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 4900

Article: 4900
Subject: Re: FPGA market overview
From: williamv@pacbell.net (William Vollrath)
Date: Thu, 26 Dec 1996 16:01:59 -0800
Links: << >>  << T >>  << A >>
In article <32B9E9F8.4690@aspect.com>, tjones@aspect.com says...
> Alon Hazay wrote:
> > 
> > I'm doing an overview of the FPGA market.
> > Does anyone know a review which compares the FPGA's of the leader
> > vendors in the market today(ALTERA,XILINX,ACTEL,ORCA etc.)?
> > 
> > Alon Hazay
> 
> 
> Please post any findings.  I would be particularly interested 
> in objective benchmarks for all the traditional parameters 
> (routability, etc.) as well as non-traditional, but otherwise 
> equally useful metrics such as ease-of-use of tools, changeability, 
> etc.
> 
> Thanks,
> Tom Jones
> 
Programmable Electronics Performance Corp. at http://www.prep.org are 
supposed to be an objective comparision of some of this.  
Their benchmarks may or may not compare everything you need and there is 
some controversy in the FPGA world about how well they compare the parts.  
It might aide your research, though.


-- 
William Vollrath
Do you want 1TByte in a 3490 cartridge at 15MBytes/sec?
LOTS Technology, Inc.
http://www.lasertape.com
Article: 4901
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 27 Dec 1996 00:02:14 GMT
Links: << >>  << T >>  << A >>
In article <32C2CDB6.2DD7@erols.com>,
   "Romuald K. Andraka" <randraka@erols.com> wrote:
>Wayne Turner wrote:

>> I find it hard to believe that all of these engineers designing flight and
>> space equipment have been hoodwinked all of these years.  When you are
>> talking about failure rates that must be calculated over a 20-year product
>> lifespan (i.e. aircraft), it is pretty accepted that ASICs and OTP devices 
>> are more reliable than SRAM.
>
>I'm not sure of your reference here.  Yes, a hardwired connection is 
>inherently more stable than a bit stored in SRAM.  There is no reason the 
>SRAM cell itself would be any less reliable than an ASIC or OTP device!  

We are talking about configuration data here, which an ASIC has none of, and 
therefore inherently more stable.  As for EEPROM or anti-fuse, as I've said, 
it generally requires a great deal more external energy to cause a change in 
state than in the equivalent design in an SRAM device.  I am not talking about 
RAM bits in each of the devices, I am talking about the CONFIGURATION DATA for 
each of these.

>Data (configurations) stored in EEPROM, EPROM and Antifuse based devices are 
>all susceptible to degradation over time and to incomplete programming.  I've 
>seen several instances of OTP and EPROM 
>type devices that verify correctly on the programmer then exhibit programming 
>failures within a few months or years.  The vendor analysis of these usually 
>indicate a weakly programmed bit.  ASICs are hardwired, so they will not 
>suffer from configuration data going bad.  However, they are susceptable to 
>process problems 

My discussion is relegated to failure rates (or potential failure rates) for 
these devices, and in a flight application where external radiation is a major 
factor, I still maintain that SRAM devices are not as reliable (due to being 
more susceptible to this radiation) than the same design in any other 
technology.  Is this not true?
Article: 4902
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 27 Dec 1996 00:05:39 GMT
Links: << >>  << T >>  << A >>
In article <32C2CFE8.1681@erols.com>,
   "Romuald K. Andraka" <randraka@erols.com> wrote:
>Wayne Turner wrote:
>
>> Please offer data to support the above statement.  Why would Xilinx SRAM 
>> parts be 10x more reliable than another SRAM part?

>I'm not sure about the numbers, but his statement is true regarding the 
>integrity of the data stored in the XILINX cell relative to data stored 
>in SRAM used in a microprocessor.  The XILINX SRAM cells are 
>deliberately designed as slow cells so that more energy is required to 
>cause an upset than a conventional SRAM memory (which is designed for 
>speed).

Are we going to compare apples to apples here?  Why would a Xilinx device be 
10X more reliable than an Altera device, for example?  We already went through 
the discussion of relative SRAM cell sizes between SRAM memory deives and SRAM 
-based logic devices.  My point of contention was it sounded as though Xilinx 
were 10X more reliable than other SRAM-based PROGRAMMABLE DEVICES.

Wayne
Article: 4903
Subject: Re: PCI Bus Based designs using FPGA's
From: williamv@pacbell.net (William Vollrath)
Date: Thu, 26 Dec 1996 16:08:32 -0800
Links: << >>  << T >>  << A >>
> > In our Fast Ethernet/ATM product we have decided to use the PCI bus as
> > our local interconnect. Have any you folks implemented PCI bus based
> > designs using some of the dense FPGA's(Altera 10K, Xilinx 4K)?  The
> > design is fairly intensive and we ruled out CPLD's due to its limited
> > gate count. 
> > 
> > I appreciate you sharing your experience.
> > 
> > Thank you
> > 
> > Sundar
> > 
> 
Altera has PCI macros available, too.  Lookm at their web site.  They've 
got free macros with limited functionality, and can connect you with 3rd 
parties that will sell you full-featured PCI interfaces that can 
synthesize into their parts.

(I suppose Xilinx can refer you to 3rd parties also, but my recent 
experience is with Altera).
-- 
William Vollrath
Do you want 1TByte in a 3490 cartridge at 15MBytes/sec?
LOTS Technology, Inc.
http://www.lasertape.com
Article: 4904
Subject: Help: Divider
From: SHIU PUN-HANG MR <mcshian@ee.mcgill.ca>
Date: Thu, 26 Dec 1996 23:04:59 -0500
Links: << >>  << T >>  << A >>
I am looking for a serial-input and serial-output divider.
Ideally,  such divider could give the either partial quotient or partial
reminder when the partial divisor and partial dividend is given serially
bit by bit. If there is such algorithm (close one), I would like to get
some reference for that divider.

Thank you 

pun
Article: 4905
Subject: Re: Exemplar's Leonardo on Linux
From: Zoltan Kocsi <zoltan@bendor.com.au>
Date: 28 Dec 1996 10:57:25 +1100
Links: << >>  << T >>  << A >>
phony@see.sig.for.real writes:

> 
> David Emrich <emrich@exemplar.com> wrote:
> 
> :Phil Ptkwt Kristin wrote:
> :> I've heard that Exemplar will release a Linux version of their Leonardo
> :> synthesis tool for Linux early next year.  Has anyone else heard this?
> :> It would be good if someone from Exemplar could comment on this.
> :
> :The port is complete for Leonardo 4.0.3.  Available in January 97.
> :

Any other info is available ?
The Exemplar webpage does not mention linux at all and if I remeber correctly,
on a PC platform Win3.1, Win95 or WinNT is listed as supported OS.
Search for the word linux retrieves no do documents.

Zoltan
Article: 4906
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Romuald K. Andraka" <randraka@erols.com>
Date: Fri, 27 Dec 1996 23:15:12 -0500
Links: << >>  << T >>  << A >>
Wayne Turner wrote:
> 
> In article <32C2CDB6.2DD7@erols.com>, 
> We are talking about configuration data here, which an ASIC has none of, and
> therefore inherently more stable.  As for EEPROM or anti-fuse, as I've said,
> it generally requires a great deal more external energy to cause a change in
> state than in the equivalent design in an SRAM device.  I am not talking about
> RAM bits in each of the devices, I am talking about the CONFIGURATION DATA for
> each of these.
> 

Yes, this is true, but as I stated if the configuration does get 
corrupted, it can be reloaded easily with the existing configuration 
logic for an SRAM based device.  If the OTP configuration gets upset, 
you're looking at a considerably more difficult recovery.  My point was 
that *I have* seen instances of OTP device programs being corrupted some 
time after a successful manufacturing test.

> >Data (configurations) stored in EEPROM, EPROM and Antifuse based devices are
> >all susceptible to degradation over time and to incomplete programming.  I've
> >seen several instances of OTP and EPROM
> >type devices that verify correctly on the programmer then exhibit programming
> >failures within a few months or years.  The vendor analysis of these usually
> >indicate a weakly programmed bit.  ASICs are hardwired, so they will not
> >suffer from configuration data going bad.  However, they are susceptable to
> >process problems
> 
> My discussion is relegated to failure rates (or potential failure rates) for
> these devices, and in a flight application where external radiation is a major
> factor, I still maintain that SRAM devices are not as reliable (due to being
> more susceptible to this radiation) than the same design in any other
> technology.  Is this not true?

Yes it is true.  However, the recovery from a corruption of 
the configuration is more difficult or even impossible in some 
situations for the OTP devices.  All in all, It really comes down to 
selecting the device that best fits the application, the designer's 
experience and proper design of fault tolerance.  Based on my comfort 
level and experience with both OTP and SRAM devices, I would choose the 
SRAM device with carefully designed fault tolerant logic over the OTP 
for safety critical applications.

-Ray Andraka P.E.  via remote
Chairman, the Andraka Consulting Group
401/884-7930  FAX 401/884-7950
mailto:randraka@ids.net
http://www.ids.net/~randraka
Article: 4907
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: Scott Kroeger <Scott.Kroeger@e-field.com>
Date: Sat, 28 Dec 1996 00:13:12 -0600
Links: << >>  << T >>  << A >>
Romuald K. Andraka wrote:
> 
> Wayne Turner wrote:
> >
> > In article <32C2CDB6.2DD7@erols.com>,
> > We are talking about configuration data here, which an ASIC has none of, and
> > therefore inherently more stable.  As for EEPROM or anti-fuse, as I've said,
> > it generally requires a great deal more external energy to cause a change in
> > state than in the equivalent design in an SRAM device.  I am not talking about
> > RAM bits in each of the devices, I am talking about the CONFIGURATION DATA for
> > each of these.
> >
> 
> Yes, this is true, but as I stated if the configuration does get
> corrupted, it can be reloaded easily with the existing configuration
> logic for an SRAM based device.  If the OTP configuration gets upset,
> you're looking at a considerably more difficult recovery.  My point was
> that *I have* seen instances of OTP device programs being corrupted some
> time after a successful manufacturing test.
> 
> > >Data (configurations) stored in EEPROM, EPROM and Antifuse based devices are
> > >all susceptible to degradation over time and to incomplete programming.  I've
> > >seen several instances of OTP and EPROM
> > >type devices that verify correctly on the programmer then exhibit programming
> > >failures within a few months or years.  The vendor analysis of these usually
> > >indicate a weakly programmed bit.  ASICs are hardwired, so they will not
> > >suffer from configuration data going bad.  However, they are susceptable to
> > >process problems
> >
> > My discussion is relegated to failure rates (or potential failure rates) for
> > these devices, and in a flight application where external radiation is a major
> > factor, I still maintain that SRAM devices are not as reliable (due to being
> > more susceptible to this radiation) than the same design in any other
> > technology.  Is this not true?
> 
> Yes it is true.  However, the recovery from a corruption of
> the configuration is more difficult or even impossible in some
> situations for the OTP devices.  All in all, It really comes down to
> selecting the device that best fits the application, the designer's
> experience and proper design of fault tolerance.  Based on my comfort
> level and experience with both OTP and SRAM devices, I would choose the
> SRAM device with carefully designed fault tolerant logic over the OTP
> for safety critical applications.


I agree with Ray, especially when you consider that some(many?) of the
EPROM, EEPROM
and Flash based devices simply copy the configuration data into internal
SRAM on powerup.
This isn't always mentioned in the documentation.  The end result is
device that's just as
susceptible to upsets as an SRAM only device, but without the inherent
configuration verification.

I had problems with Altera Classic (EP900) devices many years ago
because short spikes on Vcc
or too-slow Vcc ramp-up produced improper configuration.  That was
enough to send me running to Xilinx.

Regards,
Scott
Article: 4908
Subject: Re: I2C Bus Interface in FPGAs
From: "Bertrand" <ALSE@CSERVE>
Date: 28 Dec 1996 21:57:46 GMT
Links: << >>  << T >>  << A >>
Will :
I searched the philips site at www.semiconductors.philips.com for the
suggested 
keyword -and alikes-, but did not find the mentioned information...

Thanks in advance for any further detail.
-- 
Return address is invalid to defeat junk mail.
Please reply to : alse@compuserve.com.

--------------------------------------------------------
Wil Taphoorn <wil@tms.hobby.nl> wrote in article
<6NX7yRRJ9FB@tms.hobby.nl>...
> John L. Smith wrote on 20.12.96 about "Re: I2C Bus Interface in FPGAs":
> 
>  >I think that I2C may be of sufficient complexity and commercial
>  >interest that even if people have developed one, they are not willing
>  >to share.
> 
> Not if it's from the makers of...
> 
>  > However, if someone has developed something in the meantime, pls
>  > let me know too.
> 
> Yes, Aloys Schatorje (Philips The Netherlands) did... His examples for  
> both receiver and transmitter on a PLC42CA12 are written in SNAP
> 
> Try www.semiconductors.philips.com, try to search for 'iic_plc4'.
> 
> regards,
> wil
> --
> 
> 
Article: 4909
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 29 Dec 1996 04:22:35 GMT
Links: << >>  << T >>  << A >>
In article <32C49ED0.70CE@erols.com>,
   "Romuald K. Andraka" <randraka@erols.com> wrote:
>Wayne Turner wrote:
>> 
>> In article <32C2CDB6.2DD7@erols.com>, 
>> We are talking about configuration data here, which an ASIC has none of, 
and
>> therefore inherently more stable.  As for EEPROM or anti-fuse, as I've 
said,
>> it generally requires a great deal more external energy to cause a change 
in
>> state than in the equivalent design in an SRAM device.  I am not talking 
about
>> RAM bits in each of the devices, I am talking about the CONFIGURATION DATA 
for
>> each of these.
>> 
>
>Yes, this is true, but as I stated if the configuration does get 
>corrupted, it can be reloaded easily with the existing configuration 
>logic for an SRAM based device.  If the OTP configuration gets upset, 
>you're looking at a considerably more difficult recovery.  My point was 
>that *I have* seen instances of OTP device programs being corrupted some 
>time after a successful manufacturing test.
>
>> >Data (configurations) stored in EEPROM, EPROM and Antifuse based devices 
are
>> >all susceptible to degradation over time and to incomplete programming.  
I've
>> >seen several instances of OTP and EPROM
>> >type devices that verify correctly on the programmer then exhibit 
programming
>> >failures within a few months or years.  The vendor analysis of these 
usually
>> >indicate a weakly programmed bit.  ASICs are hardwired, so they will not
>> >suffer from configuration data going bad.  However, they are susceptable 
to
>> >process problems
>> 
>> My discussion is relegated to failure rates (or potential failure rates) 
for
>> these devices, and in a flight application where external radiation is a 
major
>> factor, I still maintain that SRAM devices are not as reliable (due to 
being
>> more susceptible to this radiation) than the same design in any other
>> technology.  Is this not true?
>
>Yes it is true.  However, the recovery from a corruption of 
>the configuration is more difficult or even impossible in some 
>situations for the OTP devices.  All in all, It really comes down to 
>selecting the device that best fits the application, the designer's 
>experience and proper design of fault tolerance.  Based on my comfort 
>level and experience with both OTP and SRAM devices, I would choose the 
>SRAM device with carefully designed fault tolerant logic over the OTP 
>for safety critical applications.

Boeing and Honeywell (I worked at Honeywell on the 777) seem to disagree with 
you.  If you do the NUMBERS, there is an obvious MTBF advantage of the EEPROM 
or anti-fuse device, or better yet, an ASIC, given equivalent flight 
environment conditions.

BTW, I liked your web page.  I happened to catch a link to it whilst surfing 
around on FPGA stuff...

Wayne
Article: 4910
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 29 Dec 1996 04:26:15 GMT
Links: << >>  << T >>  << A >>
In article <32C4BA75.79AF@e-field.com>,
   Scott Kroeger <Scott.Kroeger@e-field.com> wrote:
>Romuald K. Andraka wrote:
>> 
>> Wayne Turner wrote:
>> >
>> > In article <32C2CDB6.2DD7@erols.com>,
>> > We are talking about configuration data here, which an ASIC has none of, 
and
>> > therefore inherently more stable.  As for EEPROM or anti-fuse, as I've 
said,
>> > it generally requires a great deal more external energy to cause a change 
in
>> > state than in the equivalent design in an SRAM device.  I am not talking 
about
>> > RAM bits in each of the devices, I am talking about the CONFIGURATION 
DATA for
>> > each of these.
>> >
>> 
>> Yes, this is true, but as I stated if the configuration does get
>> corrupted, it can be reloaded easily with the existing configuration
>> logic for an SRAM based device.  If the OTP configuration gets upset,
>> you're looking at a considerably more difficult recovery.  My point was
>> that *I have* seen instances of OTP device programs being corrupted some
>> time after a successful manufacturing test.
>> 
>> > >Data (configurations) stored in EEPROM, EPROM and Antifuse based devices 
are
>> > >all susceptible to degradation over time and to incomplete programming. 
 I've
>> > >seen several instances of OTP and EPROM
>> > >type devices that verify correctly on the programmer then exhibit 
programming
>> > >failures within a few months or years.  The vendor analysis of these 
usually
>> > >indicate a weakly programmed bit.  ASICs are hardwired, so they will not
>> > >suffer from configuration data going bad.  However, they are susceptable 
to
>> > >process problems
>> >
>> > My discussion is relegated to failure rates (or potential failure rates) 
for
>> > these devices, and in a flight application where external radiation is a 
major
>> > factor, I still maintain that SRAM devices are not as reliable (due to 
being
>> > more susceptible to this radiation) than the same design in any other
>> > technology.  Is this not true?
>> 
>> Yes it is true.  However, the recovery from a corruption of
>> the configuration is more difficult or even impossible in some
>> situations for the OTP devices.  All in all, It really comes down to
>> selecting the device that best fits the application, the designer's
>> experience and proper design of fault tolerance.  Based on my comfort
>> level and experience with both OTP and SRAM devices, I would choose the
>> SRAM device with carefully designed fault tolerant logic over the OTP
>> for safety critical applications.
>
>
>I agree with Ray, especially when you consider that some(many?) of the
>EPROM, EEPROM
>and Flash based devices simply copy the configuration data into internal
>SRAM on powerup.
>This isn't always mentioned in the documentation.  The end result is

Other than the Altera FLEXLogic (previously named FlashLogic when 
Intel owned the product) devices, what devices do that?  I think they are the 
only combination FLASH/SRAM devices I know of, and I don't know of any EEPROM 
devices that do that at all.  Are there any?

Wayne
Article: 4911
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: jeff@wa1hco.mv.com (Jeff Millar)
Date: Sun, 29 Dec 1996 21:58:06 GMT
Links: << >>  << T >>  << A >>
waynet@indirect.com (Wayne Turner) wrote:

>In article <32C2CFE8.1681@erols.com>,
>   "Romuald K. Andraka" <randraka@erols.com> wrote:
>>Wayne Turner wrote:
>>
>>> Please offer data to support the above statement.  Why would Xilinx SRAM 
>>> parts be 10x more reliable than another SRAM part?
>
>>I'm not sure about the numbers, but his statement is true regarding the 
>>integrity of the data stored in the XILINX cell relative to data stored 
>>in SRAM used in a microprocessor.  The XILINX SRAM cells are 
>>deliberately designed as slow cells so that more energy is required to 
>>cause an upset than a conventional SRAM memory (which is designed for 
>>speed).
>
>Are we going to compare apples to apples here?  Why would a Xilinx device be 
>10X more reliable than an Altera device, for example?  We already went through 
>the discussion of relative SRAM cell sizes between SRAM memory deives and SRAM 
>-based logic devices.  My point of contention was it sounded as though Xilinx 
>were 10X more reliable than other SRAM-based PROGRAMMABLE DEVICES.
>
>Wayne

From my experience with high technology computing and electronics for
military aircraft systems, one need to take into account the size of
the susceptible portion of a chip from the point of view of cross
sectional area.  The most prevalent source of single event upsets
(SEU) comes from high energy cosmic rays, having enough energy to
upset even large SRAM cells.  So, designers have to look at the cross
sectional area of the cell to assess the probability of a cell getting
hit.  Some work one of my colleagues performed came to the conclusion
that DRAM cells, although upset by lower energies, the smaller cell
size make the chance of an upset much less likely.

Another factor to consider in this discussion...All components have
susceptibility  in the form of registers, local memories, state
machines, etc.  ASIC's aren't immune, especially those with large on
chip memory structures. Again one has to look at the overall cross
sectional area involved in storage.

BTW, I recall the cross sectional area is measured in units of barns
(as in couldn't the broadside of one).

jeff
Article: 4912
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: gherbert@crl.com (George Herbert)
Date: 29 Dec 1996 15:19:59 -0800
Links: << >>  << T >>  << A >>
I actually think that the best solution for this argument is to
point to the Xilinx GA products that they manufacture with the
programming done as a metal layer in the fab process... i.e.,
one time programmed during fab versions of their programmable
devices.  You work out the design with loadable ones and then
send them the program, and you get dedicated ASICs back.

I think we've seen that both OTP and FPGA have failure modes
which make flight-critical people nervous.  I certainly am 8-)


-george william herbert
Retro Aerospace
gherbert@crl.com

Article: 4913
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: Scott Kroeger <Scott.Kroeger@e-field.com>
Date: Sun, 29 Dec 1996 22:35:36 -0600
Links: << >>  << T >>  << A >>
Wayne Turner wrote:
> 
> In article <32C4BA75.79AF@e-field.com>,
>    Scott Kroeger <Scott.Kroeger@e-field.com> wrote:

> >I agree with Ray, especially when you consider that some(many?) of the
> >EPROM, EEPROM
> >and Flash based devices simply copy the configuration data into internal
> >SRAM on powerup.
> >This isn't always mentioned in the documentation.  The end result is
> 
> Other than the Altera FLEXLogic (previously named FlashLogic when
> Intel owned the product) devices, what devices do that?  I think they are the
> only combination FLASH/SRAM devices I know of, and I don't know of any EEPROM
> devices that do that at all.  Are there any?
> 
> Wayne

Wayne,

I think I overspoke (shame on me).  Now, if I stick to what I KNOW, I
should have said:

... consider that some of the EPROM and Flash based devices copy the ...

I was referring to Altera's FLEXLogic and Classic EPLD families.  At the
time
I had my problem with the EP900, Altera indicated that the EPROM storage
cell couldn't
be used to actually control a logic gate (at that time any such
structure would have been
much too slow).  This is no longer the case, but you can't tell how the
part works from the data sheet (you will not see any mention of the SRAM
based nature of Altera Classic EPLD's in the datasheet).  In some data
sheets you will see reference to power on reset functions that may take
microseconds to complete after Vcc stabilizes (and perhaps requirements
for monotonically increasing Vcc).  In the Altera Classic devices this
delay (and the requirement for monotonic Vcc rise) was caused by
EPROM->SRAM load.  Things may have changed since then, but
power-on-delays and monotonic Vcc rise specs still make me wonder.  I
suppose I've had my head buried in the Xilinx sand long enough to be
worth ignoring for my knowledge of the physics of other devices.

Regards,
Scott
Article: 4914
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Alvin E. Toda" <aet@lava.net>
Date: Sun, 29 Dec 1996 23:29:22 -1000
Links: << >>  << T >>  << A >>
On Fri, 27 Dec 1996, Romuald K. Andraka wrote:
.
.
> Yes it is true.  However, the recovery from a corruption of 
> the configuration is more difficult or even impossible in some 
> situations for the OTP devices.  All in all, It really comes down to 
> selecting the device that best fits the application, the designer's 
> experience and proper design of fault tolerance.  Based on my comfort 
.
.
> -Ray Andraka P.E.  via remote
> Chairman, the Andraka Consulting Group
> 401/884-7930  FAX 401/884-7950
> mailto:randraka@ids.net
> http://www.ids.net/~randraka
> 
> 
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). 
The treatment of the loading of the FPGA seems to be different in my 
opinion according to how you stand on the usefulness of the device. 
Those who deem the device less reliable seem to treat the loading of the 
device as part of the analysis of the reliability of the device-- it is hard to 
deny that having to insure almost perfect loading on every occasion is 
harder to justify than testing the other part just once to ensure proper 
programming.

On the other hand, those who are more aggressive in using the device 
seem treat the loading of the device as part of the system function of the 
board. Certainly it is hard to deny here that adding a lot of possible 
failure modes to system operation will contribute to less system 
reliability. Parts-wise the board they deem at least as safe as the SRAM's 
which may be on the board. Function-wise the re-programming of the 
part may be more cost effective than simply using a bigger one time 
programmable part-- the decrease in system reliability is worth the 
savings.

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. 



########################################################################
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: 4915
Subject: vga output
From: Christopher Price <pricecg@netlink.co.nz>
Date: 30 Dec 1996 11:11:48 GMT
Links: << >>  << T >>  << A >>
I've been searching for a while for vga signal generation 
implementations.

I have seen 
   i) IBM DAC/palette chip products (as the example)
   ii)  a vga hsynch, vsynch signal setup as part of a
        digital oscilloscope electronics article.
   iii) many video card descriptions/setups
 
But no general 'how to interface' information for 
setting up a basic framebuffer and vga signal generator.

  

Article: 4916
Subject: Re: I2C Bus Interface in FPGAs
From: Mika Iisakkila <iisakkil@alpha.hut.fi>
Date: 30 Dec 1996 13:52:10 +0200
Links: << >>  << T >>  << A >>
jimmiew@sprynet.com (James West) writes:
> About a year ago, I considered the same thing. I really looked, and I
> didn't find anything. But the more I thought about it, the more I
> realized it shouldn't be too hard to do from scratch. Good luck.

Sounds familiar, I've asked the same question here too :-)
No replies, so I did some dabbling on the project. Like you say, the
actual interface would not have been too difficult, but the system
will end up using quite a lot of registers because it involves bit
counters etc. I was targeting for Lattice ispLSI 1016 and dropped the
project after I realized that once the other functions and the (still
vaguely sketched) I2C interface were on the chip, there wouldn't be
registers left to store the data that was to be controlled via I2C.

Depending on the application, this would need a larger FPGA, or you
must be willing to ditch compatibility and cut some corners in the I2C
protocol (like forget about real ACKs, just clock the stuff in/out). A
Microwire-type clock/data/CS scheme would be much easier to implement
(not much more than a shift register, with maybe double buffering).
-- 
http://www.hut.fi/~iisakkil/                  - Rebi siit
Article: 4917
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 30 Dec 1996 15:17:54 GMT
Links: << >>  << T >>  << A >>


George Herbert <gherbert@crl.com> wrote in article
<5a6uav$4tu@crl4.crl.com>...
> I actually think that the best solution for this argument is to
> point to the Xilinx GA products that they manufacture with the
> programming done as a metal layer in the fab process... i.e.,
> one time programmed during fab versions of their programmable
> devices.  You work out the design with loadable ones and then
> send them the program, and you get dedicated ASICs back.
> 
> I think we've seen that both OTP and FPGA have failure modes
> which make flight-critical people nervous.  I certainly am 8-)
> 
> 
> -george william herbert
> Retro Aerospace
> gherbert@crl.com
> 
>

A couple questions/comments about this solution:

Many of the manufacturers offer this type of conversion, primarily for
economic reasons and other companies will convert designs for you.  

But, while the metal mask will get rid of the configuration memory
problems, latchup is still a concern (primarily for space systems).  Data
on the Xilinx XC3090, for instance, shows latchup threshold of ~4-7 which
is really low and the device failed during testing.  I haven't yet seen any
data on the 'hard-wired' products and don't know if they use a process and
design rules similar to the regular XC3090.

rk







 
Article: 4918
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 30 Dec 1996 15:31:48 GMT
Links: << >>  << T >>  << A >>


Jeff Millar <jeff@wa1hco.mv.com> wrote in article
<32c7e6a0.4519776@news.mv.net>...
> waynet@indirect.com (Wayne Turner) wrote:
> 
> >In article <32C2CFE8.1681@erols.com>,
> >   "Romuald K. Andraka" <randraka@erols.com> wrote:
> >>Wayne Turner wrote:
> >>
> >>> Please offer data to support the above statement.  Why would Xilinx
SRAM 
> >>> parts be 10x more reliable than another SRAM part?
> >
> >>I'm not sure about the numbers, but his statement is true regarding the

> >>integrity of the data stored in the XILINX cell relative to data stored

> >>in SRAM used in a microprocessor.  The XILINX SRAM cells are 
> >>deliberately designed as slow cells so that more energy is required to 
> >>cause an upset than a conventional SRAM memory (which is designed for 
> >>speed).
> >
> >Are we going to compare apples to apples here?  Why would a Xilinx
device be 
> >10X more reliable than an Altera device, for example?  We already went
through 
> >the discussion of relative SRAM cell sizes between SRAM memory deives
and SRAM 
> >-based logic devices.  My point of contention was it sounded as though
Xilinx 
> >were 10X more reliable than other SRAM-based PROGRAMMABLE DEVICES.
> >
> >Wayne
> 
> From my experience with high technology computing and electronics for
> military aircraft systems, one need to take into account the size of
> the susceptible portion of a chip from the point of view of cross
> sectional area.  The most prevalent source of single event upsets
> (SEU) comes from high energy cosmic rays, having enough energy to
> upset even large SRAM cells.  So, designers have to look at the cross
> sectional area of the cell to assess the probability of a cell getting
> hit.  Some work one of my colleagues performed came to the conclusion
> that DRAM cells, although upset by lower energies, the smaller cell
> size make the chance of an upset much less likely.
> 
> Another factor to consider in this discussion...All components have
> susceptibility  in the form of registers, local memories, state
> machines, etc.  ASIC's aren't immune, especially those with large on
> chip memory structures. Again one has to look at the overall cross
> sectional area involved in storage.
> 
> BTW, I recall the cross sectional area is measured in units of barns
> (as in couldn't the broadside of one).
> 
> jeff

while large SRAM cells may be upset by cosmic rays, making SRAM cells
effectively immune to SEU is a solved problem - there are a number of
techniques used to accomplish this including cross-strapped resistors,
epi-layer thickness control, and the use of different processes such as SOI
and SOS.  The voltage the cells are run at also have a large impact on what
the SEU performance will be.  The threshold LET of these hardened devices
is very, very high.  Of course, these devices are expensive and are, at
present to the best of my knowledge, limited to a 1 MBit density.  relating
this to ASICs, these same techniques have been applied to ASICs and are
available from a number of different manufacturers while also offering
hardness to total ionizing dose and latchup.

the comparison of sram vs. dram memory needs to be made very carefully as
it depends on the sram and dram selected.  most drams have a very low let
threshold - srams vary all over the place from very low to effectively
immune.  as mentioned above, there are techniques for hardeneing them. a
lot of the high density srams don't use a 6 transistor cell which hurts seu
performance.  another consideration is hard errors, which many drams and
some commercial srams are susceptible to.  for a good comparison, the
x-section vs. let curve should be used along with an environment model. 
i'd differ from your conclusion in general and from a design point of view,
i can easily design sram memories that are immune to seu - the comparable
dram design would be more complex since the dram controller is more complex
than a simple sequencer for srams, takes more power, is not always
available, and would need edac or some error-correcting/scrubbing mechanism
to achieve the same level of seu performance.  for very high density, large
memory arrays the drams are worth the extra trouble with perhaps
reed-solomon encoding.

also, for high energy cosmic rays, latchup is also a consideration.

rk
 
Article: 4919
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 30 Dec 1996 15:54:29 GMT
Links: << >>  << T >>  << A >>
> > Yes it is true.  However, the recovery from a corruption of 
> > the configuration is more difficult or even impossible in some 
> > situations for the OTP devices.  All in all, It really comes down to 
> > selecting the device that best fits the application, the designer's 
> > experience and proper design of fault tolerance.  Based on my comfort 
> .
> .
> > -Ray Andraka P.E.  via remote
> > Chairman, the Andraka Consulting Group
> > 401/884-7930  FAX 401/884-7950
> > mailto:randraka@ids.net
> > http://www.ids.net/~randraka

sram fpga's are interesting, potentially offer a lot of advantages, and
here's a couple of questions:

how are the effects of a reconfiguration blocked from affecting the
function of the board/system?

how is reconfiguration detected and how often is this done?  where is the
master copy stored?  is this in an "evil" :-) otp device such as a PROM
(pick your technology, fuse (nicrome, polysilicon), antifuse (dielectric,
amorphous silicon), eprom, eeprom)?  is the PROM used to store the original
program that is used for reload?  if it's an sram, what if the power
blinks?

can a configuration error result in any of the following and what are the
effects:
	
	1. turning an input cell into an output, overstressing drivers;
	2. turning a tri-state driver into a driver that is always enabled, again
overstress question;
	3. causing two fpga drivers onto the same routing track, again overstress
question;
	4. complete reboot of the fpga, causing it to reload and think it's in the
unprogrammed state, 
		at a very unfortunate time, negating any redundancy designed into the
fpga
	5. if jtag is used (and many manufacturers do not implement the optional
hard reset pin) and
		the tap controller is upset, can all of the pins turn to outputs,
preventing reloading?
	6. can a seu during the load process cause an illegal configuration to be
loaded, resulting
		in major problems? (equivalent to getting a parity error in the parity
error handling
		routine).
	7. turn a output into an input, having floating nodes on the board, with
perhaps high-speed
		oscillation and excessive currents generated, along with obvious system
faults?
	8. change the delay in an input (like the xilinx has) or output slew rate,
resulting in infrequent
		timing problems and system glitches?  this would be hard to detect if a
mechanism
		like a system watchdog timer is used to detect system failure.
	9. how does the "system" operate during a reload?  do external "safing
circuits" need to be
		added to the board design?

can a vcc glitch cause the fpga to think the power went off and go into a
initialization state but the system board did not? can an incorrect vcc
ramp up cause the fpga to be incorrectly initialized and not load properly?

rk
Article: 4920
Subject: Re: I2C Bus Interface in FPGAs
From: David <david@microcomm.co.nz>
Date: Mon, 30 Dec 1996 11:46:18 -0800
Links: << >>  << T >>  << A >>
Philips did do some PLD code for some of there older chip.
Look at Application Notes    AN036 , AN038 , AN039 
 
They are in the old PROGRAMMABLE LOGIC DEVICES DataBook " IC13" 1994 
Hope this helps  and let me know how it goes.

David Payne  " microcomm.co.nz "  

>Bertrand wrote:
> 
> Will :
> I searched the philips site at www.semiconductors.philips.com for the
> suggested
> keyword -and alikes-, but did not find the mentioned information...
> 
> Thanks in advance for any further detail.
> --
> Return address is invalid to defeat junk mail.
> Please reply to : alse@compuserve.com.
> 
> --------------------------------------------------------
> Wil Taphoorn <wil@tms.hobby.nl> wrote in article
> <6NX7yRRJ9FB@tms.hobby.nl>...
> > John L. Smith wrote on 20.12.96 about "Re: I2C Bus Interface in FPGAs":
> >
> >  >I think that I2C may be of sufficient complexity and commercial
> >  >interest that even if people have developed one, they are not willing
> >  >to share.
> >
> > Not if it's from the makers of...
> >
> >  > However, if someone has developed something in the meantime, pls
> >  > let me know too.
> >
> > Yes, Aloys Schatorje (Philips The Netherlands) did... His examples for
> > both receiver and transmitter on a PLC42CA12 are written in SNAP
> >
> > Try www.semiconductors.philips.com, try to search for 'iic_plc4'.
> >
> > regards,
> > wil
> > --
> >
> >
Article: 4921
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 30 Dec 1996 23:11:11 GMT
Links: << >>  << T >>  << A >>
In article <5a6uav$4tu@crl4.crl.com>, gherbert@crl.com (George Herbert) wrote:
>I actually think that the best solution for this argument is to
>point to the Xilinx GA products that they manufacture with the
>programming done as a metal layer in the fab process... i.e.,
>one time programmed during fab versions of their programmable
>devices.  You work out the design with loadable ones and then
>send them the program, and you get dedicated ASICs back.
>
>I think we've seen that both OTP and FPGA have failure modes
>which make flight-critical people nervous.  I certainly am 8-)

Altera has the same with their MPLD (mask-programmed-logic-device) that makes 
a masked version of your programmable device.  This is for both SRAM and 
EEPROM devices.

Wayne
Article: 4922
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 30 Dec 1996 23:11:39 GMT
Links: << >>  << T >>  << A >>
In article <5a6uav$4tu@crl4.crl.com>, gherbert@crl.com (George Herbert) wrote:
>I actually think that the best solution for this argument is to
>point to the Xilinx GA products that they manufacture with the
>programming done as a metal layer in the fab process... i.e.,
>one time programmed during fab versions of their programmable
>devices.  You work out the design with loadable ones and then
>send them the program, and you get dedicated ASICs back.
>
>I think we've seen that both OTP and FPGA have failure modes
>which make flight-critical people nervous.  I certainly am 8-)

Altera has the same with their MPLD (mask-programmed-logic-device) that makes 
a masked version of your programmable device.  This is for both SRAM and 
EEPROM devices.

Wayne
Article: 4923
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: waynet@indirect.com (Wayne Turner)
Date: 30 Dec 1996 23:15:46 GMT
Links: << >>  << T >>  << A >>
In article <Pine.BSI.3.91.961229231119.12091A-100000@malasada.lava.net>,
   "Alvin E. Toda" <aet@lava.net> wrote:
>On Fri, 27 Dec 1996, Romuald K. Andraka wrote:
>..
>..
>> Yes it is true.  However, the recovery from a corruption of 
>> the configuration is more difficult or even impossible in some 
>> situations for the OTP devices.  All in all, It really comes down to 
>> selecting the device that best fits the application, the designer's 
>> experience and proper design of fault tolerance.  Based on my comfort 
>..
>..
>> -Ray Andraka P.E.  via remote
>> Chairman, the Andraka Consulting Group
>> 401/884-7930  FAX 401/884-7950
>> mailto:randraka@ids.net
>> http://www.ids.net/~randraka
>> 
>> 
>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). 
>The treatment of the loading of the FPGA seems to be different in my 
>opinion according to how you stand on the usefulness of the device. 
>Those who deem the device less reliable seem to treat the loading of the 
>device as part of the analysis of the reliability of the device-- it is hard 
to 
>deny that having to insure almost perfect loading on every occasion is 
>harder to justify than testing the other part just once to ensure proper 
>programming.
>
>On the other hand, those who are more aggressive in using the device 
>seem treat the loading of the device as part of the system function of the 
>board. Certainly it is hard to deny here that adding a lot of possible 
>failure modes to system operation will contribute to less system 
>reliability. Parts-wise the board they deem at least as safe as the SRAM's 
>which may be on the board. Function-wise the re-programming of the 
>part may be more cost effective than simply using a bigger one time 
>programmable part-- the decrease in system reliability is worth the 
>savings.
>
>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
Article: 4924
Subject: Re: I2C Bus Interface in FPGAs
From: mma@rt66.com (Mark Aaldering)
Date: Tue, 31 Dec 1996 01:55:34 GMT
Links: << >>  << T >>  << A >>
     Dear John & Wil  (& others with interest)...
     
     My appologies for the difficulty in finding the application note
for   I2C in our 42VA12. We have recently launched a new family of
High  Speed (6nS), zero power (>100uA) CPLDs, and have replaced our
most  direct link to PLD information with our new CPLD web page
(www.coolpld.com). The application note for the 42VA12 I2C is still
obtainable - although via a circuitous route:
     
     
        Go To www.semiconductors.philips.com
     
        Then select 'Technical Documentation'
     
        Then Select 'PLDs'
     
     You are looking for AN036 "I2C Bus Expander" which is available
in Adobe  Acrobat format.
     
     Good Luck with your design, and don't hesitate to let us know if
we can  help!
     
     Mark Aaldering - Applications & Architecture Development Manager
     
     Philips Programmable Products Group
     9201 Pan American Freeway NE
     M/S - 08
     Albuquerque, NM, USA  87113
     Phone: 505-822-7835   Fax: 505-822-7804 
     Email: Mark.Aaldering@abq.sc.philips.com
     
          CoolRunner Support: coolpld@scs.philips.com 
          CoolRunner Web Site: www.coolpld.com
----------------------------------------------------------------------------
On Mon, 30 Dec 1996 11:46:18 -0800, David <david@microcomm.co.nz>
wrote:

>Philips did do some PLD code for some of there older chip.
>Look at Application Notes    AN036 , AN038 , AN039 
> 
>They are in the old PROGRAMMABLE LOGIC DEVICES DataBook " IC13" 1994 
>Hope this helps  and let me know how it goes.
>
>David Payne  " microcomm.co.nz "  
>
>>Bertrand wrote:
>> 
>> Will :
>> I searched the philips site at www.semiconductors.philips.com for the
>> suggested
>> keyword -and alikes-, but did not find the mentioned information...
>> 
>> Thanks in advance for any further detail.
>> --
>> Return address is invalid to defeat junk mail.
>> Please reply to : alse@compuserve.com.
>> 
>> --------------------------------------------------------
>> Wil Taphoorn <wil@tms.hobby.nl> wrote in article
>> <6NX7yRRJ9FB@tms.hobby.nl>...
>> > John L. Smith wrote on 20.12.96 about "Re: I2C Bus Interface in FPGAs":
>> >
>> >  >I think that I2C may be of sufficient complexity and commercial
>> >  >interest that even if people have developed one, they are not willing
>> >  >to share.
>> >
>> > Not if it's from the makers of...
>> >
>> >  > However, if someone has developed something in the meantime, pls
>> >  > let me know too.
>> >
>> > Yes, Aloys Schatorje (Philips The Netherlands) did... His examples for
>> > both receiver and transmitter on a PLC42CA12 are written in SNAP
>> >
>> > Try www.semiconductors.philips.com, try to search for 'iic_plc4'.
>> >
>> > regards,
>> > wil
>> > --
>> >
>> >



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