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
2017JanFebMarApr2017

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 4925

Article: 4925
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Alvin E. Toda" <aet@lava.net>
Date: Mon, 30 Dec 1996 23:25:35 -1000
Links: << >>  << T >>  << A >>
On 30 Dec 1996, Wayne Turner wrote:

> >It seems to be it's difficult to get something for nothing in this case. 
> >Because of the additional programing requirement for SRAM FPGA's, some 
> >decrease in system reliability should be expected. Perhaps each case
> >needs to have a separate analysis. 
> 
> This still does not address the fact that while SRAM memory devices are more 
> susceptible to upset, they also allow for design practices that can include 
> real-time error detection and correction, where this is not possible with the 
> configuration data in a SRAM programmable device.  The only option there (on 
> those devices that have the capability) is to constantly do a running CRC on 
> the config data, which may or may not catch an error before your functionality 
> has been compromised.
> 
> Wayne
> 
> 
I guess you may want to do what you would want to do with an SRAM fPGA
what you would like to do with an SRAM-- I am suggesting that what 
do for a flipped bit in SRAM you also do for an FPGA. If you want to
just detect an error when you read from ram, than you may just want to
detect an error on every clock cycle from the FPGA (this may require 
a redundent FPGA).

I'm not suggesting the above to be consistent. Rather, it's to avoid 
overkill and wasted effort. For example, it would be wasted effort to
try to develop a fail operational circuit for the FPGA, when we only 
detect an error in SRAM. In another example (I'm reaching), we see non-
parity rams now on PC's-- making it difficult to detect errors 
immediately, but perhaps later on writing to disk. In this case, why 
bother to design in circuits to detect FPGA loading errors-- why not
just let the system detect functional errors and warn the operator so
he or she can re-boot the PC?


########################################################################
Alvin E. Toda				aet@lava.net
sr. engineer				Phone: 1-808-455-1331

2-Sigma			  	WEB: http://www.lava.net/~aet/2-sigma.html
1363-A Hoowali St.
Pearl City, Hawaii, USA

Article: 4926
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 31 Dec 1996 14:51:23 GMT
Links: << >>  << T >>  << A >>
<somebody wrote>

> >I like the argument for SRAM FPGA's, such SRAM circuits are 
> >designed to be more stable than SRAM so they should be acceptable to 
> >the extent that SRAM is acceptable (if I may paraphrase this argument). 

i think there's a bit of over generalization here.  SRAM as a technology is
acceptable.  But it's application depends on the SRAM selected,
environment, and reliability levels (let's assume high).  SRAM performance
in radiation (concern brought up earlier) can vary from horrible to
essentially bulletproof.

architecturally, sram is easy to make relatively seu hard - this can be
done with hamming codes, reed-solomon codes, TMR, memory scrubbing, etc. 
Configuration memories in fpga's do not permit this.  also, memory elements
in fpga i/o cells are troublesome too in this regard and the cells are
important in many fpga's as they offer fast clock -> out performance and
small setup times and zero hold times, parameters critical to high
performance pipelined systems and pci interfaces, which are becoming more
popular.  

from a system performance point of view, errors in sram memories can be
detected and corrected during an access cycle; an error in an fpga
configuration memory would have a larger system impact (i.e., fault needs
to be blocked, system reloaded, find out what was missed and deal with it
which is often complex or impossible, etc).  Note that "stuck-bit" errors,
thought by some to be caused by a local total dose effect from a single
heavy ion, needs to be dealt with also.  for sram w/ reed-solomon, this is
a don't care since the error correcting power of the code is so strong. 
for hamming code or tmr-based systems, the system would continue to run
normally but with a loss of seu protection.  not so easy w/ fpga
configuration memories.  

another effect to be concerned with in these systems in general is single
event, multiple bit upset.  for srams, many commercial devices are designed
with the logical bits in a byte located physically next to each other; well
designed military or hi-rel memories will have the bits in the same byte
physically separate, permitting simple edac schemes to be very effective.

rk
Article: 4927
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 31 Dec 1996 15:11:31 GMT
Links: << >>  << T >>  << A >>
some comments are embedded and denoted w/ '*****'

rk

Alvin E. Toda <aet@lava.net> wrote in article
<Pine.BSI.3.91.961230231701.16021A-100000@malasada.lava.net>...
> On 30 Dec 1996, Wayne Turner wrote:
> 
> > >It seems to be it's difficult to get something for nothing in this
case. 
> > >Because of the additional programing requirement for SRAM FPGA's, some

> > >decrease in system reliability should be expected. Perhaps each case
> > >needs to have a separate analysis. 
> > 
> > This still does not address the fact that while SRAM memory devices are
more 
> > susceptible to upset, they also allow for design practices that can
include 
> > real-time error detection and correction, where this is not possible
with the 
> > configuration data in a SRAM programmable device.  The only option
there (on 
> > those devices that have the capability) is to constantly do a running
CRC on 
> > the config data, which may or may not catch an error before your
functionality 
> > has been compromised.
> > 
> > Wayne
> > 
> > 
> I guess you may want to do what you would want to do with an SRAM fPGA
> what you would like to do with an SRAM-- I am suggesting that what 
> do for a flipped bit in SRAM you also do for an FPGA. If you want to
> just detect an error when you read from ram, than you may just want to
> detect an error on every clock cycle from the FPGA (this may require 
> a redundent FPGA).


***** i just put up a note explaining, architecturally, that you can't do
the same
***** for fpga's that you do for srams for a number of different classes of
errors
***** which will affect the configuration memory.  for user-configured
memory elements
***** there are a number of techniques which can work and have been built
into space
***** flight hardware and tested in particle accelerators.  these
techniques are NOT
***** bulletproof but are very good but do present testability challenges,
which make
***** the device inferior to a true rad-hard memory cell, as can be found
in rad-hard/
***** hi-rel asics, for example.  
*****
***** a redundant fpga, used to detect an error on each clock cycle,
presents some
***** issues that need to be addressed:
*****
*****      1. what circuit is responsible for the comparison?  this needs
to be fast and
*****          compact and adds even more overhead.
*****      2. what about errors in the error checking circuitry if it's
implemented in, lets
*****          let's say, an fpga to be compact?  this just moves the
problem but we can
*****          make these circuits redundant too ...
*****      3. let's say you have two fpga, voters, the whole thing, how do
you test this
*****         stuff to ensure that all of the modes work and all?  how are
errors generated?
*****         this adds more circuitry and overhead and complexity.
*****      4. now, suppose we've solved 1, 2, and 3, and you get an error. 
what do you do
*****          about it?  which one is right?  which fpga do we reload?  do
we have to restart
*****          the entire system?  how do we keep the system running
(assume it's flying an
*****          aircraft full of people, or a spacecraft in critical
maneuvers, or it's doing something
*****          important with nasty weapons to either get someone or keep
you from getting
*****          squished)?  how do we get the fpga's back in sync?  

> 
> I'm not suggesting the above to be consistent. Rather, it's to avoid 
> overkill and wasted effort. For example, it would be wasted effort to
> try to develop a fail operational circuit for the FPGA, when we only 
> detect an error in SRAM. In another example (I'm reaching), we see non-
> parity rams now on PC's-- making it difficult to detect errors 
> immediately, but perhaps later on writing to disk. In this case, why 
> bother to design in circuits to detect FPGA loading errors-- why not
> just let the system detect functional errors and warn the operator so
> he or she can re-boot the PC?

***** i think this is a bit of a reach - i don't think the pc would qualify
in a 
***** safety critical application.  and 'only an error on sram' may have
***** serious effects that need to be blocked from affecting system 
***** performance.  and, if it is possible to configure, based on a
flipped-bit,
***** an input cell as an output, or a tri-state output as an always-on
output,
***** (and there's other examples) than the h/w set may be damaged and
***** overstressed (ex., metal migration) which can shorten the electronics
***** lifetime or perhaps immediately damage them.

rk
Article: 4928
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 31 Dec 1996 16:34:30 GMT
Links: << >>  << T >>  << A >>
In article <Pine.BSI.3.91.961230231701.16021A-100000@malasada.lava.net>,
   "Alvin E. Toda" <aet@lava.net> wrote:
>On 30 Dec 1996, Wayne Turner wrote:
>
>> This still does not address the fact that while SRAM memory devices are 
>> more susceptible to upset, they also allow for design practices that can 
>> include real-time error detection and correction, where this is not 
>> possible with the configuration data in a SRAM programmable device.  The 
>> only option there (on those devices that have the capability) is to 
>> constantly do a running CRC on the config data, which may or may not catch 
>> an error before your functionality has been compromised.
>> 
>> Wayne
>> 
>> 
>I guess you may want to do what you would want to do with an SRAM fPGA
>what you would like to do with an SRAM-- I am suggesting that what 
>do for a flipped bit in SRAM you also do for an FPGA. If you want to
>just detect an error when you read from ram, than you may just want to
>detect an error on every clock cycle from the FPGA (this may require 
>a redundent FPGA).

This would also work, and is similar to having redundant lock-step processors.

[...]

Wayne
Article: 4929
Subject: CFP Memory Technology Workshop
From: fmeyer@cs.tamu.edu (Jackie Meyer)
Date: 1 Jan 1997 03:25:43 GMT
Links: << >>  << T >>  << A >>
                    CALL FOR PAPERS AND PARTICIPATION

                    1997 IEEE INTERNATIONAL WORKSHOP
                ON MEMORY TECHNOLOGY, DESIGN AND TESTING

                           August 11-12, 1997

Submission deadline:  January 15, 1997

Send submissions to:

TECHNICAL PROGRAM CHAIR
Thomas Wik
LSI Logic, MS E-194
1501 McCarthy Blvd
Milpitas CA  95035, USA
408/954--4471; trw@lsil.com

Address general inquiries to:

GENERAL CHAIR
Fabrizio Lombardi
Computer Science MS 3112
Texas A&M University
College Station TX  77843, USA
409/845--5464; fax 847--8578
lombardi@cs.tamu.edu

The 1997 IEEE International Workshop on Memory Technology, Design and
Testing will be held at the Hilton Hotel and Towers, 300 Almaden Blvd,
San Jose, California, USA (408/287--2100), on August 11-12, 1997.

The workshop will include all aspects of memory design, process
technologies and testability related topics.  Memory circuit designs,
cell structures, fabrication processes, design architectures as
related to testing, verification and test methods for SRAM, DRAM,
Flash and Non-Volatile memories, EPROM, EEPROM, embedded memories,
logic-enhanced and FIFO memories, 3-D memories and content addressable
memories.  Some representative topics are:

  -  Memory fault modeling and test generation
  -  Built-in test and testable designs for memories
  -  Concurrent checking and memory fault diagnosis
  -  Quality and reliability issues
  -  Space applications and radiation hardening issues
  -  Memory failure and yield analysis
  -  High-speed, innovative designs
  -  Fault isolation, reconfiguration and repair
  -  Multiported, multibuffered memories
  -  Logic-enhanced and programmable memories
  -  Application-specific and embedded memories
  -  Multimegabit SRAMs and DRAMs
  -  CMOS, BiCMOS and bipolar designs for high yield and reliability

Authors please submit five (5) copies of an extended abstract of about
1000 words of original work on any aspect of memory technology, design
and testing to the Technical Program Chair.  Submissions should
include full names and affiliations of authors, contact information
and should indicate the intended presenter.

Submissions are due January 15, 1997.  Authors will be notified of
acceptance on March 31, 1997.  Final papers will be due May 15, 1997.
Presentations will be 30 minutes, inclusive of discussion.

Sponsored by:
  IEEE Computer Society
  Technical Committee on Test Technology
  Technical Committee on VLSI
In cooperation with:
  IEEE Solid-State Circuit Society
Article: 4930
Subject: Re: I2C Bus Interface in FPGAs
From: cong shiping <vn5s-cng@asahi-net.or.jp>
Date: Wed, 01 Jan 1997 20:28:35 +0900
Links: << >>  << T >>  << A >>
Stephan Gick wrote:
> 
> Hi all,
> does anybody know application notes of FPGA realizations of a I2C Bus
> interface?
> Stephan
> 
> ------------------------------------------------
> Stephan Gick
> Fraunhofer Institute for Integrated Circuits IIS
> Dep. Electronic Systems ESY
> Am Weichselgarten 3
> D-91058 Erlangen
> EMail: gck@iis.fhg.de
> -------------------------------------------------

I use BrookTree's video I/F chips develped some video capture cards . 
All of these chips have I2C bus I/F .As the I2C bus is very slowly , 
I always use software to control it . Use 2 bit of data bus , one for
SCL and one for SDA . 


Shiping Cong

vn5s-cng@asahi-net.or.jp
Article: 4931
Subject: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: "Austin Franklin" <darkroom@ix.netcom.com>
Date: 1 Jan 1997 21:09:17 GMT
Links: << >>  << T >>  << A >>
I just sat down to compile a clients design I made some changes to...and
WIR2XNF said everything was out of date.  Symbols, wir files, schematics,
everything!  It thought that 1/1/97 was earlier than 12/30/96!  Some
programmer bumbled....

Since the new Viewdraw doesn't write out a symbol or schematic unless it's
actually changed, I had to go through the entire project, symbol by symbol,
schematic by schematic, and 'change' something in each of them so it would
write out a new symbol/schematic/wir file.  For some reason, the check
program wasn't able to do this, probably cause it can't write symbols.

Anyone else have this problem?

Austin Franklin
..darkroom@ix.netcom.com.

Article: 4932
Subject: Re: What Does ASIC Stand For?
From: eswar@mega.megamed.com (Subramnyeswar Saladi)
Date: 1 Jan 1997 22:53:09 GMT
Links: << >>  << T >>  << A >>
John Cooley (jcooley@world.std.com) wrote:
: simonson@skopen.dseg.ti.com (Kevin M Simonson) writes:

: >     I'm feeling really ignorant right now.  Is there anyone who can tell
: >me what ASIC stands for?  Thanks.
: > 
: >                                     ---Kevin Simonson

: Aren't "ASICs" a brand of tennis shoe?   :^)

Why were the shoes named ASIC's ?


------------------------------------------------------------------------------
-- Eswar Saladi          email : eswar@megamed.com       pager : 408-381-6316
------------------------------------------------------------------------------
-- Home Page : http://home.aol.com/Anln
------------------------------------------------------------------------------
-- Favorite Quote : The essence of advancement is simplicity. --Dr. Zarkov
------------------------------------------------------------------------------
-- ALL VIEWS EXPRESSED ARE MINE AND MINE ALONE, NO COPYRIGHT VIOLATIONS
------------------------------------------------------------------------------
Article: 4933
Subject: Re: What Does ASIC Stand For?
From: gmiller@inca.co.nz (Gaven Miller)
Date: Thu, 02 Jan 1997 01:55:36 GMT
Links: << >>  << T >>  << A >>
Subramnyeswar Saladi (eswar@mega.megamed.com) wrote:
> John Cooley (jcooley@world.std.com) wrote:
> : simonson@skopen.dseg.ti.com (Kevin M Simonson) writes:

> : >     I'm feeling really ignorant right now.  Is there anyone who can tell
> : >me what ASIC stands for?  Thanks.
> : > 
> : >                                     ---Kevin Simonson

I'm not a hardware guy, but I have come to understand that ASIC means
"Application Specific Integrated Circuit". IOW, some electronics dedicated
to a specific task. 

> : Aren't "ASICs" a brand of tennis shoe?   :^)

> Why were the shoes named ASIC's ?


> ------------------------------------------------------------------------------
> -- Eswar Saladi          email : eswar@megamed.com       pager : 408-381-6316
> ------------------------------------------------------------------------------
> -- Home Page : http://home.aol.com/Anln
> ------------------------------------------------------------------------------
> -- Favorite Quote : The essence of advancement is simplicity. --Dr. Zarkov
> ------------------------------------------------------------------------------
> -- ALL VIEWS EXPRESSED ARE MINE AND MINE ALONE, NO COPYRIGHT VIOLATIONS
> ------------------------------------------------------------------------------


--

Quote For The Month:

"I wanna buy you some sunglasses - the kind you can see through"


Article: 4934
Subject: Re: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: fliptron@netcom.com (Philip Freidin)
Date: Thu, 2 Jan 1997 09:34:25 GMT
Links: << >>  << T >>  << A >>
In article <01bbf827$de5c8940$4cf65ecf@drt1> "Austin Franklin" <darkroom@ix.netcom.com> writes:
>I just sat down to compile a clients design I made some changes to...and
>WIR2XNF said everything was out of date.  Symbols, wir files, schematics,
>everything!  It thought that 1/1/97 was earlier than 12/30/96!  Some
>programmer bumbled....
>
>Since the new Viewdraw doesn't write out a symbol or schematic unless it's
>actually changed, I had to go through the entire project, symbol by symbol,
>schematic by schematic, and 'change' something in each of them so it would
>write out a new symbol/schematic/wir file.  For some reason, the check
>program wasn't able to do this, probably cause it can't write symbols.
>
>Anyone else have this problem?

Well, I didnt have the problem till I read your article, :-)

>
>Austin Franklin
>..darkroom@ix.netcom.com.

Bug confirmed, with the wir2xnf that comes with the newer sw, such as the
WV0 7.2 and I assume XACT 6.0.1 or 6.0.0

The bug does not exist in the older (and much faster) wir2xnf V 5.1 ( the
pre windows release of the WV tools, for DOS)

My quick fix is to do some global date changing. I use XTREE, which will
work unless you have long file name ( bigger than 8.3 style), but that's
another story of woe.

Anyway, went into the SYM directory, and changed all dates to current
date and time. Then went into SCH and did the same (but a few seconds 
later), and finally went into the WIR directory and changed dates and 
times (a few seconds later than the SCH dates/times. Whole thing took me 
about 10 seconds.

I hope you have a similar utility to do the same. Maybe there is a way to 
do global date change with the NT 4.0 explorer, but I don't know.

Philip Freidin.

P.S.  I don't know how much longer I can keep out of the ASIC versus FPGA
for configuration SRAM reliability fire fight :-) (it would be so much
more interesting if the protagonists actually knew something about the
configuration memory in these chips)

Article: 4935
Subject: 6809 VHDL MODEL
From: disacson@aol.com (DISACSON)
Date: 2 Jan 1997 15:39:48 GMT
Links: << >>  << T >>  << A >>

Does anyone know of a company that has the 6809 microprocessor in VHDL, or
even schematic form.?

Thanks
Daniel Isacson

Article: 4936
Subject: Foundation X-Block Support
From: Peter Drescher <peterd@pu.nz.fh-koeln.de>
Date: Thu, 02 Jan 1997 08:56:51 -0800
Links: << >>  << T >>  << A >>
I want to use a 64 Bit X-Block bus. 
Ther are only BUS_IF Interfaces to 32 Bit, so I try to edit a BUS_IF32
symbol. There is a problem with the x-block bus, because there are no 
brackets. (normal bus use brackets to define bus members B[0:31])

How can I edit a X-Block symbol ?

Peter
Article: 4937
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: Henry Spencer <henry@zoo.toronto.edu>
Date: Thu, 2 Jan 1997 19:43:49 GMT
Links: << >>  << T >>  << A >>
In article <Pine.BSI.3.91.961230231701.16021A-100000@malasada.lava.net> "Alvin E. Toda" <aet@lava.net> writes:
>...In another example (I'm reaching), we see non-
>parity rams now on PC's-- making it difficult to detect errors 
>immediately, but perhaps later on writing to disk...

Or perhaps never.  Non-parity PCs are a testament to the fact that many
people don't care whether their spreadsheets give correct results, so
long as they run fast.
-- 
"We don't care.  We don't have to.  You'll buy     |       Henry Spencer
whatever we ship, so why bother?  We're Microsoft."|   henry@zoo.toronto.edu
Article: 4938
Subject: Re: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: dando@phish.nether.net (Mann`y)
Date: 2 Jan 1997 19:51:01 GMT
Links: << >>  << T >>  << A >>

Austin Franklin (darkroom@ix.netcom.com) wrote:
: I just sat down to compile a clients design I made some changes to...and
: WIR2XNF said everything was out of date.  Symbols, wir files, schematics,
: everything!  It thought that 1/1/97 was earlier than 12/30/96!  Some
: programmer bumbled....

: Since the new Viewdraw doesn't write out a symbol or schematic unless it's
: actually changed, I had to go through the entire project, symbol by symbol,
: schematic by schematic, and 'change' something in each of them so it would
: write out a new symbol/schematic/wir file.  For some reason, the check
: program wasn't able to do this, probably cause it can't write symbols.


Hello:
    This is true especially when you are transferring the 
schematics and symbols from machine to another whose dates may
not be in sync. And wir2xnf will give you errors , a simple
way to get rid of this is to first go to the directory sch
and touch *.* and touch * and do the same in the sym directory
Now when you do a check appropriate wir files will be generated.
Because, the files are touched it means that the files are changed.
The program *touch* may be in the mks toolkit.

Hope this helps.

Article: 4939
Subject: Re: What Does ASIC Stand For?
From: "Geremy F." <geremy@ee.gatech.edu>
Date: Thu, 02 Jan 1997 16:03:26 -0500
Links: << >>  << T >>  << A >>
Application Specific Integrated Ciruit

-- 
>>>-Geremy-->
Article: 4940
Subject: Re: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: peter@nowhere.com
Date: Thu, 02 Jan 1997 21:30:47 GMT
Links: << >>  << T >>  << A >>
Philip,

Am I right in saying that if one uses the Remake All option (xmake -r
IIRC) then the problem does not arise, even with XACT6?

I always use remake-all, since 95% of the time is spent in APR/PPR
anyway.

Otherwise, one needs the little touch.exe utility which is available
among the many "unix-lookalike" DOS utility kits, or used to come with
DR-DOS.

Peter.

Peter.

Return address is invalid to help stop junk mail.
E-mail replies to z80@digiserve.com.
Article: 4941
Subject: Re: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: "Austin Franklin" <darkroom@ix.netcom.com>
Date: 2 Jan 1997 23:42:56 GMT
Links: << >>  << T >>  << A >>
Peter,

I don't think remake -all will do it.  You need to update the symbol files
first, then the schematics then the wir files.  I don't believe remake -all
will re-write the symbol files.

Austin Franklin
darkroom@ix.netcom.com

Article: 4942
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Alvin E. Toda" <aet@lava.net>
Date: Thu, 2 Jan 1997 15:09:55 -1000
Links: << >>  << T >>  << A >>
<Rich Katz wrote>

> > <Al Toda wrote>
> >I like the argument for SRAM FPGA's, such SRAM circuits are=20
> >designed to be more stable than SRAM so they should be acceptable to=20
> >the extent that SRAM is acceptable (if I may paraphrase this argument).=
=20

>i think there's a bit of over generalization here.  SRAM as a technology i=
s
>acceptable.  But it's application depends on the SRAM selected,
>environment, and reliability levels (let's assume high).  SRAM performance
>in radiation (concern brought up earlier) can vary from horrible to
>essentially bulletproof.

>architecturally, sram is easy to make relatively seu hard - this can be
>done with hamming codes, reed-solomon codes, TMR, memory scrubbing, etc.=
=20
>Configuration memories in fpga's do not permit this.  also, memory element=
s
>In fpga i/o cells are troublesome too in this regard and the cells are
>important in many fpga's as they offer fast clock -> out performance and
>small setup times....

I agree that in terms of robustness is the design, it=92s very different.
But, it=92s still too early to tell-- the jury is still out.

I think that when you talk of commercial over the counter type of=20
FPGA parts, you don=92t really concern your self with high rel=20
concerns. It=92s up to the persons that use these parts to address
these concerns for high rel designs.=20

The goals may be the same (error detection or failing
operational) but the means of getting there will certainly be up to
the designers. Commercial parts certainly don't have these
high rel requirements-- since it seems that consumer items which=20
have massive production volumes are pushing the technology
at the present time.=20

The technology is still quite young-- if designers think that
these SRAM FPGA parts have a role in high rel applications
than the onus of coming up with "bulletproof"-ness is on them.
It seems to me that simply using a bigger OTP part is more
cost effective that adding redundancy, error detection, or=20
recovery schemes to the design.=20

But ( as has been mentioned before by someone), it may be a=20
requirement that it may necessary to fix un-expected operational=20
problems in flight with programmable h/w changes as well=20
transmitting fixes to the flight program. However it may just be=20
cheaper to do another flight for the cost of launching a system=20
with all this wazoo stuff.


########################################################################
Alvin E. Toda=09=09=09=09aet@lava.net
sr. engineer=09=09=09=09Phone: 1-808-455-1331

2-Sigma=09=09=09  =09WEB: http://www.lava.net/~aet/2-sigma.html
1363-A Hoowali St.
Pearl City, Hawaii, USA

Article: 4943
Subject: Usb Cores ( synthesisable ) and ( simulation models )
From: dando@phish.nether.net (Mann`y)
Date: 3 Jan 1997 01:18:27 GMT
Links: << >>  << T >>  << A >>
Hello :
Do any of you know of a USB host interface (UHCI) design available for
an FPGA (any vendor will do)? 

I'd also appreciate any information on approximate gate count
implementations of the USB host interface.

Also, does anybody know of a peripheral side USB controller other than
Intel's 82930 and the impending Atmel part?

Thanks,
Manish
Article: 4944
Subject: Re: Usb Cores ( synthesisable ) and ( simulation models )
From: "Steven K. Knapp" <optmagic@ix.netcom.com>
Date: 3 Jan 1997 05:28:29 GMT
Links: << >>  << T >>  << A >>
I believe that CAE Technology has created a USB interface in Xilinx FPGA
devices.  See http://www.xilinx.com/products/logicore/alliance/cae/cae.htm.

Mann`y <dando@phish.nether.net> wrote in article
<5ahmp3$jm0@news.cic.net>...
> Do any of you know of a USB host interface (UHCI) design available for
> an FPGA (any vendor will do)? 

-- 
Steven K. Knapp
E-mail:  optmagic@ix.netcom.com
Programmable Logic Jump Station:  http://www.netcom.com/~optmagic

Article: 4945
Subject: Re: NT 4.0, ViewOffice 7.2 and Xilinx tools...96->97 problem...
From: fliptron@netcom.com (Philip Freidin)
Date: Fri, 3 Jan 1997 09:36:44 GMT
Links: << >>  << T >>  << A >>
In article <32d1192d.105679338@news.alt.net> peter@nowhere.com writes:
>Philip,
>
>Am I right in saying that if one uses the Remake All option (xmake -r
>IIRC) then the problem does not arise, even with XACT6?

No. This problem is sepparate from the behaviour of Remake All.

The intent of Remake All is to bypass the (not actually fuctional)
behaviour of XMAKE to avoid rebuilding things that are already built. In
the past (XC3000 and APR) this made some sense, as there were several
steps that could be skipped for pieces of the sub-hierarchy that had not
changed. With the current design flow, almost all the compute time
occurs in PPR, and because of the way that Wir2Xnf runs, bypassing stuff
can't even be done without a lot of (wasted) effort. 

The problem that Austin found (and I've confirmed) is a sepparate date 
checking system, local to wir2xnf itself, and manifested in the current, 
but not previous version.

When wir2xnf runs, it checks the following

WIR files are newer than the matching SCH file
SCH files are newer than the matching SYM file
All SYMs in an SCH file are older than the SCH file.

Unfortunately, the date handling for 1/1/97 (versus 12/31/96) seems to
be broken, and so wir2xnf thinks that there is a need for updating a
whole set of things. (I have no idea of the range of dates that are 
misshandled. )

Whenever things get really messy, I wipe the WIR directory, and
run:

	check -p projectname

which seems to a fairly good job of getting everything back the
way it should be.

There are multiple touch type programs available for dos that are 
shareware/public-domain.

Philip Freidin

>
>I always use remake-all, since 95% of the time is spent in APR/PPR
>anyway.
>
>Otherwise, one needs the little touch.exe utility which is available
>among the many "unix-lookalike" DOS utility kits, or used to come with
>DR-DOS.
>
>Peter.

Article: 4946
Subject: Re: What Does ASIC Stand For?
From: mwdf1@eng.cam.ac.uk (Mike Froggatt)
Date: 3 Jan 1997 10:28:04 GMT
Links: << >>  << T >>  << A >>
In article <32CC229E.5A3E@ee.gatech.edu>,
Geremy F. <geremy@ee.gatech.edu> wrote:
> Application Specific Integrated Ciruit

Or:

Arrgh! SPICE Isn't Converging!

Mike Froggatt
Thin Film Microelectronics Group
Cambridge University Engineering Department
Article: 4947
Subject: posting test
From: htytus@iglou1.iglou.com (Hul Tytus)
Date: Fri, 3 Jan 1997 10:45:36 GMT
Links: << >>  << T >>  << A >>
this is a test - please ignore.

Article: 4948
Subject: EPX880 & 8160 to Become Obsolete
From: Paul Walker <paul@walker.demon.co.uk>
Date: Fri, 3 Jan 1997 14:50:11 +0000
Links: << >>  << T >>  << A >>
Altera's PDN9625, dated 19 December 1996, announces last order and
delivery dates for the FX880 and 8160 devices (30 June, 97 and 98
respectively).

An excuse is offered that the flash process is at end of life. 

No apology (or cash) is offered for apparently causing esisting
customers to re lay out their PCBs to use a different device.

And what other Altera device has the flexible clocking, the ample output
enables, the 10-bit blocks because you always need a bit more than eight
bits for a byte, or the free and excellent development tools?

And what other CPLD from anywhere offers the RAM/ROM use of blocks in
the way these devices do? Logic that will go into a single look-up table
in this ROM would not fit into the whole of most 84-pin PLDs.

Perhaps the spec could be simplified a bit: I've not used the compare
function, and only used the RAM as a look-up table, so it could be just
flash or E^2, without the SRAM, as long as it's ISP.

On the other hand a parity function would be nice, and the fan-in of 26
was a few short on some designs. The clocking and output enabling are
superb and should be kept. And so should the pinout so we don't have to
redesign the PCBs.

Between now and June 98 gives Altera ample time to produce such a family
of devices.

Will they, or will they ditch their customers?

Follow-ups from Altera (and from the old INTEL PLD group) will be most
welcome.
-- 
Paul Walker                      4Links                      phone/fax
paul@walker.demon.co.uk          P O Box 816, Two Mile Ash    +44 1908
http://www.walker.demon.co.uk    Milton Keynes MK8 8NS, UK      566253
Article: 4949
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 3 Jan 1997 15:08:31 GMT
Links: << >>  << T >>  << A >>


Alvin E. Toda <aet@lava.net> wrote in article
<Pine.BSI.3.91.970102150915.21847A-100000@malasada.lava.net>...

>I think that when you talk of commercial over the counter type of 
>FPGA parts, you donít really concern your self with high rel 
>concerns. Itís up to the persons that use these parts to address
>these concerns for high rel designs. 

Note that many of the major FPGA vendors are selling their parts not as
commercial devices, but as qualified military/space grade devices.  And if
a particular technology such as SRAM is chosen, the basic question is can
the device be made to perform to the requirements of a 'critical
application' without too much pain or risk.  For example, pulling a Xilinx
data book off of the shelf , they have a chapter on "Military Logic Cell
Arrays" and process their devices to MIL-STD-883 Class B and participate in
the DESC SMD (standard military drawing) program and have assigned SMD
numbers for their parts.  Also, Harris had a program to design/build
radiation-hardened Xilinx compatibles (HS-XC3020MS) with guaranteed
radiation performance.  These are not being sold as commercial devices

>The goals may be the same (error detection or failing
>operational) but the means of getting there will certainly be up to
>the designers. Commercial parts certainly don't have these
>high rel requirements-- since it seems that consumer items which 
>have massive production volumes are pushing the technology
>at the present time. 

That is correct - the commercial world for the most part is driving things.
 But designers of military and space electronics will note that there is a
really, really strong push to use more modern, commercial off the shelf
(cots) devices.  And the original question was what technology should we
use: mask programmed asics, sram, antifuse (pick your kind), lpga?  and
architecturally, how does SRAM-based devices fit into a system with hi-rel
requirements?

>The technology is still quite young-- if designers think that
>these SRAM FPGA parts have a role in high rel applications
>than the onus of coming up with "bulletproof"-ness is on them.
>It seems to me that simply using a bigger OTP part is more
>cost effective that adding redundancy, error detection, or 
>recovery schemes to the design. 

At present I agree - although others have talked about SRAM-based devices
as a "solution."  And is there a possible scheme for making SRAM-based
FPGA's "bulletproof" for these types of applications?

>But ( as has been mentioned before by someone), it may be a 
>requirement that it may necessary to fix un-expected operational 
>problems in flight with programmable h/w changes as well 
>transmitting fixes to the flight program. However it may just be 
>cheaper to do another flight for the cost of launching a system 
>with all this wazoo stuff.

While it *may* be a requirement to fix un-expected operational problems in
flight, it is a requirement that what is launched/flown/etc. is reliable,
otherwise it wouldn't be a 'critical' application.  And there are many
architectural means of getting there with either mask programmable devices,
antifuse-based one's, etc.  And computers are built with many of the
functions in s/w (and with back-up power supplies, error-correcting
memories, parity on busses, etc., etc.) so that things can be modified
later.  FPGA-based designs can include programmable state machines,
micro-sequencers, etc. without having to use SRAM-based devices.  And user
memory can easily be protected.  Clearly, there are some high-performance
applications where these approaches won't work.  But, putting a ton of
patches on SRAM-based devices may clearly be much too complicated/expensive
in return for the benefits of reprogrammability.

So,

I am interested in designers coming up with the ways to make them
"bulletproof" as you mentioned.

rk

########################################################################
Alvin E. Toda				aet@lava.net
sr. engineer				Phone: 1-808-455-1331

2-Sigma			  	WEB: http://www.lava.net/~aet/2-sigma.html
1363-A Hoowali St.
Pearl City, Hawaii, USA


----------



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
2017JanFebMarApr2017

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