Lamar, Sounds like a very good idea. I haven't read the SCSI spec before so would you pardon a few questions?
How many pins does Ultra320 require, is it 32x2 (differential) = 64? Also what are the max cabling lengths at this speed? What is the clock rate on those pins? You mention the ability to send data between hosts without CPU intervention. Does that mean its possible to send data from the USRP to multiple hosts connected on the same SCSI bus? By the way I think that gigabit is 1Gbps full-duplex ie: 2Gbps. Keep the ideas coming, David Carr Lamar Owen wrote: >After a brainstorming session at PARI, I've come up with an alternative to >either USB or GigE, with up to three times the throughput of GigE. > >Ultra320 SCSI. > >Pros: >1.) Standard software interface from the Linux side (libscg like used for >CD >burners); >2.) Scalable throughput with PCMCIA UltraWide SCSI available; any given >SCSI >device has to be able to scale from 5MB/s SCSI-1 up through the top end for >the device; Ultra320 gives 320MB/s of throughput, potentially, which would >increase our bandwidth ten times (three time faster than GigE) and would be a >good fit for the chips involved. The SCSI layer could even be done on an >FPGA, since it's standard LVDS levels for LVD-SCSI. >3.) Interface boards are available for any throughput range needed without >changing any hardware at the USRP. Need 10MB/s? Get a $5 eBay special. >Need the full rate? get a 64-bit PCI-X U320 screamer. >4.) External and internal interfaces available. >5.) Ability to process and redirect streams from one USRP to another one >without CPU intervention over the SCSI bus. > >Comments? > > _______________________________________________ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org http://lists.gnu.org/mailman/listinfo/discuss-gnuradio
