Netgear XAV5001 500Mb/s powerline Ethernet adaptor The quest for faster home networks continues, and with 802.11n Wi-Fi failing to deliver promised 300-450Mb/s peak speeds to a lot of people - blame sneaky manufacturers slipping cheap, single-antenna, 72Mb/s adaptors into their devices - some punters, even those without wireless blackspots in their homes, are looking to …
So I have been using power networking for about 3 years, I started to use it to get broadband to a second floor with out the cabling mess.
It seems that power networking is not getting the development that I think it should, as it is a great way to get networking to parts not covered by WiFi or having to lay miles of cable. But it is good to see that this are moving :)
This post has been deleted by its author
No, not subjective at all. It is objective and measurable. A lot of power line kit violates the standards it is required to respect, and Ofcon are doing Sweet FA about it.
Microwave ovens operate at a specific frequency, and the unlicensed ISM band that they could interfere with was created precisely because the interference from ovens could render it useless. It therefore isn't saleable spectrum.
If people pay for the use of certain parts of the RF spectrum, they are entitled to expect other users not to interfere with that use. What's so subjective about that?
"why would you want to wire your house with cat 5 and not either cat 6 or fibre optic ?"
Cat 6 is a very good (technical) option but pricey, probably worth it for signal integrity and future-proofing for higher bandwidth - but fibre optic ?... i'm really not sure that's an option for most folk. The high cost of materials and end-points/adaptors is one (quite large IMO) deterrent but the clincher for me would be making sure all the splices and joins were good. I think maybe it might be a bit of a stretch beyond splicing and crimping UTP - which pretty much what any reasonably competent DIYer could handle - and horrendously expensive to get people in to install it.
Not that the idea of FO in the house isn't incredibly appealing mind you....
PS. I increasingly see people use 'Cat 5' as a short-hand for 'Cat 5e' so i'm inclined to give the OP the benefit of the doubt on that. That said, these days if you going to wire up a house then, I agree, Cat 6 would be the way to go (not that there is really such a thing as 'certified' or 'approved' Cat 6 cabling last I looked, but that's a generally minor issue)
CAT5e cabling is pretty damn good at radiating when used with GigE.
I find that unless you use STP (ideally with a centre earth wire + overall screen) and terminate the screens then 20m of cabling makes a lovely antenna. An earth and one end but not the other makes it an even BETTER antenna - could still pick it up 50+ metres away :D
It was enough to bugger up our DVB-T receivers on some multiplex; admittedly we're still waiting for analogue to be switched off but it radiates a good bit more than you'd imagine when running a GigE network fairly hard.
One cable run (unterminated) knocked 3Mbps off our ADSL (17Mbps down/1.7Mbps up normally) and that's how I noticed. Took a SW radio to find the problem and I still find it strange but terminate both ends and back to normal.
tl;dr DIY GigE cabling can be almost as bad if not done right and I haven't even mentioned earth loops :P
Even unshielded twisted pair (UTP) shouldn't be that noisy.
Unless your telly reception is marginal anyway, there shouldn't be much noise from even GigE over Cat5e UTP - for a start, all the energy is down in HF and VHF bands.
I would suggest that it's wired wrongly (mixing up pairs), except that I wouldn't expect GigE to work with mismatched pairs.
If you are using STP or FTP, earthing the shield at both ends is best, but only if you have a decent bonding network (earth grid) there already. Having a shield and not connecting either end (either leaving it hanging or connecting to something which isn't properly earthed) will be slightly worse than using UTP.
NB if you're getting earth loops, it's a sign that you don't have a good bonding network.
GigE radiates harmonics way up the spectrum - that's the nature of the signal it carries (dear gods I'm starting to remember transmission line theory - noooo! ).
UTP is generally pretty crap in long runs - short cabling (patch cables etc) is fine but long runs are trouble IME. Picks up all sorts of EMI rubbish along the way too and probably does a decent job of re-radiating it to the other end of the run. Makes no odds to the GigE right enough.
Our TV signal is normally OK but its now appalling since an adjacent transmitter switched to digital and Waltham had to retune. Oh well never mind, only another 3 months (and three more retunes) to go - says volumes about the way the UK is now that its taken 13 YEARS to get to the stage of switching over to transmitting "DVD-quality" video (heh and half the multiplexs aren't even that!).
Oh and I don't have any earth loops because I know what they are - the average DIYer won't ;)
You have to keep in mind that a lot of your total cost is going to be driven by labor and possibly repairs to walls that have to be destroyed rather than the cost of the cable itself. I opted for a bundled cable myself when I had my house wired during construction. The cable bundle was mainly an accounting gimmick so that I could get multiple cables dropped for one price rather than 2x or 5x.
At the time when I was shopping for cable, the bundle with the fiber was nearly the same as the one without. So now I have a fiber wired house. Not really sure yet what to do with it though.
Cat6 might have been an improvement but I'm not sold on the idea.
Another thing about the "structured cabling" is that it kind of serves as a bit of mass protection (like a motorcycle group on the highway) that resists bad handling. It simply won't bend easily in ways that are contrary to how fiber and network cat5 should be handled.
Handy too... I still get shivers any time I recall what they do to the poor cat5 they put in for phones. Ugly. Scary. Ugly. Ugly.
When our net connections are finally up to the gigabit sort of speed it won't make much of a difference to your browsing, though your internal network says it's that fast. No, if you are trying to get gigabit speeds it will look like goodbye to Freeview as well. Some kit interferes up to 300MHz and are they aiming to knock out telly as well? That'll make them really popular?
Plastic fibre is coming, Molex Precise Networks showed us some prototype kits a couple of years back, you can lose it down the edge of the carpet, and it'll be cheap.
my experience with powerline networking has been great. I'm pretty sure mine are 85mbps ones, and they are unbranded from ebuyer, but i have great pings to UK servers for most games, and if im playing on a central EU server, i get around 60-90 ping which is pretty good compaired to my Abit WIfi-AP wireless card which i used to get around 120-150 ping.
I didnt even have to configure anything with them and they ended up being plug and play.
"I don't get any interference using a FM radio".. what the hell? You DO know that FM uses a phase-locked-loop on a carrier frequency right? FM is designed to disguise radio spectrum noise. Try an AM radio.. far more indicative of potential noise issues. Try checking with your local radio HAM before making such a misguided statement.
...it buggers up DAB nicely if its on the same mains ring and sometimes even if its on a different phase (ie someone elses house), never mind different mains ring. Oh and yes I do understand the difference between conducted & radiated EMI ta, the point is that these abominations are guilty of breaking the rules on BOTH in every single installation, bar none.
Since the days of FM are apparently limited then best not get too attached to the idea of listening to any broadcast radio other than that found on LW :)
Funny how every major user of spectrum from the CAA to GCHQ* to the BBC all state that it IS a problem.
The exception? Yes its Ofcom who strangely enough don't actually employ much in the way of RF engineers, unlike all the other organisations who disagree with them.
*yes they do, they just weren't allowed to say so in public
Wouldn't have thought so. But it set me thinking about the socket arrangements
These things are obviousy useful if you are not in a position (or don't want to) to fit cable runs - eg rented or older properties. Or to use a laptop with a faster wired connection.
But not every house has loads of wall sockets, and extension leads are as aesthetically untidy as running cabling (in the case of older properties that aren't red brick, sockets are generally fewer)
But these adaptors take up a lotof space around the socket. So why don't they do a passthrough for the mains - a socket on the adapter itself to plug your laptop PSU into?
Apparently there are speed limits with regards to the protocol used as well.
For example, on my 1000Mbps LAN I get transfer rates of around 350Mbps using ftp, but only about 180Mbps using SMB.
Can anyone explain why ? I can't believe protocol overheads are to blame because I'm measuring the data speed on the router (so no chance of ignoring header overhead if measuring on the same system as generates the traffic).
SMB is a bad protocol, there is a lot of overhead and pauses. There are things that you can do about it though, depending on whether you are using samba or windows - google is your friend.
I'm a bit disappointed that the test in the review was 'copy one file over the lan once'. This tests a lot more than just the networking, it tests the read speed of the NAS, the write speed on the laptop, etc. A far better test would have been netperf.
Sir Runcible Spoon,
Try linking your ethernet cable directly between two machines and see how fast it goes (you will problably have to set up ip manually though as no DHCP server and you will have to connect to each other by typing the ip address. This takes out all equipment just to see what your best speed is.
Then also change on both cards your MTU to 9000 (providing your cards support jumbo frames). Doing this on between 2 macs usually get's around 80 to 90 MB/s. Do note sometimes when altering MTU connection goes a bit flakey though usually unplugging the cable and replugging it gets them to sync up a bit quicker - note setup has to be done on both machines.
If the switch also supports jumbo frames then afterwards can be left like that. If they are transmitting to each other over IP then those speeds can be obtainable.
350Mbps though is around what I'd expect a gig network to do without changing to jumbo frames etc... it has to do with what buffer the card has and how many packets per second it is able to process. When MTU is at 1500 which is the normal default for 100Mbps a packet has 1500 bytes in total. Of this 53 bytes is overhead so actual packet is 1447 bytes.
If the card/pc combo can process say 40000 packets per second then that becomes 55 Megabytes/second. This is raw ethernet though and the packets will additionally be wasted by the protocol - ftp usually being the leanest hence the fastest. Packets per second on Gb max is 1.4 million per second though it's not just the card but the OS, cpu speed and bus speed the card is plugged into...
if MTU is upped to 9000 though then the packet should be around 8947 bytes x 40000 in theory would get 341 Megabytes/Second but the ceiling is 100 Megabytes/second.
Of note most home routers cannot come anywhere near such packet switching speeds (usually they max out at around 4000 pps as the processor clock is the limiting factor though this would only effect routing to the internet or between ethernet and wireless).
ps I'm using rough figures the numbers are ballpark not spec accurate... It's mostly figures I see by observing ping flood responses when I put in a network to see if there are any bottlenecks anywhere. I'm not a routing engineer...
I've got the 200Mb version of these, with a built-in filtered plug.
I've noticed that plugging in other devices nearby (i.e. not into the filtered plug on the powerline adapter) can cause serious degradation of acheivable speeds. Particularly phone chargers, laptops, routers etc - i.e. the things that you are mostly likely to plug in nearby.
My speeds improved by around 30-40% just by making sure that everything was plugged in through the filter. Extension cables can have a similar effect if you don't have filters. e.g. if you have a double wall socket, plug everything into a 4 or 6 socket extension cable that is itself plugged into one of the double sockets on the wall, and plug the powerline into the other half of the socket.
I use DAB radio and have seen no signal problems whilst using these adapters.
This post has been deleted by its author
"a mean speed of 9.96MB/s or 79.67Mb/s"
9.96MB/s is actually 99.6Mb/s - in serial data there are (usually) 10 bits in a byte, not 8. Has been this way ever since the RS232 standard stumbled out of the primordial ooze, complete with a start bit and a stop bit framing each character sent. These days we use something called 4B5B, which is a bit more sophisticated, but the "8 becomes 10" ratio is the same. So many people grumble about not getting full line speed due to this simple error.
Ironicly, the only place where the "8" actually gives the correct reslt is on 100 meg Ether, because that actually runs at 125MHz, so the two wrong calculations cancel each other out.
Highly improbable in general (though there are some real examples).
Even back in the days of synchronous serial protocols there were generally only 8 bits in a byte.
Definitely not applicable to modern Ethernet-like situations.
Try again.
There's nothing magic about 8 bits. Most serial protocols added 1 start and 1 or 2 stop bits to every 8-bits of data, so it's perfectly correct to do async serial calculations on the basis of 10 transmitted bits per byte of data.
As for "back in the days" of serial synchronous protocols, you might like to take a look at ADSL...
I'm quite familiar with G.DMT and the various other layers between what's on the phone wire and what the user's Ethernet port sees, thank you (not up to speed on VDSL though). At each of those layers, a byte of user data consists of 8 bits of user data, though obviously the symbol representing that byte of data may have been expanded into more than 8 bits by the time it hits the wire. So what. You want to add the relevant share of the framing overhead into what you're calling a byte too? ATM headers? The header bits in an IP packet? Don't think so. For all practical purposes, a byte is 8 bits.
An excellent exercise in not answering the points raised, I think.
My ADSL comment was intended to note that synchronous serial protocols are still in widespread current use, certainly not "back in the days".
As for the 8-bits comment, the original commenters point was that carrying 8-bits (1 byte) of user data over an async serial conenction generally requires 10 bits of data on the wire. Which is a more concise way of saying what you have finally accepted "obviously the symbol representing that byte of data may have been expanded into more than 8 bits by the time it hits the wire". So, where's the problem?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
