Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search

Messages from 7275

Article: 7275
Subject: Unbonded Pad Resources
From: johna@dvorak.amd.com (John Archambeault)
Date: 20 Aug 1997 22:25:25 GMT
Links: << >>  << T >>  << A >>

Does anyone know how to instantiate the use of the flip-flops / buffers in the
unbonded pads in OrCAD?  In a previous desgin I used verilog / synopsys to tell
XACT to use those flip-flops to instantiate a 100 bit FIFO.

However, I cannot figure out how to do it when starting from OrCAD.

Thanks,
John
-- 
John Archambeault
Article: 7276
Subject: Re: ISP Stories
From: jim granville <Jim.Granville@xtra.co.nz>
Date: Wed, 20 Aug 1997 17:03:18 -0700
Links: << >>  << T >>  << A >>
Steven J. Ackerman wrote:
> 
<snip>
> If you're using Lattice parts, you have to fully complete and debug
> your design before you lay out a circuit board. And the signal pinout
> will be completely arbitrary, not easily circuit board routable. If
> you think that you can go in and re-assign pinouts or make logic
> changes after the fact you may be in for a nasty suprise - the Lattice
> router may no longer be able to fit your design into the part ! I have
> found this to be the case if the part is over 50% utilized.

Yikes!
We have just completed a design where we routed the PCB FIRST !
( using ATMEL ATF1500 PLDs, 100% & 87% packed )

This gave very close to single sided routing between RAM, 2 x PLCC44
PLDs, 1 x PLCC44 uC.

The other side is now 95% ground plane, and we have a very quiet, 2
layer, design that
is a VGA Terminal.

I shudder to think how a 'PLD determined' design would have performed
:-(


This 'my fitter knows best' philosophy is too common, but with pressure,
things will
get better :-)

- jim G.

> 
> The Synario software is easy to use, much better than their own PDS.
> 
> Other than these discrepancies we find the Lattice parts to be useful
> and have shipped several thousand units of various designs with no
> manufacturing problems.
> 
> Steven J. Ackerman, Consultant
> ACS, Sarasota, FL
> sja@gte.net
> http://www.acscontrol.com

-- 
======= Manufacturers of Serious Design Tools for uC and PLD  =========
= Optimising Modula-2 Structured Text compilers for ALL 80X51 variants
= Reusable object modules, for i2c, SPI and SPL bus interfaces
= Safe, Readable & Fast code - Step up from Assembler and C
= Emulators / Programmers for ATMEL 89C1051, 2051, 89C51 89S8252 89C55
= *NEW* Bondout ICE for 89C51/89C52/89C55 
= for more info, Email : DesignTools@xtra.co.nz  Subject : c51Tools

Article: 7277
Subject: Programme: Reed-Muller 97 -- Int. Workshop on Function Representations
From: Jon.Saul@comlab.ox.ac.uk (Jon Saul)
Date: Wed, 20 Aug 97 18:15:38 BST
Links: << >>  << T >>  << A >>



			Provisional Programme

           3rd International Workshop on Applications of the
                Reed-Muller Expansion in Circuit Design 
                           (Reed-Muller 97)

                  In Cooperation with IFIP WG 10.5.

                                                                          
Place: Oxford University, Oxford, UK.
Date:  September 19-20, 1997
        just after ESSCIRC 97 (European Solid-State Circuits Conference)
                Southampton, UK, September 16-18 . 

This workshop focuses on the application of new techniques in the
representation and realization of discrete functions. AND-EXOR based
representations are often simpler than AND-OR based representations,
and have other important properties. Decision diagrams are being
extensively studied, and have offered powerful new techniques for
verification and synthesis.  The goal of the workshop is to bring
together researchers in these and related fields to discuss new
approaches and results. The first workshop was held in September 1993,
in Hamburg, and the second in August 1995, in Tokyo.

Call for posters: As well as the refereed work listed below we are
also making space available for the display of posters describing very
recent work. If you'd like to take advantage of this, simply contact
Frances Page at the address below with the title of your poster and a
list of authors. Space will be allocated on a first come first served
basis.

Enquiries :
Miss Frances Page
Oxford University Computing Laboratory
Wolfson Building
Parks Road
Oxford OX1 3QD, UK.
Tel: +44 1865 283505/273838
Fax: +44 1865 273839
Email: Frances.Page@comlab.ox.ac.uk

Further details are available on the website: 
http://www.comlab.ox.ac.uk/oucl/users/jon.saul/ReedMuller97.html

*************************************************************************

FRIDAY 19 SEPTEMBER 1997
************************

08.45-09.10 Registration
09.10-09.15 Welcome 

09.15-09.40
Decision Diagrams and AND/OR Graphs for Design Automation Problems
Rolf Drechsler, Wolfgang Kunz, Dominik Stoffel and Alenka Zuzek

09.40-10.05
Reordering Based Synthesis
Andreas Hett, Rolf Drechsler and Bernd Becker

10.05-10.30
ETDD-based Generation of Complex Terms for Incompletely Specified Boolean
Functions
Gueesang Lee

******************
10.30-11.30
Coffee and Posters
******************

11.30-11.55
Minimized Generalized Partially Mixed Polarity Reed-Muller Expansion
M Marek-Sadowska, G Guner, T Krishnamurthy, S Gargeshwari and Ch Sharma

11.55-12.20
Exclusive-OR of Two Sum-of Products Expressions: Simplification and an
Upper Bound on the Number of Products
Debatosh Debnath and Tsutomu Sasao

12.20-12.45
New Fast Approach to Approximate ESOP Minimization for Incompletely
Specified Multi-Output Functions
Ning Song and Marek Perkowski 

************
12.45-02.00
Buffet Lunch
************

02.00-02.25
Low Power Aspects of XOR based Circuit Design
Yibin Ye, Kaushik Roy and Rolf Drechsler

02.25-02.50
Lattice Diagrams Using Reed-Muller Logic
Marek Perkowski, Malgorzata Chrzanowska-Jeske and Yang Xu

02.50-03.15
Hardware Acceleration of Two-level ESOP Minimization Using CAMs
Jonathan Saul

***************
03.15-04.00
Tea and Posters
***************

04.00-04.25
Two Hierarchies of Generalized Kronecker Trees, Forms, Decision Diagrams
and Regular Layouts
Marek Perkowski, Lech Jozwiak and Rolf Drechsler

04.25-04.50
Spectral Transforms and Word-Level Decision Diagrams
Rolf Drechsler, Radomir Stankovic and Tsutomu Sasao

***********************
WORKSHOP DINNER	
***********************

Day 1 Posters
*************


In addition to posters relating to all the above papers the following
will be displayed during Friday:

Non-Abelian Groups in Optimization of Decision Diagrams Representations
of Discrete Functions
Radomir Stankovic

Reed-Muller Transform and Wavelets Theory:  An Alternative Look at
Reed-Muller Expansions
Radomir Stankovic and Yasushi Endow

Test Set Generation for Functional Decision Diagram Circuits using
Genetic Algorithm
A Bystrov and A Almaini

A General Data Structure for XOR-Decomposition of Sets of Switching
Functions
Bernd Steinbach and Christian Lang

Compact Testing of AND-EXOR Programmable Logic Arrays
Roustam Latypov

****************************************************************************

SATURDAY 29 SEPTEMBER 1997
**************************

09.15-09.40
Complexity Measures for AND-EXOR Expressions
Tsutomu Sasao

09.40-10.05
Minimizing Polynomial Implementation of Weakly Specified Logic Functions
and Systems
Arkadij Zakrevskij

10.05-10.30
The Complexity of Symmetric Functions in the Polynomial Forms
Julia Manstivoda and Nikolay Peryazev 

******************
10.30-11.30
Coffee and Posters
******************

11.30-11.55
Case Study:  Manipulating EXOR-OBDDs by Means of Signatures
Christoph Meinel and Harald Sack

11.55-12.20
Compilation of Fast Manipulation Algorithms for K*BMDs
Stefan Horeth and Rolf Drechsler

12.20-12.45
Combinational Logic-Level Verification using Boolean Expression Diagrams
Henrik Hulgaard, Poul Williams and Henrik Andersen

************
12.45-02.00
Buffet Lunch
************

02.00-02.25
A Heuristic Procedure for Finding AND-OR-XOR Expansions of Incompletely
Specified Boolean Functions
Elena Dubrova, Michael Miller and Jon Muzio

02.25-02.50
A Canonical AND/EXOR Form that includes both the Generalized Reed-Muller
Forms and Kronecker Reed-Muller Forms
Marek Perkowski, Lech Jozwiak and Rolf Drechsler

02.50-03.15
A Critique of Mixed Exclusive-/Inclusive-OR Logic Synthesis for the
Xilinx XC6200 FPGA
Paul Metzgen and Jonathan Saul

***************
03.15-04.00
Tea and Posters
***************

04.00-04.25
A New Linearly Independent, Zhegalkin Galois Field Reed-Muller Logic
Karen Dill, Konika Ganguly, Robert Safranek and Marek Perkowski

04.25-04.50
Universal and Robust Testing of Stuck-Open Faults in Reed-Muller
Canonical CMOS Circuits
D Das, S Chakraborty and B Bhattacharya

04.50-05.15
Reed-Muller-Fourier Representations Versus Galois Field Representations
of Four-Valued Logic Functions
Radomir Stankovic, Dragan Jankovic and Claudio Moraga

05.15-05.20 Closing

Day 2 Posters
*************

Decision Diagrams for Discrete Functions Representations
Radomir Stankovic

On Self-dual Complements of Fixed Polarity Reed-Muller 2-forms
Ken Fok, Malgorzata Marek-Sadowska and Simone Boehringer

Evolutionary Minimization of Generalized Reed-Muller Forms
Karen Dill and Marek Perkowski

Decomposition of Logical Functions in Reed-Muller Logic
Elena Zaitseva and D Popel

Parallel and Distributed Algorithms for Minimization of Incompletely
Specified Logic Functions in Reed-Muller Domain
Svetlana Yanushkevich, Grzegorz Holowinski, Vlad Shmerko and
Elena Zaitseva

************************************************************************ 



-- 
Jonathan Saul,                             Tel: +44 1865 273842
Oxford University Computing Laboratory,    Fax: +44 1865 273839
Wolfson Building, Parks Road,            Email: Jon.Saul@comlab.ox.ac.uk
Oxford OX1 3QD,	UK.						
Article: 7278
Subject: Re: LogiBLOX components in VHDL?
From: "Anil T.L.N." <anil@xilinx.com_spam_prevent>
Date: 21 Aug 1997 01:49:16 GMT
Links: << >>  << T >>  << A >>
Hi! Logiblox components are to be viewed as black boxes in Synopsys. The
vhi file
provided is for helping in instantiating the component in your design. 
However,
you should attach a dont_touch attribute on these instances in VHDL. Please
note that dont_touch attribute must be attached on the instance name and
not the component name. 

For e.g.,
if you have a statement as UADD: ADDER
you have to attch the dont_touch attribute in Synopsys to UADD
dont_touch {UADD}

This will tell Synopsys not to look at this module. Synopsys will issue a
warning to this effct which you can ignore. Later in M1, ngdbuild will
merge
your Logiblox netlist ADDER.ngo with the other netlists and create the
merged
.ngd file. 

Anil.


Exjobbare Joachim Strombergson <qmwchim@emw.ericsson.se> wrote in article
<33F9781E.1CFBAE39@emw.ericsson.se>...
> Hi!
> 
> I'm having problems using LogiBLOX components as component instances in
> my VHDL-code. I'm using the LBGUI tool to customize the blocks to match
> my need. I then try to include these block in my code by inserting the
> generated VHI-file into my code and completing the port assignments. But
> when I try to read in the design into Synopsys Design Analyzer the tool
> does not recognize the component.
> 
> I either would like to inlude the LogiBLOX like any other library, and
> then customize the component with attributes (but don't know how), or
> somehow make Synopsys realize that the implementation of the component
> is given by the generated netlist in the NGO-file.
> 
> I've read through the material in the dltext documentation, but it
> doesen't help me out.
> 
> Any help greatly appreciated.
> Joachim Strömbergson
> 
Article: 7279
Subject: new radix-8 division algorithms
From: "Prof. Vitit Kantabutra" <kantviti@isu.edu>
Date: Wed, 20 Aug 1997 22:06:57 -0600
Links: << >>  << T >>  << A >>
I just finished a new paper describing two new radix-8 division algorithms
that are based on comparisons of relatively narrow operands rather than on
confusing, bug-inducing P-D plots.  One of the algorithms is maximally-
redundant and uses two's complement partial remainders, while the other
algorithm is minimally-redundant and uses redundant partial remainders
in signed-digit format (requiring no assimilation for each iteration).

These algorithms are similar to the ones I presented at ICCD '96, but are
fixed-radix (radix-8).  That is, the new algorithms no longer retire an
unpredictable number of bits per iteration, but instead retire exactly
three bits.

For a copy of the paper or questions please contact me at

Vitit Kantabutra
Department of Mathematics
Idaho State University
Pocatello, ID 83209, U.S.A.
vkantabu@howland.isu.edu
kantviti@fs.isu.edu
(208) 236-3405 (voice)
(208) 236-2636 (fax)


Article: 7280
Subject: Re: ISP Stories
From: "Graham Rhodes" <grahamr@mmtech.co.uk>
Date: 21 Aug 1997 08:04:07 GMT
Links: << >>  << T >>  << A >>


Joel W. Kolstad <Joel.Kolstad@Techne-Sys.com> wrote in article
<01bcadaa$d72d3140$0307e38f@zimbo>...
> Steven J. Ackerman <sja@gte.net> wrote in article
> <5tc6sr$6l9$1@gte2.gte.net>...


> Cypress likes to pride themselves on the ability to use completely
> arbitrary pin assignments when they're trying to convince you to use
their
> stuff over Lattice's.  In general I think they're correct, but one thing
> they don't mention is that even their pin assignments start becoming
> non-arbitrary when you start utilizing over about 75% of the device. 
> Better than Lattice, but definitely not as cool as they'd have you
believe.

Certainly Cypress does reasonably well with user-defined pin assignments
even when the device is pretty full. I have been running very close to the
wind with one of their devices approx 98% full, and still able to do the 
odd tweaks to the design.

It is noticeable that I had to hand assign the pins originally on this
device
because Cypresses fitter could not handle the part.

The routability of these parts is better than average.

Graham


 
Article: 7281
Subject: Re: ISP Stories
From: Steve Darby <" sdarby"@hmi.com>
Date: Thu, 21 Aug 1997 08:18:18 -0500
Links: << >>  << T >>  << A >>
Terry Harris wrote:
> 

> Lattice parts have a GRP (global routing pool) which can route any
> signal anywhere but unfortunately with some undocumented limitations,
> when you lock down pins and the fitter can't fit (which sounds like
> the problem the previous poster was getting) it is because of these
> GLB restrictions. Being undocumented it is pretty hard to know what to
> do about it.
> 

The Lattice parts also have a Output Routing Pool (ORP), which can route
any the output of a GLB to any of a set of pins.  This is where I found
the fitter was running out of resources.  Each GLB has a set of
associated pins where the ORP can be bypassed, but the fitter was too
dumb to assign GLBs to bypass the ORP (even though I assigned the GLBs,
the fitter would move them and it wouldn't fit).  After some searching,
I found I could designate the output as CRIT, the fitter finally
bypassed the ORP, and it routed.

My experience has been that the Lattice router is the hardest to work
with to get pins locked.  It clearly takes the "fitter knows best"
philosophy.  

I have also used XILINX XACT fitter, and it is better, but still has
problems.  I recently had a design with all my pins locked and only one
free pin lest on the chip.  When I added one more output, I assumed the
fitter would know where to put it (the one free pin seemed like a good
choice).  To my chagrin, the fitter couldn't route this last output. 
When I locked the pin to the only free output, it routed fine.

With fitters like these, it is kind of scarry that Vantis is designing
HW around their fitter.

Steve Darby
Huntsville Microsystems, Inc.
Article: 7282
Subject: (no subject)
From: Sufyaan Kazi <jab@mediatecnicsplc.demon.co.uk>
Date: Thu, 21 Aug 1997 15:53:31 +0100
Links: << >>  << T >>  << A >>
Since the Samsung site is being hacked into pieces can anyone identify these chips. 

Chip 1:
Model No KM41C1600AJ-6

Chip 2:
Model No KM41C1600AJ-6

I believe that these are 72 pin chips for pc's but how many megs are they and what speed, etc.

Chip3:
Model No KMM3144C213AT-6

Chip4:
Model No KMM51441000BTG-8

I believe that these chips are for Sun workstations, They are 200 pins but again can anyone supply me with any further details.

please send replies to jab@mediatecnicsplc.demon.co.uk as we don't read the newsgroups regularly.

Many thanks in advance,

Sufyaan Kazi
c/o Mediatecnics Plc
United Kingdom

Article: 7283
Subject: Re: Unbonded Pad Resources
From: Jim Chase <Jim.Chase@LucasControls.Com>
Date: Thu, 21 Aug 1997 10:33:04 -0500
Links: << >>  << T >>  << A >>
John Archambeault wrote:
> 
> Does anyone know how to instantiate the use of the flip-flops / buffers in the
> unbonded pads in OrCAD?  In a previous desgin I used verilog / synopsys to tell
> XACT to use those flip-flops to instantiate a 100 bit FIFO.
> 
> However, I cannot figure out how to do it when starting from OrCAD.
> 
> Thanks,
> John
> --
> John Archambeault

I have never used Orcad for Xilinx design, however, with any schematic
capture program you can put a location constraint on the IFD or OFD
primitive to specify the location.  All unbonded I/O are designated U?
instead of P?.  So use a constraint similar to LOC=U52.  You can find
the unused I/O names in XDE.

Jim Chase
Article: 7284
Subject: ISP Stories
From: Martin Vorbach <Martin.Vorbach@SCRAP.de>
Date: Thu, 21 Aug 1997 18:28:30 +0200
Links: << >>  << T >>  << A >>
> We are considering the use of ISP CPLD's.  Anyone have any amusing
> or helpful anecdotes regarding their use?
> Thanks much.
	[M.Vorbach]  

	We crashed a design using Lattice. Don´t use it.

	Now we use Altera MAX9K and they are OK. We have good experience
using 9320 and 9560 and the result is very well up to an utilization of
80% cells and 80% pins. 
	The MaxIIPlus Version 7.2 fitter contains a bug and the result
is a very very bad fitting: If you´ll get this problem ask ALTERA for
the internal release 7.22. This version works very well!

	Also we have very good experience using Altera 8K and 10K, but
they work with SRAM cells. Please don´t use 7K, they are not very
satisfying.

Article: 7285
Subject: Altera Flex10K I/O setup time
From: tom@focus-systems.on.ca (Tom Pawelko)
Date: Thu, 21 Aug 1997 17:14:37 GMT
Links: << >>  << T >>  << A >>
On an Altera Flex 10k130-4 fpga there is a 12.6ns input delay between
an external I/O pin and the IOE input register. I'm told this was
added to improve the Input Hold time and that the only way to bypass
it is to use a register in the LE. This contradicts the 1996 data book
recommendation to use the "input register for external data that
requires a fast setup time". Adding the 3.8ns FF setup time and
subtracting the 5.8 clk distribution, results in a 10.6ns setup time
of data to the external clock. 

Since we advance our fpga clock with an external PLL, this Tinreg=12.6ns
is a direct waste of useful clock budget. Needless to say, at 40 MHz,
it amounts to half of one's clock cycle. 

Has anyone dealt with this? Is there some way to get around it? On
smaller Flex parts, it is faster using an internal LC register, but
doesn't appear to be on the 10K100 and 10K130 parts.

Tom 

-- 


------------------------------------------------------------------------------
Thomas P. Pawelko                           phone: (519) 746-4918
Article: 7286
Subject: Re: Unbonded Pad Resources
From: Tom Burgess <tburgess@drao.nrc.ca>
Date: Thu, 21 Aug 1997 10:27:29 -0700
Links: << >>  << T >>  << A >>
John Archambeault wrote:
> 
> Does anyone know how to instantiate the use of the flip-flops /
> buffers in the
> unbonded pads in OrCAD?  In a previous desgin I used verilog /
> synopsys to tell
> XACT to use those flip-flops to instantiate a 100 bit FIFO.
> 
> However, I cannot figure out how to do it when starting from OrCAD.
> 

Don't recall where or if this is documented, but you just put the
INTERNAL attribute into the part options_1 name field. If you want
to lock it to a particular unbonded IOB, I think some variation of
"LOC=unnn,INTERNAL" will work. The "u" indicates unbonded, and you
need to go into XDE to find out what "nnn" is for a particular
unbonded IOB.

	regards, tom (tburgess@drao.nrc.ca)
Article: 7287
Subject: make problem
From: "Michael R. Palmer" <m_palmer@edocombat.com>
Date: Thu, 21 Aug 1997 10:56:58 -0700
Links: << >>  << T >>  << A >>
I am using OrCAD Express to design into a Xilinx FPGA.  My opeations are 
mostly defined by VHDL.  The project simulates perfectly with no 
problems.  When I try to compile so I can move into the Xilinx tools I 
receive the error message below.

c:\orcadwin\express\samples\vhdl\stave_select.vhd(64) : Failure : 
[Load061]   Unsynthesizable element of design: use of generic symbol 
stave
ERROR [SYN0003]   The synthesis operation failed.
Compile operation failed

When I double click on that line in the session log the line below is 
highlighted in the VHDL source file.

else delay_num := (abs(Beam - stave)) mod mod_num;

Beam and delay_num are variables.
mod_num is a constant.
stave is a generic.

Thank you for your help.

****************************************************
*Michael R. Palmer				   *
*Scientific Programmer				   *
*e-mail: m_palmer@edocombat.com			   *
*Ph. (757) 424-1004                                *
*Fax (757) 424-1602				   *
****************************************************

Sincerely,

Michael R. Palmer
Article: 7288
Subject: Re: ISP Stories
From: Ed Barrett <ed.barrett@worldnet.att.net>
Date: Thu, 21 Aug 1997 14:23:12 -0500
Links: << >>  << T >>  << A >>
Steve Darby wrote:
> 
> Terry Harris wrote:
> >
> 
> > Lattice parts have a GRP (global routing pool) which can route any
> > signal anywhere but unfortunately with some undocumented limitations,
> > when you lock down pins and the fitter can't fit (which sounds like
> > the problem the previous poster was getting) it is because of these
> > GLB restrictions. Being undocumented it is pretty hard to know what to
> > do about it.
> >
> 
> The Lattice parts also have a Output Routing Pool (ORP), which can route
> any the output of a GLB to any of a set of pins.  This is where I found
> the fitter was running out of resources.  Each GLB has a set of
> associated pins where the ORP can be bypassed, but the fitter was too
> dumb to assign GLBs to bypass the ORP (even though I assigned the GLBs,
> the fitter would move them and it wouldn't fit).  After some searching,
> I found I could designate the output as CRIT, the fitter finally
> bypassed the ORP, and it routed.
> 
> My experience has been that the Lattice router is the hardest to work
> with to get pins locked.  It clearly takes the "fitter knows best"
> philosophy.
> 
> I have also used XILINX XACT fitter, and it is better, but still has
> problems.  I recently had a design with all my pins locked and only one
> free pin lest on the chip.  When I added one more output, I assumed the
> fitter would know where to put it (the one free pin seemed like a good
> choice).  To my chagrin, the fitter couldn't route this last output.
> When I locked the pin to the only free output, it routed fine.
> 
> With fitters like these, it is kind of scarry that Vantis is designing
> HW around their fitter.
> 
> Steve Darby
> Huntsville Microsystems, Inc.

Maybe you just didn't know how to best utilize the fitter! Pins can be
fixed in the source file (schematic or VHDL), with a seperate pin file
or in the most recent Lattice fitter by point and click. In the last
case a window opens showing the signal names - you click on the signal
name than click on the pin shown on the package diagram. I don't know
what could be easier. I have not seen any issues with Lattice parts
pin-lock performance even at very dense utilization. They also have an
app note benchmarking pin locking performance on their web page at
http://www.latticesemi.com.

Ed
Article: 7289
Subject: X84 FPGA Board $199.00
From: Richard Schwarz <aaps@erols.com>
Date: Thu, 21 Aug 1997 15:50:45 -0400
Links: << >>  << T >>  << A >>
The X84 FPGA development board is available in the
Month of September for $199.00. The price goes back up
to $275.00 at the end of September.

*The X84 is a XILINX FPGA development
*Board  which can run in a PC or  stand alone.
*It feature:
*An on board FPGA,
*Prom socket,
*Oscillator socket,
*Timer circuit,
*Status LEDs,
*Download software,
*C code,
*8255 chip,
*Decode pal,
*Strapable PC address,
*Accepts XILINX Download cable.
*Four 20 pin IDC IO connectors.
*VHDL examples
*C code boilerplates.

The FPGA is socketed and the board
can take the following FPGAs:

                                X84 BOARD FPGA          Typical Gate
Count
                                          XC5202PC84
2-3K
                                          XC5204PC84
4-6K
                                          XC5206PC84
6-10K
                                          XC5210PC84
10-16K
                                          XC4003PC84
2-5K
                                          XC4005PC84
3-9K
                                          XC4006PC84
4-12K
                                          XC4008PC84
6-15K
                                          XC4010PC84
7-20K

The board is also available with the XILINX Foundation Software
See the details of the board and the software kits that go with it at:

http://www.associatedpro.com/aps/x84.html

A free on line tutorial for the board is available at :

http://www.associatedpro.com/aps/x84lab

and a free EDA on line newsletter is available at:

http://www.associatedpro.com/aps

--
__/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/

    Richard Schwarz, President
    Associated Professional Systems Inc. (APS)
    email: aaps@erols.com
    web site: http://www.associatedpro.com
    Phone: 410-569-5897
    Fax:   410-661-2760

__/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/


Article: 7290
Subject: Re: FPGA prototyping board
From: "Steven K. Knapp" <sknapp @ optimagic.com>
Date: 22 Aug 1997 03:25:20 GMT
Links: << >>  << T >>  << A >>
Aptix 'http://www.aptix.com' is probably your best bet.   There is also a
fairly comprehensive list of FPGA/CPLD boards listed on The Programmable
Logic Jump Station at 'http://www.optimagic.com/boards.html'.
-- 
Steven Knapp
OptiMagic(tm) Logic Design Solutions
E-mail:  sknapp @ optimagic.com
Programmable Logic Jump Station:  http://www.optimagic.com

Bjorn Sihlbom <emwbs@emw.ericsson.se> wrote in article
<33F9B003.1381@emw.ericsson.se>...
| Hi!
| 
| I need to prototype a board incl. a complex ASIC.
| Do anyone know about of-the-shelf boards with space for a lot of big
| FPGAs and perhaps configurable routing?
| 
| Any other hints?
| 
| Regards,
| 
| Bjorn.
| 
Article: 7291
Subject: Re: ISP Stories
From: terry.harris@dial.pipex.com (Terry Harris)
Date: Fri, 22 Aug 1997 03:50:04 GMT
Links: << >>  << T >>  << A >>
Martin Vorbach <Martin.Vorbach@SCRAP.de> wrote:

>> We are considering the use of ISP CPLD's.  Anyone have any amusing
>> or helpful anecdotes regarding their use?
>> Thanks much.
>	[M.Vorbach]  
>
>	We crashed a design using Lattice. Don´t use it.

Lots of people have crashed BMW's wouldn't stop me guying one <grin>

>	Now we use Altera MAX9K and they are OK. We have good experience
>using 9320 and 9560 and the result is very well up to an utilization of
>80% cells and 80% pins. 
>	The MaxIIPlus Version 7.2 fitter contains a bug and the result
>is a very very bad fitting: If you´ll get this problem ask ALTERA for
>the internal release 7.22. This version works very well!

And how is 8.0 that's been out for a couple of months? 
>
>	Also we have very good experience using Altera 8K and 10K, but
>they work with SRAM cells. Please don´t use 7K, they are not very
>satisfying.

Must be thousands of not very satisfied Altera 7k users out there
then? I don't suppose anyone wanting faster then 15ns would find the
9k series very satisfying either. 


Cheers Terry...
Article: 7292
Subject: Re: ISP Stories
From: terry.harris@dial.pipex.com (Terry Harris)
Date: Fri, 22 Aug 1997 03:52:31 GMT
Links: << >>  << T >>  << A >>
Ed Barrett <ed.barrett@worldnet.att.net> wrote:


>Maybe you just didn't know how to best utilize the fitter! Pins can be
>fixed in the source file (schematic or VHDL), with a seperate pin file
>or in the most recent Lattice fitter by point and click. In the last
>case a window opens showing the signal names - you click on the signal
>name than click on the pin shown on the package diagram. I don't know
>what could be easier.

Duhh I think we figured out how to ask it to lock pins, we didn't
figure out how to make it fit after locking pins. Having a pair of
16bit data buses randomly scattered around a 208pin package is pretty
ugly.  Slightly less ugly is as close as I got. 

>I have not seen any issues with Lattice parts
>pin-lock performance even at very dense utilization. 

Well you can't have looked too closely because there are. The way
product terms are shared within GLBS mean a complex function only
leaves room for simple ones so there is some arbitrary gouping into
GLBS based on function product term requirements. Then there is the
limited number of inputs to GLBS which again can cause grouping. Then
product term clocks feed all 4 flipflops in a GLB so one flipflop with
a product term clock means the rest of the GLB can only be used for
non-registered functions. 

Within a Megablock you have the output routing pool which can help a
lot but there are also undocumented restrictions in the GRP which
means even if you have things nicely fitting into GLBS (or twin GLBS
in the 3k series) you can't move the GLBS around in the device. You
try it and the fitter just takes longer, prints a few more dots and
barfs saying it can't route (or something like that).  When you do get
something the fitter manages to route you don't get any indication of
GRP usage except perhaps how many dots the fitter printed while
trying, just cross your fingers and hope that it will take more than a
few 'straws' to break the fitter again. 

That said I don't think the Lattice parts are significantly worse than
anyone else they have a larger number of small restrictions (like I
said quirky) so some designs will fit and run better in a Lattice part
compared to the competition. 

 
Cheers Terry...
Article: 7293
Subject: division paper
From: "Prof. Vitit Kantabutra" <kantviti@isu.edu>
Date: Thu, 21 Aug 1997 22:56:11 -0600
Links: << >>  << T >>  << A >>
My division paper is now available at the following site:

http://math.isu.edu/~vkantabu/radix8.pdf

I'd really appreciate any comments you may have about these new results.
Article: 7294
Subject: Re: ISP Stories
From: Tim Forcer <tmf@ecs.soton.ac.uk.nojunk>
Date: Fri, 22 Aug 1997 09:32:12 +0100
Links: << >>  << T >>  << A >>
Terry Harris started a thread:

>>> Lattice parts .. global routing pool ... undocumented limitations,
>>> when you lock down pins and the fitter can't fit ....

To which Steve Darby added comments including:

>> The Lattice parts also have a Output Routing Pool (ORP), which can route
>> any the output of a GLB to any of a set of pins.  This is where I found
>> the fitter was running out of resources.  Each GLB has a set of
>> associated pins where the ORP can be bypassed, but the fitter was too
>> dumb to assign GLBs to bypass the ORP (even though I assigned the GLBs,
>> the fitter would move them and it wouldn't fit).  After some searching,
>> I found I could designate the output as CRIT, the fitter finally
>> bypassed the ORP, and it routed.

Then Ed Barrett commented:

> Maybe you just didn't know how to best utilize the fitter! Pins can be
> fixed in the source file (schematic or VHDL), with a seperate pin file
> or in the most recent Lattice fitter by point and click. In the last
> case a window opens showing the signal names - you click on the signal
> name than click on the pin shown on the package diagram. I don't know
> what could be easier. I have not seen any issues with Lattice parts
> pin-lock performance even at very dense utilization. They also have an
> app note benchmarking pin locking performance on their web page at
> http://www.latticesemi.com.

Read Steve's post more carefully.  Steve DID know all about locking
pins, his problem was that he had to lock the GLB, the pin, AND the
interconnect between them before the fitter would come up with the right
answer.  A different problem to the one you comment on, but very
significant given the architecture.  Using Lattice's TERRIBLE pDS
software I ended up fixing just about everything, so the fitter had
almost nothing to do.  Even most of the logic was sorted outside pDS. 
That way I knew what I was getting and how it was derived.  (Note that
when Lattice recently updated pDS, they didn't even bother to tell me,
despite my pDS being a full-price, registered copy.  Unless the new one
is a vast improvement, I'm not sure it would be worth upgrading anyway.)

A secondary issue with Lattice (and many other CPLDs) is that despite
all these routing pools, the silicon is arranged in blocks of blocks, so
there is no way (with the bigger parts at least) that you can have
totally free pin assignment.  I'm not knocking the architecture - it has
a lot of good points - but it isn't, and can't be, a total solution.

Tim Forcer               tmf@ecs.soton.ac.uk
Department of Electronics & Computer Science
The University of Southampton, UK

The University is not responsible for my opinions
Article: 7295
Subject: VHDL model for VME Slave Interface
From: Patrick Jones
Date: 22 Aug 1997 09:00:10 GMT
Links: << >>  << T >>  << A >>
Does anyone know of a supplier of tested & documented
VHDL code for a VME slave interface?

Thanks in advance,

Patrick



Article: 7296
Subject: Re: Xilinx PCI simulation problem...
From: "Austin Franklin" <dark9room@ix.netcom.com>
Date: 22 Aug 1997 15:00:46 GMT
Links: << >>  << T >>  << A >>
I finally found someone at Xilinx who 'explained' what was going on in the
simulations...

He said that the timing numbers that are in the simulation file are
'extremely conservative' and they have a perl script that I have to run on
the back annotated .xnf file to put in the 'databook' numbers.

So, I ran this script, and now my simulation works fine....but the actual
silicone still doesn't work behind a DEC bridge...and when we try to probe
it...it works...

So, next question is has anyone tested out their Xilinx PCI design behind a
PCI bridge?

Austin Franklin
darkroom@ix.netcom.com


Austin Franklin <dark9room@ix.netcom.com> wrote in article
<01bcacd2$220a6ef0$f249bacd@drt1>...
> The PCI bus spec guarantees 7ns setup and 0ns hold.  In order to meet the
> PCI spec, the 4k IOB flip flops have to be set to delay...which claims to
> require a 7ns setup, and 0ns hold.
> 
> Ok, all that seems fine...but when I generate a .vsm file and simulate
this
> using ViewSim and a command file that gives 7ns setup and 0ns hold, I get
> setup violations..claiming setup is 6.2, 10.2 is required.  Interestingly
> enough, it's not the 7ns setup that is causing the problem, it's the 0ns
> hold.  When I change the 0ns hold to 11, it works fine.
> 
> Now...my first guess is something is wrong with the timing model that
> Xilinx provides.  I have not gone into the .vsm file to see what it is
that
> is wrong, but I am about to.
> 
> What my question is, has anyone simulated a Xilinx PCI design using
ViewSim
> and using 7ns setup and 0ns hold....successfully?
> 
> Thanks,
> 
> Austin Franklin
> darkroom@ix.netcom.com
> 
> To reply to this post, remove the number from the reply address.
> 
> 
Article: 7297
Subject: Re: MaxPlusII from Altera.
From: spp@bob.eecs.berkeley.edu
Date: 22 Aug 1997 16:19:27 GMT
Links: << >>  << T >>  << A >>
MaxPlus2 works well in general but I have a few complaints:

(1) It has a habit of overwriting your design input files
(e.g. *.acf, maxplus2.ini).   So, for consistent results you 
must keep extra copies of these and write a script
that copies them into the relevant directories.

(2) After invocation, it will chug along for minutes before
it tells you "license in use".

(3) It does not always reserve device pins as you would
expect based on your design input.  (I think if you have
an undriven output pin, it will not reserve it.  Things
like that.)

Steve


Article: 7298
Subject: Re: ISP Stories
From: Botond Kardos <nobody@nowhere.com>
Date: Fri, 22 Aug 1997 13:13:27 -0400
Links: << >>  << T >>  << A >>
Steven J. Ackerman wrote:
> 
...
> 
> If you're using Lattice parts, you have to fully complete and debug
> your design before you lay out a circuit board. And the signal pinout
> will be completely arbitrary, not easily circuit board routable. If
> you think that you can go in and re-assign pinouts or make logic
> changes after the fact you may be in for a nasty suprise - the Lattice
> router may no longer be able to fit your design into the part ! I have
> found this to be the case if the part is over 50% utilized.
> 
...

   I have the bad habit that I design the PCB fisrt and assume that all
PLDs would fit. Altera FPGAs cause no trouble. But the puzzle game I
have to play with Lattice GLBs to fit in the given part can be very
annoying.

   Cheers,
   Botond

-- 
Kardos, Botond  -  at Innomed Medical Ltd. in Hungary
eMail: kardos@mail.matav.hu
phone/fax: (0036 1) 351-2934
fax: (0036 1) 321-1075
Article: 7299
Subject: Re: ISP Stories
From: Botond Kardos <nobody@nowhere.com>
Date: Fri, 22 Aug 1997 13:19:23 -0400
Links: << >>  << T >>  << A >>
Oops, I've almost forgot to mention that besides their routing
problems I do like Lattice PLDs. They're fast (consume a lot :< ) and
the ISP is quite easy with them.

-- 
Kardos, Botond  -  at Innomed Medical Ltd. in Hungary
eMail: kardos@mail.matav.hu
phone/fax: (0036 1) 351-2934
fax: (0036 1) 321-1075


Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search