HomePlug AV2 Head-To-Head

[krellboy]

My examination suggests that the Zyxel/Allnet is an un-optimized copy of the reference design, while the Linksys is reworked. The Zyxel/Allnet is slightly "crummier" in choice of components as well.

I'm looking at the Zyxel now. It's 99.8% identical to the Allnet and is clearly made by the same Taiwanese company.

Looking at the design more closely, it appears to be missing an optoisolator between the "line" section and the "analog" section that the Linksys has, in addition to a bunch of components in that area. If I was going to make a guess as to how to save money from the reference design it might be to take out the zero-crossing detector. If that is what they did exchanging the firmware between the Linksys and the Zyxel would be a really bad idea, the configuration of the zero crossing detector is one of the things set in the firmware.

 

[ekerazha]

TRENDnet's HomePlug AV2 MIMO adapters ship in June 2014... so late...

 

[thoma]

Someone reads your review. They say "huh? uplink and downlink? It's a piece of CABLE, how could those numbers be different, I don't understand" So you send him a link to your "test methodology".

Ah, but HomePlug is not a piece of cable. It is a complex protocol that runs over an extremely noisy medium with complex channel characteristics.

To quote from HomePlug AV and IEEE 1901: A Handbook for PLC Designers and Users:

Because much of the noise experienced by each node may be highly localized due to attenuation, the noise profile seen by each PLC device may be significantly different. Therefore, powerline channels are not typically symmetric.

Suppose HomePlug A is on a dedicated circuit straight to the breaker, and HomePlug B is on a windy circuit with some appliances plugged in. At B, the uplink will be faster than the downlink. At A, the downlink will be faster than the uplink.

Even without any appliances plugged in on the two circuits, the link still won't be perfectly symmetrical. Too many variables. Suppose that circuit B runs parallel to circuit C for the first 50 feet, and then circuit C branches off to serve the kitchen. The parallel run will transfer refrigerator noise from circuit C to circuit B through capacitive coupling.

You might be able to get symmetrical rates for HomePlug in the lab. But not in the real world, which is much more chaotic.

 

[thoma]

I'm looking at the Zyxel now. ... If I was going to make a guess as to how to save money from the reference design it might be to take out the zero-crossing detector.

This would be a strange thing to remove, as Zyxel advertises QoS functionality for the PLA5205. See: http://www.zyxel.com/us/en/products_services/pla5205.shtml

Zero-crossing detection is used for QoS in a HomePlug network. The central coordinator sends out a beacon that is synchronized to the zero-crossings of the AC waveform. This is important because the zero-crossing is usually the quietest part of the AC cycle. HomePlug units with QoS traffic are assigned TDMA time slots, which guarantees them a certain amount of bandwidth every two AC cycles.

HomePlug AV2 does have provision to ignore the zero-crossings, and send out the beacon at a fixed 25/30 Hz rate. The point of making it configurable in the firmware is to allow HomePlug to run over coax, where there is no AC line frequency. Coax is a high-quality shielded medium with little noise, so it really doesn't matter when the beacon is sent out.

This is probably not intended to save the OEM a few pennies by removing the zero-crossing detector when running HomePlug over powerlines! That rather defeats the design of the QoS system. If the beacon is missed because it was sent at the wrong time, then other HomePlug units could miss it, and then stomp over another transmission.

 

[krellboy]

This would be a strange thing to remove, as Zyxel advertises QoS functionality for the PLA5205. See: http://www.zyxel.com/us/en/products_services/pla5205.shtml

Zero-crossing detection is used for QoS in a HomePlug network. The central coordinator sends out a beacon that is synchronized to the zero-crossings of the AC waveform. This is important because the zero-crossing is usually the quietest part of the AC cycle. HomePlug units with QoS traffic are assigned TDMA time slots, which guarantees them a certain amount of bandwidth every two AC cycles.

Yes, it would be strange, but that was my guess from looking at the boards.

Maybe I can take some photos and upload them.

 

[stevech]

It may be (is likely?) that attenuation (by appliances, plug strip filters, etc) of these high frequencies is more of a problem than is noise.

 

[thoma]

It may be (is likely?) that attenuation (by appliances, plug strip filters, etc) of these high frequencies is more of a problem than is noise.

Yes, it would reduce speeds, but it would not result in asymmetric speeds. Pure attenuation (without any noise) should have the same effect on signal strength in both directions.

Noise is different, because it matters where that noise is introduced onto the circuit. Noise introduced near the sender has less impact on transmission rate than noise introduced near the recipient.

 

[lazna]

Signal measurement

You can measure signal level and possible attenuation with HomePlug monitor. See http://forum.ethernetovercoax.eu/index.php?topic=126.0