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 43725

Article: 43725
Subject: Lattice .ldf to VHDL help
From: Nicolas Matringe <nicolas.matringe@ipricot.com>
Date: Fri, 31 May 2002 12:36:27 +0200
Links: << >>  << T >>  << A >>
Hello all
I have a reference design which uses a Lattice CPLD to implement some glue logic
and I plan to use a Xilinx FPGA (which will have other functions) in my design.
I have translated the .ldf netlist in VHDL but I don't know how to verify that
it works as I expect (in fact I don't know what to expect). Does anyone know of
an ldf simulator?

-- 
Nicolas MATRINGE           IPricot European Headquarters
Conception electronique    10-12 Avenue de Verdun
Tel +33 1 46 52 53 11      F-92250 LA GARENNE-COLOMBES - FRANCE
Fax +33 1 46 52 53 02      http://www.IPricot.com/

Article: 43726
Subject: Re: State machine synthesis
From: "Robert O. Taniman" <bobchen74@yahoo.com>
Date: Fri, 31 May 2002 12:46:13 +0200
Links: << >>  << T >>  << A >>
Thanks for all replies.

Anyway, can  I just set this timing constraints in Leonardo Spectrum?
Actually I have tried to use it there, but I didn't notice a significant
improvement.

The stranger thing is I have synthesized and done place-and-route for my
receiver with nice result (Quartus reported that the max freq was 137 MHz, I
need 125 MHz). Then I switched to my transmitter, and it also went well. The
problem arose when I swtiched back to my receiver (coz I forgot to put a
shift register just for delaying purpose), but now it only achieved 100 MHz,
though I used the same procedure as far as I can remember.

With headache,
Robert

"Phil Connor" <philip_john_connor@hotmail.com> wrote in message
news:5204909a9ebad1ce3e59a429d49d05f1.89140@mygate.mailgate.org...
> "Robert O. Taniman" <bobchen74@yahoo.com> wrote in message
> news:abjqkg$ck0$1@dinkel.civ.utwente.nl
>
> Hi Robert,
>
> A second thought....
>
> Have you got some sort of timing constraint on the synthesis
> in the UCF file. If you're new to FPGAs its possible to overlook
> this requirement.
>
> Use something like...
>
> ####TIMING CONSTRAINTS #####
> NET "Clk" TNM_NET = "Clk";
> TIMESPEC "TS_Clk" = PERIOD "Clk" 50 MHz HIGH 50 %;
> ############################
>
> Naturally there are a mass of more complicated ways to put timing
> constraints on the synthesis. These can be found in the literature.
>
> If you are nearly there with your current timing then simply do
> multiple Place&Route with the constraints until the chip passes.
>
> If not follow the more radical Xilinx tips, previously referenced, to
> speed it up, the biggest improvement being "reduce the number of
> layers of logic".
>
> Any serious design needs to have all timing problems resolved. That is,
> if it's only for your own use as an amusement then you could live with
> the odd glitch, otherwise I'd fix it.
>
> Best regards
>
> Phil
>
>
>
>
>
>
> --
> Posted via Mailgate.ORG Server - http://www.Mailgate.ORG



Article: 43727
Subject: Re: IO simulations
From: rickman <spamgoeshere4@yahoo.com>
Date: Fri, 31 May 2002 09:02:33 -0400
Links: << >>  << T >>  << A >>
Austin,

Thanks for the reply.  But I don't see how Hspice using IBIS files
helps.  Hspice is a licensed product and is not cheap.  I was referring
to Spice simulators being available as freeware.  

So it is not likely that the Spartan II chips will get spice models
anytime soon?  



Austin Lesea wrote:
> 
> Rick,
> 
> The spice models are encrypted because they contain the foundry's
> technology files (highly proprietary).
> 
> A smart technologist can figure out exactly how a foundry is building
> their transistors from examining the details of the technology models.
> 
> We have encrypted spice models only for customers who sign NDAs, and
> who have the latest Avant! Hspice.
> 
> The IBIS models are developed from these spice models.
> 
> The latest version of Hspice accepts IBIS models, so the need for
> spice models has actually diminished.
> 
> Virtex II ushered in a new age of high speed designs, and as such, an
> effort was made to support such designs with better modeling
> resources.
> 
> Austin
> 
> 
> rickman wrote:
> 
> > Checking the Xilinx web site it looks like you can download IBIS
> > models
> > for various devices including the Spartan IIe devices.  However,
> > these
> > files do you no good if you don't have IBIS simulator which seem to
> > cost
> > even more than a logic or HDL simulator.
> >
> > I also found that there are spice models available for the VirtexII
> > parts.  But not for any other devices.  Since Spice is much more
> > available, any idea why the Spice models are not available for parts
> >
> > other than the VII?
> >
> > --
> >
> > Rick "rickman" Collins
> >
> > rick.collins@XYarius.com
> > Ignore the reply address. To email me use the above address with the
> > XY
> > removed.
> >
> > Arius - A Signal Processing Solutions Company
> > Specializing in DSP and FPGA design      URL http://www.arius.com
> > 4 King Ave                               301-682-7772 Voice
> > Frederick, MD 21701-3110                 301-682-7666 FAX

-- 

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Article: 43728
Subject: LFSR with 2^n instead of (2^n)-1
From: jetmarc@hotmail.com (jetmarc)
Date: 31 May 2002 07:06:01 -0700
Links: << >>  << T >>  << A >>
Hi.

I'm working on an address generator for a DMA controller in FPGA.
To save logic resources, I want to use non-sequential addresses.
Instead, I want to use an LFSR.

The range is 512 byte, or 2^9.  The great program "LFSR Testbench"
calculates a variety of possible solutons for this problem.  However,
the silicon-efficient solutions have a cycle of (2^n)-1.  I need
the whole 2^n to move all 512 bytes.

Since I am new to LFSRs, I checked the theory with pencil and
paper with a 2-bit LFSR.  "LFSR Testbench" suggests:

      feedback := LFSR(1) xor not(LFSR(0));
      LFSR(0) <= feedback;
      LFSR(1) <= LFSR(0) xor feedback;

My pencil & paper work suggests:

      bit1 <= bit0;
      bit0 <= not(bit1);

to count through 00,01,11,10.  The latter solution uses less
logic gates, and more important: less WIDE logic inputs.

With 9 bits, LFSR Testbench suggests a solution with 8 inputs
in one logic equation.  This kills FPGA logic area.

Is it possible to come up with a similar solution for 2^9 like
I did for 2^2?  I ask because I don't want to spend 2 days
trying, when it's known to be impossible.

Marc

Article: 43729
Subject: Re: about Configure FLEX10K10 with 89c51
From: wq00001@163.com (wuqiang)
Date: 31 May 2002 07:22:40 -0700
Links: << >>  << T >>  << A >>
I use .ttf file and give additional ten clocks cycles after all bits were sent.
Now I want to use .rbf file to try. if the .ttf file has some troubles?

Article: 43730
Subject: Re: about Configure FLEX10K10 with 89c51
From: wq00001@163.com (wuqiang)
Date: 31 May 2002 07:28:21 -0700
Links: << >>  << T >>  << A >>
Yes, I have given the additiional ten clock cycles, but the
Configure_Done signal was never to High.


"Falk Brunner" <Falk.Brunner@gmx.de> wrote in message news:<ad5ior$ue5ed$1@ID-84877.news.dfncis.de>...
> "wuqiang" <wq00001@163.com> schrieb im Newsbeitrag
> news:32ebf098.0205291650.25752ccf@posting.google.com...
> > I use mcu 89c51 to PS configure Flex10k10. I have followed the timing
> > in the altera pdf file, used the content of .ttf(Byte) file as
> > unsigned char array. Now NStatus signal is ok, but Conf_Done is never
> > to become hign to inform me to configure success. I check the board
> > and program again, but no error is found.
> > Who have finished similar work?
> 
> Did you generate additional configuration clock cycles after all bits have
> been loaded into the FPGA(4..6)? There are required to initiate the
> startup-sequence.

Article: 43731
Subject: Re: LFSR with 2^n instead of (2^n)-1
From: Dilip V. Sarwate <sarwate@uiuc.edu>
Date: Fri, 31 May 2002 14:36:23 GMT
Links: << >>  << T >>  << A >>
Look for de Bruijn sequences.  If all else (a.k.a. a search on the
Internet) fails, the book "Shift Register Sequences" by S. Golomb,
Holden-Day 1967; Aegean Press 198? discusses these in Chapter 6 .

--
 .-.     .-.     .-.     .-.     .-.     .-.     .-.
/ D \ I / L \ I / P \   / S \ A / R \ W / A \ T / E \
     `-'     `-'     `-'     `-'     `-'     `-'

Article: 43732
Subject: Re: IO simulations
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Fri, 31 May 2002 07:50:09 -0700
Links: << >>  << T >>  << A >>

--------------2DDC8936630EA09CDC8B2AD0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Rick,

I know exactly what you are referring to.  Rather than just say, "nope" I
took the opportunity to explain why the answer is what it is.

And freeware versions do not support encrypted files.

Sorry,  we support Hspice for spice for a very good, and sound business
reason.  The hotline has Hspice licenses, and we can solve problems, and
support customers this way (in addition to protecting our IP).  We also have
licenses for Cadence SpectraQuest (for the multi-gigbit transceivers), and
HyperLynx.  These are our recommended SI simulation tools, and are Xilinx
partners.

If you have a simple "what if" question, the hotline can do a simulation for
you (commercial customers please!  students are supported thru the Xilinx
University program http://www.xilinx.com/univ/index.htm) and email you the
results (similar to what I have done here on this board).

Austin


rickman wrote:

> Austin,
>
> Thanks for the reply.  But I don't see how Hspice using IBIS files
> helps.  Hspice is a licensed product and is not cheap.  I was referring
> to Spice simulators being available as freeware.
>
> So it is not likely that the Spartan II chips will get spice models
> anytime soon?
>
> Austin Lesea wrote:
> >
> > Rick,
> >
> > The spice models are encrypted because they contain the foundry's
> > technology files (highly proprietary).
> >
> > A smart technologist can figure out exactly how a foundry is building
> > their transistors from examining the details of the technology models.
> >
> > We have encrypted spice models only for customers who sign NDAs, and
> > who have the latest Avant! Hspice.
> >
> > The IBIS models are developed from these spice models.
> >
> > The latest version of Hspice accepts IBIS models, so the need for
> > spice models has actually diminished.
> >
> > Virtex II ushered in a new age of high speed designs, and as such, an
> > effort was made to support such designs with better modeling
> > resources.
> >
> > Austin
> >
> >
> > rickman wrote:
> >
> > > Checking the Xilinx web site it looks like you can download IBIS
> > > models
> > > for various devices including the Spartan IIe devices.  However,
> > > these
> > > files do you no good if you don't have IBIS simulator which seem to
> > > cost
> > > even more than a logic or HDL simulator.
> > >
> > > I also found that there are spice models available for the VirtexII
> > > parts.  But not for any other devices.  Since Spice is much more
> > > available, any idea why the Spice models are not available for parts
> > >
> > > other than the VII?
> > >
> > > --
> > >
> > > Rick "rickman" Collins
> > >
> > > rick.collins@XYarius.com
> > > Ignore the reply address. To email me use the above address with the
> > > XY
> > > removed.
> > >
> > > Arius - A Signal Processing Solutions Company
> > > Specializing in DSP and FPGA design      URL http://www.arius.com
> > > 4 King Ave                               301-682-7772 Voice
> > > Frederick, MD 21701-3110                 301-682-7666 FAX
>
> --
>
> Rick "rickman" Collins
>
> rick.collins@XYarius.com
> Ignore the reply address. To email me use the above address with the XY
> removed.
>
> Arius - A Signal Processing Solutions Company
> Specializing in DSP and FPGA design      URL http://www.arius.com
> 4 King Ave                               301-682-7772 Voice
> Frederick, MD 21701-3110                 301-682-7666 FAX



Article: 43733
Subject: Re: Do I have metastability issues?
From: John_H <johnhandwork@mail.com>
Date: Fri, 31 May 2002 15:26:29 GMT
Links: << >>  << T >>  << A >>
Without losing data or inserting junk in your 100 MHz multiplexed stream, the 25
MHz has to be related to the 100 MHz in *some* fashion.  If the relationship
between these clocks allows good clk-to-out at 25 MHz relative to the setup and
hold at 100 MHz, accounting for the skew and jitter between the two domains,
everything works.  If you don't know the relationship, only that they're phase
locked, a short FIFO would be the cleanest implementation with a "half full" as
the startup state so the FIFO doesn't over or under fill.


Ken Mac wrote:

> Hello folks,
>
> Recently I did a design (Xilinx Spartan-II) that required data to be passed
> from a 33MHz domain to a 425kHz domain.
>
> I did this using an Asynchronous FIFO and all was well.
>
> Now for another design:
>
> If I have 4 registers being clocked at 25MHz (with data coming in on each
> clock) and these 4 registers are to be multiplexed in turn into one register
> at 100MHz (i.e. the MUX and storage register are being clocked at 100MHz),
> do I need to worry about metastability?
>
> I would guess that I do, unless I used one of the CLKDLLs to divide the
> 100MHz clock down by 4 to 25MHz (thus making the 25MHz synchronised to the
> 100MHz in some way?) - would this ensure no metastabilty problems and allow
> me to just wire up the MUX to the 4 registers without any intervening
> Asynchronous FIFO or whatever?
>
> Thanks for your time,
>
> Ken


Article: 43734
Subject: Re: Nets in multiple schematics?
From: Kamal Patel <kamal.patel@xilinx.com>
Date: Fri, 31 May 2002 09:36:35 -0600
Links: << >>  << T >>  << A >>
Hi Loi,

The trick here is not to double-click on the net,
and then try to name it.  This will end up in the
error you have seen.  What you should do is to be
in net naming mode, then type in a net name in the
options box, and finally click on the net you want
named.

I hope this helps.

Regards,
Kamal

Loi Tran wrote:

> Hi,
> 
> I'm working with WEBpack 4.1 and I was wondering if there is a way to work 
> with multiple schematics on same level sheets instead of hierarchy sub-level 
> format.  Example, I want 2 separate sheets on the same level to have a common 
> signal line, and therefore a common net.  I tried it and a query tells me 
> they're two different nets altogether.  Please tell me if this can be done?
> 
> Thanks
> 
> LT
> 
> For the Xilinx guys who've been helping me out, thank you very much.
> 


Article: 43735
(removed)


Article: 43736
Subject: Re: LFSR with 2^n instead of (2^n)-1
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Fri, 31 May 2002 18:22:26 +0200
Links: << >>  << T >>  << A >>
"jetmarc" <jetmarc@hotmail.com> schrieb im Newsbeitrag
news:af3f5bb5.0205310606.49667136@posting.google.com...
> Hi.
>
> I'm working on an address generator for a DMA controller in FPGA.
> To save logic resources, I want to use non-sequential addresses.
> Instead, I want to use an LFSR.
>
> The range is 512 byte, or 2^9.  The great program "LFSR Testbench"
> calculates a variety of possible solutons for this problem.  However,
> the silicon-efficient solutions have a cycle of (2^n)-1.  I need
> the whole 2^n to move all 512 bytes.

I think there are two problems.

1.) You would like to save ressources.

2) You would like to find a special LFSR sequence.

The second one is a nice brain teaser. So far my comment on this.
For the first one, I dont think the problem (saving ressources) is worth the
trouble.
A 9 bit counter just takes 9 FlipFlops and a carry chain. Not much to save.
Even in a CPLD (which dont have carry chain by now) you still need 9
macrocells.

--
MfG
Falk





Article: 43737
Subject: Re: place and route simulation time
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Fri, 31 May 2002 18:28:59 +0200
Links: << >>  << T >>  << A >>
"Allan Herriman" <allan_herriman.hates.spam@agilent.com> schrieb im
Newsbeitrag news:3cf6e314.346890862@netnews.agilent.com...

> I still find synthesiser bugs even in purely synchronous designs.  I
> don't think gate level sims are going to go away any time soon.

Yes, there are still synthesizer bug. But would you hunt down such a bug
with a timing simulation of a full FPGA?
I woudnt.
Ray Andraka proposed to do post-mapping simulation to veryfy the sythesis
result, since a post-maping is (much?) faster than post P&R.

--
MfG
Falk





Article: 43738
Subject: Re: LFSR with 2^n instead of (2^n)-1
From: jsavard@ecn.aSBLOKb.caNADA.invalid (John Savard)
Date: Fri, 31 May 2002 16:31:14 GMT
Links: << >>  << T >>  << A >>
On Fri, 31 May 2002 14:36:23 GMT, Dilip V. Sarwate <sarwate@uiuc.edu>
wrote, in part:

>Look for de Bruijn sequences.  If all else (a.k.a. a search on the
>Internet) fails, the book "Shift Register Sequences" by S. Golomb,
>Holden-Day 1967; Aegean Press 198? discusses these in Chapter 6 .

Unfortunately, what a de Bruijn sequence does is it inserts an extra
zero into a standard LFSR sequence by adding a 0 after the 000...0001
state is reached, and that requires a logic gate as wide as the shift
register, which is exactly what the poster would like to avoid.

But perhaps there is a simpler way to generate them.

His example, with a NOT gate, strictly speaking is _not_ an LFSR; it
is provably impossible to construct an LFSR of n stages with a period
longer than 2^n-1 (specifically because the allowed operations in
LFSRs are such that once it's loaded with all zeroes, it stays there
forever).

Of course, he could always use an LFSR with one extra cell, and only
use the values in the other cell when that cell contained a 1. But
that would be extremely inefficient, allowing only half the bandwidth
to be used.

John Savard
http://home.ecn.ab.ca/~jsavard/index.html

Article: 43739
Subject: Re: LFSR with 2^n instead of (2^n)-1
From: John_H <johnhandwork@mail.com>
Date: Fri, 31 May 2002 16:46:36 GMT
Links: << >>  << T >>  << A >>
To implement a 9-bit LFSR for the DMA address generator, you need 9
register bits.  To implement a 9-bit counter you need 9 register bits.
No difference in logic resources from the flop standpoint.   Since you
need all 9 bits in parallel you can't use a Xilinx SRL16 primitive to
help save resources.

So...  What device architecture are you working with such that the
counter takes up more resources?  Saving 8 LUTs in an FPGA (using the
direct-in to the registers and leaving the LUTs for other combinatorial
functions) doesn't seem like something to worry about.

Efficiency is great but at what cost?


jetmarc wrote:

> Hi.
>
> I'm working on an address generator for a DMA controller in FPGA.
> To save logic resources, I want to use non-sequential addresses.
> Instead, I want to use an LFSR.
>
> The range is 512 byte, or 2^9.  The great program "LFSR Testbench"
> calculates a variety of possible solutons for this problem.  However,
> the silicon-efficient solutions have a cycle of (2^n)-1.  I need
> the whole 2^n to move all 512 bytes.
>
> Since I am new to LFSRs, I checked the theory with pencil and
> paper with a 2-bit LFSR.  "LFSR Testbench" suggests:
>
>       feedback := LFSR(1) xor not(LFSR(0));
>       LFSR(0) <= feedback;
>       LFSR(1) <= LFSR(0) xor feedback;
>
> My pencil & paper work suggests:
>
>       bit1 <= bit0;
>       bit0 <= not(bit1);
>
> to count through 00,01,11,10.  The latter solution uses less
> logic gates, and more important: less WIDE logic inputs.
>
> With 9 bits, LFSR Testbench suggests a solution with 8 inputs
> in one logic equation.  This kills FPGA logic area.
>
> Is it possible to come up with a similar solution for 2^9 like
> I did for 2^2?  I ask because I don't want to spend 2 days
> trying, when it's known to be impossible.
>
> Marc


Article: 43740
Subject: Re: Engineering Samples for free?
From: "Leon Heller" <leon_heller@hotmail.com>
Date: Fri, 31 May 2002 17:10:06 +0000 (UTC)
Links: << >>  << T >>  << A >>

"Rajat Karol" <rajatkarol@softhome.net> wrote in message
news:a78f7c6d.0205292217.78041f01@posting.google.com...
> Hi,
>
> I wanted to know whether Xilinx or Altera provides engineering samples
> for free?

I got a sample CPLD (one of the little ones) from Altera. I don't think I
requested it. I haven't got round to trying it yet, though. I might make a
PCB for it at home. I don't think Xilinx or Altera give away samples of
their larger chips on a routine basis, although I have got Altera samples
from my Altera distributor in the past.


>
> Also whether there is any difference in function/ quality ( in terms
> of perfomance or features or testability ) between engineering samples
> and production samples.

They should be identical, unless they are pre-production units. I've been
caught out by this, before now.

>
> Also are engineering samples available at a lesser rate ??

Not generally. They often cost more, because of the small quantities.

Generally, if you want free samples you are best off approaching your
distributor and seeing if they can help, or a manufacturer's  FAE, if you
know one.

Leon
--
Leon Heller, G1HSM leon_heller@hotmail.com
http://www.geocities.com/leon_heller
Low-cost Altera Flex design kit: http://www.leonheller.com









Article: 43741
Subject: Re: IO simulations
From: "Steve Casselman" <sc_no_spam@vcc.com>
Date: Fri, 31 May 2002 17:30:32 GMT
Links: << >>  << T >>  << A >>
Can't Xilinx write a behavioral spice model that gives 99% accuracy? It does
not have to be the 100% netlist version of the I/O cell just something
someone can use to model the I/O pin and do some simulation. Xilinx should
get used to putting out information and just saying "we don't support this
but..." If someone wants to reverse engineer the part they can do it. That
is no reason to withhold (or make expensive) information someone needs to do
a proper design with Xilinx parts. I think there is some information that
can be released and supported by the users at large. I just wish Xilinx
would start to understand the difference between what they need to keep
secret, what they can release that needs full support and information they
can give out and not support. For example you have the bitstream specs for
lookup tables, BRAMs and flops. But not routing. I can however get this
information from the XDL tools but only at great expense in time. I guess
I've just heard the "if we tell you we'll have to support you" or "if you
know this your going to hurt yourself" lines a little too often.


Steve


"Austin Lesea" <austin.lesea@xilinx.com> wrote in message
news:3CF78DA1.8E0ECD98@xilinx.com...
Rick,
I know exactly what you are referring to.  Rather than just say, "nope" I
took the opportunity to explain why the answer is what it is.
And freeware versions do not support encrypted files.
Sorry,  we support Hspice for spice for a very good, and sound business
reason.  The hotline has Hspice licenses, and we can solve problems, and
support customers this way (in addition to protecting our IP).  We also have
licenses for Cadence SpectraQuest (for the multi-gigbit transceivers), and
HyperLynx.  These are our recommended SI simulation tools, and are Xilinx
partners.
If you have a simple "what if" question, the hotline can do a simulation for
you (commercial customers please!  students are supported thru the Xilinx
University program http://www.xilinx.com/univ/index.htm) and email you the
results (similar to what I have done here on this board).
Austin

rickman wrote:
Austin,
Thanks for the reply.  But I don't see how Hspice using IBIS files
helps.  Hspice is a licensed product and is not cheap.  I was referring
to Spice simulators being available as freeware.
So it is not likely that the Spartan II chips will get spice models
anytime soon?
Austin Lesea wrote:
>
> Rick,
>
> The spice models are encrypted because they contain the foundry's
> technology files (highly proprietary).
>
> A smart technologist can figure out exactly how a foundry is building
> their transistors from examining the details of the technology models.
>
> We have encrypted spice models only for customers who sign NDAs, and
> who have the latest Avant! Hspice.
>
> The IBIS models are developed from these spice models.
>
> The latest version of Hspice accepts IBIS models, so the need for
> spice models has actually diminished.
>
> Virtex II ushered in a new age of high speed designs, and as such, an
> effort was made to support such designs with better modeling
> resources.
>
> Austin
>
>
> rickman wrote:
>
> > Checking the Xilinx web site it looks like you can download IBIS
> > models
> > for various devices including the Spartan IIe devices.  However,
> > these
> > files do you no good if you don't have IBIS simulator which seem to
> > cost
> > even more than a logic or HDL simulator.
> >
> > I also found that there are spice models available for the VirtexII
> > parts.  But not for any other devices.  Since Spice is much more
> > available, any idea why the Spice models are not available for parts
> >
> > other than the VII?
> >
> > --
> >
> > Rick "rickman" Collins
> >
> > rick.collins@XYarius.com
> > Ignore the reply address. To email me use the above address with the
> > XY
> > removed.
> >
> > Arius - A Signal Processing Solutions Company
> > Specializing in DSP and FPGA design      URL http://www.arius.com
> > 4 King Ave                               301-682-7772 Voice
> > Frederick, MD 21701-3110                 301-682-7666 FAX
--
Rick "rickman" Collins
rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.
Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX



Article: 43742
Subject: Re: State machine synthesis
From: Mike Treseler <mike.treseler@flukenetworks.com>
Date: Fri, 31 May 2002 11:03:19 -0700
Links: << >>  << T >>  << A >>


William Wallace wrote:


> However, my experience based on a few simple experiments is that using
> a single process based on transitions and clocks works as well
> real-estate-wise, and is much more readable.  This is with XST and
> FPGAExpress synthesizers.
> 
> Anybody have a different experience?


No. Single process machines work fine. It's style issue.

  -- Mike Treseler




Article: 43743
Subject: Re: How to add delay in fpga(spartan)?
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Fri, 31 May 2002 20:13:57 +0200
Links: << >>  << T >>  << A >>
"samlin" <gammalin@seed.net.tw> schrieb im Newsbeitrag
news:ad8bp5$t07$1@news.seed.net.tw...
> hi...
> I have a problem...
> if I use a delay in fpga...
> How can I do?

What kind of delay do you need? How much delay do you need?

--
MfG
Falk





Article: 43744
Subject: Re: IO simulations
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Fri, 31 May 2002 11:43:53 -0700
Links: << >>  << T >>  << A >>
Steve,

Sorry you feel this way.

But as you well point out, keeping our hard earned IP confidential is only part
of the issue.  The other part is the support.

Right now our support is 'world class'.  It doesn't happen by accident.  What
you suggest would fragment resources, cause more customer issues.

We are daily trying to make FPGAs a "pushbutton" experience for the average
user.  My holy grail is to reduce the ever more difficult signal integrity
issues to something that is closer to the experience of the HDL to bitstream
flow, and less something accomplished by "experts" with magic wands.

There is no such thing as 'unsupported' documentation.  Even these emails have
led to support issues.

If this philosophy (in part) helps create a market for Xilinx experts, who make
a living from consulting in the (more) optimal use of the FPGAs, I would hardly
lobby to have Xilinx put them out of business.....

Austin


Steve Casselman wrote:

> Can't Xilinx write a behavioral spice model that gives 99% accuracy? It does
> not have to be the 100% netlist version of the I/O cell just something
> someone can use to model the I/O pin and do some simulation. Xilinx should
> get used to putting out information and just saying "we don't support this
> but..." If someone wants to reverse engineer the part they can do it. That
> is no reason to withhold (or make expensive) information someone needs to do
> a proper design with Xilinx parts. I think there is some information that
> can be released and supported by the users at large. I just wish Xilinx
> would start to understand the difference between what they need to keep
> secret, what they can release that needs full support and information they
> can give out and not support. For example you have the bitstream specs for
> lookup tables, BRAMs and flops. But not routing. I can however get this
> information from the XDL tools but only at great expense in time. I guess
> I've just heard the "if we tell you we'll have to support you" or "if you
> know this your going to hurt yourself" lines a little too often.
>
> Steve
>
> "Austin Lesea" <austin.lesea@xilinx.com> wrote in message
> news:3CF78DA1.8E0ECD98@xilinx.com...
> Rick,
> I know exactly what you are referring to.  Rather than just say, "nope" I
> took the opportunity to explain why the answer is what it is.
> And freeware versions do not support encrypted files.
> Sorry,  we support Hspice for spice for a very good, and sound business
> reason.  The hotline has Hspice licenses, and we can solve problems, and
> support customers this way (in addition to protecting our IP).  We also have
> licenses for Cadence SpectraQuest (for the multi-gigbit transceivers), and
> HyperLynx.  These are our recommended SI simulation tools, and are Xilinx
> partners.
> If you have a simple "what if" question, the hotline can do a simulation for
> you (commercial customers please!  students are supported thru the Xilinx
> University program http://www.xilinx.com/univ/index.htm) and email you the
> results (similar to what I have done here on this board).
> Austin
>
> rickman wrote:
> Austin,
> Thanks for the reply.  But I don't see how Hspice using IBIS files
> helps.  Hspice is a licensed product and is not cheap.  I was referring
> to Spice simulators being available as freeware.
> So it is not likely that the Spartan II chips will get spice models
> anytime soon?
> Austin Lesea wrote:
> >
> > Rick,
> >
> > The spice models are encrypted because they contain the foundry's
> > technology files (highly proprietary).
> >
> > A smart technologist can figure out exactly how a foundry is building
> > their transistors from examining the details of the technology models.
> >
> > We have encrypted spice models only for customers who sign NDAs, and
> > who have the latest Avant! Hspice.
> >
> > The IBIS models are developed from these spice models.
> >
> > The latest version of Hspice accepts IBIS models, so the need for
> > spice models has actually diminished.
> >
> > Virtex II ushered in a new age of high speed designs, and as such, an
> > effort was made to support such designs with better modeling
> > resources.
> >
> > Austin
> >
> >
> > rickman wrote:
> >
> > > Checking the Xilinx web site it looks like you can download IBIS
> > > models
> > > for various devices including the Spartan IIe devices.  However,
> > > these
> > > files do you no good if you don't have IBIS simulator which seem to
> > > cost
> > > even more than a logic or HDL simulator.
> > >
> > > I also found that there are spice models available for the VirtexII
> > > parts.  But not for any other devices.  Since Spice is much more
> > > available, any idea why the Spice models are not available for parts
> > >
> > > other than the VII?
> > >
> > > --
> > >
> > > Rick "rickman" Collins
> > >
> > > rick.collins@XYarius.com
> > > Ignore the reply address. To email me use the above address with the
> > > XY
> > > removed.
> > >
> > > Arius - A Signal Processing Solutions Company
> > > Specializing in DSP and FPGA design      URL http://www.arius.com
> > > 4 King Ave                               301-682-7772 Voice
> > > Frederick, MD 21701-3110                 301-682-7666 FAX
> --
> Rick "rickman" Collins
> rick.collins@XYarius.com
> Ignore the reply address. To email me use the above address with the XY
> removed.
> Arius - A Signal Processing Solutions Company
> Specializing in DSP and FPGA design      URL http://www.arius.com
> 4 King Ave                               301-682-7772 Voice
> Frederick, MD 21701-3110                 301-682-7666 FAX


Article: 43745
Subject: Re: Time for a new computer. Suggestions?
From: Ray Andraka <ray@andraka.com>
Date: Fri, 31 May 2002 19:32:24 GMT
Links: << >>  << T >>  << A >>
Even the PCMCIA stuff doesn't do too well under NT.

rickman wrote:

> Instead of disabling it in the BIOS, why not make it removable with a
> slide in cartdrige or use a PCMCIA adapter?  I guess with the slide in
> cartdrige you still have to take it out of the BIOS unless they provide
> software to automatically do that as with the the PCMCIA interfaces.
>
> Ray Andraka wrote:
> >
> > I currently do my backups onto a drive on another machine.  The hard part is
> > remembering to do the copy.   I don't trust one machine to hold the data
> > uncorrupted.  I had a case a year or so ago where my raid card bellied up and took
> > out the data on the raid array.  I am still concerned about a virus getting into the
> > network and wiping both, so I keep not-frequent-enough-backups of the accounting and
> > project files onto CDs.  I've dabbled with tapes a couple of times, but frankly have
> > never had much success with them.  I had attempted to use an IDE drive on my machine
> > which I did disk images too on a frequent basis, and intended to disable in the BIOS
> > except when doing the back-up.  Good intention, but too inconvenient to keep it off
> > line.
> >
> > Nicholas Weaver wrote:
> >
> > > In article <3CEE74F0.1CC15C51@yahoo.com>,
> > > rickman  <spamgoeshere4@yahoo.com> wrote:
> > > >Hal Murray wrote:
> > > >>
> > > >> >Agreed,  The raid is in the current system for reliability.  I've suffered 3
> > > >> >or 4 disk failures over the years and the time spent recovering from backups
> > > >> >is well worth the extra cost of having a raid system.
> > > >>
> > > >> Don't forget to back things up (somehow) anyway.  RAID won't protect
> > > >> you from software/mushware/operator errors.
> > > >
> > > >Just to add my two cents worth.  I use a manual disk mirror approach for
> > > >protection against HW, SW and "mushware" issues.  The second HD is an
> > > >exact duplicate of the first and is updated on a regular basis.  We do
> > > >this once a week, but the weak point in the system is the operator
> > > >reliability.  If I forget, I have lost some of my protection.
> > > >
> > > >By not having automatic duplication, we have some protection against
> > > >accidentally deleted file and other issues from software installation
> > > >and such.  If we decide the backup is eaiser to work with than the
> > > >messed up current drive, we swap the drives and copy the backup to the
> > > >old original.
> > > >
> > > >This has saved us in some situations where a software crash wiped out a
> > > >significant part of the OS, once when we were infected with a virus and
> > > >many times when we did something to a file that we regretted.
> > > >
> > > >So RAID may not be the best option if you have the discipline to do your
> > > >backups.
> > >
> > > My 2 cents worth:
> > >
> > > Retrospect.  It is a wonderful mac backup tool, doing both full and
> > > incremental backups, both on the machine and to other client machines
> > > over the network.  My dad has had it save his bacon on his production
> > > mac system many times: he uses both a nightly incremental backup of
> > > everything and a nightly copying of critical files onto other drives.
> > >
> > > They finally have a Windows version as well, which if it works half as
> > > good as the Mac version, makes it worth its weight in gold.
> > > --
> > > Nicholas C. Weaver                                 nweaver@cs.berkeley.edu
> >
> > --
> > --Ray Andraka, P.E.
> > President, the Andraka Consulting Group, Inc.
> > 401/884-7930     Fax 401/884-7950
> > email ray@andraka.com
> > http://www.andraka.com
> >
> >  "They that give up essential liberty to obtain a little
> >   temporary safety deserve neither liberty nor safety."
> >                                           -Benjamin Franklin, 1759
>
> --
>
> Rick "rickman" Collins
>
> rick.collins@XYarius.com
> Ignore the reply address. To email me use the above address with the XY
> removed.
>
> Arius - A Signal Processing Solutions Company
> Specializing in DSP and FPGA design      URL http://www.arius.com
> 4 King Ave                               301-682-7772 Voice
> Frederick, MD 21701-3110                 301-682-7666 FAX

--
--Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930     Fax 401/884-7950
email ray@andraka.com
http://www.andraka.com

 "They that give up essential liberty to obtain a little
  temporary safety deserve neither liberty nor safety."
                                          -Benjamin Franklin, 1759



Article: 43746
Subject: Re: LFSR with 2^n instead of (2^n)-1
From: newman5382@aol.com (newman)
Date: 31 May 2002 13:10:04 -0700
Links: << >>  << T >>  << A >>
> Is it possible to come up with a similar solution for 2^9 like
> I did for 2^2?  I ask because I don't want to spend 2 days
> trying, when it's known to be impossible.
> 
> Marc

A discussion of 2^N sequences is on page 181 of the HDL Chip Design
book by Douglas J Smith.  If you are on a tight budget, I once saw
someone reading this book at Barnes and Nobles if you get my drift.

Newman

Article: 43747
Subject: Re: How to add delay in fpga(spartan)?
From: to ja <adres@email.com>
Date: Fri, 31 May 2002 20:55:52 GMT
Links: << >>  << T >>  << A >>
On Fri, 31 May 2002 20:13:57 +0200, "Falk Brunner"
<Falk.Brunner@gmx.de> wrote:

>> I have a problem...
>> if I use a delay in fpga...
>> How can I do?
>
>What kind of delay do you need? How much delay do you need?

not much... just a couple of days...

Article: 43748
Subject: Re: IO simulations
From: "Steve Casselman" <sc_no_spam@vcc.com>
Date: Fri, 31 May 2002 22:45:42 GMT
Links: << >>  << T >>  << A >>
First I want to say I really like most of the people I meet at Xilinx but as
the systems you try and field become more complicated your going to need to
give up more information or grind to a halt. I look at it as a lack of over
all confidence in the ablities of your users. It reminds of Clive whats his
name form Altera who stood up and said "we make our tools for morons because
morons use our tools." I've had this talk with lots of people and it is like
your trying to make a pushbutton automobile. Imagine a car with only one
button and all the software is written by microsoft but they won't tell
anyone how it works or how to fix it when it breaks... I had to laugh when a
high ranking Xilinx exec said "we are proud to have the best place and route
for our parts in the industry." Last time a company "broke the code" they
came up with a place and route tool 10x better than the Xilinx in house
tool. The Xilinx president at the time bought the company and was eventually
fired for it.   But it was this tool devloped outside of Xilinx that gave
Xilinx a big jump over Altera. It is like this I/O issue you don't give up
anything by making a behavioral spice model for the I/Os.  You know it's not
that I dislike anyone or am mad I just feel like Richard Stallman sometimes,
just give me the code and I'll fix my own problem. Ever think about open
sourcing the code for your EDA tools. What do you have 500 software
engineers? Open source your tools and you'll have 5000 engineers working for
you and you don't have to pay them.  Just do it like Sun did with Java.
Anyone who wants to steal your IP can do it anyway.

I think just the little stuff you've done so far has been good. Look at XDL
that is great stuff. The docs on that have not been changed since the 4000
series.  And of course the advanced user guide for Virtex. The guy who first
wrote that up almost got fired for it.

So here is what I think. The company that starts to make it easy for other
companies to create tools for it's parts will be the winner. If Altera or
Xilinx would publish everything people would use that information to make
better tools. Yes the FPGA company might loose 2% in EDA sales but I think
they would make this up in part sales by expanding their market into areas
where they don't have the man power to address issues.

Have a great weekend!!

Steve


"Austin Lesea" <austin.lesea@xilinx.com> wrote in message
news:3CF7C469.98890A30@xilinx.com...
> Steve,
>
> Sorry you feel this way.
>
> But as you well point out, keeping our hard earned IP confidential is only
part
> of the issue.  The other part is the support.
>
> Right now our support is 'world class'.  It doesn't happen by accident.
What
> you suggest would fragment resources, cause more customer issues.
>
> We are daily trying to make FPGAs a "pushbutton" experience for the
average
> user.  My holy grail is to reduce the ever more difficult signal integrity
> issues to something that is closer to the experience of the HDL to
bitstream
> flow, and less something accomplished by "experts" with magic wands.
>
> There is no such thing as 'unsupported' documentation.  Even these emails
have
> led to support issues.
>
> If this philosophy (in part) helps create a market for Xilinx experts, who
make
> a living from consulting in the (more) optimal use of the FPGAs, I would
hardly
> lobby to have Xilinx put them out of business.....
>
> Austin
>
>
> Steve Casselman wrote:
>
> > Can't Xilinx write a behavioral spice model that gives 99% accuracy? It
does
> > not have to be the 100% netlist version of the I/O cell just something
> > someone can use to model the I/O pin and do some simulation. Xilinx
should
> > get used to putting out information and just saying "we don't support
this
> > but..." If someone wants to reverse engineer the part they can do it.
That
> > is no reason to withhold (or make expensive) information someone needs
to do
> > a proper design with Xilinx parts. I think there is some information
that
> > can be released and supported by the users at large. I just wish Xilinx
> > would start to understand the difference between what they need to keep
> > secret, what they can release that needs full support and information
they
> > can give out and not support. For example you have the bitstream specs
for
> > lookup tables, BRAMs and flops. But not routing. I can however get this
> > information from the XDL tools but only at great expense in time. I
guess
> > I've just heard the "if we tell you we'll have to support you" or "if
you
> > know this your going to hurt yourself" lines a little too often.
> >
> > Steve
> >
> > "Austin Lesea" <austin.lesea@xilinx.com> wrote in message
> > news:3CF78DA1.8E0ECD98@xilinx.com...
> > Rick,
> > I know exactly what you are referring to.  Rather than just say, "nope"
I
> > took the opportunity to explain why the answer is what it is.
> > And freeware versions do not support encrypted files.
> > Sorry,  we support Hspice for spice for a very good, and sound business
> > reason.  The hotline has Hspice licenses, and we can solve problems, and
> > support customers this way (in addition to protecting our IP).  We also
have
> > licenses for Cadence SpectraQuest (for the multi-gigbit transceivers),
and
> > HyperLynx.  These are our recommended SI simulation tools, and are
Xilinx
> > partners.
> > If you have a simple "what if" question, the hotline can do a simulation
for
> > you (commercial customers please!  students are supported thru the
Xilinx
> > University program http://www.xilinx.com/univ/index.htm) and email you
the
> > results (similar to what I have done here on this board).
> > Austin
> >
> > rickman wrote:
> > Austin,
> > Thanks for the reply.  But I don't see how Hspice using IBIS files
> > helps.  Hspice is a licensed product and is not cheap.  I was referring
> > to Spice simulators being available as freeware.
> > So it is not likely that the Spartan II chips will get spice models
> > anytime soon?
> > Austin Lesea wrote:
> > >
> > > Rick,
> > >
> > > The spice models are encrypted because they contain the foundry's
> > > technology files (highly proprietary).
> > >
> > > A smart technologist can figure out exactly how a foundry is building
> > > their transistors from examining the details of the technology models.
> > >
> > > We have encrypted spice models only for customers who sign NDAs, and
> > > who have the latest Avant! Hspice.
> > >
> > > The IBIS models are developed from these spice models.
> > >
> > > The latest version of Hspice accepts IBIS models, so the need for
> > > spice models has actually diminished.
> > >
> > > Virtex II ushered in a new age of high speed designs, and as such, an
> > > effort was made to support such designs with better modeling
> > > resources.
> > >
> > > Austin
> > >
> > >
> > > rickman wrote:
> > >
> > > > Checking the Xilinx web site it looks like you can download IBIS
> > > > models
> > > > for various devices including the Spartan IIe devices.  However,
> > > > these
> > > > files do you no good if you don't have IBIS simulator which seem to
> > > > cost
> > > > even more than a logic or HDL simulator.
> > > >
> > > > I also found that there are spice models available for the VirtexII
> > > > parts.  But not for any other devices.  Since Spice is much more
> > > > available, any idea why the Spice models are not available for parts
> > > >
> > > > other than the VII?
> > > >
> > > > --
> > > >
> > > > Rick "rickman" Collins
> > > >
> > > > rick.collins@XYarius.com
> > > > Ignore the reply address. To email me use the above address with the
> > > > XY
> > > > removed.
> > > >
> > > > Arius - A Signal Processing Solutions Company
> > > > Specializing in DSP and FPGA design      URL http://www.arius.com
> > > > 4 King Ave                               301-682-7772 Voice
> > > > Frederick, MD 21701-3110                 301-682-7666 FAX
> > --
> > Rick "rickman" Collins
> > rick.collins@XYarius.com
> > Ignore the reply address. To email me use the above address with the XY
> > removed.
> > Arius - A Signal Processing Solutions Company
> > Specializing in DSP and FPGA design      URL http://www.arius.com
> > 4 King Ave                               301-682-7772 Voice
> > Frederick, MD 21701-3110                 301-682-7666 FAX
>



Article: 43749
Subject: Re: Virtex2 placement problem
From: "MikeJ" <mikejNO SPAM@freeuk.com>
Date: Sat, 1 Jun 2002 00:55:02 +0100
Links: << >>  << T >>  << A >>
Have not been following this thread, but I think vII numbering in general
makes more sense, complete pain in rear before.

my 2 euros
Mike

<Martin> wrote in message news:ee76d2b.0@WebX.sUN8CHnE...
> Hi Marcus,
>
> tzhe first idea here that comes to my mind is the difference for the
numbering in Virtex-II compared to Virtex.
> In the 'old' Virtex architechture you started in the top left corner and
counted down and to the right in rows and columns. Now in the V-II
architecture you start in the bottom-left corner and count in a kind of
x-y-diagram to the right and upwards.
>
> Check if you still have the old syntax. Previously the syntax would have
been something like ...RxCy.S0.
> This must be now something like ...XaYb.
>
> My 2 Euro-Cent.
>
> Martin





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