Vendor supplied symbol/part models?

Hi folks,

Doing a board design with a 456 pin Xilinx FPGA, I find myself in the 
laborious and potentially error-prone process of building a symbol, 
footprint and part model from scratch.   I am aware that commercial part 
libraries are available, but we are a university department and don't 
have those sort of $$$ to throw around for small-run custom designs.

Anyway it seems to me that it would be in vendors' interests (Xilinx in 
this case) to provide verified symbol and footprint models for major 
design tools (Mentor, Protel etc)?  A quick search of the Xilinx web 
site didn't turn up anything.

Is there some point I'm missing here, or are my expectations unreasonable?

Regards,

John

>Doing a board design with a 456 pin Xilinx FPGA, I find myself in the 
>laborious and potentially error-prone process of building a symbol, 
>footprint and part model from scratch. ...

Most people want the schematics to be somewhat readable.  That
usually means breaking a 456 pin part up into several boxes.
So I doubt if a vendor could provide a generic library part
that would work for your design.  (Yes, they could do the
footprint.)

It would be nice if there were a script/program that would
take an intermediate file and make both the schematic library
parts and the pin constraints file for the FPGA.  Or something
like that - the idea is to make sure they were kept in sync.


My suggestion would be to just do it by hand and carefully
check things.  Then get a couple of friends to help you check
it again.  It would be worth bribing with beer/pizza and/or
offering to trade roles when their design needs checking.

What I've done on many occasions is to collect paper copies of
all the data sheets and net lists (both by net and by part/pin)
and all the other info you think might be interesting, and take
over a conference room with a big table and do a check-everything
level design review just before the board goes out.  And gerber
plots and ...  Get somebody to check everything you can think
of to check.  They don't have to know much about your design,
just have enough experience and common sense to read the data
sheets and schematics and see if the connections make sense.
(Double-double check the bubbles/inversions.)


There is a lot of regularity in the footprint.  Assuming your
board level CAD system has some sort of scripting, you can
probably write a program/script to generate a script that
will make the part.

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

"Hal Murray" wrote:

> >Doing a board design with a 456 pin Xilinx FPGA, I find myself in the
> >laborious and potentially error-prone process of building a symbol,
> >footprint and part model from scratch. ...
>
> Most people want the schematics to be somewhat readable.  That
> usually means breaking a 456 pin part up into several boxes.
> So I doubt if a vendor could provide a generic library part
> that would work for your design.  (Yes, they could do the
> footprint.)
>
> It would be nice if there were a script/program that would
> take an intermediate file and make both the schematic library
> parts and the pin constraints file for the FPGA.  Or something
> like that - the idea is to make sure they were kept in sync.

Well ... about 18 months ago a project I was working on was seriously
stalled due to this very issue.  An engineer working with me spent no less
than 18 hours working on defining schematic symbols and pcb footprints for
just ONE large device.  Part of the problem was that the EDA tool --maybe I
should say "most EDA tools"-- failed miserably at having any intelligence
about this process.

The real problem, however, was that, if you ever need to edit, modify,
re-use the part and, maybe, slice the pin set in a different way, you would
be up for another frustrating EDA library maintenance marathon.  A typical
examplet that I had to face was migrating a design from something like a
2V1000 to a 2V500, where the footprints have a downward migration path.

Anyhow, to make a long story (about three months work) short, I wrote a
Windows application that allowed me to fully describe any component quickly.
From that it's pushbutton to generate footprints, schematic symbols and a
constraint file.  Once all the IO for a component is entered, it is an easy
matter to re-slice it any way you want.  You can, for example, slice an FPGA
to create separate schematic symbols for each bank, or merge banks, etc.  As
a benchmark, I can do an 456 pin FPGA in about 20 minutes, starting from
scratch and finishing with pcb and schematic symbols.  More imporantly,
moving pins around or even totaly redoing the assortment of pins on the
schematic symbols is a no-brainer after the intial work is done.

In general terms, I developed a very low opinion of EDA tools, as it seems
that those writing the code seldom have to use it for anything more than the
few trivial examples that ship with the tools.  I think it's fair to say
that, today, you can spend $10K on a tool and fully expect to get unusable
libraries and crippled library creation and maintenance tools.

Maybe the OP can grab students from the Computer Science department and put
them to work on a generalized component creation utility like what we built.
I wish I could provide you with this utility, but it cost a lot of money to
develop and I now consider it both a proprietary tool and a competitive
advantage in many ways.

I wish chip vendors would agree upon a component decription
language/database format of some sort.  These files could be published and
CAD data very easily derived from them.  That would be very useful.


-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Martin Euredjian

To send private email:
0_0_0_0_@pacbell.net
where
"0_0_0_0_"  =  "martineu"

Martin Euredjian wrote:
> "Hal Murray" wrote:
> 
> 
>>>Doing a board design with a 456 pin Xilinx FPGA, I find myself in the
>>>laborious and potentially error-prone process of building a symbol,
>>>footprint and part model from scratch. ...
>>
>>Most people want the schematics to be somewhat readable.  That
>>usually means breaking a 456 pin part up into several boxes.
>>So I doubt if a vendor could provide a generic library part
>>that would work for your design.  (Yes, they could do the
>>footprint.)

Of course - with a Xilinx part, breaking along IO bank lines, plus one 
for power, and one for config etc, is a pretty sensible approach.  But I 
do take your point.

I actually did find a symbol for the part buried deep in the 
mentor/expedition libraries, however it was monolithic, and the pin 
layout/ordering was, as best i could tell, random.

> maybe I
> should say "most EDA tools"-- failed miserably at having any intelligence
> about this process.

And doesn't the "A" in EDA stand for "automation"?? ...

> In general terms, I developed a very low opinion of EDA tools, as it seems
> that those writing the code seldom have to use it for anything more than the
> few trivial examples that ship with the tools.  I think it's fair to say
> that, today, you can spend $10K on a tool and fully expect to get unusable
> libraries and crippled library creation and maintenance tools.

I'm pretty new to PCB design and so on, but have been repeatedly warned 
by several experienced designers "do not trust the library parts"...

> Maybe the OP can grab students from the Computer Science department and put
> them to work on a generalized component creation utility like what we built.

Ha - student labour!  This approach can be a bit hit and miss...

> I wish I could provide you with this utility, but it cost a lot of money to
> develop and I now consider it both a proprietary tool and a competitive
> advantage in many ways.

That's cool.  It's not looking too bad - I'm about 4 hrs in, have done 
the schematic symbols, now doing the pin numbering.  I should be done by 
the end of today.  Happily my design is cleanly fractured along IO bank 
lines.

> I wish chip vendors would agree upon a component decription
> language/database format of some sort.  These files could be published and
> CAD data very easily derived from them.  That would be very useful.

Some existing standard like EDIF might be able to support this already.

Regards,

John

> I'm pretty new to PCB design and so on, but have been repeatedly warned 
> by several experienced designers "do not trust the library parts"...

Yup.  I relax a bit after I have successfully used a part on a design.

It helps to check the raw gerber files to verify that they make
sense.  (The idea is to catch pin numbering errors and/or right/left
top/bottom mirroring errors.)

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.

>
> It would be nice if there were a script/program that would
> take an intermediate file and make both the schematic library
> parts and the pin constraints file for the FPGA.  Or something
> like that - the idea is to make sure they were kept in sync.
>

Apologies if this is of no use (my third post in FPGA land).  Is
http://sourceforge.net/projects/pinout/  of any use?

Cheers

Mike

hmurray@suespammers.org (Hal Murray) writes:

> >Doing a board design with a 456 pin Xilinx FPGA, I find myself in the 
> >laborious and potentially error-prone process of building a symbol, 
> >footprint and part model from scratch. ...
> 
<snip>
> It would be nice if there were a script/program that would
> take an intermediate file and make both the schematic library
> parts and the pin constraints file for the FPGA.  Or something
> like that - the idea is to make sure they were kept in sync.
> 

Mentor have this - it's called Boardlink.  It reads the PAR tool
output and then can automatically, or with guidance, create a
fragmented set of symbols for use in your schematics.

They also have Boardlink Pro which allows you to do pin assignment
upfront in a nicer place than either PACE or Altera's pin assigment
GUI (that's their claim, I haven't used it :-)...  

<snip>
> 
> What I've done on many occasions is to collect paper copies of
> all the data sheets and net lists (both by net and by part/pin)
> and all the other info you think might be interesting, and take
> over a conference room with a big table and do a check-everything
> level design review just before the board goes out.  And gerber
> plots and ...  Get somebody to check everything you can think
> of to check.  They don't have to know much about your design,
> just have enough experience and common sense to read the data
> sheets and schematics and see if the connections make sense.
> (Double-double check the bubbles/inversions.)
> 

And RS232 Tx and Rx :-)

<snip>

-- 
martin.j.thompson@trw.com
TRW Conekt, Solihull, UK
http://www.trw.com/conekt

Hi All,
    I also use bits of Perl script to produce a text file which lists 'nets
connected to the device' next to 'device pin names'. The Perl get the data
from the CAD tool netlist output and Xilinx's pinlists. (I used to cut and
paste pinlists from the PDF but I think Xilinx provide these files
nowadays). This makes it easier to go through the design and look for
typos/errors. The Perl also lists unused pins, which catches mistakes too.
    Also, the script makes the pin LOC bit of the UCF file, the net names on
the schematic must match the top level Entity port names, of course.
            cheers, Syms.


>
> My suggestion would be to just do it by hand and carefully
> check things.  Then get a couple of friends to help you check
> it again.  It would be worth bribing with beer/pizza and/or
> offering to trade roles when their design needs checking.
>
> What I've done on many occasions is to collect paper copies of
> all the data sheets and net lists (both by net and by part/pin)
> and all the other info you think might be interesting, and take
> over a conference room with a big table and do a check-everything
> level design review just before the board goes out.  And gerber
> plots and ...  Get somebody to check everything you can think
> of to check.  They don't have to know much about your design,
> just have enough experience and common sense to read the data
> sheets and schematics and see if the connections make sense.
> (Double-double check the bubbles/inversions.)
>
>
> There is a lot of regularity in the footprint.  Assuming your
> board level CAD system has some sort of scripting, you can
> probably write a program/script to generate a script that
> will make the part.
>

Hi Hal,

Hal Murray wrote:

> My suggestion would be to just do it by hand and carefully
> check things.  Then get a couple of friends to help you check
> it again.  It would be worth bribing with beer/pizza and/or
> offering to trade roles when their design needs checking.

It didn't end up taking very long, maybe 5 hours I suppose.  In my youth 
(!) I entered pages and pages of hex by hand from computer magazines - 
those old skills came flooding back to me yesterday as I trolled one 
line at a time through the pinout sheet, checking off each pin and so on...

> There is a lot of regularity in the footprint.  Assuming your
> board level CAD system has some sort of scripting, you can
> probably write a program/script to generate a script that
> will make the part.

I should look into this - I'm using Mentor's DesignView and 
ExpeditionPCB, presumably they have some scripting capabilties.

Thanks for your reply.

John

John Williams <jwilliams@itee.uq.edu.au> writes:

> Hi Hal,
> 
> Hal Murray wrote:
> 
> > My suggestion would be to just do it by hand and carefully
> > check things.  Then get a couple of friends to help you check
> > it again.  It would be worth bribing with beer/pizza and/or
> > offering to trade roles when their design needs checking.
> 
> It didn't end up taking very long, maybe 5 hours I suppose.  In my
> youth (!) I entered pages and pages of hex by hand from computer
> magazines -
> those old skills came flooding back to me yesterday as I trolled one
> line at a time through the pinout sheet, checking off each pin and so
> on...
> 
> > There is a lot of regularity in the footprint.  Assuming your
> > board level CAD system has some sort of scripting, you can
> > probably write a program/script to generate a script that
> > will make the part.
> 
> I should look into this - I'm using Mentor's DesignView and
> ExpeditionPCB, presumably they have some scripting capabilties.
> 

Lots of luck - the Expedition flow has zero scripting capability :-)

We use the same flow, and I ended up writing some perl which read the
Altera .pin file, created a symbol (using acmaker), then created a PDB entry
in hkp, then HKP2partsdb it into the local parts database.  Not
pleasant, but it meant that changing the pinout of our 2x356 BGAs (it
was a while ago - I thought they were scary then :-) was a whole lot
less hassle.

If you're using that flow, I'd look into Boardlink (which you have
included with your DesignView license - or maybe Boardlink Pro, which
costs money.

Cheers,
Martin

-- 
martin.j.thompson@trw.com
TRW Conekt, Solihull, UK
http://www.trw.com/conekt