I'm hoping some people knowledgable in Wi-Fi can give me a hand as this is driving me nuts with the inconsistency. Basically I have a project where I'm trying to compare 802.11n to 802.11g by looking at its advancements in turn. What I've done is got 2 linksys e3000 routers with dd-wrt and am using one as an AP at the other as a client bridge connected to it. On each side of the wireless bridge are devices using jperf to measure the throughput.
Using 802.11g jperf on UDP mode showed a bandwidth of around 25Mb/s, +- 3Mb/s. Which sounds reasonable for 802.11g with WPA2 which i've read can slow things down.
Now when setting it to 802.11n, on a 5 GHz channel, with the MCS index set to auto and a 20 MHz channel width I get around 38 to 46 MB/s, with a 40 MHz channel width increase that figure by only a one or two MB/s on average. Surely that is not right as on paper doubling the channel width should double the throughput, yet the increase here is negligable. And if I force the MCS index to a specific value to force 1 or 2 spatial streams the routers won't connect to eachother.
One suggestion has been to move the routers greatly apart as their close proximity in this test setup (25cm to 100cm) could actully be causing major issues, and that a greater distance may actully improve everything to figures closer to my estimations. This is what I will give a try tommorow.
Another issue that proberly doesn't help things, especialy in TCP tests is that the connection seems to be one way, my PC can communicate through the client router, to the AP router, to my home LAN/internet, but devices on my LAN can't ping my PC (I think its because the router is perhaps not sending the data back over the wireless bridge), which means I can't use iperf in TCP dual mode. and TCP mode even in 1 way gives realy inconsisten results (such as <1 Mb/s for 802.11n).
Any help would be very welcome as this just seems so inconsisten and I'm stuck with the hardware I have and need to try and make the most of it for this project.
Using 802.11g jperf on UDP mode showed a bandwidth of around 25Mb/s, +- 3Mb/s. Which sounds reasonable for 802.11g with WPA2 which i've read can slow things down.
Now when setting it to 802.11n, on a 5 GHz channel, with the MCS index set to auto and a 20 MHz channel width I get around 38 to 46 MB/s, with a 40 MHz channel width increase that figure by only a one or two MB/s on average. Surely that is not right as on paper doubling the channel width should double the throughput, yet the increase here is negligable. And if I force the MCS index to a specific value to force 1 or 2 spatial streams the routers won't connect to eachother.
One suggestion has been to move the routers greatly apart as their close proximity in this test setup (25cm to 100cm) could actully be causing major issues, and that a greater distance may actully improve everything to figures closer to my estimations. This is what I will give a try tommorow.
Another issue that proberly doesn't help things, especialy in TCP tests is that the connection seems to be one way, my PC can communicate through the client router, to the AP router, to my home LAN/internet, but devices on my LAN can't ping my PC (I think its because the router is perhaps not sending the data back over the wireless bridge), which means I can't use iperf in TCP dual mode. and TCP mode even in 1 way gives realy inconsisten results (such as <1 Mb/s for 802.11n).
Any help would be very welcome as this just seems so inconsisten and I'm stuck with the hardware I have and need to try and make the most of it for this project.