Further UW2453 testing: My tests were downloading a 36mb file over the internet from a fast host (HTTP). When repeating the tests, I unloaded the module, replugged the device, and loaded the module again inbetween each test. I made sure not to change the location or orientation of the device, etc.
I measured the average transfer rate of the download using wget, and the number of failed transmissions by counting the retry_failed interrupts. It may sound odd to count failed transmissions in a download test but keep reading anyway... First comparison: zd1211rw vs vendor driver. In all cases the PLL locked onto table 5, so table 6 was used for radio programming. zd1211rw: 1. 120 fails, 773kb/sec 2. 313 fails, 593kb/sec 3. 659 fails, 602kb/sec zydas 2.16.0.0: 1. 1051 fails, 816kb/sec 2. 1799 fails, 643kb/sec 3. 2186 fails, 381kb/sec There may be some heat-related issues here: all the numbers steadily increase with time. Further testing showed that they decrease after a while, then increase again, etc. I repeated the test from cold with the vendor driver first, and found the opposite results (the vendor driver from cold had a few hundred fails, then zd1211rw from warm had thousands of fails). Second comparison: calibrating the radio on channel 1 (the hard-coded default) vs calibrating the radio on channel 11 (the channel where my network is running). All tests done with zd1211rw. calibrated on channel 1 (PLL locked on table 5): 1. 120 fails, 773kb/sec 2. 313 fails, 593kb/sec 3. 659 fails, 602kb/sec calibrated on channel 11 (PLL locked on table 4): 1. 2545 fails, 486kb/sec 2. 2572 fails, 487kb/sec 3. 2334 fails, 513kb/sec The 2nd set of results were consistent even from cold -- calibrating the radio on channel 11 produced ~2500 failures for each test. This is a shame because one obvious way of attempting to fix calibration problems would be to recalibrate the radio on the target channel, but this test indicates it doesn't improve anything... Third comparison: calibrating on channel 1 vs using autocal autocal is used when the PLL doesn't lock on any RF configuration tables. I though I'd try it as the other RFs work like this - a "one size fits all" radio programming scheme. (I don't actually know that autocal is designed to work like this, I'm just speculating based on its name) Results were not pretty: couldn't test anything as when programmed with autocal values, I get no scan results so can't even associate. Same results with the vendor driver. Conclusion: zd1211rw is no better/worse than vendor driver with the UW2453 RF (they both perform badly), and my 2 initial guesses at finding ways to improve reception/transmission success rate have failed miserably. Daniel ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Zd1211-devs mailing list - http://zd1211.ath.cx/ Unsubscribe: https://lists.sourceforge.net/lists/listinfo/zd1211-devs
