cc the list

The short answer is you cannot. You might modify b43 to increase that number,
but your device might not last very long. The reason is that every Broadcom chip
contains an SPROM that contains the maximum power to be sent to the radio, among
other things. Assuming that Broadcom sacrificed a number of devices to learn a
safe value, increasing the power might damage your chip. You might have one that
can stand more power than most, but it might stand less. Are you willing to risk
your hardware?

In b43, the MAC layer sends a power of 27 dBm to b43, but the driver reduces the
value to maximum based on that chip's programming. The reduced number is what
iwconfig reports. I have no idea what the Windows driver is doing, but I doubt
that it is sending more than the SPROM max to the chip, no matter what it is
reporting.

Why does ndiswrapper get better results? First of all, ndiswrapper only converts
the Windows system calls to the Linux equivalent. It is the Windows driver that
does the work. Remember that Broadcom software personnel wrote that Windows
driver with full access to the device specifications. The code in b43 was
written from "specifications" deduced by disassembling Broadcom code. Any
mistakes in the disassembly, or errors in the original Broadcom code that have
since been corrected might hurt the performance of b43 on that device relative
to the Windows driver.

The performance of the Windows driver relative to b43 varies a lot depending on
the particular chip you have. The BCM4306 chips are the worst - the BCM4311's
are the best. What chip do you have?

There are some differences between 2.6.27 and 2.6.28 in the behavior of the MAC
layer. The newer code does much better at recovering from low signals. If you
want the best, then you should use the kernel from the wireless-testing git
tree. That code will not be in mainline until 2.6.29. Alternatively, you could
use the code from compat-wireless. It is the same as the wireless-testing tree.

Larry

_______________________________________________
Bcm43xx-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/bcm43xx-dev

Reply via email to