Rolf van Kleef

Rolf van Kleef

Software engineer

© 2020

Dark Mode

Shortfall of Bandwidth with Powerline Ethernet Adaptor
2020-12-25

In episode 34 of the Self-Hosted podcast, one of the hosts (Chris) talks about his experiences from using a TP-Link AV1000 powerline adaptor in an RV, wherein he notes that whilst his adaptors are rated for gigabit speeds, he is only able to achieve around 300mbps. There are a multitude of factors which might impact this performance, and this article will explore some of them in an attempt to explain why this shortcoming exists, and how it might be mitigated.

HomePlug AV2

The adaptor used and linked does indeed appear to advertise supporting gigabit speeds, but in actuality, this only refers to the fact that the ethernet socket on the adaptors support 1000Base-T as specified in IEEE 802.3ab. The actual carrier protocol used is HomePlug AV2 (linking to Wikipedia as I cannot access their sources, which would be extremely useful to take away the doubt hereafter stated), which appears to be a single-duplex 1300mbps protocol. Due to overhead, the actual speed one would see is approximately 1000mbps (in single-duplex) for normal TCP/IP loads, or 500mbps in full-duplex.

Extrapolation from HomePlug AV1

It appears that HomePlug AV2 regularly underperforms in real-world uses, as is made evident by a presentation published on the homeplug website. If we extrapolate these numbers to AV2, we end up at an expected total bandwidth of 500mbps in single-duplex, without calculating any overhead. This is likely not a correct extrapolation as a different transfer encoding is used in AV2, but this shows that there is at least a chance that AV2 might underperform quite significantly.

Interference from Solar Installations

HomePlug AV2 can choose from a wide spectrum of frequencies, anywhere between 2MHz and 86 MHz (with the bands between 30MHz and 86MHz added by AV2 since AV1). According to Pieter-Tjerk de Boer in his technical notes on wide-spectrum radio observations during and near a solar eclipse , a significant reduction in noise is observed in the radio spectrum between 0MHz and 29Mhz. According to another article posted on the website of VERON on disruptions in the frequency spectrum due to solarpanels, significant interference is observed between 100KHz and 450MHz. This article also writes about how a significant amount of interference can directly be attributed to the use of a non-compliant power-optimizer. We can clearly see that both sources observe interference in frequencies that overlap with frequencies used by HomePlug AV2, but we cannot use this as an indication that this would cause interference for HomePlug devices, as the solar installation we are looking at has significant differences to the installations observed in these articles. Additionally, Chris stated in a Telegram chat room that he (quite obviously) connected his solar panels directly to a charge controller that charges a battery, which somewhat isolates his power network to which he has connected his powerline adaptors, and the solar panel and it’s support circuitry. Additionally, there is assumed to be an inverter connected in between the battery and the power circuitry to which the powerline adaptors are connected.

Unconventional Circumstances

The aforementioned inverter might possibly be a source of interference for the powerline adaptors, as many inverters don’t produce perfect sine-waves, but either square waves or near-sine-waves. Both of these types of waves contain significant amounts of high-frequency components that might cause disruptions in the powerline adaptors. The reason we don’t expect to see this on normal grid-connected systems is because the grid is able to introduce a significant amount of stability in such a system.

Conclusion

As the research found, a large part of the bandwidth drop can potentially be attributed to the advertised speed only being single-duplex. We might be able to find this out more definitively by benchmarking single-duplex, as well as full-duplex loads on the installation. Additionally, the chosen protocol might also underperform quite significantly, as can be indicated by an extrapolation from measurements on an old version of the protocol. The solar installation might also negatively impact the powerline ethernet installation. We can confirm or rule this out by disconnecting the solar installation from the battery, and covering the solar panels. Moreover, the inverter may be a cause of problems. We can get an indication on whether this is the case by testing the adaptors in a normal grid-connected installation.

Additional notes

In the podcast, Chris discusses his wish to be able to use multiple powerline ethernet adaptors. According to what I can read on Wikipedia, IEEE 1901 (which HomePlug AV2 is a part of) requires certain interoperability. Whilst I cannot guarantee it will work, and actual experience would only be able to provide a definitive answer, this appears to be a strong indicator that HomePlug AV2 adaptors are indeed able to work in multiple pairs. Whether two different sets of plugs can interoperate is not talked about, and might be a vendor-specific feature.

Sources