip CAN reliably "get the bits". But there is this extra jitter that is a consequence of the longer cables.

 

As Alex mentioned this is not something unique to the ER, ALL Ethernet PHY chips from all manufacturers do this. It is just how Ethernet works.

 

John S.

 

 

[Meridimac]

On 9/19/2023 at 1:22 PM, Superdad said:

It would seem the easiest would be to connect your other endpoint to the Cisco switch, move the EtherREGEN, and remove the ENO "filter" from the 'B' side.

 

OK, here’s the new topography from Theater to Electrical Closet to Living Room for endpoints #1, #2, #3 and #4: (CAT5 are longer in-wall runs over 8’)

Catalyst (T) > ENO to #1, AQ to #2, and CAT5 to Meraki (E)

Meraki (E) > CAT5 to #3 for whole house audio system and CAT5 to Meraki OCXO (L)

Meraki OCXO (L) > singlemode fiber to ER (L)

ER (L) > Shunyata Venom to #4

 

[dancause]

Meridimac : interested in your experience with placing the ENO before or after the ER in the chain : what is the impact on SQ.  I'm juggling with the same situation with a weaversa lan filter (The filter works pretty well just before the DAC but I want to test what would happen before the ER.