Low tranfere rate from WAN site to PC's.

This is probably close to the max you can get - there's an extra overhead in wireless data compared to fixed-line, plus the encryption will slow stuff down a little.

Mark McIntyre

Not so much.

I have compared with to others running wireless on a 2 mbit. Also encrypted. 218-220 KB/sek. One of them even with the exact same router/AP.

Palle Jensen hath wroth:

Well, as long as you don't attach any cartoons, we're safe. Just where did all those protesters find Danish flags to burn?

As Mark said, that's fairly close to normal. Let's play with the numbers.

2Mbits/sec on an ADSL connection will yield about 1780Kbits/sec or about 222KBytes/sec. That's what the online performance tests should show. The 10-15% overhead is the "ATM cell tax" penalty for using ridiculously small packets (53 byte packets with 48 bytes of data) to move potentially huge IP packets (64KBytes). This sorta explains how it works: | you have PPPoE or PPPoA, you'll be at the high end of the "ATM cell tax".

Numbers, not generalizations please. What bandwidth are you getting with FTP from your ISP? Some ftp clients (i.e. Windoze 98SE and ME) report wrong numbers for transfer statistics. Some Windoze FTP applications (i.e. CuteFTP) seems to report ridiculously high numbers because they seem to do a running average. If the initial value is very high, later values will be affected. You can also see that in the IE6 file transfers, that start out with spectacularly high transfer rates and eventually slow down. If the files are too small, they never reach a stable average and tend to be toward the high side.

Also, if you're measuring thruput, be sure to test both UDP (streaming video) and TCP (file transfer). UDP does not require an acknowledgement and therefore goes faster than TCP.

So much for the ADSL part of the puzzle. When you get to the wireless end, you'll find that it's MUCH faster than your 2MBit/sec. No way should the wireless be a limiting factor unless something is broken. I've had the stupid ethernet cable drive me nuts. It was miswired but worked well enough to establish a connection. However, the error rate was very large and the thruput was terrible. Replacing the cable instantly fixed the problem. Don't remind me that I was the one that built the cable.

Your maximum wireless performance should follow this table: |

the wide variations in wireless thruput depending on how the wireless router is setup. For example, a 54Mbit/sec wireless association can yield 24.4Mbits/sec in thruput. However, if you enable the 802.11b compatibility mode, the maximum thruput drops to about 14.4Mbits/sec. However, at most speeds, your wireless will be running MUCH faster than your ADSL connection, so it's not an issue.

Intel Proset had some problems with the speed going down to 1Mbit/sec in the presence of interference, and never bothering to go back up when the interference subsided. Upgrade to the latest version.

Via wireless or wired? 2.4Mbits/sec is still rather slow for either a wired or wireless connection.

Model number of the Linksys PCI card. Hardware version number of the PCI card. Software version that you claim is the latest. This is probably un-necessary for solving this problem, but if you ask questions later, it's a good thing to describe the hardware exactly.

Same as above. Need hardware model, hardware version, and driver version. Model number of the laptop and exact operating system is also helpful.

Linksys PCI adapter has a decent antenna. The 3com PCMCIA card does not. The difference is in the antenna.

Netstumbler will only find access points that broadcast their SSID. It cannot find client radios or non-802.11 devices. Instead of looking for interference, just try changing the channel (1, 6, or 11). If the performance improves, then you have interference on the old channel. If you really want to find sources of interference, you should either use Kismet (under Linux) or a spectrum analyzer.

Amazing. My war driving through Santa Cruz shows that about 75% of the networks heard are on channel 6.

How far is the "range"? In general, a wireless network will not cause problems unless it's very close, very strong, or someone is using a power amplifier.

I don't think you can. The only numbers I can find wrong is the relatively slow 2.4Mbits/sec transfer rate. However, I have no idea how you are testing this, so even that may be correct.

Incidentally, your English is quite good.

actually, this is incorrect. V 0.4.0 will find SSIDless APs too. At least, it does for me.

Mark McIntyre

I'm using v0.3.30 on my Windoze ME laptop. It will find access points that respond with a blank SSID, but not those that don't respond at all. Netstumbler is an active scanner and requires that an access point responds to the probe requests. If it doesn't, then Netstumbler can't see it. Mostly, I use Kismet these daze.

Eww.. Don't mention the war.... Don't know about the flags? I know that some muslims demonstrants mistakenly picked the swiss flag instead :-)

OK then. I'll read carefully...

I can't deny that you are perfectly right in this. But mayby I/we should stick to the "inside" of the network. PC AP PC. If both computers are connected to the AP wit an exellent connection at

54 Mbps, then I would believe that I should be able to yield something better than ~ 300-400 KB/Sek (~3 Mbps). (As you write later)

I undestand. But it shouldn't really mean a lot since I compare with speeds meassured with the cable plugged in the network adapter instead og the wireless.

I have used FileZilla. I have used Firefox. I have used BySoft network monitor

All with same results.

My ISP has a FTP test site constructed for speed test purposes. With cable plugged in I have never ever seen anything else than 220 KB/Sek. Rock steady and always. But again: I believe all you tell me about overhead, encryption overhead, Packet loss etc..

Let's take a look at the inside of the network. Why do I get such lousy performance? I was hoping for something like 10 Mbps. Instead I get one third of ths instead..

Noted.

It really isn't.

Notice the wide variations in wireless thruput depending on how the

Thanks for the link! Interessting reading. The problem in this particular case is that I am almost limited by the the wireless performance! I am sitting 30 ft from the acces point, but are still offered 54 Mbps with exellent connection. Still I don't get much more that approximately 3 Mbps. Annoying.

This is wireless.

//Here I am asked for more accurate hardware info, so I try...//

############################################ Acces point: WRT54GC Firmware: 1.03.0. Antenna: Linksys High Gain Antenna HGA7S (7 dBi)

Ref:

Network adapter: WMP54G Used on Stationary PC Driver version: 3.0.3.0

Don't use vendors software. Only installed drivers and letting XP control the wireless.

Ref:

Network adapter: 3COM Office Connect Used on Laptop PC.

REF:

Driver version: 3.3.0.156

Vendors software installed. ############################################

Laptop PC (Amitech FreeNote 4602). XP Professional.

Staionary PC: Homebuild. XP Home.

I live in the middle of Copenhagen, compacted in large old buildings. I live on 4'th floor. The nearest network I know of is on 1'st floor. A guess would be that the network that I see is placed from 50-100 ft away. But I can't really know. If I browse after networs with my Dell laptop (with build in wireless), then I see approximately 10 AP. With my stationary PC I rarely see more the 2.

:-/

Thanks again :-)

on a /WHAT/ ? Good heavens!!

So either Netgear, SMC and 3com routers don't actually turn off SSID broadcast (I have one of each and with it disabled, I can see their MAC in Netstumbler), or else 0.4.0 has some extra ability.

Mark McIntyre

I have so many laptops (all with dead batteries) that I have a specific function for each one. In this case, my wireless sniffer and tester laptop is a Micron Something Pentium One/300 with a fabulous

128MBytes of RAM. I tried to run W2K on it and the lack of RAM made it incredibly slow. I'm slowly transfering my wireless junk to a Panasonic Toughbook CF-M34 (PIII/600 with 256MBytes) which is currently running W2K. Netstumbler 0.4.0 only runs on W2K and XP. I tried 0.4.0 on ME and found that it overscribbled some important file that I had to manually restore. Netstumbler 0.3.30 works just fine on Win 98SE and ME, but only with a very limited set of wireless devices.

I really don't know. I guess I could sniff, capture some packets, and feed them to Ethereal to see what they're doing. As I vaguely recall, most of the routers made up to about a year ago simply responsed with blank SSID's. Only recently have there been routers that simply don't respond.

I think it's easy enough to tell for sure. Fire up Netstumbler and watch the flashing wireless light on your access point. If it's like mine, it only flashes when the access point transmits. Or, you can use some kind of RF signal detector like a diode and a volts-guesser. Pound on the access point with Netstumbler and see if the light flashes. If it does, then it's probably transmitting something in response to the probe request. If it just sits there, it's not.

Palle Jensen hath wroth:

Yes, it should be faster. First try it with one computer connected to the AP via ethernet, the other wireless. If you have: - a fast computer. - 100baseT ethernet connection. - a 54Mbits/sec wireless connection. then you should get about half the wireless rate or 25Mbits/sec thruput.

However, if your ethernet connection is 10baseT instead of 100, then it will be the limiting factor and reduce your speed to either about

8Mbits/sec for full duplex, or 5Mbits/sec full duplex. Again this is with only one computer via wireless. Since you have two wireless computers, try each computer, one at a time, over wireless to identify if either computer is a problem.

Then, try it with both computers via wireless. Under ideal conditions, you should get about 12Mbits/sec with a 54Mbit/sec wireless association. The speed is cut in half for each wireless link.

I prefer to use Iperf 1.7:

For internet testing, I like the Web100 NDT tools such as: Denmark, I think the closest is in Switzerland: sure to check the result details. Also note that a version of the NDT server works with IPerf.

You're back to the ISP again. Do NOT do your local benchmarks via the ISP. Do it between your computers. What you're doing is benchmarking the ADSL line, and not the speed of the wireless. The wireless should be MUCH faster than the ADSL and your benchmarks will be limited by the speed of the ADSL.

I don't know. The easiest way to tell is to only test one section at a time. The last time I had a similar performance problem, it was a miswired ethernet cable *AND* a defective 10/100baseT switch. It's very difficult to find problems when there are two or more things broken.

Hmmmm.... At 30 ft, I would NOT expect you to get a 54Mbits/sec connection. Please double check your connection speed. Most of my testing shows that with conventional antennas (rubber ducky on the access point and MiniPCI internal laptop card), I can go to about 15ft before the connection speed drops to below 54Mbit/sec due to reflections and increased noise. You probably have a better antenna.

Suggestion: Temporarily configure your wireless access point for a fixed speed instead of "automatic". That will end the issue of the connection speed. If you're going to do your testing at 30ft (not recommended), then I suggest a speed such as 24Mbits/sec OFDM which faster than 4 times your ADSL speed and should be useful for the two wireless computer test.

Thank you for supplying the links. That makes it much easier for me.

I have no experience with that model router. Careful. That points to the US web site that reports the latest firmware is 1.02.8. The Denmark web site at: |

1.03.0. You have the latest firmware.

Ok, That works. I've used a few of these, but with the supplied drivers. Nothing unusual here.

I've noticed that products with Super-G tend to cause problem in systems that do not support Super-G. It's usually at the access point, where I have to turn OFF Super-G in order to get things working. However, in this case, this is the only client radio in your system that supports Super-G. If there's a setting in the driver that allows you to turn OFF Super-G, try it.

Are these reasonably fast machines (PIII/600 or better) that can support speeds of 25Mbits/sec? Do both computers have a 100baseT-FDX (full duplex) ethernet port? If the machines are slow, for one reason or other such as insufficient memory, slow CPU, cheap junk ethernet card, etc..., then benchmarking may be affected.

That's a good indication of the relative sensitivity of each computer. The more systems you see, the more sensitive the radio. It might also be the position and location of the radios. Do any of the 10 or 2 access points show a very strong (4-5 bars) signal strength indication? Are they on the same channel as what you're using?

Anyway, try to seperate the testing so that you're only testing one wireless link at a time. If there is a problem, it should be obvious.