Hi John,
Part numbers:
XC5VSX95T-1FF1136C - this is the lowest speed grade, with lead, commercial
temperature range. Available today for purchase from Avnet ($2,735.00, 4 wk
lead time) or Nu Horizons ($ 2,789.70, 22 in stock in NA).
Lead-free is possible and the speed grade could also be a -2. -3 is not
available in SX95T (or SX240T).
Does anyone have a requirement for lead-free?
-Francois
John Ford wrote:
Hi John,
The ROACH can take the LX110T, LX155T, or SX95T. The LX series has very
few multipliers, even the new LX155T only has 128, which is still fewer
than on the IBOB. For sheer size of FFT, we're usually multiplier-limited.
I tried building a PFB for the LX110T on the ROACH prototype and ran out
of multipliers at only about 25% logic utilization.
That build was a 2^13 real PFB (2^12 complex channels) with 4 taps and 4
parallel inputs. Using 8 parallel inputs to get the full bandwidth of the
ADCs, I think that dropped the design down to something like 1K channels.
I think the 64 DSP48E's of the LX110T is definitely not the way to go for
DSP applications. The 128 on the LX155T is a slightly better balance, but
you're still losing 80% of the multipliers for a 65% increase in logic.
Thanks, Henry. The SX95T is available for a reasonable price (~$2795.00),
so I will go with those. Do you have an exact part number for that chip?
Just for fun I priced an SX240 that's considered for roach-II. They're
$15,500 each! Still cheaper than a BEE-2, though...
John
Thanks,
Henry
John Ford wrote:
Hi all. What are the recommended FPGA's for the ROACH? I want the one
that will allow me to build the biggest PFB/FFT's in the chip. It's not
clear to me that the SX is the best one, but I really have no idea.
What say you all?
John
