We want to estimate the power consumption of the FPGA only on a ROACH
board. The idea is to isolate the FPGA's rail currents by looking at
differences with the FPGA in an active state, and then suspended (i.e.
starting and then killing a borph hardware process). The Actel
monitoring system has rather coarse voltage resolution, so we have
instrumented the current sense resistors on our ROACH with a
multichannel DVM. Using a large design, for example a CASPER 4 Meg
Spectrometer, the measurements seems sensible, that is, currents on
all rails are substantially higher with the hardware process running.
When running our own smaller test designs (various small correlators)
there is at least one anomaly. With the FPGA suspended the 1.8V rail
has greater current, and lower with the hardware process running.
Can anyone suggest an explanation, or avenue of investigation, for
this? We realize, of course, that the power rails run more than just
the FPGA, so even sensible looking estimates may have systematic
errors. Any suggestions as to how to better isolate the FPGA current
would be appreciated. (One thought we had would be to bypass borph
and program the Virtex 5 with JTAG.)
Thanks for any help the ROACH aficionados can offer.
Jonathan
