discuss-gnuradio
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Discuss-gnuradio] GSOC ideas [802.11]


From: address@hidden
Subject: Re: [Discuss-gnuradio] GSOC ideas [802.11]
Date: Wed, 17 Apr 2013 10:10:18 +0200 (CEST)

Hi all,
I'm still alive,don' t worry. Im reading more about 802.11.
About 802.11a/g: i thing that the best thing is write few block with a lot of 
code inside,bacause gnu radio is stream oriented,while these wifi protocol 
don't have a stream conception,but rather a structure frame(for istsance we 
have ofdm symbol with different mean and speed in the same "stream")
In addition,the best thing is have demodulators of subcarrier incorporeted in 
the same block which also perform the ifft,because the speedrate(and 
consequently the kind of modulation) can change dinamically.The stream outgoing 
this block will go inside another block,which initially insert data in .pcap 
files,and after it put the stream inside wireshark.
If i understand correctly,the work of bastian don't work with any speedrate.


I'm agree that moving symbol detection on fpga could improve performance,but 
unfortunaly i don't have fpga for testing.
Another thing: i have some knowledge about cuda(only about high level language,
not "assembler", and architecture,i followed a bit a coursera course).If you 
thing that is interesting,i colud try to move ifft(there are library ready) and 
demodulation on it,but we can have a high bottleneck during transfer data 
between cpu and gpu(except if you have cuda core inside cpu,and memory shared 
with graphics).Instead i don't think to move symbol detection on cuda core,
because in my opinion it couldn't provide more performance.cuda cores could 
return stream demodulate to insert into wireshark block
Another question: is more important latency or performance?
Or i could try to implement 11n(it needan highly optimisation).which do you 
prefer?(maybe i could implement both,if I'll have time in surplus)

At the beginning i'll work about the first section,and after about the second

>Something obvious (although you might not actually want it :) is a 2D
>equalizer and higher modulation schemes.
>MB
sorry,but i didn't understand..why an equalizer?ofdm don't requeire it,is it?



reply via email to

[Prev in Thread] Current Thread [Next in Thread]