Just wanted to say that I'm having that very issue as well - but thought at first it would be due to my (probably) bad wlan connection itself.

But after reading JaXXoN's message, I conducted several tests and have to confirm that's not signal quality related. There has to be another issue producing the problem.

Hi!

Just to keep you in the loop, I tried it out with a USB WLAN dongle (attached to the carambola) and the output is just as smooth as with console or LAN. This is a D-Link DWL G-122 which uses rt73mod.ko (also a Ralink part). So it really just boils down to the rt2800pci.ko module!

I have attached the kernel .config with rt73 support for the carambola board.

regards

Bernhard

Can you repeat test with decreased power down to 0dBm ?

BTW: it's quite less choppy in the other direction, means, when logging into the Linux PC (from the carambola board) via ssh/wlan and then executing the choppydemo on the PC, making the output visible on the serial console of the carambola board. But there is still some notable choppiness.

regards

Bernhard

I also have theses lags on a wlan connection between two carambolas. Ping looks like this:

64 bytes from 192.168.42.2: seq=30 ttl=64 time=3.327 ms
64 bytes from 192.168.42.2: seq=31 ttl=64 time=1.577 ms
64 bytes from 192.168.42.2: seq=32 ttl=64 time=1.427 ms
64 bytes from 192.168.42.2: seq=33 ttl=64 time=106.175 ms
64 bytes from 192.168.42.2: seq=35 ttl=64 time=115.179 ms
64 bytes from 192.168.42.2: seq=36 ttl=64 time=2.573 ms
64 bytes from 192.168.42.2: seq=37 ttl=64 time=1.995 ms
64 bytes from 192.168.42.2: seq=38 ttl=64 time=103.806 ms


Tx-power is 10dBm, signal streng is 42/70 -68dBm

Any hints what can be done to improve this?

I've made some ad-hoc tests and got better results with a shield (delivered by 8devices) soldered on the carambola board.

Carambola setting: txpower on 10, Carambola on the DevBoard
PC-Setting: Ubuntu 12.04 running on VMWare (running on Win7 host)
Both connected via WLAN.

Executing (as root) on PC: ping -c 10000 -f <ip-address of carambola>
several times in sequence manually.

Summary Results:
Without shield:
40%: ~70ms (68-75)ms
20%: >220 (228)ms
40%: ~1000ms (993-1000)ms
best time (max):68.030ms
worst time (max): 1000.753ms

With shield:
70%: 14-31ms (this low time was never reached without shield)
10%: ~80ms (without shield this was 40% best time)
20%: >210ms (with shield the worst time, without shield 20%)
best time (max): 14.139ms
worst time (max): 248.230ms

It looks like the shield improves the max-time significantly.
Conclusion: in my setting, a shield improves ping time significantly.