Re: [Madwifi-devel] continuous tx mode

[Michael T. <mik...@ap...>]

> I have tried out continuous transmit mode and was able to look at the ouput
> on a R&S FSL spectrum analyzer. When I put my card (5413 chipset) into
> continuous transmit mode energy is shown underneath the spectrum mask (for a
> 20MHz channel), however, no bursts are decoded and the modulation is always
> shown as 6 Mbps BPSK regardless of me changing the txcontrate. I have a few
> questions about the observed behavior:
>
> 1) when in cont tx mode, is the card expected to transmit 802.11 frames, or
> just puts out random symbols?
>   
It is sending a management frame with an invalid type, but the timing 
values have been cranked way down to out-of-spec levels.  The testing 
labs wanted to see 100% duty cycle.  This is really not possible and 
stay within 802.11 specs as far as I could tell.  I did everything I 
could to eliminate all delays and timing and just blast frames.  They 
would not be decipherable in that form though.  The test labs wanted to 
see 100% duty cycle for the radio, not 802.11 signals.  The txcont 
basically blasts trash on the RF at the specified power level.
> 2) should rate adjustment work? If so, the 6Mbps BPSK may be an artifact of
> the FSL not being able to decode bursts
>   
Rate adjustment is trashed all over to maximize carrier use and keep the 
carriers active all the time.  Our test labs had a very expensive scope 
up and were complaining any time there were normal time between frames, 
etc.  Continuous transmit mode is not supposed to be a usable signal as 
far as I know.
> 3) should power control work? Regardless of what value I write with the
> txcontpower command the output power does not change; also, the driver seems
> to accept any insanely high value (e.g. 1000).
>   
The reported output power (such as iwconfig / iwlist ) will not change.  
You should be able to see the change with the SA.  The driver clips.  
Units are 0.5dBm and you can specify as much as your card will handle.
> I should say that I am working with an older version of madwifi (trunk rev
> 25xx) with local modifications. If needed, I can re-run the test with any
> other version you suggest to rule out problems that may exist in that older
> rev, or problems that I may have introduced.
>   
Latest txcont stuff is in HEAD of trunk.  And it worked last time I was 
in the lab, but could have regressed.  Happy to help troubleshoot it if 
necessary but my SA went back into the field with our service team.

Thanks,

Mike