[ltp] Wifi range: Cisco Aironet 350 vs IBM a/b/g combo

Ivarsson, Torbjorn (T) linux-thinkpad@linux-thinkpad.org
Wed, 7 Jan 2004 07:18:50 -0600


> -----Original Message-----
> From: honey@gneek.com [mailto:honey@gneek.com]
> Sent: Tuesday, January 06, 2004 7:50 PM
> 
> On Tue, 6 Jan 2004, Ivarsson, Torbjorn (T) wrote:
> 
> > I am not an expert in WiFi-technology, but I am a working 
> professional in the field of RF (Radio Frequencies). Your 
> statement does not surprise me. 
> 
> Thanks: but which one!  I'm left a bit confused by your comments...

All of them I guess... :) Lower range and "disappointing" (!?) rates...

> OK: but remember I'm not using 802.11a at all here: I'm comparing
> coverage of an 802.11b card (the Cisco) with an 802.11b-compatible
> card (the IBM/Atheros) that happens to be doing it via 802.11g - the
> same wavelength, as you say.  The IBM also supports 802.11a, but I've
> never tried it and don't have an 802.11a access point.

Then it may be a function of hardware. Maybe Cisco have better components than Atheros. Maybe your antenna connections aren't optimium and some loss has been introduced. Maybe there's some setting you need to change to increase transmit power. Maybe you need to upgrade the firmware to get some better filter algorithms in the card. I don't know. I may even have left out some trivial things.

I haven't seen any technical studies comparing performance between different cards. It'd be interesting if you tested with different PCMCIA cards. The antennas in the lid should be a better solution than the antennas on a PCMCIA card, but if you get better performance from PCMCIA you may want to return the MiniPCI...

> Yep, understood: I quickly realised that with 802.11b a year or so
> ago, when I found that 5Mb/s was about the practical maximum for
> 802.11b: so that means I expect about 20Mb/s maximum from 802.11g.
> It's the disparity between this figure and the 7Mb/s I actually got,
> and what iwconfig tells me, that confused me.

Have you tried to see what throughputs you get when you're next to the AP vs. very far away? (it should be better close by) It may also be of some interest to se whether or not you get a lot of dropped TCP packets...

> Thanks, didn't know that: useful!  Although, hard to test as TCP/IP
> is a two way protocol?  I may get lost here, but if the ACK's don't
> arrive back in time from the receiver then transmission is throttled
> - at least in TCP?  I'd better stop here before I embarrass myself.

Right. TCP/IP was not meant for wireless connections since TCP/IP desires near error-free transmission medium. The RF environment is not even near error-free, therefore some tricks are done to deliver as low error-rate as possible to the TCP/IP protocol layers (you may hear about stuff like Medium Access Control [MAC], Radio Link Protocol [RLP], etc.). Hence the need for overhead information in the bit-stream.

Testing the performance of data network, especially wireless networks, can be tricky. The performance also varies depending on what kind of data you are transmitting - FTP being prone to much higher rates than HTTP, UDP higher than TCP (because UDP is not reliable, i.e., no ack is needed).

Being a recent newbie to Linux, and being somewhat unfamiliar with traditional data networks, I couldn't tell you what tools to use or how to troubleshoot the networking part. Other people on this list can probably help you with that.

> If anyone has swapped these cards and can confirm or deny this big
> difference I'd be very grateful.

Maybe you can abuse your local computerstore's return policy. Get a PCMCIA card, try it, return it...

T.
-----
Obviously, opinions expressed in this mail are personal and do not reflect the opinions of my employer. Blah, blah, blah...