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

[Thomas Tsou]

On Sun, May 9, 2010 at 11:28 PM, zzw.1012 <address@hidden> wrote:
> Hi,
>     I'm studying benchmark_tx.py now. I find that the packet size is not
> right (at least I'd like to think so) in the process of making packet, which
> can be seen in pkt.py and packet_utils.py. In the packet consist of 2 bytes
> packed_preamble, 8 bytes packed_access_code, 4 bytes header, 4 bytes
> outband_crc with default 1500 bytes size, padding bytes and endbyte"\x55" .
> I use default gmsk modulatiom. So, the packet have 2 + 8 + 4 + 1504 + 1 + 1
> = 1520bytes.
>     However, in the function of _npadding_bytes(pkt_bytes_len,
> samples_per_symbol, bits_per_symbol) , there have such description:
> "generate sufficient padding such that each packet ultimately ends up being
> a multiple of 512 bytes when sent across the USB, we send 4-byte samples
> across the USB (16-bit I and 16-bit Q), thus we want to pad so that after
> modulation the resulting packet is a multiple of 128 samples". Also , in the
> function "int usrp_basic_tx::write(const void *buf, int len, bool
> *underrun)"in the usrp_basic.cc,  there have similar code like "if (len < 0
> || (len % 512) != 0){fprintf(stderr, "usrp_basic_tx::write: invalid length
> =  %d\n", len}".
>     they both tell me that the data across the USB must be a multiple of 512
> bytes .but in the example of benchmark_tx.py, the packet size is 1520
> bytes.  what's wrong ?

1520 bytes only refers to the packet size. The transmitted sample
stream at the physical layer includes a number of other factors
including conversion to bits, samples per symbol, and sample size. So
for the default case, a 1520 byte packet using 4 samples per symbol
yields 48640 samples or 194560 bytes, which is a multiple of 512.

  Thomas