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

[JohnSwenson]

As Alex mentioned I am just finishing up a complex test system for actually doing an end to end clock analysis (digital data stream to DAC output) in order to really nail down what is going on in our systems. There are several hypothesis for what is going on, and several possible schemes for suppressing these, I will be testing all of these out, and the one that works the best will go in the switch.

 

This is not JUST a development project, there is a lot of fundamental research going on. I have been doing a bunch already, enough to know that SOMETHING is going on, but not enough detail to figure out what it really is. That's what this stage of testing is all about.

 

So even if I wanted to spill the beans, I can't, we don't exactly know what beans are going to be yet!

 

The result of the above is that there is no way to speed this up, I HAVE to go through the research first before we can get to the details. We know what seats we are going to have, what the dashboard will look like, probably what tires, but not sure what engine is going to be in there.

 

John S.

[JohnSwenson]

3 hours ago, warpeon said:

This could be a game changer for me... I was considering to send my wifi router to get it modded to get to a cleaner network. Perhaps EtherRegen would already do the trick for me.

 

looking forward to the details on:

 

1) recommended setup: I guess the typical setup for most people is - Modem --> Wifi Router --> switch --> a) NAS, b) Computer (probably a Roon Server), c) streamer. Where should EtherRegen(s) be placed

 

2) some technical explanations:

a) why the clean port is 10/100 not 10/1000?

b) the presence of the SFP, is it for bridging 2 EtherGen?

c) LPS1.2 can power EtherRegen?

d) best placement in the recommended setup and why?

 

The new switch does two main things:

1: blocks leakage current from upstream so no leakage current travels through the Ethernet cable to the renderer etc.

2: blocks clocking issues from upstream network devices and provides its own very low phase noise clocking on the "clean" port.

 

Because of the no leakage current on the downstream cable you don't have to worry about audio cables picking up noise from the leakage current. BUT the upstream cables still have leakage current running through them. Because of this it is a BAD idea to place the switch right in with your audio equipment. You want to keep it somewhat away from the audio system so noise from the upstream cable doesn't get picked up. Do not set this (or ANY) switch on top of your DAC etc.

 

You can use this as a general purpose switch to plug in routers, NAS etc and feed the clean port to the streamer, OR can use as a "clean up" device near the audio system with just one upstream connection and the clean connection. It's entirely up to you how you want to configure that. The extra ports are there to make it convenient for you to use it in many different configurations.

 

The added expense of the extra ports is quite small, so we decided that if we can fit them on the case somehow to go ahead and include them.

 

Why is the clean output 10/100 not gigabit? Because it is much cleaner to do so. A significant amount of jitter on a Ethernet cable come from noise on the power/ground (PG) networks inside the chip. The more stuff is going on and the faster it is doing it, the more noise gets generated on the PG network. Gigabit has way more stuff going on inside, thus generates a lot more noise, which causes significantly more jitter. By keeping it down to 100 the clean port has much lower jitter.

 

The SFP port is there because it is very easy to do and very cheap to add. Some people already have optical networks in their home, this just makes it a lot easier to use for them. There is no specific purpose for it, other than another "dirty side" port which makes it easier to use for some people.

 

Yes an LPS-1.2 can power it, but you have to be careful there. As with what I have been talking about in other threads, part of the leakage blocking needs a ground connection to the switch. The LPS-1.2 specifically isolates the ground so you will need another way to ground the switch. We will most likely be adding some form of grounding connection on the case for this purpose. The UASMPS provided with the switch already has the DC output grounded so no separate grounding needs to be done. The internal power network in the switch will be extremely good, the result is that there should be no advantage to using an LPS-1.2.

 

John S.

[JohnSwenson]

As to the questions on PoE, it will not support PoE in any form. PoE works by running  very high voltage low current DC over the Ethernet wires (the wires are thin so it has to be low current), this requires very high ratio switching DC/DC converters on both ends. The converters are NOT known for being low noise. The whole purpose of this switch is to make things as low noise as possible, throwing in PoE would completely destroy that.

 

John S.

[JohnSwenson]

4 hours ago, octaviars said:

 

If the upstream cables uses your ground solution (JSSG) will the radiated noise still be a problem to consider in a system?  

Yes if the cables are shielded properly with my shielding techniques then the radiated leakage current is hugely reduced so it will not a be a problem.

 

John S.

[JohnSwenson]

2 minutes ago, R1200CL said:

@JohnSwenson

 

Now, I must have been missing or mixing something from the other treads. 

If a LPS-1/1.2 is powering on those present mentioned switches, and the Meanwell is modified with the shunt or I use the new UASMPS, do I still need to add JSGT on the DC comming from the LPS-1/1.2 powering that switch (Cisco SG108D-08 / Netgear FS105 as an example). ?

 

Yes,

the blocking of upstream leakage only happens when the ground plane in the switch is actually grounded. The easiest way to make sure that happens is to ground the negative of the power supply. Some switches have a ground screw, but that just grounds the CASE, NOT the ground plane of the PCB.

 

The LPS-1(.2) deliberately blocks the ground connection from the input to output, so even if the INPUT is grounded, the output is NOT grounded. Thus a separate ground connection on the output is required if you want to get rid of the leakage from upstream network devices.

 

The new switch will contain a similar connection scheme as is used on the "special" switches so the upstream leakage reduction works in the same way.

 

John S.

[JohnSwenson]

1 hour ago, rickca said:

By 'upstream network devices' does that mean only modem/router or any device connected to one of the input ports such as a NAS.

It is NOT just devices connected to the switch, it is the entire set of devices connected with wired Ethernet cable, of any type.

 

John S.

[JohnSwenson]

11 hours ago, R1200CL said:

 

Shall this be an active or passive device ?

 

If active, you purchase two EtherRegen. If passive, I’m not sure if it requires a gigabyte port in with 8 wires active, as a spitter can only use 4 wires and thus is either 10 or 100. 

 

 

Ø

This is not really a splitter. Note all the wires stay separate. 10/100 takes 2 pairs and standard Ethernet cable has 4 pairs, thus two of the pairs are not used. This device interleaves the two separate 10/100 circuits and puts them on one cable, they stay two independent circuits.  The only reason for doing this is if you already have a single Ethernet cable in the wall and you want to run two 10/100 circuits, you have one of these on each end of the single cable, thus you get two separate circuits without having to run a new cable in the wall.

 

Note this only works for 10/100, gigabit uses all 4 pairs, thus there are no unused pairs in the cable.

 

John S.

[JohnSwenson]

11 hours ago, R1200CL said:

 

Well only John can answer if the clean port will use 4 or 8 wires, but we know it already is a 100, so I assume it’s 4, and then a spitter can’t be used. 

 

Maybe also you can can create other unwanted issues as well, as you must be 100% sure the equipment you connect to the splitter doesn’t interfere with each other. 

 

I think you rather purchase 2 switches ?

There is no such thing as a passive Ethernet Splitter, the only way to get data from one port to two ports is a switch. And even that doesn't logically send the same data to the endpoints, the software must send the data to one, then to the other. (It CAN be done if your protocol is using multicast packets, that is pretty rare and the software on both ends has to be written for it.)

 

The upshot is, there is no easy way to "duplicate" the single clean port, it needs to be a seperate port on the switch fabric. Doing this for our switch would essentially double the electronics, you couldn't even keep the same power supply, since it would take double the current to do that, there would be too much stuff to fit in the same box, so it would take a bigger case. The net result is that this switch with two clean ports would cost double what the single one costs, and force everybody to pay double when very few people really need two clean ports. If you really need two clean ports, get two. I'm certainly not going to spend the time and money to make TWO separate designs, one with 1 clean port and one with 2 clean ports. That is three times the effort of just doing the single clean port project.

 

John S.

[JohnSwenson]

13 hours ago, R1200CL said:

@JohnSwenson

Alex has already stated that the SFP port will be on the ‘’dirty’’ side. 

 

Will adding better clock to the SFP port also raise the cost quite much ?

And if so, could it the be looked at as equal to the clean RJ45 port ?

 

Or is it just as simple that your post above with raised cost and more current applies to the SFP port as well ?

I guess a clean expensive optinal SFP module that can be ordered separately and field mounted is not an option either ?

 

Of cause there may not be any reason for doing this unless your others designs at a later stage will be upgraded with SFP ports. 

 

And we dont yet know if propper designed fiber interfaces would add much SQ. 

 

The switch fabric is already getting the same clock as the clean side, so I'm not sure what this is all about. The SFP cage in this switch was never designed to be a "clean" side port. Functionally it is the same as the other "dirty" side ports.

 

Since the switch fabric is getting a VERY low phase noise clock anyway, taking a signal in and out of two dirty side ports is essentially the same as other companies that are doing "audiophile switches". The SFP port is no different in this regard.

 

Sticking a copper module into the SFP port does NOT make it a clean port. The big special part about the clean port is reduction of clocking noise from upstream network devices, this ONLY applies to the clean port, NOT to any of the dirty ports and since the SFP port is just another dirty port, it does not apply to the SFP port.

 

Again I'm not quite sure what you are asking, adding another SFP port would again raise the price significantly, but not as much as another clean port. If you add a second clean port, then adding a second dirty SFP port would be trivial since you already have a bigger case and beefier PS etc.

 

Obviously I'm not explaining this properly, but I don't know how better to do it.

 

John S.

[JohnSwenson]

4 hours ago, R1200CL said:

@thyname

Yes that’s what John has been saying in another post as well. 

 

If you decide to use any of ultracaps LPS from Uptone, you will have to do some JSGT in addition.

 

Either earth the minus barrel or use a special supplied ground point on the chassis. 

 

Actally I had misunderstood the JSGT myself with those named switches in the other thread. So it turned out in addition to internally do the shunt of the Meanwell supplied withe the LPS-1, I also had to ground the minus of the input voltage of my Cisco SG100D-08. 

 

In addition it seems that my switch may not be the same John tested, so I may not have any blockages at all ?

 

Mine has plastic cover. 

 

 

The Cisco switch I'm using has a V2 in the upper part and a V02 in the lower part.

 

I tested a newer model (I don't remember the version number) and it did not work. I have not tested the V01 model.

 

John S.

[JohnSwenson]

5 hours ago, thyname said:

@JohnSwenson and @Superdad: 

 

Here is what a guy I know from our local FB group said when I posted a link to this thread:

 

------------------------------

 

OK, I work in the computer networking world...I have my hands and finger logged into routers, VPNs, firewalls, and switches every day. And I troubleshoot those devices down to the packet level, yes I know how to read a TCP/IP and/or Ethernet packet. And just my to cents, the idea of speeding $400+ on an eight port copper (CAT5 or CAT6 RJ45 cable) switch which is move a digital packet from device A to device B is a waste of funds. Please understand that I think it is super cool that they are building this super cool switch...both all those ones and zeros will still get to the place they are going. And I understand how to build networks to avoid discards/resend/jitter/too many devices doing... and any $100 8 port switch will do the same job as that super cool $400 ...

 

-----------------------------

 

His words. Not mine. 

 

Thoughts? Could you explain the benefits of an "audiophile" Ethernet Switch in layman's terms?

@diecaster is correct, it is not about getting the ones and zeros across, it is all about other things that come along the cable, specifically leakage current, and phase noise of the clocks that produce the data streams.

 

First off note that I have not fully proven all of the below, I am in the process of doing the fundamental research on the details of what is happening, so the details may change, but I have enough information for the overall concepts.

 

Leakage current, this comes from all AC powered power supplies, it travels through cables to other power supplies and back to the AC line. Ethernet devices wired with copper cables contain transformers, most people think that these block leakage, well yes and no. They DO block the low impedance leakage that most people measure when they talk about "leakage". But it turns out that SMPS power supplies generate a different type of leakage I call high impedance leakage, which goes right through these transformers. I've written in great detail about this in other posts. So what is bad about leakage? It travels through the ground plane of your devices, even all the way down to the DAC, and flows through the ground planes of the devices. This creates a noise on the ground plain with the same spectrum as the leakage, which is right in the middle of the audio range. The ground plane noise increases the phase noise of the clocks sitting on it, in particular the clock in your DAC. This added phase noise causes distortions in the audio signals coming out of the DAC. The cheap SMPS supplied with most consumer LAN equipment generates huge amounts of this leakage current.

 

The other is phase noise of clocks in the network. Again consumer LAN equipment does not usually have clocks with very low phase noise. These clocks are used to clock out the data going over the wire, so they too have the same phase noise as the internal clock. When this data stream goes into a receiver it generates noise on the ground plane which in turns generates extra phase noise on any clock connected to the ground plane. This can cascade down through connected devices into the DAC, again increasing the phase noise of the local clock.

 

Leakage and phase noise can interact, the leakage on the Ethernet cable, can go through the motherboard on a renderer, adding phase to the clock that winds up driving a USB port which winds on on the ground plane of the DAC etc. The exact details of how this works is what I'm trying to work out now.

 

And BTW there is a good probability that I was a designer of chips that are in those switches and routers etc that he mentions, so I have some idea of how they work.

 

John S.

 

 

[JohnSwenson]

24 minutes ago, thatguy said:

Thanks for the reply. I am trying to understand his post and the word clock or clocks was used a number of times in the quoted post. Example: 'phase noise of clocks in the network' and 'These clocks are used to clock out the data going' and 'adding phase to the clock that winds up driving a USB port' and 'increasing the phase noise of the local clock'  

 

 

Almost all digital devices have some form of oscillator, the generic term is a "clock" since its "tick" is used to control the flow of data through the device. It may be a separate oscillator which contains a quartz crystal and the electronics or the electronics may be embedded in another chip and just the quartz crystal external.

 

In a common switch the electronics are embedded in the switch chip and an external 25MHz Quartz crystal is connected to the switch chip. Inside the switch chip there is at least one PLL (Phase Locked Loop)  which takes the 25MHz from the oscillator and synthesizes 125MHz which is the symbol rate in both 100Mbit and gigabit Ethernet.

 

A USB receiver chip will have something similar except it is frequently 24 or 12 MHz.

 

A DAC will frequently have two oscillators, one for each sample rate "family" (44.1, 88.2, 176.8, 352.6) , (48, 96, 192, 384). Common frequencies are 22.5792 MHz  and 24.576 MHz. A few use twice those frequencies.

 

A commonly held belief is that the phase noise (ie jitter) of these clocks in the DAC are what is important, so designers spend money to get low phase noise oscillators for DACs. The flip side of this is the belief that no other clock any where in the data chain is important, so cheap high phase noise oscillators are used for the USB receiver, Ethernet chip, etc.

 

Recent study is starting to show that this is not the case, the phase noise from other clocks can overlay the noise from the local clock. Thus phase noise form other clocks can contaminate the local clock. Even if this "bleed through" is small, the phase noise from devices such as consumer switches and routers is many orders of magnitude higher than that of really low phase noise oscillators used in better DACs. The result is that you are not getting what you paid for in that DAC. In order for the DAC to sound as good as it should, this clock contamination needs to be dealt with.

 

That's what this switch does, it has a very low phase noise oscillator to clock out the symbols onto the cable and special circuitry that prevents phase noise on the incoming packets from contaminating THAT clock.

 

John S.

 

 

[JohnSwenson]

1 hour ago, thatguy said:

Thanks for the reply, to confirm: you are saying that any clock that connects to the circuit board of the DAC chip is creating "phase noise" of the DAC's clock.

All oscillators have their own phase noise. The issue here is that the phase noise of the clocks that generate a data stream can wind up increasing the phase noise of a local clock, this is true for all digital devices. Each pass through a device attenuates this influence, so say you have a router, then two switches before the audio system, the phase noise from the clock in the router will significantly less affect on the DAC clock than the clock in the last switch.

 

This happens all the time, even in normal LAN gear, the noise from the upstream device gets attenuated by say an order of magnitude (exactly how much this is I have to determine), then overlayed on the noise from the next device, but since they are all pretty much the same level, you'll never see the very small increase. But in a DAC the clock has so much less noise than what is in common LAN gear this overlaying becomes significant.

 

As an example using completely arbitrary numbers, lets say you have a switch with a clock of 1000units, and a DAC with a clock of 10 units. If you put one of the switches after the other, the phase noise of the second switch will be 1000+(1000/10) = 1100, you'll never notice the difference. But if you feed the data from the switch into the DAC you get the 10+(1000/10) = 100, the local clock has become MUCH worse because of the contamination from the clock in the switch. Even if the attenuation is 1/100 the the local clock noise still doubles.

 

John S.

[JohnSwenson]

9 hours ago, audiojerry said:

I hope discussion of this topic hasn't stopped yet, because I have a question that may not have been covered. 

 

Is this upcoming switch going to be wireless?

 

Is this a dumb rookie question? I'm asking because my audio system is not on the same floor as my modem and router. My PS Audio DSJ has a network bridge that has ethernet and i2s input. The only way to use ethernet would be from a wireless bridge. Would I need to have my cable company run a second cable to my audio room?  

Our new switch will NOT be wireless.

 

You CAN buy inexpensive wireless bridge boxes that have WiFi and a regular Ethernet jack on the same box. They do not cost very much and should work very well with the new switch. Just plug a regular Ethernet cable between the bridge and the switch.

 

A bunch of companies make these, I personally have used the TrendNet ones and they work well. They frequently come as multiple use boxes so may not be called a "bridge" but that is one of the modes they can be used in. a lot of these are called a WAP (Wireless Access Point) and the bridge mode may be called "client mode" in some of them.

 

John S.

[JohnSwenson]

The switch is designed to be used with the "clean port", the 10/100 one connecting directly to the critical device for audio, (ultraRendu etc) and all other connections, such as servers, NAS, compute engines etc. go into one of the other ports which are full gigabit.

 

Putting some other standard device in between our switch and the critical audio device (ultraRendu etc) just puts back in leakage and clock noise, negating what this switch is all about.

 

John S.

[JohnSwenson]

Yes,

the SFP port is Gigabit. It supports SFP speed negotiation, but what you plug into it has to also support speed negotiation if you want to run slower than Gigabit. Not all modules do. When it powers up it is running Gigabit, if the module supports speed negotiation and wants to run slower it will do so.

 

Note that the SFP port is not the "clean port", it is essentially the same as the other Ethernet ports.

 

John S.

[JohnSwenson]

6 hours ago, cat6man said:

I guess I missed that the other 'dirty' ports are full gigabit.

 

Is there any concern about the amount of activity in the switch?

Is it better to have less traffic on the other (non-10/100 'clean')  ports?

Note we have not actually built this yet, so I can't say for sure, but this is being designed such that what happens with the other ports does not affect the "clean" port.

 

John S.

[JohnSwenson]

3 hours ago, nonesup said:

I currently have an AQVox switch, a Melco N1ZH60-2 (double RJ45 port) and a Lumin A1. Connection is from AQVox to Melco and from the second port RJ 45 to Lumin. I deduce from your answer that using your switch you should connect Melco to the dirty port of Uptone and the Clean Port of Uptone to Lumin, better than Clean Port from Uptone to Melco and second port from Melco to Lumin?

Correct, you want the clean port to be the last thing going into the Ethernet jack of whatever is the last Ethernet point in the audio chain.

 

John S.

[JohnSwenson]

Not yet, that is the next phase, the clock research.

 

I'm building a test setup with a board with 5 possible ways of dealing with clocks, a test signal generator which starts with an ultra low phase noise clock and modulates it to generate a specific phase noise marker that will never exist naturally. This clock will be used to clock various parts of the digital audio chain feeding into a DAC, whose output goes into a 24 bit ADC which is clocked with yet another ultra low phase noise clock. I've been working on a suite of software to analyze the output of the DAC which should show any actual analog differences coming out of the DAC.

 

Everything is all up and running except for the test board, I have the raw boards but have not stuffed it yet, there are a LOT of parts on this board and I don't really want to do it by hand. I've been building a DIY pick and place machine kit and decided to place the board with this. It is all built and wired up and I'm going through the setup and calibration phase right now. I should have it up and running in a couple days and I can then assemble the last board and get this test setup actually working.

 

So barring some unforeseen situation I should be getting results in a couple months. There is a LOT to do to get the test setup working properly.

 

John S.

[JohnSwenson]

 

19 hours ago, austinpop said:

 

Come on... TedAlex. This is not an attack. We're just asking for a reasonable option. There are constraints for customers too.

1. older Cybersahfts have only 50Ω outputs
2. SOtM OCX-10 impedances have to be selected at time of purchase
3. Ref 10 has 6x 75Ω and 2x 50Ω

For any number of reasons, including the cables they own, customers may want to select the impedance of the input on the EtherRegen they buy. The competition provides that option too: http://www.sotm-audio.com/sotmwp/english/shop/snh-10g/

 

I know @JohnSwenson is more aware than laypeople that impedance matching involves BNC connectors as well as the internal and external cable, which is why we would trust you guys to do it right. Is an impedance switch such a bad idea? And it would save you from building and stocking 2 different versions.

 

If Cybershaft can implement an impedance switch, I'm sure John can too.

The problem is that in order to have both a 50 and 75 external input you need both a 50 and 75 BNC jack. There is no way to properly electrically implement both with one jack. Any such approach will degrade the signal, I have a feeling those that want to use multi thousand dollar external clocks will NOT want to have the circuit deliberately degrade the signal! Thus in order to do it we will need to have both jacks. The big problem is there is no room for both jacks on the case we have been planning for, to do it would mean going to a larger enclosure which will cost more money, not just for the extra parts, but for the enclosure and larger board.

 

There is a good possibility that traces on the board will wind up being longer because of the larger board, possibly slightly degrading performance.  Is it worth it?

 

My personal leanings are against it, when we go with a different case we have not used before, there are ALWAYS several iterations of case and board that have to be done in order to get everything just right. This adds a lot of development cost AND a couple months to the delivery time. Do you guys REALLY want that?

 

John S.

[JohnSwenson]

Back on the "grounding" requirements.

 

The high impedance leakage from upstream components which use SMPS still needs to get taken care of. Whether you need a ground idepends on your connections.:

 

Upstream is optical: no leakage can get through the optical so no need for grounding, it doesn't hurt if it is grounded, it just doesn't need it.

 

Upstream is electrical, either copper SFP or one of the 4 ports: If ANY of the devices on the network connected with copper cables uses a SMPS (which includes almost all computers and network devices) then you do need grounding. If you use the SMPS supplied with the switch, it is already grounded, so no extra grounding is required. If you use an LPS-1 or LPS-1.2 then you WILL require extra grounding, we are supplying some form of ground connector to connect a ground. If you use a third party supply you will need to determine if it is not ground and provide a ground if it is not.

 

Whether it is grounded or not the switch will still work, the grounding is needed to get the best performance possible out of it.

 

The ports on the "network side" (the 4 Ethernet jacks and the SFP port) are designed so IF you have grounding, no leakage (high or low impedance) will make it between any ports.

 

John S.

[JohnSwenson]

On the SFP modules, it is gigabit only. Most of the optical modules  are already gigabit only, but the Ethernet interface SFP modules come in two flavors, gigabit only OR 10/100/1000. The 10/100/1000 will not work with this switch. The protocol is different, I chose to use the protocol that works with the optical SFP modules since most people want to use it that way.

 

The devices with SFP cages from some companies will only work with "compatible" SFP modules. There is no electrical difference, the modules come encoded with the name of the company (Cisco etc) and the device checks that name and if it doesn't match it refuses to connect.

 

I refuse to play such games so I don't check for the name at all, so you can use any module that meets the protocol.

 

John S.

 

 

[JohnSwenson]

5 minutes ago, Em2016 said:

 

Hi John

 

If someone were to use a pair of EtherREGEN's as a pair of FMC's... with an LPS-1.2 on the most downstream FMC... 

 

...How does this leakage compare with someone using a single EtherREGEN + Uptone grounded SMPS, in your own personal 'preliminary' measurements so far... I'm not asking for published measurements yet, but I'm sure you've compared in your own personal measurements?

 

i.e. is the isolation as good as optical, in your early findings?

 

I don't quite understand the question.

 

The high impedance leakage comes in from an electrical connection network connection. With optical there is none.

 

At this point we don't know how much high impedance leakage attenuation occurs through the EtherRegen circuits. There will be some, but it will not be absolute, the high impedance leakage is very difficult to block. The low impedance leakage attenuation will be very high (I don't know for sure what it will be, but it will be quite high, I expect over 100dB)

 

Since you are talking about an optical connection, there is none coming over the network connection, so the only leakage is coming in over the power supply. So your question is really about, "what is the leakage difference between the UA SMPS and an LPS-1.2".

 

The ground shunt on the UA SMPS drops the high impedance leakage a lot, about 80 dB but does not touch the low impedance leakage. The LPS1-2 attenuates the low impedance leakage by about 120dB (factor of a million) and significantly less than that for high impedance leakage. The combination of the shunted SMPS and the LPS-1.2 attenuates BOTH high and low impedance leakage by at least 120dB.

 

So for an optical network connection, the total leakage attenuation through the EteherRegen (including network and PS) is going to be very large for both supplies, at least 120dB attenuation for both types of leakage. With the LPS-1.2 there is theoretically higher attenuation, but in practice there will always be enough "sneak paths" in any electronic circuit that there will probably not be that much difference.

 

i hope that answers the question.

 

John S.

 

 

[JohnSwenson]

7 minutes ago, diecaster said:

 

@JohnSwenson Please tell us that stacking EtherREGEN switches won't improve things...please!

I hope not. The purpose of this switch is that it attenuates both leakage and clocking issues by such a large amount that there is no audible influence from anything upstream of the switch, period.

 

Time will tell whether that is actually achieved, but that is what I am working towards, and quite confident that will happen.

 

John S.

[JohnSwenson]

50 minutes ago, Em2016 said:

 

Hi John, this answers my question (I think).

 

So the isolation method of the EtherREGEN (looking at the 'clean' output) won't be quite as good as optical?

 

I guess the next question is did you consider optical isolation inside the EtherRegen itself, to isolate the upstream electrical network connections? As this totally blocks the leakage of the upstream electrical network connection (the same way an expensive pair of EtherRegen's + fibre optic cable connection would). 

The problem is not the signal isolators, they are extremely good, it is the power supplies. We didn't really want to build two LPS-1.2s into every switch, this would make it SO expensive, large, power hungry etc that it wouldn't be a viable product.

 

We are working on a separate isolating supply between the two sides. This has not been finalized at this point, there is a lot of work to go on this. We can get extremely high isolation for low impedance leakage, but the high impedance leakage is tough to block in a power supply. That is why we are using the shunted supply, to shunt the high impedance leakage before it gets to circuits.

 

John S.