Here's a strange one...

I've just built five DiscFerret (<http://www.discferret.com/>) boards.=20
All four were built using Multicore LF318 paste in a reflow oven. Peak=20
temperature was 240C, checked with a K-type thermocouple. The reflow=20
profile was Kester's "standard Lead-Free profile", which matches one of=20
Multicore's standard profiles (although the datasheet implies that it's=20
hard to screw up the reflow profile for LF318).

All the parts on all the boards are REV 0x03 (as reported by my=20
PICKit2), which (it seems) would make them Rev A3 or A4. Date codes are=20
as follows:
   07374AT -- fitted to two working boards
   1028YWW -- fitted to one working board and two faulty boards

All boards are running my variant of Microchip's HID bootloader, and the=20
same firmware revision is loaded.


My ATE tests run a bunch of tests on the FPGA-to-PIC comms link:
   1) PIC =3D> FPGA Comms
      The PIC uses two I/O pins (RE0 and RE1) to communicate with a small
      shift register / FSM on the FPGA. This FSM latches the state of the
      other ten FPGA I/Os, then feeds this data back to the PIC over a
      two-wire "SPI-like" link. The PIC reads 23 bits: 20 zeroes, a one,
      then the 10 data bits.

   2) FPGA =3D> PIC Comms
      The PIC uses the same two I/O pins to shift data into a shift
      register on the FPGA (it's a different config bit-stream). The FPGA
      dumps this data onto the ten I/O lines, then the PIC reads the
      state of the pins.

   3) PMP Data Bus Test
      The PIC is remotely reset, forcing re-enumeration and Parallel
      Master Port re-initialisation (tests 1 and 2 disable the PMP).
      The FPGA is reprogrammed, then the PMP is then used to read two
      FPGA registers: FIXED55, which always returns 0x55, and FIXEDAA,
      which always returns... (drumroll please)... 0xAA.
      There's also a Scratchpad register to which the PIC can read and
      write (this is part of the Data Write test).


Here's the issue. Tests 1 and 2 pass on all four boards. If I shove a=20
piece of wire between a pin and ground, I get a FAIL on whichever test=20
is running at the time. This is good...


Problem is, only two of the four boards can actually communicate with=20
the FPGA using the PMP. The other two always return a value of 0x00 when=20
reading FIXED55 and FIXEDAA.


Can anyone think why the PMP might do this?


The code is online at <http://hg.discferret.com/firmware/>. Click=20
"files" to get the file list, then the "file" link next to a file to=20
view its contents. Almost all of the interesting code is in=20
HardwareProfile.h and main.c; the rest of the code is the Microchip USB=20
Framework.

I'm starting to think "defective chips", but these were new-out-of-tape,=20
ordered via Microchip Samples to replace three others that were DOA...=20
(doesn't bode well, does it?)

Thanks,
--=20
Phil.
piclist@philpem.me.uk
http://www.philpem.me.uk/
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.