[Discuss-gnuradio] Update: USRP, etc

[Eric Blossom]

Hi Folks,

I know it's been a bit quite around here, so I thought I'd give you a
bit of an update.  Matt & I have been diligently working on the
"USRP", the Universal Software Radio Peripheral.  The specs for it are
on the wiki at http://comsec.com/wiki?UniversalSoftwareRadioPeripheral.

For a sneak preview of one of the prototypes, take a look here:
http://comsec.com/images/usrp.jpg

At this point, we can load firmware into the Cypress FX2 USB
controller, send commands and status back and forth between the host
and the board & load FPGA configuration bit strings into the FPGA.

We've done preliminary testing of the digital upconverter FPGA code
(Verilog), and it appears to be working as expected.

Right now we're working a problem related to streaming data at high
speed between the host, the FX2 and the FPGA.  We've got data moving
at 24MB/sec, but there's a problem we're struggling with having to do
with the FX2 clocking out the last word in a USB packet twice to the
FPGA.  This is not good.  For those of you who enjoy screen shots from
logic analyzers, take a look at these:

http://comsec.com/images/head.png The head of a 256 word packet on its
way to the FPGA.  The data is clocked into the FPGA on the rising edge
of IFCLK when WR is high.

http://comsec.com/images/tail.png The tail of the same packet, showing
the extra word being clocked out (two writes of 0x00ff).  The
immediate problem is that the write line is asserted for 257 clocks,
while we only want to transfer 256 words.  You can see this by looking
at the X to O time measurement in the screen shot.  The clock runs at
48MHz, and X to O (the width of WR) is 5.352 uS, or 257 cycles.

http://comsec.com/images/head2.png  The head of the following packet,
showing the 0x0100 data in its correct place, the first word of the
second packet.

If any of you have any experience with the FX2 "GPIF", please drop me
a line.

That's it for now!

Eric