Hacker News Reader

Comment by hamtastic | Hacker News Reader

The possibilities here with sufficiently sophisticated software and tunable hardware (deep learning of RF propagation characteristics in a building using data from portable sensors) are fascinating! But honestly, MOCA already solves this problem. The MOCA 2.5 standard (https://en.wikipedia.org/wiki/Multimedia_over_Coax_Alliance#...) already delivers 2.5gbit total (time-multiplex) throughput which is more than enough for real-world no-compromise bidirectional gigabit ethernet on isolated runs or distributed backhaul to multiple wifi APs. I'm using this in my (115 y.o.) house now where new cable runs are impractical and the coax plant is unused.

View on HN

Happy to see MoCA get a shout- out. I spent most of 2003-2015 working on the design of MoCA chips, starting before the v1.0 spec release up through v2.5. AMA

How is MoCA 2.5 different from G.hn Wave 2?

The main difference is that MoCA is designed to run over coaxial cable, and G.hn is meant for use over power lines.

They have similar use cases ("route data over wires that already exist in your homes"), but different problems (MoCA: avoid interfering with your cable/satellite/cable modem feed, G.hn: deal with the absolutely abysmal signal quality on power lines, avoid accidentally broadcasting in the FM bands,...)

From a modulation perspective, G.hn wave 2 and MoCA 2.0 are broadly similar. They both use OFDM as the main modulation and LDPC as the FEC. They both are scheduled networks (like 802.11, unlike wired ethernet) where a master node allocates transmission time slots to other nodes. The devil is in the details though and I'm not as familiar with the G.hn wave2 spec as MoCA, so the detail I can give you is limited.

You alluded to it but the other difference is MoCA is shielded, EoP turns your house wires into a ton of radiating antennas across the frequency bands. I still have no idea how the FCC even approved the damn things after seeing what they do to the RF spectrum nearby.

Correct. Coax is nice and quiet in both directions. In home power wiring is basically a giant, poorly tuned antenna.

I can tell you MOCA 1.0 can be made to work over powerline (at reduced bandwidth, if you rip out the RF frontend and run at baseband), and it can pass emissions, but it didn't work well enough (by % of households able to achieve >= target data rate) to be worth commercializing.

Why is MoCA still so rare and expensive? Is there a technical reason or is it just some combination of business reasons?

It seems like it's been the obvious choice for no-new-wires home networking for years but you basically can't find it at retail and have to look for it even online.

I think there were a couple of contributing factors:

- WiFi is still easier to install (where you can get a good signal), and built into phones/laptops/etc. so mostly all you have to do to get wireless networking is buy the router. For MoCA, you have to buy & install a box for every endpoint. Also, MoCA is limited to places you have coax, which is usually bedrooms & living rooms. If you need connectivity in the garage/attic/closet/kitchen, MoCA might require new wires anyway.

- Lack of interest from MoCA developers. The main "customer" for MoCA chips was cable set-top-box vendors (Scientific Atlanta, etc.) for "multi-room PVR" products. That's where almost all MoCA networking chips made ended up. The cable/satellite vendors needed the deterministic performance of MoCA (vs WiFi) and they were always planning on putting a box in every room with a TV - which generally already aligns with where the cable taps are. They could also pre-configure the MoCA so it wouldn't interfere with some other stuff they might want to put on the cable (Example: DirecTV put its downlink from the dish squarely in the middle of the MoCA frequency band, so you had to configure MoCA to a different channel in DirecTV houses than e.g. Verizon). The market for bare Ethernet-Coax Bridges (ECB) was always tiny in comparison.

- The chicken and egg problem. The consumer market for bare Ethernet-Coax gateways was smaller (see above) as it has to compete with both WiFi and "just run some new CAT-5" (as well as niche things like HPNA/G.hn) so it didn't get a lot of focus or advertising. In turn, this means most people have no idea that MoCA even exists, so they don't go looking for it. D.Link, Netgear, Linksys, etc. then decided that lack of demand meant it's not worth developing/advertising new/improved versions of the products, etc.

MoCA was a very targeted solution for adding IP connectivity to things which were already wired together on a Coax network, and it did a great job at that. We sold 100s of millions of chips. MoCA was never meant to be all things to all people though, and while I personally use MoCA in my home I never got my own parents to use it - they just have the one computer hooked directly to their router, and WiFi for their iPad.

For what it’s worth, you can buy the DirecTV DECA adapters for like $20 on Amazon. They’re actually just MoCA adapters, but they run at a lower frequency than regular MoCA so they don’t interfere with the DirecTV signal. They work fine by themselves.

The only downside is that they’re 100 Mbit/s.

That said, they really do give you a rock solid 100 Mbit/s.

If you need more bandwidth, I think Verizon sells a MoCA 2.5 adapter for like $60 which should give you GigE.

They're actually ~91Mbit/s, but yes, very, very stable. They also seem to add about 4ms of latency.

The 4ms of latency is a product of the MOCA protocol.

As a fully scheduled network, each packet must wait for a timeslot to send a reservation packet, wait for the schedule to be updated (map packet), and then wait for the actual scheduled time.

The reservation timeslots and mail packets are on a fixed schedule (approximately- there are cases where the timing changes if you have a poor link) with a consequence of an unloaded MOCA network having a transit time of ~2ms going from the master node, or ~2.5ms going anywhere else (master node transmit is faster since it gets to skip the reservation step. Times are averages, as the exact time depends on the alignment of the time of arrival to the scheduling period). Round- trip ping times should go up by roughly 4.5ms.

Under ideal conditions, it is possible to get 100Mbps UDP throughput on a pair of MoCA nodes (1518 byte packets). The physical media can support up to 110Mbps (MOCA 1.0), 140Mbps (MOCA 1.1) or 450Mbps (MOCA 2.0) user throughput per channel (up to 5 channels in MOCA 2.5), but that's shared bandwidth (all traffic summed together). Throughout will fall off in bad channels (minimum 40Mbps), or if you use smaller packets (higher scheduling requirements per packet) so YMMV

Yes - that's been my experience with MoCA too. I needed to add an AP in one room and putting a new ethernet cable would have been a huge hassle. Luckily there was an existing TV extension into the room and a pair of MoCA adaptors works perfectly over the run. It's significantly better than Powerline options.

The other option to be aware of is G.hn on 2-wire telephone extension cable (that wasn't relevant in my case) - the adaptors for this seem to be more expensive and harder to find though than MoCA.

What hardware are you using? I looked briefly into this since I also have coax run throughout the house unused, but it was hard for me to know what to look for.

I'm running a MoCA 1.4 network at home with a mixed set of ActionTec ECB2500C and Netgear MCA1001 Ethernet-Coax bridges. The interoperated w/ no problems.

You can get newer versions that support MoCA 2.5, but I haven't tried them.

> The other option to be aware of is G.hn on 2-wire telephone extension cable (that wasn't relevant in my case) - the adaptors for this seem to be more expensive and harder to find though than MoCA.

If you've got (unneeded) telephone wire in your walls near where you want Ethernet, there's a good chance you have at least two pairs, which you can use for 100BaseTX, either point to point (ugh) if cable was run in bus formation as used to be common, or more usefully in more modern wiring where phone lines ran to a central location (hopefully somewhere that's appropriate to terminate ethernet near, but often at the telco DMARC on the outside of the home)

Yes. You can often run Gigabit ethernet successfully over short runs of 4 pairs too, even if it's not Cat 5e.

I was thinking about this for a minute, I don't think this technology is valuable because MOCA already exists and can be used instead, I suspect it fits a sort of different use case. It allows antennas to be separately installed from the radios, and allowing existing coax cabling to be used for the distribution.

So my read is instead of having to install power, MOCA, and various endpoints around, I can centrally install my radios/APs, and distribute the antennas if I have existing cabling, such as cabling for security cameras.

FWIW the only commercially-available MoCA 2.5 product I've found that you can actually buy as an end user is the one at gocoax.com (not affiliated.)

I've been very happy with their product.

Reply on news.ycombinator.com