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

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search

Messages from 29400

Article: 29400
Subject: Re: Configuration of FPGA using SPROM
From: Chris Dunlap <cdunlap@xilinx.com>
Date: Mon, 19 Feb 2001 10:20:44 -0700
Links: << >>  << T >>  << A >>
Make sure to keep your program pulse short.  It should not be greater than .5ms.
You should need to pulse program at all though, the FPGA will autoconfigure on powerup.  I have never heard of DONE going high right at the start of configuration.  I would very your board connections and make sure the pullup resistor is not too small.  Use an o-scope and charachterize DIN, INIT, and DONE.  It is rare for one of those FPGAs to be bad but if you have verified no board level problems, it could be possible that
you have a bad device.  You may want to try another one just to be sure.  Of course, you may want to write out a raw bits file from the prom file formater to verify the data on din when o-scoping it.  The .rbt file is easy to read in terms of figuring out the data being shifted into the device.

Good luck,

Chris

radhika wrote:

> hi chris,
>  I tried calling the support hotline and they would not answer my questions probably cos i am  a student and they just asked me to send an email to xup@xilinx.com which I did thats about it .
> I have connected my init signal to Vcc through a 4.7K and now the scenario has changed my init is low and when i apply a 0V pulse at my prog pin and remove the pulse init is going high along with prog pin but ldc adn done are going high at the same moment but i think atleast 3 to 4 clk pulses are required to start confn and how should the cclk be after the end of confn and my hdc is high both during and after configuration.
> regards
> radhika


Article: 29401
Subject: Re: what
From: Chris Dunlap <cdunlap@xilinx.com>
Date: Mon, 19 Feb 2001 10:27:46 -0700
Links: << >>  << T >>  << A >>
Those are constraint attributes you can place on your RAM block if you wanted to
initialize the RAM on powerup.  It looks like you have elected not to initialize
it so it will default to all zeros.

Chris

"W.Turk" wrote:

> Hi :
>   The message occured when i use Foundation3.1 :
> WARNING:NgdBuild:526 - On the RAMB4_S8_S8 symbol "p2/u1/bram4", the following
>    properties are undefined: INIT_00, INIT_01, INIT_02, INIT_03, INIT_04,
>    INIT_05, INIT_06, INIT_07, INIT_08, INIT_09, INIT_0A, INIT_0B, INIT_0C,
>    INIT_0D, INIT_0E, INIT_0F. A default value of all zeroes will be used.
>    what's meaning?
>    How do i?
>    Thanks a lot


Article: 29402
Subject: Re: ALtera CPLD
From: Andy Peters <"apeters <"@> noao [.] edu>
Date: Mon, 19 Feb 2001 10:48:18 -0700
Links: << >>  << T >>  << A >>
ccu wrote:
> 
> I found CPLD no internal tri-state buffer to share the same bus.
> If there are more than 100 registers need to access via external cpu
> bidirection data port(8051) .
> How can i avoid the huge100:1  mux output?

Choose an FPGA with internal tristates.

--a

Article: 29403
Subject: Infering DPRAM with both outputs
From: "Max L." <maxlarin@_spam_visto.com>
Date: Mon, 19 Feb 2001 19:22:18 +0100
Links: << >>  << T >>  << A >>
Hello,

The question refers to dual ported memory inferring using
Leonardo Spectrum for synthesis and targeting Xilinx' Spartan XL device:

How to infer both outputs of a dual ported memory's element
(namely SPO and DPO of RAM16X1D) from plain Verilog.

An attempt to write:
//+++
module memdp64x1( raddr, waddr, do, do_at_wa, di, we, clk );
....
always @(posedge clk)
   if (we) mem[waddr] = di;
assign do = mem[raddr];
assign do_at_wa = mem[waddr];
//---
results in 'plain' logic (sea of DFFs and LUTs).

Any clues are welcome.



Article: 29404
Subject: Re: Fine Phase Shift in VirtexII
From: Peter Alfke <palfke@earthlink.net>
Date: Mon, 19 Feb 2001 18:23:51 GMT
Links: << >>  << T >>  << A >>
Heinrich,
what you describe cannot be done with one DCM. It requires a fixed input
frequency and can generate a fixed phase difference set by configuration,
or it can step the phase in increments of ~50 ps forwards or backwards.

Tell us what you want to achieve, and there might be a different way to do
that with two DCMs if your pattern is fixed.

Gruß
Peter Alfke, Xilinx Applications
P.S. Today is a holiday in the US:
Presidents' Day ( Washington's and Lincoln's birthdays combined).

Heinrich Fonfara wrote:

> We intend to use the VirtexII device for inserting periods each x
> nanoseconds  in a clock with a period of y nanoseconds (for example 3
> periods with 8ns each, then 1 or 2 periods of 10ns, then some of 8ns and
> so on......) so that the phase is shifted dynamically by 2 ns in a
> certain direction. The shifting should be controlled by an appropriate
> pulse.
> In the data sheets I did not find enough information to know if that is
> possible to realize it with DCM of VirtexII.
> It would be great if someone could tell me where to get more information
> about phase shifting applications.
>
> Thanks in advance
>
> Heinrich Fonfara


Article: 29405
Subject: Re: ALtera CPLD
From: Muzaffer Kal <muzaffer@dspia.com>
Date: 19 Feb 2001 20:16:06 GMT
Links: << >>  << T >>  << A >>
"ccu" <d837928@oz.nthu.edu.tw> wrote:

>I found CPLD no internal tri-state buffer to share the same bus.
>If there are more than 100 registers need to access via external cpu
>bidirection data port(8051) .
>How can i avoid the huge100:1  mux output?
>

If you can't change your device to one which has internal tri-states,
you can design a tree of muxes. You can do a three-level 4:1 mux tree
with a 2:1 mux at the end to generate a 128:1 mux. This needs 42 4:1
muxes and a 2:1 mux.

Muzaffer

FPGA DSP Consulting
http://www.dspia.com

Article: 29406
Subject: Re: ALtera CPLD
From: Rick Filipkiewicz <rick@algor.co.uk>
Date: Tue, 20 Feb 2001 00:34:20 +0000
Links: << >>  << T >>  << A >>


Andy Peters wrote:

> ccu wrote:
> >
> > I found CPLD no internal tri-state buffer to share the same bus.
> > If there are more than 100 registers need to access via external cpu
> > bidirection data port(8051) .
> > How can i avoid the huge100:1  mux output?
>
> Choose an FPGA with internal tristates.
>
> --a

With up to 800 register bits he's going to need an FPGA anyway.


Article: 29407
Subject: meeting high hold time input requirement with Virtex
From: Arrigo Benedetti <arrigo@vision.caltech.edu>
Date: 19 Feb 2001 18:26:51 -0800
Links: << >>  << T >>  << A >>
Dear all,

I am interfacing a RAMDAC (IBM RGB526) to a Virtex device and I have
found problems in meeting the DAC input hold time requirement with
respect to its load clock, that can be as high as 4 ns. Since the
Virtex min clock to out is 1.5ns this is clearly a problem if the DAC
load clock is the same clock driving the output registers inside the
FPGA.
The solution that I have implemented is to have the Virtex generate
the DAC clock by delaying the main pixel clock with a linear array of
LUT1's configured as buffers.
Is there any better way to solve this without resorting to an external
clock management device like the Roboclock?

Thanks,
-Arrigo

Article: 29408
Subject: 5 Clocks in a spartan-II
From: Kent Orthner <korthner@hotmail.nospam.com>
Date: 20 Feb 2001 12:57:02 +0900
Links: << >>  << T >>  << A >>
Okay, so I know that this topic has been covered before,
and I've read through the Deja News (It's now google, 
with a much better interface), so I'm really only looking 
for some reassurance.

I have 4x 77.76 clocks, not necessarily in phase or with 
exactly the same frequency, driving the four major blocks 
on the chip.  I then have an processor interface block 
running at 52MHz.  This bock doesn't really do much 
more than push stuff in a FIFO, and pop stuff off of it.

I plan to have each 77.76MHz clock on a primary clock line,
from a PLL. 

This leaves the 52MHz clock on a secondary clock line.  I
plan to keep all of the logic driven by the secondary clock 
line on the right side of the chip, with the exception of the 
BlockRAMs for the FIFO; they'll be on both sides.

This leaves me with two questions:

1. Using the primary clock lines, I'm guaranteed not to have
race conditions between FF inputs and CLK inputs.  If I use 
the secondary clock lines, will I run into problems with this?
Does trace handle this fairly well?

2. Where should I connect the pin for the secondary clock line?
I figure it should be on the top, in the middle of the right half.

Unfortunately, I have to choose the pin-out and get the PCB 
into fabrication, and *then* do the FPGA design, so I can't 
simply decide later when it's all said and done.

Any comments, suggestions, and sweet nothings will be 
greatly appreciated.

-Kent

Article: 29409
Subject: Re: meeting high hold time input requirement with Virtex
From: Phil Hays <spampostmaster@home.com>
Date: Tue, 20 Feb 2001 04:49:38 GMT
Links: << >>  << T >>  << A >>
Arrigo Benedetti wrote:

> Is there any better way to solve this without resorting to an external
> clock management device like the Roboclock?

An idea is to put out the data on the other edge of clock.  If the external
device latches on the rising edge of clock, put out data after the falling edge
of clock.  Hold time is then a half a clock period.  Setup time might be an
issue, as it is less than a half a clock period.  However, many devices with
large hold times have small setup times.

 ____      ____      ____      ____      ____      
/    \____/    \____/    \____/    \____/    \____

______ _________ _________ _________ _________
______X_________X_________X_________X_________X


-- 
Phil Hays

Article: 29410
Subject: Re: meeting high hold time input requirement with Virtex
From: Kent Orthner <korthner@hotmail.nospam.com>
Date: 20 Feb 2001 15:03:03 +0900
Links: << >>  << T >>  << A >>

> Arrigo Benedetti wrote:
> > Is there any better way to solve this without resorting to an external
> > clock management device like the Roboclock?

Phil Hays <spampostmaster@home.com> writes:
> An idea is to put out the data on the other edge of clock.  If the external
> device latches on the rising edge of clock, put out data after the falling edge
> of clock.  Hold time is then a half a clock period.  Setup time might be an
> issue, as it is less than a half a clock period.  However, many devices with
> large hold times have small setup times.

Another variation on Phil's solution is to use the output of the 
Virtex DLL to drive your logic inside the FPGA.  By using the +90,
+180, or +270 delay output options from the DLL, you can skew your
output data in 1/4 clock cycle increments.

There's also fast vs. slow (Slew rate limited) output driver 
options that you can fiddle with.

HTH,
-Kent

Article: 29411
Subject: UCF problem "- Could not find NET "
From: "Helen Long" <madisonfff@usa.net>
Date: Tue, 20 Feb 2001 01:25:21 -0600
Links: << >>  << T >>  << A >>
Hi I am trying to generate bit file and download to XS40 board
I use XC40005XLPC84, but I can successfully implement it without
my own ucf file, it I connet my ucf file it fails
What is wrong? My code and my ucf file and error message are listed
below thanks a lot!

Qian

Here is my code
library IEEE;
use IEEE.std_logic_1164.all;
use IEEE.std_logic_unsigned.all;

entity noGSR is
port(
CLOCK: in STD_LOGIC;
--RESET: in STD_LOGIC;
--HEX: in STD_LOGIC_VECTOR(3 downto 0);
LED: out STD_LOGIC_VECTOR(6 downto 0));
end noGSR;

architecture SIMPLE of noGSR is
signal HEX: STD_LOGIC_VECTOR(3 downto 0);

begin
      HEX<="0000";
  UP_COUNTER:process(CLOCK)
  begin
--    if(RESET='1') then

    if(CLOCK'event and CLOCK='1') then
      HEX <= HEX + 1;
    end if;
    end process;
SHOW:process(HEX)
begin
....
end process;

end SIMPLE;


Here is my ucf file
NET CLOCK LOC=P13;
NET LED LOC=25,26,24,20,23,18,19;

And I use Xilinx Foundation 2.1l to implement it
it shows error

Annotating constraints to design from file "led.ucf" ...
ERROR:NgdHelpers:18 - Could not find NET "CLOCK" in design "led".  NET entry
is
   "NET CLOCK LOC=P13;
   "
ERROR:NgdHelpers:18 - Could not find NET "LED" in design "led".  NET entry
is
   "NET LED LOC=25,26,24,20,23,18,19;


"
ERROR:Parsers:11 - Encountered unrecognized constraint while parsing.
ERROR:NgdBuild:19 - Errors found while parsing constraint file "led.ucf".

One or more errors were found during NGDBUILD.  No NGD file will be written.

Writing NGDBUILD log file "led.bld"...

==================================================






Article: 29412
Subject: Re: UCF problem "- Could not find NET "
From: Richard Meester <rme@quest-innovations.com>
Date: Tue, 20 Feb 2001 08:38:56 +0100
Links: << >>  << T >>  << A >>
Helen,

Best thing to do is to open the constraint editor, or the edif file with an
editor, and look how the net is called. Probably the net is renamed by the
ngdbuild or the mapper. Once you find the net you need, change the name in the
UCF file.

Richard

Helen Long wrote:

> Hi I am trying to generate bit file and download to XS40 board
> I use XC40005XLPC84, but I can successfully implement it without
> my own ucf file, it I connet my ucf file it fails
> What is wrong? My code and my ucf file and error message are listed
> below thanks a lot!
>
> Qian
>
> Here is my code
> library IEEE;
> use IEEE.std_logic_1164.all;
> use IEEE.std_logic_unsigned.all;
>
> entity noGSR is
> port(
> CLOCK: in STD_LOGIC;
> --RESET: in STD_LOGIC;
> --HEX: in STD_LOGIC_VECTOR(3 downto 0);
> LED: out STD_LOGIC_VECTOR(6 downto 0));
> end noGSR;
>
> architecture SIMPLE of noGSR is
> signal HEX: STD_LOGIC_VECTOR(3 downto 0);
>
> begin
>       HEX<="0000";
>   UP_COUNTER:process(CLOCK)
>   begin
> --    if(RESET='1') then
>
>     if(CLOCK'event and CLOCK='1') then
>       HEX <= HEX + 1;
>     end if;
>     end process;
> SHOW:process(HEX)
> begin
> ....
> end process;
>
> end SIMPLE;
>
> Here is my ucf file
> NET CLOCK LOC=P13;
> NET LED LOC=25,26,24,20,23,18,19;
>
> And I use Xilinx Foundation 2.1l to implement it
> it shows error
>
> Annotating constraints to design from file "led.ucf" ...
> ERROR:NgdHelpers:18 - Could not find NET "CLOCK" in design "led".  NET entry
> is
>    "NET CLOCK LOC=P13;
>    "
> ERROR:NgdHelpers:18 - Could not find NET "LED" in design "led".  NET entry
> is
>    "NET LED LOC=25,26,24,20,23,18,19;
>
> "
> ERROR:Parsers:11 - Encountered unrecognized constraint while parsing.
> ERROR:NgdBuild:19 - Errors found while parsing constraint file "led.ucf".
>
> One or more errors were found during NGDBUILD.  No NGD file will be written.
>
> Writing NGDBUILD log file "led.bld"...
>
> ==================================================

--
Quest Innovations
tel: +31 (0) 227 604046
http://www.quest-innovations.com



Article: 29413
Subject: RSA on FPGA
From: "ajd" <ajadu76@hotmail.com>
Date: Tue, 20 Feb 2001 09:56:08 -0000
Links: << >>  << T >>  << A >>

Anyone ever done it? Which bits of the algorithm were implemented?

Andrew





Article: 29414
Subject: Re: UCF problem "- Could not find NET "
From: Laurent Gauch <laurent.gauch@amontec.com>
Date: Tue, 20 Feb 2001 12:01:34 +0100
Links: << >>  << T >>  << A >>
Helen,

The best way is to generate the .ucf automatically using Xilinx Design 
Manager tool (menu Design -> Lock Pins...).
You will see directly what are your real pad names.

Laurent for Amontec
www.amontec.com
------------------------------------------------------
Amontec introduces a new PCI core development system

Helen Long wrote:

> Hi I am trying to generate bit file and download to XS40 board
> I use XC40005XLPC84, but I can successfully implement it without
> my own ucf file, it I connet my ucf file it fails
> What is wrong? My code and my ucf file and error message are listed
> below thanks a lot!
> 
> Qian
> 
> Here is my code
> library IEEE;
> use IEEE.std_logic_1164.all;
> use IEEE.std_logic_unsigned.all;
> 
> entity noGSR is
> port(
> CLOCK: in STD_LOGIC;
> --RESET: in STD_LOGIC;
> --HEX: in STD_LOGIC_VECTOR(3 downto 0);
> LED: out STD_LOGIC_VECTOR(6 downto 0));
> end noGSR;
> 
> architecture SIMPLE of noGSR is
> signal HEX: STD_LOGIC_VECTOR(3 downto 0);
> 
> begin
>       HEX<="0000";
>   UP_COUNTER:process(CLOCK)
>   begin
> --    if(RESET='1') then
> 
>     if(CLOCK'event and CLOCK='1') then
>       HEX <= HEX + 1;
>     end if;
>     end process;
> SHOW:process(HEX)
> begin
> .....
> end process;
> 
> end SIMPLE;
> 
> 
> Here is my ucf file
> NET CLOCK LOC=P13;
> NET LED LOC=25,26,24,20,23,18,19;
> 
> And I use Xilinx Foundation 2.1l to implement it
> it shows error
> 
> Annotating constraints to design from file "led.ucf" ...
> ERROR:NgdHelpers:18 - Could not find NET "CLOCK" in design "led".  NET entry
> is
>    "NET CLOCK LOC=P13;
>    "
> ERROR:NgdHelpers:18 - Could not find NET "LED" in design "led".  NET entry
> is
>    "NET LED LOC=25,26,24,20,23,18,19;
> 
> 
> "
> ERROR:Parsers:11 - Encountered unrecognized constraint while parsing.
> ERROR:NgdBuild:19 - Errors found while parsing constraint file "led.ucf".
> 
> One or more errors were found during NGDBUILD.  No NGD file will be written.
> 
> Writing NGDBUILD log file "led.bld"...
> 
> ==================================================


Article: 29415
Subject: Re: UCF problem "- Could not find NET "
From: Dave Vanden Bout <devb@xess.com>
Date: Tue, 20 Feb 2001 07:40:47 -0500
Links: << >>  << T >>  << A >>


Helen Long wrote:

> Hi I am trying to generate bit file and download to XS40 board
> I use XC40005XLPC84, but I can successfully implement it without
> my own ucf file, it I connet my ucf file it fails
> What is wrong? My code and my ucf file and error message are listed
> below thanks a lot!

You might try expanding your constraint into individual constraints as follows:

NET LED<0> LOC=P25;
NET LED<1> LOC=P26;
...

>
>
> Qian
>
> Here is my code
> library IEEE;
> use IEEE.std_logic_1164.all;
> use IEEE.std_logic_unsigned.all;
>
> entity noGSR is
> port(
> CLOCK: in STD_LOGIC;
> --RESET: in STD_LOGIC;
> --HEX: in STD_LOGIC_VECTOR(3 downto 0);
> LED: out STD_LOGIC_VECTOR(6 downto 0));
> end noGSR;
>
> architecture SIMPLE of noGSR is
> signal HEX: STD_LOGIC_VECTOR(3 downto 0);
>
> begin
>       HEX<="0000";
>   UP_COUNTER:process(CLOCK)
>   begin
> --    if(RESET='1') then
>
>     if(CLOCK'event and CLOCK='1') then
>       HEX <= HEX + 1;
>     end if;
>     end process;
> SHOW:process(HEX)
> begin
> ....
> end process;
>
> end SIMPLE;
>
> Here is my ucf file
> NET CLOCK LOC=P13;
> NET LED LOC=25,26,24,20,23,18,19;
>
> And I use Xilinx Foundation 2.1l to implement it
> it shows error
>
> Annotating constraints to design from file "led.ucf" ...
> ERROR:NgdHelpers:18 - Could not find NET "CLOCK" in design "led".  NET entry
> is
>    "NET CLOCK LOC=P13;
>    "
> ERROR:NgdHelpers:18 - Could not find NET "LED" in design "led".  NET entry
> is
>    "NET LED LOC=25,26,24,20,23,18,19;
>
> "
> ERROR:Parsers:11 - Encountered unrecognized constraint while parsing.
> ERROR:NgdBuild:19 - Errors found while parsing constraint file "led.ucf".
>
> One or more errors were found during NGDBUILD.  No NGD file will be written.
>
> Writing NGDBUILD log file "led.bld"...
>
> ==================================================

--
|| Dr. Dave Van den Bout   XESS Corp.               (919) 387-0076 ||
|| devb@xess.com           2608 Sweetgum Dr.        (800) 549-9377 ||
|| http://www.xess.com     Apex, NC 27502 USA   FAX:(919) 387-1302 ||



Article: 29416
Subject: Re: 5 Clocks in a spartan-II
From: chris@cgschneider.com (Chris G. Schneider)
Date: 20 Feb 2001 18:31:49 +0100
Links: << >>  << T >>  << A >>
Kent Orthner <korthner@hotmail.nospam.com> writes:
 
> I have 4x 77.76 clocks, not necessarily in phase or with 
> exactly the same frequency, driving the four major blocks 
> on the chip.  I then have an processor interface block 
> running at 52MHz.  This bock doesn't really do much 
> more than push stuff in a FIFO, and pop stuff off of it.
> 
> I plan to have each 77.76MHz clock on a primary clock line,
> from a PLL. 
> 
> This leaves the 52MHz clock on a secondary clock line.  I
> plan to keep all of the logic driven by the secondary clock 
> line on the right side of the chip, with the exception of the 
> BlockRAMs for the FIFO; they'll be on both sides.
> 
> This leaves me with two questions:
> 
> 1. Using the primary clock lines, I'm guaranteed not to have
> race conditions between FF inputs and CLK inputs.  If I use 
> the secondary clock lines, will I run into problems with this?
> Does trace handle this fairly well?

Clock skew may be a problem. One time I used a secondary clock for
a clock with 39 MHz (without intension) and it worked with no problems.
There were a lot of really large counters driven by this "clock" net.

As far as I remember, trace calculates the clock skew as well.

> 2. Where should I connect the pin for the secondary clock line?
> I figure it should be on the top, in the middle of the right half.

It would be best to but in the center of yout logic. Unfortunately
the pads are at the edge of the die :-). But try to avoid long routing 
delays.

> Unfortunately, I have to choose the pin-out and get the PCB 
> into fabrication, and *then* do the FPGA design, so I can't 
> simply decide later when it's all said and done.

We do it that way, too. This saves a lot of time. It just reqiures good 
communication between FPGA and PCB designers.
 
> Any comments, suggestions, and sweet nothings will be 
> greatly appreciated.

You did not mention your application, but the four nearly identical
clock just raise the question whether you can combine them. External
phase differnce may be comensated by using the DLLs, but ask the experts
on this subject. One other thig is that 52Mhz * 1.5 = 78 MHz approx 
77.76 MHz. Can you derive the 52 MHz from 77.76 MHz? Or why don't you 
use 77.76 MHz for your processor interface?  

-- 
Chris

Article: 29417
Subject: Re: UCF problem "- Could not find NET "
From: chris@cgschneider.com (Chris G. Schneider)
Date: 20 Feb 2001 18:32:00 +0100
Links: << >>  << T >>  << A >>
"Helen Long" <madisonfff@usa.net> writes:

> Hi I am trying to generate bit file and download to XS40 board
> I use XC40005XLPC84, but I can successfully implement it without
> my own ucf file, it I connet my ucf file it fails
> What is wrong? My code and my ucf file and error message are listed
> below thanks a lot!
> 
> Qian
> 
> Here is my code
> library IEEE;
> use IEEE.std_logic_1164.all;
> use IEEE.std_logic_unsigned.all;
> 
> entity noGSR is
> port(
> CLOCK: in STD_LOGIC;
> --RESET: in STD_LOGIC;
> --HEX: in STD_LOGIC_VECTOR(3 downto 0);
> LED: out STD_LOGIC_VECTOR(6 downto 0));
> end noGSR;
> 
> architecture SIMPLE of noGSR is
> signal HEX: STD_LOGIC_VECTOR(3 downto 0);
> 
> begin
>       HEX<="0000";
>   UP_COUNTER:process(CLOCK)
>   begin
> --    if(RESET='1') then
> 
>     if(CLOCK'event and CLOCK='1') then
>       HEX <= HEX + 1;
>     end if;
>     end process;
> SHOW:process(HEX)
> begin
> ....
> end process;
> 
> end SIMPLE;
> 
> 
> Here is my ucf file
> NET CLOCK LOC=P13;
> NET LED LOC=25,26,24,20,23,18,19;
> 
> And I use Xilinx Foundation 2.1l to implement it
> it shows error
> 
> Annotating constraints to design from file "led.ucf" ...
> ERROR:NgdHelpers:18 - Could not find NET "CLOCK" in design "led".  NET entry
> is
>    "NET CLOCK LOC=P13;
>    "
> ERROR:NgdHelpers:18 - Could not find NET "LED" in design "led".  NET entry
> is
>    "NET LED LOC=25,26,24,20,23,18,19;
> 
> 
> "
> ERROR:Parsers:11 - Encountered unrecognized constraint while parsing.
> ERROR:NgdBuild:19 - Errors found while parsing constraint file "led.ucf".
> 
> One or more errors were found during NGDBUILD.  No NGD file will be written.
> 
> Writing NGDBUILD log file "led.bld"...
> 
> ==================================================
> 
> 
> 
> 
> 

Did you connect LED <= HEX ?

-- 
Chris


Article: 29418
Subject: Re: Fine Phase Shift in VirtexII
From: "Erik Widding" <widding@birger.com>
Date: Tue, 20 Feb 2001 13:51:14 -0500
Links: << >>  << T >>  << A >>
"Heinrich Fonfara" <fonfarah@ibmt.fhg.de> wrote in message
news:3A9102B2.D72DC65F@ibmt.fhg.de...
> We intend to use the VirtexII device for inserting periods each x
> nanoseconds  in a clock with a period of y nanoseconds (for example 3
> periods with 8ns each, then 1 or 2 periods of 10ns, then some of 8ns and
> so on......) so that the phase is shifted dynamically by 2 ns in a
> certain direction. The shifting should be controlled by an appropriate
> pulse.
> In the data sheets I did not find enough information to know if that is
> possible to realize it with DCM of VirtexII.

You couldn't do this in one DCM.  But you could do this with three DCMs, two
LUTs and a BUFGMUX.

Use one DCM to create a clock with period Y, and 50% duty cycle, call this
signal Y1.
Use a second DCM to phase shift Y1 by -Y/2 + 8nS, call this signal X1.
Use a second DCM to phase shift Y1 by -Y/2 + 10nS, call this signal X2.
Use two LUTs to produce the following signals:
    XP1 = Y1 and X1
    XP2 = Y1 and X2
Use a BUFGMUX to switch between XP1 and XP2.

I make no claims that this is a reasonable solution to the problem.  I only
claim that it probably won't glitch if:
    X is less than 1/2Y - a couple of nS.
    You floorplan the circuit, or atleast define max delays to ensure
somewhat matched skew.

The reason for this is that any two input circuit represented in a LUT will
not glitch as long as only one input changes at a time.  Hence the reason
for subtracting a couple extra nS.

DCM and BUFGMUX are definitely pretty cool elements in the Virtex-II
architecture.  Thuimbs up to Xilinx on this one...


Regards,
Erik Widding.

--
Birger Engineering, Inc.  --------------------------------  781.481.9233
38 Montvale Ave #260; Stoneham, MA 02180  -------  http://www.birger.com




Article: 29419
Subject: Re: Altera process change....
From: Jean-Paul Smeets <jpsmeets@xs4all.nl>
Date: Tue, 20 Feb 2001 22:02:37 +0100
Links: << >>  << T >>  << A >>
On Mon, 19 Feb 2001 08:35:17 +0000, Rick Filipkiewicz
<rick@algor.co.uk> wrote:

>> Arcane question, but...
>> Has anyone had trouble with Altera's 10K50V devices following the
>> process change from E50 to F51 over the course of last year?
>> We have 10-for-10 bad pcbs that have F51s, that work with E50s.
>> They went from 3 to 4-layer metal, same fabs/geometry.
>> Thanks!
>> Joe Curren
>> Printrex, Inc.
>
>Don't know Altera specifically but I've come across similar situations
>before. Causes were either:
>
>o The new process is faster & so is leading to hold time problems
>somehwhere.
>
>o The metatstability characteristics have changed - got worse. Unlikely
>if the first possibility is true.

I have seen something similar when we used a faster version of an
FPGA. These were also hold time problems.

However there is a trick you can use to verify this.
We normally use a cooling spray to cool the FPGA from 25 C (room
temperature) to well below  0C. Simply cool the FPGA till the outside
turns completely white (water in the air forms a thin layer of snow).
This can be used to simulate faster versions of the device. If you
just meet your hold time, cooling the device will cause it to fail.

J.P. Smeets


J.P. Smeets
business:
Ellips
Woenselsestr 352A
5623 EG Eindhoven
tel: +31-40-2456540
fax: +31-40-2467183
email: jeanpaul@ellips.nl
home:
Loondermolen 23
5612 MH Eindhoven
tel: +31-40-2465105
email: jpsmeets@xs4all.nl

Article: 29420
Subject: Re: Fine Phase Shift in VirtexII
From: Falk Brunner <Falk.Brunner@gmx.de>
Date: Tue, 20 Feb 2001 22:52:42 +0100
Links: << >>  << T >>  << A >>
Erik Widding schrieb:
> 
> You couldn't do this in one DCM.  But you could do this with three DCMs, two
> LUTs and a BUFGMUX.
> 
> Use one DCM to create a clock with period Y, and 50% duty cycle, call this
> signal Y1.
> Use a second DCM to phase shift Y1 by -Y/2 + 8nS, call this signal X1.
> Use a second DCM to phase shift Y1 by -Y/2 + 10nS, call this signal X2.
> Use two LUTs to produce the following signals:
>     XP1 = Y1 and X1
>     XP2 = Y1 and X2
> Use a BUFGMUX to switch between XP1 and XP2.
> 
> I make no claims that this is a reasonable solution to the problem.  I only
> claim that it probably won't glitch if:
>     X is less than 1/2Y - a couple of nS.
>     You floorplan the circuit, or atleast define max delays to ensure
> somewhat matched skew.

Hmmm, I dont think thats a really FINE resolution, this can be also done
with a high speed clock (say 200MHz).
This gives you a resolution of 5ns (2.5ns if you can manage to use both
edges.)
I did this for a digital PLL in a Spartan II -5 device @150 MHz (Yeahh,
I know, sounds like DC for the high speed guys ;-))

A question to the Xilinx guys.

Why dont you let the user (me ;-)) access the delay line directly, so
that the user defined logic can control on which tap the delay line is
taped?? This would be VERY nice for frequency synthesizer/PLLs based on
DDS (I saw a chip on PLAIN OLD 0.8u technology which had a 32 tap delay
line running at 100MHz with a delay of 150ps per tap, which gives you a
so called virtual frequency resolution of 3.2 GHz) With the 45ps of
Virtex II this would be MUCH better.

-- 
MFG
Falk

Article: 29421
Subject: to you sir Peter Alfke...
From: erika_uk@hotmail.com ("erika churchil")
Date: 20 Feb 2001 23:11:50 +0100
Links: << >>  << T >>  << A >>
Hello,

Last week, i asked if a disabled FDC still consumes power or now.
The answer of  PETER said that a disabled FDC consumes as much power as a 
clocked FDC but with a constant input.

could you tell me how,generally ,we compute  a power ?, and what's the ratio 
between the power consumed by a disabled FDC and the power consumed by an 
FDC  fed by an input that toggles continuosely between 1 and 0 ?

Anticipated thanks

--Erika
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.


-- 
Posted from [143.117.60.33] by way of f223.law4.hotmail.com [216.33.149.223] 
via Mailgate.ORG Server - http://www.Mailgate.ORG

Article: 29422
Subject: Re: Fine Phase Shift in VirtexII
From: "Erik Widding" <widding@birger.com>
Date: Tue, 20 Feb 2001 18:14:52 -0500
Links: << >>  << T >>  << A >>
"Falk Brunner" <Falk.Brunner@gmx.de> wrote in message
news:3A92E72A.AE9D5F54@gmx.de...
> Erik Widding schrieb:
> >
> > I make no claims that this is a reasonable solution to the problem.  I
only
> > claim that it probably won't glitch if:
> >     X is less than 1/2Y minus a couple of nS.
> >     You floorplan the circuit, or atleast define max delays to ensure
> > somewhat matched skew.
>
> Hmmm, I dont think thats a really FINE resolution,  this can be also done
> with a high speed clock (say 200MHz).
> This gives you a resolution of 5ns (2.5ns if you can manage to use both
> edges.)

I must admit writing the post a little too quickly, and possibly leaving too
much for the reader to conclude.  I made no statements about resolution,
which happens to be as good as that of the DCM, i.e. 100pS range, BUT you
can only switch between two pulse widths.  So if you need a third pulse
width you would need to spend enough time outputting the pulses X1 while you
change the pulse width X2, using a state machine to control phase shift on
the DCM.

> I did this for a digital PLL in a Spartan II -5 device @150 MHz (Yeahh,
> I know, sounds like DC for the high speed guys ;-))


Technically, this would be a DLL.  And it is an option.  If one used the 0
and 180 outputs of a DCM to drive two state machines one should easily be
able to create output pulses with a resolution of ~1400pS ( 1/(360MHz*2) the
max speed of DCM in -4) and an accuracy of better than 500pS (error
dominated by max delay through LUT for 4 input function in -4).  Using the
three DCMs merely gets you about 10X the resolution, and about the same
accuracy, with some serious limitations of course.


Regards,
Erik Widding.


--
Birger Engineering, Inc.  --------------------------------  781.481.9233
38 Montvale Ave #260; Stoneham, MA 02180  -------  http://www.birger.com




Article: 29423
Subject: Re: to you sir Peter Alfke...
From: Chris Dunlap <cdunlap@xilinx.com>
Date: Tue, 20 Feb 2001 16:16:13 -0700
Links: << >>  << T >>  << A >>
The power calculators should take that into account for you.  You will say in
the power calculator how many flops you have, then seperately what the toggle
rate is on those flops.

That seems to be the obvious answer although I missed the first question and the
answer so you could be going down a completely different route than what I
answered :)

Chris

erika churchil wrote:

> Hello,
>
> Last week, i asked if a disabled FDC still consumes power or now.
> The answer of  PETER said that a disabled FDC consumes as much power as a
> clocked FDC but with a constant input.
>
> could you tell me how,generally ,we compute  a power ?, and what's the ratio
> between the power consumed by a disabled FDC and the power consumed by an
> FDC  fed by an input that toggles continuosely between 1 and 0 ?
>
> Anticipated thanks
>
> --Erika
> _________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
>
> --
> Posted from [143.117.60.33] by way of f223.law4.hotmail.com [216.33.149.223]
> via Mailgate.ORG Server - http://www.Mailgate.ORG


Article: 29424
Subject: Re: ALtera CPLD
From: "luigi funes" <fuzzy8888@hotmail.com>
Date: Wed, 21 Feb 2001 00:04:45 GMT
Links: << >>  << T >>  << A >>

ccu ha scritto nel messaggio <96rb23$oo7$1@thccy25.nthu.edu.tw>...
>I found CPLD no internal tri-state buffer to share the same bus.
>If there are more than 100 registers need to access via external cpu
>bidirection data port(8051) .
>How can i avoid the huge100:1  mux output?


If you don't need speed, you could access it serially, as
a shift register.

Luigi





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

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search