Prev: ModelSim XE III error
Next: Xilinx Synthesis Tool generates clock signals from combinatorial logic
From: Ed McGettigan on 18 May 2010 15:50
On May 18, 10:45 am, rickman <gnu...(a)gmail.com> wrote:
> On May 18, 11:59 am, Ed McGettigan <ed.mcgetti...(a)xilinx.com> wrote:
>
>
>
>
>
> > On May 17, 11:47 pm, rickman <gnu...(a)gmail.com> wrote:
>
> > > On May 17, 3:17 pm, Ed McGettigan <ed.mcgetti...(a)xilinx.com> wrote:
>
> > > > On May 17, 1:49 am, "Nial Stewart"
>
> > > > <nial*REMOVE_TH...(a)nialstewartdevelopments.co.uk> wrote:
> > > > > >> Come on Altera (and the rest), give us a standard.
> > > > > > don't count on them to "give" a "good" standard ;-)
> > > > > > they're in for the money, they follow the money...
>
> > > > > Aye, but it's getting to the stage where an FPGA and programming
> > > > > memory footprint is matched by the programming header!
>
> > > > > It could start to affect device selection.
>
> > > > Are you serious?
>
> > > > I don't think that Xilinx, Altera or ARM really care what header is
> > > > used on the target board.  Each of use picked something that we think
> > > > makes sense and provided a ribbon cable that mates the JTAG cable to
> > > > the target board.
>
> > > > Nothing prevents you from using an alternative connector on the target
> > > > board and creating an adapter that connects to the JTAG cable.
>
> > > > Ed McGettigan
> > > > --
> > > > Xilinx Inc.
>
> > > Yes, I have seen that done, but it is often a PITA to come up with the
> > > adapter cable.  Going from a 0.1" connector to a 2 mm connector is not
> > > really a big advantage, especially when the problem is connector
> > > height.  The best connector is a two piece design like the Conan
> > > series from FCI or similar.  The pins are fully supported by the
> > > housing and so are not delicate, they have a small footprint, they are
> > > cheap, but most of all, they can be very, very short, even when mated
> > > to the cable.  The only down side is that they are not so easy to use
> > > if not supported as a standard.  Ideally the cable would be a flex
> > > cable.  That costs a lot to make in small quantity.  If the FPGA
> > > vendors used a common connector the flex cable and connector could be
> > > a standard programmer component.
>
> > > I guess I don't agree with your other post that adapters are "simple
> > > and easy" to make.  I guess they are if you are just going from one
> > > large connector to another large connector.  But then, what is the
> > > point?
>
> > > Is this an example of vendors listening to their customers?
>
> > > Rick- Hide quoted text -
>
> > > - Show quoted text -
>
> > When I said that adapters are simple, this is an example of what I
> > talking about.http://farm3.static.flickr.com/2598/4167273762_609f88c0e1_o.jpg
>
> > This converted from the Xilinx 2x7 2mm header to the previous flying
> > wire style. Doing the same for another connector would not involve
> > creating a custom cable with two different connectors. The simple PCB
> > adapter would have the Xilinx/Altera/ARM connector on one side and
> > your favorite connector mate on the other. The ribbon/flex cable would
> > have your favorite connector on both ends and would likely be a
> > standard product if you are using a popular connector.
>
> I'm not sure what you are saying exactly, but it sounds like you are
> suggesting that I make my own custom cable adapter which is exactly
> what I am saying is the problem.  The image you provide shows a large
> bulky header/pcb/header which is exactly what I am saying is a very
> poor solution in many applications.  If I need to test a board in a
> stack that only has 5 mm clearance between boards, how am I supposed
> to get that monster in there?
>
> > > Is this an example of vendors listening to their customers?
>
> > We listen all of the time, but we often hear mutually exclusive
> > desires.  There was a big uproar when we went from the 1x7 100mil
> > flying lead style to current 2x7 2mm header.  In order to ease the
> > transition the adapter above was created.
>
> It would seem to me that Xilinx did a "Classic Coke" change where the
> customers weren't allowed a choice until they yelled.  Sure, if you
> change your programming cables to drop compatibility with the existing
> way of connecting, there will be plenty of people who will yell.  The
> product mentioned above is a good example of how to offer an
> improvement and maintain compatibility.  It's only a shame that they
> didn't go to a low profile connector.
>
> There are a lot of apps for low profile devices.  Just look at your
> packaging.  How do people debug those designs with a debugging header
> that is > 10 mm tall?  I have used custom approaches (at other
> companies) when the vendor approaches didn't work.  But for a one off,
> this is an expensive way to go.
>
> Rick- Hide quoted text -
>
> - Show quoted text -

You don't put the adapter on your PCB board, it goes on the cable pod.

Xilinx didn't wait until people complained. We knew that it would be
an issue and the adapter was included in the initial PC-IV and USB
cables to support legacy systems.

Ed McGettigan
--
Xilinx Inc.
From: Nial Stewart on 19 May 2010 04:17
> 2mm pitch is a good size/strength/standard compromise, but I'd not
> stop at 2x3/2mm; such a standard, should
> have more than one size, the others being super-sets.
> So the minimum common subset, is 2x3, then you add another 2x3 for
> example, to allow more test signals.
> With FT2232 and FT2232H becoming standard, there are a lot of unused
> (but paid for) signals sitting right next to the JTAG ones.


I'm really just talking about manufacturers JTAG/programming headers.

There are too many other proprietary combinations to 'standardise'.

IMHO.


Nial


From: -jg on 19 May 2010 17:13
On May 19, 8:17 pm, "Nial Stewart"
<nial*REMOVE_TH...(a)nialstewartdevelopments.co.uk> wrote:
> > 2mm pitch is a good size/strength/standard compromise, but I'd not
> > stop at 2x3/2mm; such a standard, should
> > have more than one size, the others being super-sets.
> > So the minimum common subset, is 2x3, then you add another 2x3 for
> > example, to allow more test signals.
> > With FT2232 and FT2232H becoming standard, there are a lot of unused
> > (but paid for) signals sitting right next to the JTAG ones.
>
> I'm really just talking about manufacturers JTAG/programming headers.
>
> There are too many other proprietary combinations to 'standardise'.
>

Of course, but to become a standard, it needs to be widely useful,
and if everyone expands it in different directions, then it is less
useful.

So, that's why you need to define not only the 'PGM only' sub-set,
but also the suggested supersets too.

Here's a real example we struck:
A 10 pin JTAG header than covers JTAG pgm, but not JTAG fuse override.

We found that with just one pin-change, the header could easily do
BOTH tasks, and here, a vendor define of both functions, would have
been smarter.

(and it would have helped their sales too)

-jg
First  |  Prev  | 
Pages: 1 2 3 4 5
Prev: ModelSim XE III error
Next: Xilinx Synthesis Tool generates clock signals from combinatorial logic