WireWorld Starlight CAT 8 cable review

[scan80269]

Another consideration, especially for computer based setups, is the ability to use a NIC of your choosing. There are inexpensive, Intel and HP / Dell (rebadged Intel) server NIC's out there for $30.

 

If the issue is the implementation of the PHY and it's susceptibility then just going with a dedicated NIC should suffice.

 

I'm still not buying that a short 1 or 2 meter run out of an allowable 100 meters is going to affect the power regulation of the PHY. Easily testable with a 100 meter vs a 1 or 2 meter cable.

 

Cable length can have an effect on PHY power consumption. For example, my Netgear GS108T switch has a "Short Cable Mode" configurable for each port which places that port into low power mode for cables less than 10m in length. If a short cable can help reduce PHY power consumption it should lower overall switch or NIC power consumption which can lead to a sonic improvement. Whether the PHY power reduction is realized is dependent on switch or NIC implementation and configuration.

 

For audio streaming a recent favorite is my modded "100Mbps only" (pins 1,2,3,6 conducting, pins 4,5,7,8 insulated) Ethernet cable. When a gigabit PHY is forced to work at 100Mbps there is significant power reduction. Such a cable appears to deliver comparable or better sound than using a Fast Ethernet switch fed by a linear power supply or battery.

[scan80269]

Are you sure about this? The inverse would be that since GBe has 10 times the bandwidth that it would actually be working 10 times less.

 

Here is the datasheet for Intel's 82579 GBe PHY

www.mouser.com/pdfdocs/82579datasheetvol21.pdf

 

Table 7 on page 27 of this 82579 PHY datasheet shows device power consumption at 100Mbps being about half that of 1000Mbps (gigabit), for both active and idle scenarios.

 

PHY generated noise at gigabit should be short duration but higher intensity, vs. noise at 100Mbps being much longer duration but lower in magnitude. Note that PHY power consumption is still significant even at idle, and the PHY doesn't power down unless the connection to the link partner is removed. Gigabit idle power is higher than 100Mbps idle power (by more than 2X) so the latter may have an advantage from SQ perspective.

[scan80269]

In what scenario, with cables that meet spec, does the power draw numbers listed, get modified and at what length?

 

And to take the supposition to the most logical of conclusions based on the working theory.

 

An ultimate test would be 24/192 PCM (this data rate should work) on a link that is set to 10Mbps with a short length say 3-6 feet of cable and and a 1000Mbit connection with say a 100' cable.

 

We could use the boutique cable for the short run and some Belden for the long run. What would the outcome have to be for you to admit there is going to be no audible difference?

 

I've haven't seen any claim that 10Mbps is audibly superior to 1000Mbps anywhere on any forum.

 

I don't think I can answer your question, as that Intel datasheet lists "Power Consumption Targets", which don't necessarily equate to max/worst-case, and there is no data showing power consumption variance by Ethernet cable length. Shorter cable should translate to lower PHY power, but I have no idea by how much.

 

To argue about Ethernet cables sounding different seems ridiculous, as Ethernet is not involved in digital audio timing so an Ethernet cable cannot possibly have a sound signature in and of itself. BUT, I believe the Ethernet cable type and length can affect how the Ethernet PHY on the receiving end operates (e.g. in streamer/renderer) and the extent the Ethernet PHY noise impacts the rest of the circuitry in a specific piece of equipment is VERY design dependent. I suspect there are large differences among various contemporary designs. So the Ethernet cable (among other components within the Ethernet subsystem) indirectly induces a performance change in the non-Ethernet and audio timing sensitive section of the box, e.g. D-to-A converter, leading to a SQ change. Better designs can strongly immunize against incoming Ethernet signal integrity issues, and with such a box, the expectation is little or no SQ change when Ethernet cables are changed (make, construction, length, etc.). A poorer box design will be vulnerable to a harder-working PHY generating more noise that more severely contaminates the D-to-A conversion circuitry. Such a box should more readily reveal SQ changes with Ethernet cable changes, and is in a position to be "helped" to some extent by cable tweaks.

 

Another possibility is the rest of the audio system being incapable of revealing any SQ difference with different cables. I have at least one such setup: my bedroom system based on a pair of KEF X300A speakers driven by USB. It's a very decent sounding system for a bedroom, but no good for discerning sound changes coming from USB or Ethernet cable changes. If other woes in the system are swamping the effects of the cables then obviously changing the Ethernet cables won't yield any audible difference.

 

The audible differences among Ethernet cables are not the result of BER differences, but rather how hard the Ethernet PHY of the receiving device has to work to deliver error-free data packets downstream to the rest of the circuitry. That signal integrity in transport interfaces like Ethernet & USB can have an effect on SQ is a relatively new understanding. In the ideal world signal integrity of Ethernet or USB should have no influence on SQ but to achieve this is easier said than done.

 

I still have your "unplug Ethernet cable while streaming" exercise queued up at my friend's venue (my own bedroom system is not resolving enough as a test vehicle, and my headphone setup is not ideal for assessing sound stage and instrument/vocal sizing changes). With my friend's Aries Femto capable of buffering an entire Redbook or even a high-res track, the Ethernet cable hot plug/unplug test should be a valid control test. I'll report my findings once it gets done.

 

I also have a project to attempt to correlate audio device power consumption with SQ. My theory on this needs to be backed up with measurements of controlled experiments. PHY power is a new knob to play with.

 

In a different post, I have reported listening test results (at my friend's place) favoring 100Mbps over 1000Mbps within a pseudo control experiment, using the afore-mentioned "100Mbps only" Ethernet cable. Results were with a specific set of gear, so I'm not generalizing this preference to be universal. In other words, YMMV.

[scan80269]

The argument is that since the PHY is driven by a range of voltages dependent on the quality or I should say various metrics of the cable.

 

The howl at the moon crowd is arguing that out of 328 possible feet of cable that one expensive 4/6/12 foot run of Ethernet is going to some how be less work for the PHY to decode data off of vs another 4/6/12 run of other, spec exceeding cable.

 

External EMI and RFI are out of the equation since UTP CAT6 is proven to be noise immune to 30Mhz and STP, if done correctly can give you 1000% increase in rejection and is good up to 100Mhz noise rejection.

Ethernet (1000Base-T and 100Base-TX) uses point-to-point connections. We are not talking about concatenating a short cable with a 100m cable with a passive RJ-45 coupler. The PHY in the streamer sees a signal waveform that is a function of the PHY transmitter at the other end of the cable, and the cable itself. If you connect a router to a switch with a 100m cable, then connect the streamer to the same switch with a short cable, the streamer will NOT be affected by the 100m cable, since it's not talking on the same switch port that 100m cable is plugged into.

 

I believe the SQ difference between different makes of Ethernet cable of the SAME short length (e.g. 1 foot) would be very hard, if at all possible to hear. This is because for short length, the variance of electrical characteristics among the brands should be very small to non-existent. For longer lengths, however, the difference in characteristics, especially for things like per-foot capacitance and differential impedance consistency, can be significant. The difference between and 1 foot and 100m cables of the same brand & type should be easier to hear, though as mentioned before, the audibility depends on the streamer sensitivity (or lack thereof) to Ethernet PHY noise, and the resolution of the rest of the system to reveal any difference, so there is definitely a YMMV factor.

 

This theory is essentially a parallel to the one where better USB signal integrity helps raise SQ for USB-input DACs, as reported by many users of the UpTone REGEN. To experiment on whether a different Ethernet cable can make a sonic difference, only the Ethernet cable plugged into the streamer should be changed. All other Ethernet cables involved in transporting digital audio from source to streamer (but not plugged into streamer itself) are on different Ethernet links and should have no impact on streamer SQ. As for the Ethernet switch, it should be beneficial to power it with a linear power supply, to minimize the effect of the switch being a noise source.

 

I don't have a 100m Ethernet cable, but my friend and I have listened to a 50' Cat5E and a 3' BJC Cat6A cable plugged into an Aries Femto, and the difference was quite audible. Ditto for replacing the Netgear wall wart with a small Teradak LPS for the FS105v3 Fast Ethernet switch linked to Aries. I will, however, not invest in expensive "audio-grade" Ethernet cables, as I don't believe they will make anywhere near enough difference (or at all) to justify the expense.