Wireless performance on the Axim - Page 2

Howard2k · 06-20-04, 08:12 PM

Nope. The bandwidth will still be the same. I'm less convined that this is an accurate test of throughput. But I haven't looked into it in enough detail.

For the record though the bandwidth will be the same but it may be more affected by environmental conditions. High signal to noise ratio, interference, range, presense of other devices on the same network fighting for bandwidth etc.

jackwesleyroper · 06-21-04, 07:53 AM

so the throughput does depend on the quality of the network transmission, rather than the speed?

this would explain the difference between some axims gaining throughput of over 3mb/s. I'm still confused though, why such a big difference? Environmental conditions determine bandwidth, aswell as throughput, coupled with the processing power and bus speeds, does this give the throughput?

Howard2k · 06-21-04, 08:22 AM

With wireless it's a bit more complex.

Firstly, wifi negotiates speed based on signal strength. So the negotiated speed will be between 1 - 11Mb/s. This speed represents the data rate in the air between the device and the AP. I believe this is purely signal strength - not signal quality (but I could be wrong, I had a look through the 802.11b spec and didn't see a specific description so maybe it's vendor specific). This native speed has a clear effect on throughput. For a fair number of us we probably would connect at 11Mb/s anyway. I think the signal to noise ratio has more of an impact than the actual signal strength in a fair number of home environments.

Secondly, there is the whole issue of two or more devices speaking at once. Very simplistically, if I have 11Mb/s of bandwidth and two devices (say PPC and AP) then you have half the bandwidth available. For a more detailed description run a query on google with define:csma/ca and define:csma/cd. Wifi uses csma/ca but csma/cd is much more widely documented and there is some overlap.

Thirdly, the Axim itself has many limitations. The processor on the CF card, the bus of the CF slot, the ability for that data to run through the processor etc.

The PXA270 processors (X30) are the ones seeing the higher speeds (more efficient processor AND a higher clock rate) of up to 3Mb/s (I haven't seen this personally) while for those of us with the PXA26x or PXA25x we're typically seeing no more than 2Mb/s.

Finally - the actual protocol has an impact too. TCP is a much more "robust" protocol than UDP, but this has an effect on it's throughput. TCP and ACK would be good things to look at if you want to know more. When we talk about TCP/IP we're referring to a suite of protocols. TCP is one and UDP is another. IP is also one. Perhaps it's unfair that they called it TCP/IP not TCP/UDP/IP. :)

Howard2k · 06-21-04, 08:25 AM

Oh and I didn't mention environment conditoins. They can cause packet corruption (for example) which means that you send the data and then send part of it again. Our NET throughput in this case is reduced.

This is one of the reasons why 802.11a networks will often benchmark higher than 802.11g networks. Both are 54Mb/s but the 5GHz band is a lot cleaner. (today anywway).