Re: Re: [Discuss-gnuradio] PACKET FORMAT "error"

[Thomas Tsou]

2010/5/11 zzw.1012 <address@hidden>:
>  Hi, Thomas
>     thanks for your help ! you are right!
>      I add printf code in the function of usrp_basic.cc like this:
> int usrp_basic_tx::write(const void *buf, int len, bool *underrun)
> {
>  ...
>  printf ("len = %d\n", len);
>  if (len < 0 || (len % 512) != 0)
>  {
>   fprintf(stderr, "usrp_basic_tx::write: invalid length=  %d\n", len);
>  }
> }
> then I run the example of benchmark_tx.py :
> ./benchmark_tx.py -f 2500M -s 1500 -M 0.001
> it really send 194560 bytes  = 190k for the default case using 4 samples per
> symbol and it send 380k with -M 0.002.
> but , I still have some question. first , after making packet with 1520
> bytes, then whether is it going on gmsk(default) modulation? I think the
> 1520-bytes packet changes to 190kbytes occur in the process of gmsk. Is it
> right? how to  calculate the number from 1520 to 194560?
> second, in the process of sending packet, the printf information as follow:
> (./benchmark_tx.py -f 2500M -s 1500 -M 0.001)
> len = 16384
> len = 15872
> len = 16384
> len = 16384
> len = 16384
> len = 15872
> len = 16384
> len = 512
> len = 16384
> len = 16384
> len = 16384
> len = 16384
> len = 14848
> Is there have some discipline in sending data? I find it seems unorderly.

Most of these values will be dependent on the modulator. Using the
default settings, you have 8 bits per byte, 1 bit per symbol, 4
samples per symbol, and 4 bytes per sample (16-bit I&Q).

The variability is due to the nature of streaming data and the
GNURadio runtime implementation. At the abstraction level going into
the usrp, there is no notion of fixed packet sizes or frame
boundaries, just samples. The repeated value you see of 16384 bytes is
related to the maximum size of transfer blocks in the underlying USB
layer.

  Thomas