UART, JTAG and MCSI connections to FCDEV3B

Mychaela Falconia mychaela.falconia at gmail.com
Wed May 3 02:51:45 UTC 2017


Hello FC community,

Those of you who've been following the FCDEV3B project may remember
that one of the outstanding issues with these boards involves the
header connectors for the UARTs, JTAG and MCSI.  The dual UART header
is 10 pins in 2x5 arrangement and the JTAG header is 14 pins in 2x7
arrangement, and both were originally intended for use with ribbon
cables.  But as soon as I connected a 10-wire ribbon cable terminated
with a 10-pin IDC receptacle to the dual UART header in the middle, I
discovered that the headers are too close together on the PCB and that
connecting an IDC receptacle to the middle one precludes connection to
the other two - the sides of the IDC receptacle body just barely miss
the edge pins of the adjacent headers.

At first I was thinking that we would have to forego the convenience
of ribbon cables and connect with discrete wires instead, and possibly
have to make a new board spin with increased connector spacing going
forward, which would be a huge expense in hired labor alone to make
that kind of change to our tight PCB layout.  I also exercised the
discrete wires connection method this last weekend when I was looking
into the flash boot problem (I played a little with JTAG in that
process), and found it to be OK when you are only working with one
board and only need to connect the wires once, but very problematic
when you have several boards to test and need to connect them in turn.

However, I am now exploring another approach which may be close to a
perfect solution for us: it would allow the use of ribbon cables, there
would be one solid connector to connect and disconnect (instead of a
ponytail of discrete wires) when swapping boards or taking the setup
down for transport, and it should work with the tight spacing between
the headers on our boards.  If this method works as I expect, we should
be able to keep our tight PCB design going forward, i.e., avoid the
very costly respin to change this aspect of the layout.

Here is the approach I am exploring for the dual UART connection
between an FCDEV3B and an FT2232D adapter board:

* Take two 5-wire ribbon cables, or take one 10-wire ribbon cable and
  split it down the middle.

* Partially separate the wires near the ends, but not all the way down
  the ribbon cables, so you still have two 5-wire ribbons and not 10
  loose wires.

* Crimp a Molex SL female crimp-type terminal onto each partially
  separated wire end.

* Insert these female-terminal-equipped wire ends into Molex SL crimp
  connector housings as follows:

* On the FCDEV3B end, have the wires from both 5-wire ribbons go into
  a single 10-pin (2x5) housing.

* On the FT2232D end, terminate each 5-wire ribbon with a single-row
  15-pin housing matching the pins on the PLDkit breakout board.  The
  FT2232D boards from pldkit.com have a row of 15 pins on one side
  bringing out FT2232D channel A, and another row of 15 pins on the
  other side bringing out FT2232D channel B.  The order of signal pins
  and the position of the GND pin relative to the signal pins is the
  same on each side.

The end result will be two 5-wire ribbon cables ending in single-row
15-pin connectors on one end, to be connected to the two sides of an
FT2232D board, joined together into a single 10-pin (2x5) connector on
the other end, to be plugged into an FCDEV3B.  I have just placed the
order with Digi-Key for the necessary parts, and I will try this
approach in practice when these parts arrive.

I expect this approach to work better than the IDC receptacle because
the Molex crimp connector housings I'm going to use are made to be
infinitely stackable both end to end and side to side, unlike the
bulky sides of ribbon cable IDC receptacles.  The above detailed
description of the cable construction is just for the dual UART
connection, but the exact same approach will work for JTAG and MCSI
connections.

Oh, and why specifically Molex and not some other connector vendor?
Simply because I happen to have a big Molex catalog in dead-tree form,
a legacy from the days when part catalogs and databooks were commonly
published as dead trees, so looking at their parts is super-convenient
for me. :-)

M~


More information about the Community mailing list