Re: [Discuss-gnuradio] Understanding the Data Structure for the USB transfer

[Eric Blossom]

On Wed, Nov 07, 2007 at 03:40:44PM +0100, Uwe Bonnes wrote:
> Hello,
> 
> the last day I am banging my head against the problem how the data is
> structured for the USB transfer between the FX2 and the PC. Datasheet,
> FX2 reference and the web didn't givve me the right clue.
> What I understand:
> - the FPGA pumps interleaved samples into the FX2 fifo via GPIF. E.g.
> http://www.nd.edu/~jnl/sdr/docs/tutorials/4.html talks about a sequence
> I0 Q0 I1 Q1 I2 Q2 I3 Q3 I0 Q0 I1 Q1, etc.
> - these samples are somehow grouped into 512 byte blocks and sent over USB
> - the PC receives these blocks and must somehow know where a data group
> starts
> 
> I can only think that always 512 byte blocks are transferred. With a
> full number of data groups fitting into the block, the block would always
> start with I0. But what happens if the PC gets out of sync, by not reading a
> full block with some strange event? Then the next 512 byte blocks read would
> always start somewhere in the datagroup. Any way to resync? Are short
> packets and the PKTEND signal used? 
> 
> Thanks for any hints
> 
> -- 
> Uwe Bonnes                address@hidden

We never transfer a short block.  They are always 512 bytes long.
There's no problem keeping track of the proper alignment.

Eric