The true experimental tweak thread

[scan80269]

Thanks John!

 

I have already noticed that is is not properly hi-speed with the help of @YashN and @mansr!

 

You told me about these problems ones before, but I am a quite stubborn guy when I have an "obstacle" in front of me!

 

Even though many people have posted great reviews about the Intona it is really a step back in my point of view. Since I already use unshielded twisted pair data wires, no 5v and GND lift on my USB cables plus a battery powered Regen in-line I am not all that interested to add another USB powered device in the chain...but I am still interested to know if a passive type of data only isolation could improve things. Anyway, it seems to be a dead-end without really deep knowledge in the subject.

I will concentrate my efforts on my coming experiment with a twisted pair SAT cable instead plus see if I can find some other use for my coming optocouplers (maybe on my twisted pair SAT cable? ?)

 

Hi Cornan,

 

Like John said, there is no simple way to passively isolate USB 2.0 (D+/D-) signals, due to the definition of USB 2.0 (covering Low Speed (1.5Mbps), Full Speed (12Mbps) and High Speed (480Mbps) using bi-directional mostly-differential signaling. Products like the Intona and Corning optical USB cable use custom logic on both sides of the isolation barrier, and are way too complex for DIY. These active isolation designs are the only ways to meet USB 2.0 specification compliance, and the complexity is unavoidable.

 

Ironically, USB 3.0 SuperSpeed (5Gbps) and SuperSpeed+ (10Gbps) use unidirectional true differential pairs (SSTX+/SSTX-, SSRX+/SSRX-) for signal transmission, which have similar architecture to other high speed differential links such as Ethernet, PCI Express, SATA, MIPI CSI & DSI, DisplayPort, HDMI, etc., and implementing galvanic isolation is possibly much more straightforward compared to USB 2.0, and may not require much more than pulse transformers built with the proper electrical specs. Unfortunately, digital audio streaming has yet to catch up to USB 3.0, but perhaps one day...

 

By the way, what is a SAT cable?

[scan80269]

@Cornan, I discovered my Netgear GS105v5 switch can be powered by +5V DC. At this voltage, idle current consumption with all ports unconnected is 60mA, and current with two devices attached at gigabit is 175mA, which is under 1W of power! Switch current drops further to 137mA after replacing the CAT6 cable between my Aries Femto and GS105v5 with my "100Mbps only" CAt5e cable, and to 100mA when a second 100Mbps cable is used between my Orbi satellite and GS105v5.

 

It would be interesting to know how the D-Link DGS-105 compares with Netgear GS105v5 in terms of their power consumption.

[scan80269]

Thanks for sharing scan80269! 

 

Power consumption is of great interest to me. That is why I find QoS and EEE so appealing.

I would gladly measure the current consumption on the D-Link but need to know exactly how you perform the test? I only have a battery powered multimeter available.

 

I found my 100Mbps only DIY cable to improve upon FMCs, but somehow they was'nt improving over Cat 5E with my current setup with TP Link RE450 as a wireless adapter. However, when I have got time I will make a new pair of short 100Mbps only cables using a standard Cat 6 cable to rule out if my results is due to the teflon sleeved silver wires that I am using in my current 100Base-TX version.

I've been using an HP E3631A DC power supply to set the voltage and do the current measurements. This supply has a digital voltage display with 10mV resolution and current display with 1mA resolution. Very handy. I simply set 5.00V constant voltage and read the device current consumption off the display. The same supply allows me to determine how low a DC input voltage can go for an Ethernet switch or FMC before it quits working. This is how I discovered the GS105v5 can run with +5V when the included SMPS wall wart delivers +12V. In contrast, the FS105v3 Fast Ethernet switch will absolutely not work with +5V DC input, though I did modify one of these with a linear regulator so it works with +5V.

 

If you have an RJ45 crimp tool you can make your own "100Mbps only" cables, and a really short cable, say 2-3 inches, can act as a male-to-male coupler which apparently no one makes.

[scan80269]

Ok, I will order a simple plug-in energy meter to check the power consumption. They are quite cheap anyway.

 

I do have both Deltaco tool-free RJ45 plugs plus chinease screw terminal RJ45 plugs available, so it will not be a problem (well slightly fiddly) to make short DIY versions. Very strange that there is no RJ45 male-to-male plugs available, but I guess there is just no market for it.

 

I was wondering if I actually could have use of my 4-pin (2+2) opto-couplers on the short RJ45 adapter, using one on the two rx terminals and one on the two tx terminals (ie. 100Base-TX)? I have´nt found any proof if this is would actually work. I guess it might not be possible due to the very same reasons why it do not work for USB. However, I might just try it anyway since it is quite easy to put together.

If I recall correctly, the two twisted pairs used in 100Base-TX are unidirectional (one pair up, one pair down), so if you figure out the direction of data flow for each pair, inserting opto-couplers that support the proper switching speed (100Base-TX uses 125Mega symbols per second) may just work, but you are essentially creating an equivalent of FMC.

 

USB 2.0 is different since the D+/D- pair is half-duplex and therefore bi-directional. This is why a device like the Intona isolator employs custom designed logic housed in two FPGA chips (one on each side of the isolation barrier) to achieve correct functionality.

 

Gigabit (1000Base-T) will also not work with opto-couplers since all 4 twisted pairs transmit and receive at the same time.

[scan80269]

@Cornan, I agree with what John told you that a totally passive Fast Ethernet isolator using opto-couplers cannot work! What John suggested involving a pair of RMII PHY chips along with isolators is of course an active design, requiring a power feed-in. Such an active design is at risk of undermining the isolation benefits that you sought in the first place, unless it is powered by a battery or an isolating LPS like the UpTone LPS-1.

 

Until we gain a better understanding of the frequency range of any leakage currents that can jump Ethernet pulse transformers (e.g. behind each port of a switch or host controller) and get carried down Ethernet cables, any design to provide additional isolation for Ethernet is not guaranteed to be effective.

 

For example, if you look inside an EMO Systems EN-70HD or Baaske MI 1005 Ethernet isolator you will find small pulse transformers custom designed to support the 125 Megasymbols/second symbol rate of Fast Ethernet and gigabit. A well-designed pulse transformer can be a good band-pass filter (attenuate frequencies significantly above and below the target operating frequency). Also, Ethernet (like USB, HDMI, etc.) using differential signaling is supposed to offer a high degree of common mode rejection, so any noise coupling equally to the two conductors of a differential pair should get cancelled at the PHY receiver and not continue onward to the rest of the circuitry. If there is really some kind of high-frequency noise that Ethernet can carry that is detrimental to SQ for digital audio, it needs to be carefully analyzed before an effective solution can be implemented. For the time being, something like the EN-70HD is about as good as it gets, though I haven't heard the Etalon isolator.

 

I'm glad you had good results with your DIY "100Mbps only" Ethernet cables (with connectivity in only pins 1,2,3 & 6). My friend made me two more of these cables using standard CAT5e cable stock, and I've been putting them to good use.

[scan80269]

I actually though of another fun experiment with two switches. "Balanced" (well, maybe not) signal. Ta-Dam!

There is 8 pins used for 1000Base-TX (T568B) speed. What about separating those into 4 pins+4 pins at the input/output of the switches? The tricky part would be to find out how to make those cables working properly. Intitially I would assume to use a Y-cable design with pin 1+2 & 3+6 on one leg and pin 4+5 & 7+8 on the other leg emerging to a full 8-pin plug at the RE450 & Aries Mini end.

In theory this would mean that each switch will "Regenerate" separate signals and possibly do that job more efficiant. In practise, who knows?

Anyway, I will start with the daisy-chained switches (FMC-style). In the mean time maybe someone could talk me out of the other experiment by proving it impossible in the first place?

@Cornan, I don't think your idea will work. The 4 twisted pairs of a 1000BASE-T port must work together as a group. They are handled by the same chip inside a switch. If you separate the 1/2 and 3/6 pairs with the 4/5 and 7/8 pairs to different switches, the link will no longer be gigabit, and the two switches have no way of working in tandem to complete the 4-pair gigabit functionality.

[scan80269]

Thanks scan!  How about using a Ethernet splitter at both ends? They'll use all 4 pairs, but can you see any advantages/disadvantages connecting them that way?

 

What is an Ethernet splitter?

 

If you are thinking of increasing the physical separation of the 4 twisted pairs, you should be aware that properly built Ethernet cables, especially CAT6 and above, already have the ability to minimize inter-pair crosstalk (critical for speeds above gigabit), so spacing the pairs farther apart from each other will not be of any advantage. You also need to avoid inadvertently changing the end-to-end cable lengths of the 4 pairs relative to each other, as this will cause the signals to not arrive at the receiving end at the same time, which will mess up the PHY: you will be doing more harm than good.

[scan80269]

Completely my way of thinking. It is all learning by error, have fun experimenting and enjoy the thrills when those out-of-the-boxes ideas actually work as intended! 

Thinking out of the box and thought experiments are always great and should be encouraged. Keep it up!

 

It's just that some of us with a bit more in depth knowledge of how modern high-speed interfaces (e.g. Ethernet, USB, HDMI ...) work may be in a better position to predict the outcome of a particular experiment...

 

I'm so glad I found CA, as I have learned so much good stuff from John Swenson and a few others.

[scan80269]

Here is an example of a Ethernet splitter.

 

[ATTACH]33628[/ATTACH]

 

 

Using a pair of those with same length (30cm) Cat 6A U/UTP will be a bad idea?

 

Here is how it would look like:

 

Wireless adapter>30 cm Cat6a>Ethernet splitter>2x 50cm Cat 6a to 2xD-Link switches>Ethernet splitter>30 cm Cat 6a>Aries Mini

Such a splitter is designed to support one RJ45 port carrying two 100Mbps (100BASE-TX) Ethernet connections, with each connection using two out of the four twisted pairs.

 

While the connectivity you listed (from wireless adapter to Aries Mini) should work in theory, I would be leery of the impedance bump introduced by each splitter. The wiring inside the splitter is unlikely to be impedance controlled, so the Ethernet signals will encounter two additional impedance discontinuities (bumps) with the splitters in place. I don't know how severe the impedance bumps are but they will impact signal integrity and possibly make the PHYs work harder to retrieve the data stream.

 

Why don't you go ahead and try it to hear what it sounds like? My gut feel: SQ may worsen vs. no splitters, but the proof is in the listening!

[scan80269]

Thanks! I will see if there are any GB splitters around. If there are I might give it a go since I will have two identical switches anyway. 

As far as I know there is no such thing as a gigabit splitter. 1000BASE-T uses all 4 twisted pairs, and Ethernet signals are always connected point-to-point, with exactly two ends. At gigabit, the 4 pairs are transmitting and receiving at the same time over the same physical wires.

 

The closest thing to a "gigabit splitter" is a gigabit Ethernet switch, like the ones you have.

 

Looking again at your connection topology, there may a problem. I failed to see the "2xD-Link switches". If you use the Ethernet splitter to run two cables into two separate switches, I don't think that will work. The splitters will force link speed down to 100Mbps (a bit like my "100Mbps only" Cat5e cable) so you will not be running gigabit end-to-end. Also, as explained previously, a gigabit network connection cannot function being split across two switches with each getting two of the four twisted pairs.

[scan80269]

@Cornan, since you have two identical gigabit switches, one easy thing you can do is to cascade them:

 

wireless adapter <> Switch 1 <> Switch 2 <> Aries Mini

 

This topology preserves end-to-end gigabit link speed. You may want to experiment with battery (or LPS-1) powering both switches to compare the SQ with just one switch.

[scan80269]

Looking at this Ethernet splitter gives a good idea that it might actually work, but only with 100Mbit speed between the switches.

 

[ATTACH]33633[/ATTACH]

 

It is actually wired the same way as I wanted to do with my DIY Y-ethernet cables.

Now I have a question for you. What controls the speed? If RE450 output controls the speed it will be GB to the first Ethernet splitter. If the output of the second switch also controls the speed would'nt it end up with GB speed at the Aries Mini end...or is the output speed controlled by the input speed?

 

I can see that there might be problems with speed sync with side-by-side switches, possibly affecting SQ negatively. However, I cannot see why it would'nt work at all since both switches receive identical signals?

 

Ethernet link speed is negotiated between the two link partners, i.e. the devices at the two ends of an Ethernet cable. Gigabit Ethernet switches are backward compatible with 100Mbps and 10Mbps on a per-port basis. A switch usually features LEDs that indicate the negotiated link speed for each active port. A gigabit switch has internal per-port buffering to support different ports running at different speeds, i.e. some ports at gigabit and others at 100Mbps or even 10Mbps.

 

For an audio setup involving several chains of Ethernet links, the end-to-end throughput will be determined by the link running at the lowest speed. For example:

 

music source > Gigabit switch > Aries

 

Throughput here is gigabit (assuming cables are Cat5e or higher) since all devices have gigabit capable RJ45 ports.

 

music source > Gigabit switch > Fast Ethernet switch > Aries

 

Throughput here is 100Mbps. Note that the music source and the gigabit switch are still talking at gigabit, but the Fast Ethernet switch effectively lowers the end-to-end throughput to Aries down to 100Mbps. The link between the gigabit switch and Fast Ethernet switch is 100Mbps.

 

You simply can't wire Ethernet switches side-by-side, as that creates a mode of operation completely outside of Ethernet specifications (10BASE-T, 100BASE-TX, 1000BASE-T, etc.). You are of course free to experiment, but I can predict the outcome with pretty high certainty. You can chain (cascade) switches as this is supported within the spec, though each switch added to the chain will increase latency.

[scan80269]

Hi Cornan,

 

I just stumbled into a tweak that's not necessarily going to advance SQ, but may be a step forward in facilitating USB DAC shootouts in a reasonably fair way.

 

I discovered my Aries Femto streamer can recognize more than one USB DAC, and is easily configured to play to any one of them.  I inserted a bus-powered USB3.0 hub (Unitek) between Aries and my usual Vega DAC, and plugged another USB DAC (e.g. Holo Audio Spring or Schiit Yggdrasil) into another downstream port of the USB hub.  The Lightning DS app for Aries then offered either DAC to be selected for playback.  Of course there is no way to play to more than one USB DAC at a time, since the USB link between streamer and DAC is always 1 to 1, but inserting the USB hub enables the DAC selection to be made in Lightning DS with a few taps on the iPad, and without the need to touch any cable connection already made to the DACs.

 

I suspect an UpTone REGEN would be a better device to use than this generic USB bus-powered hub, from the SQ perspective, but since the REGEN has only one downstream USB port it doesn't allow more than one DAC to be directly connected to it.  I suppose I can use an audiophile grade USB hub, or hack one up DIY style.  Several of my audio friends have acquired new DACs recently, so there are plenty of DAC shootout opportunities going forward.

 

My Auralic Vega DAC does not need +5V USB bus power at all, but many of my other DACs do, so perhaps I should also look into clean Vbus injection downstream from this USB hub.  I may also need to order more LPS-1 power supplies from UpTone...

 

I have yet to see what happens when two USB DACs are connected via a USB hub to a Windows 10 PC, though I suspect this should work in a similar fashion, and may even allow the DAC driver to be unified to Microsoft's USB Audio 2.0 unified driver (usbaudio2.sys) soon to debut with the next major Windows 10 release.  I can post a follow-up once this is checked out.

 

What do you think?