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 36125

Article: 36125
Subject: Can anyone guide me in selecting an FPGA?
From: dvdprsns36@hotmail.com (David)
Date: 30 Oct 2001 14:25:55 -0800
Links: << >>  << T >>  << A >>
I'm looking for a FPGA that will b e used with a SERDES 
device. Originally I was looking at an XCV2000 or APEX1000 
along with a SERDES device (most likely TI or Conexant 
3.125Gbps) but I've heard that Xilinx is coming out with a 
FPGA that will have high-speed SERDES functionality built in. 
Does anyone know what the story is with that? Would you 
recommend that as a better/cheaper (how much?) alternative?


Dave

Article: 36126
Subject: Re: Can anyone guide me in selecting an FPGA?
From: Mike Treseler <mike.treseler@flukenetworks.com>
Date: Tue, 30 Oct 2001 14:50:58 -0800
Links: << >>  << T >>  << A >>
David wrote:
> 
> I'm looking for a FPGA that will b e used with a SERDES
> device. Originally I was looking at an XCV2000 or APEX1000
> along with a SERDES device (most likely TI or Conexant
> 3.125Gbps) but I've heard that Xilinx is coming out with a
> FPGA that will have high-speed SERDES functionality built in.
> Does anyone know what the story is with that? Would you
> recommend that as a better/cheaper (how much?) alternative?

If you are in a hurry, just buy an off the shelf SERDES.

Both brand A and X have new parts out designed for this,
but you know how version 1.0 of anything can go . . .

   --Mike Treseler

Article: 36127
Subject: Re: Leonardo bugs
From: Russell Shaw <rjshaw@iprimus.com.au>
Date: Wed, 31 Oct 2001 10:35:16 +1100
Links: << >>  << T >>  << A >>


Mike Treseler wrote:
> 
> Russell Shaw wrote:
> 
> > I'm using leonardo-spectrum from the altera site:
> >
> > Version: v2001_1d.24_OEM_Altera
> >     (Release OEM Altera Candidate, compiled Jul  5 2001 at 23:42:12)
> >
> > When i modify vhdl code in a file and save it within leonardo, a
> > new "run flow" compilation frequently doesn't read the new file.
> > Is there a way around this problem?
> 
> Leo reloads a file every time.
> Check the file exemplar.log to see what it used.
> The file path/name is the first uncommented line.

The most recent exemplar.log on my pc isn't being
updated when i do a compile.

> Perhaps you forgot to save your project settings.

I've tried that plenty of times, but leonardo has
lots of bugs, and one of them is that most of the
device settings you do in the flow tabs don't get
saved with the project. Are you using a similar
version?

Article: 36128
Subject: Field Programmable Logic in energy poor environments
From: iokennedy@hotmail.com (Irwin Kennedy)
Date: 30 Oct 2001 15:36:39 -0800
Links: << >>  << T >>  << A >>
Hi,
I have a been thinking about the feasibility of using FPGA's in
wireless computing devices like the much talked about "3rd Generation"
cellular terminals. With plenty of different wireless interfaces vying
for market share (GSM,GPRS, UMTS, 802.11, bluetooth - I'm in the UK :)
) it would appear that having a terminal reconfigure itself to suit
its environment is a neat solution.

However, after a casual play around with the Xilinx Virtex Power
Estimate Worksheet (http://www.xilinx.com/cgi-bin/powerweb.pl) it is
clear that power consumption did not make top priority in the design
of this architecture - quiescent power consumption appears to be in
the order of 10's to 100's of milli-Watts, however portable devices
demand idle power wastage to be in the order of micro-Watts or less.

So, do you think it is possible to apply similar power saving
techniques as used in low power DSPs like the TI C55X devices in order
to make FPGAs feasible in un-tethered devices? By this I don't mean to
suggest the FPGA would replace the DSP, I simply think it could be
integrated somehow and take on certain highly compute intensive
kernels to meet timing constraints and reduce energy consumption.

To make my question a little less philosophical I will list some of
the things that I believe (somewhat naively) could be done to reduce
quiescent energy consumption at the device level in an architecture
like the Xilinx Virtex.

(I guess some of these ideas, like other engineering decisions would
incur certain overheads - like an increase in die area or a decrease
in speed).

* Use less "leaky" transistors! (?)

* Ability to switch off power to non-volatile hardware (everything
except for          configuration memory, RAM and flip/flops).

* Remove all the unnecessary circuitry - like all the different I/O
standard stuff. ("Unnecessary" in this specific high volume
application)

* Ability to prune the clock tree, hence only deliver the clock to the
parts of the device in use.

* Use very low power SRAM for configuration memory (since this cannot
be switched off without incurring a huge latency/energy penalty next
time the FPGA is required).

any comments or opinions on how far energy wastage could be reduced
and the consequences in terms of die area/speed?

Irwin Kennedy
PhD Student
University of Edinburgh

Article: 36129
Subject: what about FPGA with embedded processor?
From: "Jae-cheol Lee" <jchlee@lge.com>
Date: Wed, 31 Oct 2001 00:22:33 GMT
Links: << >>  << T >>  << A >>
There were some news about Virtex II with PowerPC cores.
Altera announced Excalibur series with ARM or MIPS cores.

What is the schedule of production of the first one?

Is there any good story on the use of the second one?

Please let me know...



Article: 36130
Subject: Re: Leonardo bugs
From: Mike Treseler <mike.treseler@flukenetworks.com>
Date: Tue, 30 Oct 2001 16:36:41 -0800
Links: << >>  << T >>  << A >>
Russell Shaw wrote:

> > I'm using leonardo-spectrum from the altera site:
> >
> > Version: v2001_1d.24_OEM_Altera
> The most recent exemplar.log on my pc isn't being
> updated when i do a compile.

Either your project directory isn't where
you think it is, or you need to reinstall the
software. 

> most of the
> device settings you do in the flow tabs don't get
> saved with the project. 

Haven't seen anything like that.

> Are you using a similar version?

I am using v2001_1a.32

You might try a released version.

 --Mike Treseler

Article: 36131
Subject: Re: Device support Foundation 3.1i SP8
From: Kamal Patel <kamal.patel@xilinx.com>
Date: Tue, 30 Oct 2001 17:50:24 -0700
Links: << >>  << T >>  << A >>
Hello VR,

Device support can be found in Answer Record 12562 in the
Answers Database.  In regards to your question, the 4.1i version
of the Foundation software does not add any new device support.

I hope this helps.

Best regards,
Kamal Patel



VR wrote:

> I didn't see this question answered, so I'm posting under a slightly
> different subject.
>
> I just would like to know if Foundation 4.1i (w/SP1?) supports any more
> parts than 3.1i w/SP8.
>
> Trying to find this information on Xilinx's site isn't easy, I've already
> tried, that's why I'm asking here.
>
> I'm also curious if Foundation 4.1i actually has any real usefulness over
> Foundation 3.1i(w/SP8)? Is "XST" synth much better than the latest
> offerings from Synplicity(Synplify Pro 7.0) or Exemplar(Leonardo 2001.1d)?
>
> I found Foundation 3.3ISE pretty non-intuitive and I didn't like how with
> Foundation 3.3ISE the Xilinx ver. of FPGA Express would overwrite a
> non-Xilinx FPGA Express (and subsequently, wouldn't call the non-Xilinx
> version!).
>
> Thanks,
> VR.


Article: 36132
Subject: Re: Can anyone guide me in selecting an FPGA?
From: Philip Freidin <philip@fliptronics.com>
Date: Wed, 31 Oct 2001 01:03:29 GMT
Links: << >>  << T >>  << A >>
The Xilinx Virtex-2 is the underlying FPGA for their upcomming
Virtex-2 Pro parts that they say will have the 3.125GBaud SerDes
blocks on it. Xilinx press releases and presentations say that the
underlying SerDes technology is the Conexant SerDes that
they have licensed. I believe that Conexant SerDes technology
has been spun out into a separate company named MindSpeed.

Given this, I would recommend doing your design with Xilinx Virtex-2
and the MindSpeed/Conexant SerDes, and use the integrated
product when it becomes available.

You can probably get more info on this by contacting your local
Xilinx sales office.

Philip Freidin

On 30 Oct 2001 14:25:55 -0800, dvdprsns36@hotmail.com (David) wrote:
>I'm looking for a FPGA that will b e used with a SERDES 
>device. Originally I was looking at an XCV2000 or APEX1000 
>along with a SERDES device (most likely TI or Conexant 
>3.125Gbps) but I've heard that Xilinx is coming out with a 
>FPGA that will have high-speed SERDES functionality built in. 
>Does anyone know what the story is with that? Would you 
>recommend that as a better/cheaper (how much?) alternative?
>
>
>Dave

Philip Freidin
Fliptronics

Article: 36133
Subject: Re: Guided Design, Xilinx Virtex-E
From: "Kevin Neilson" <kevin_neilson@removethis-yahoo.com>
Date: Wed, 31 Oct 2001 01:05:26 GMT
Links: << >>  << T >>  << A >>
I'd update to 4.1i.  There are some improvements in the placer.  Also, don't
overconstrain (i.e., set multicycle paths, etc.) and maybe do some area
constraints in the floorplanner.

"William Lenihan" <lenihan3we@earthlink.net> wrote in message
news:3BDE59A3.3C188BD5@earthlink.net...
> I have a xcv600E design that is fairly well populated (82% utilization)
> and the P&R time (starting from scratch) on our best Sun workstation is
> ~ 2 hours (using Alliance s/w version 3.1i, though it says 'v3.3.08i' in
> the "About" dialog box).
>
> I've been reading up on the concept of 'guided design', which is
> supposed to let the software leverage the last P&R in doing the next
> P&R, which presumably produces a new bitstream in much less time. ......
> my problem: I can see from the messages scrolling by in the Flow
> Engine's window, that yes the switches are activated for guided design,
> but it still takes in excess of 2 hours to do the P&R. This isn't the
> efficiency gain I was hoping for.
>
> I tried this for a design that had a small change (INIT contents of a
> Block RAM) and no change (same EDIF netlist), and there was no
> difference.
>
> What are the trick(s) to making this guided design thing actually work
> (faster)?
>
> Do I need to move up to the 4.1i release?
> Special patches?
> Does it even work at all (i.e., produce small design changes in less
> time)?
>
>
> Thanks in advance for any suggestions.
>
> --
> ==============================
> William Lenihan
> lenihan3weNOSPAM@earthlink.net
> .... remove "NOSPAM" when replying
> ==============================
>
>



Article: 36134
Subject: Re: Shift Registers with Xilinx Foundation 2.1
From: Philip Freidin <philip@fliptronics.com>
Date: Wed, 31 Oct 2001 01:11:06 GMT
Links: << >>  << T >>  << A >>
On Tue, 30 Oct 2001 19:55:43 +0100, Falk Brunner <Falk.Brunner@gmx.de> wrote:
>django625 schrieb:
>> 
>> Greetings
>> 
>> I'm using a Spartan XCS40XL part in a design which has several shift
>> registers of various lengths and I'm using schematic entry.  I'm
>> confused as to how to specify the timing.  I don't have enough clock
>> buffers to use one for each shift register and I need obviously to
>> ensure that a clock edge arrives at shift register flip-flop N before
>> data has been shifted by flip-flop N-1, and it seems that this is a
>> problem with my longer shift registers.
>
>;-))
>The global clock buffers drive a clock net for the whole chip. So all
>you need is just to connect your clock input pin to a global clock
>buffer and then use this signal for all FFs in the shift register.

To be more accurate, your clock signal from outside the chip is
connected to the input of a clock buffer, and the output of the clock
buffer is connected to the clock pin of all your flipflops.

The clock buffers have dedicated routing resources on their outputs,
called "clock nets", that have been pre-designed to have minimal
skew between all destinations. The net also has sufficient buffering
that it can drive all flipflops in the chip.

> Works
>nice, you dont have to worry about skew, fan-out, hold-time etc.
> 

Philip
Philip Freidin
Fliptronics

Article: 36135
Subject: Re: How can I design a bi-deriction bus buffer?
From: "deerlux" <deerlux@hotmail.com>
Date: Tue, 30 Oct 2001 17:21:43 -0800
Links: << >>  << T >>  << A >>
I want design bi-direction bus within the chip not on the interface of the
chip.3-state gate is only used on the interface of the chip.
I had solved the problem by using another group of pins.
3-state gate is very good.
Thank you!
"Peter Alfke" <palfke@earthlink.net>
??????:3BDE390F.FA1AC592@earthlink.net...
>
>
> deerlux wrote:
>
> > I want to implement a bideriction bus buffer controlled by "DIR". But I
> > don't want to use 3-state gate.Can I do it?
>
> No, you cannot. What's wrong with 3-state? It has served us well for 35
> years...
>
> Peter Alfke
>
>



Article: 36136
Subject: Re: Phase noise of Xilinx/Altera DLL/PLL
From: "Paul Teagle" <pteagle@bigpond.net.au>
Date: Wed, 31 Oct 2001 01:57:02 GMT
Links: << >>  << T >>  << A >>

"Ian Dedic" <ian.dedic@acg.fujitsu-fme.com> wrote in message
news:1476aea8.0110300737.70ca6884@posting.google.com...
> I doubt if any integrated DLL/PLL is good enough for this, depending
> on what level of phase noise/jitter performance you're looking for.

-100dBc/Hz from 1kHz to 10kHz, down from there...  :-(

> When we're doing our DAC evaluations we have difficulty finding any
> clock source with low enough jitter, even low-noise benchtop RF
> generators.
>
> You might want to look at PLLs based on voltage-controlled SAW
> oscillators from vendors like Sawtek -- their lowest jitter VCSOs have
> about 0.01ps rms jitter...
>

Thanks for the comments. I was pretty sure that an integrated device wasn't
going to be suitable for what we wanted (including the PLLs built in to some
of the DACs). We're currently looking at a "discrete" Colpitts oscillator,
with a ceramic resonator as an inductor, and a seperate PLL control chip, as
a clock source.

The Sawtek idea is worth looking at, but I need a non-standard frequency,
and it would probably cost megabucks for a custom VCO...

The FPGA clock is EASY compared to this analog/RF stuff....

Paul T
CAE Inc
[from home]



Article: 36137
Subject: Re: Leonardo bugs
From: Russell Shaw <rjshaw@iprimus.com.au>
Date: Wed, 31 Oct 2001 14:09:10 +1100
Links: << >>  << T >>  << A >>


Mike Treseler wrote:
> 
> Russell Shaw wrote:
> 
> > > I'm using leonardo-spectrum from the altera site:
> > >
> > > Version: v2001_1d.24_OEM_Altera
> > The most recent exemplar.log on my pc isn't being
> > updated when i do a compile.
> 
> Either your project directory isn't where
> you think it is, or you need to reinstall the
> software.

I downloaded and re-installed it, but nothing changed.
A search of exemplar.log on the whole pc didn't find
anything other than what's in the current directory.

> > most of the
> > device settings you do in the flow tabs don't get
> > saved with the project.
> 
> Haven't seen anything like that.
> 
> > Are you using a similar version?
> 
> I am using v2001_1a.32
> 
> You might try a released version.

I guess mine's released, as it came from the altera site.

Also, if i open the project in leonardo, but don't open
the .vhd text file, then "run flow" compiles it. If i
use note-pad to put some garbage in the file and save
it, leonardo doesn't show the syntax errors. If i delete
the .vhd file, leonardo still compiles it, giving the
same error messages as the last compilation. It seems
like a data-base thing. I'm using win2k.

Article: 36138
Subject: Re: Field Programmable Logic in energy poor environments
From: Peter Alfke <palfke@earthlink.net>
Date: Wed, 31 Oct 2001 03:30:24 GMT
Links: << >>  << T >>  << A >>


Irwin Kennedy wrote:

> Hi,
> I have a been thinking about the feasibility of using FPGA's in
> wireless computing devices like the much talked about "3rd Generation"
> cellular terminals. <snip>

> To make my question a little less philosophical I will list some of
> the things that I believe (somewhat naively) could be done to reduce
> quiescent energy consumption at the device level in an architecture
> like the Xilinx Virtex.
>
> (I guess some of these ideas, like other engineering decisions would
> incur certain overheads - like an increase in die area or a decrease
> in speed).

FPGA design, like all engineering, involves trade-offs.
We at Xilinx are being pressured by our customers to deliver FPGAs that
are:
faster, bigger in logic capacity, cheaper, more flexible and easier to
use, and consume less power. In my estimate somewhat in that order. And
we try to do a good job responding to all these conflicting demands. Here
are some comments:

>
>
> * Use less "leaky" transistors! (?)

Until recently, there was extremely little junction leakage or
sub-threshold leakage current. XC3000L could run on 50 microamps.
Unfortunately, as we approach 100 nm technology, subthreshold leakage
current becomes significant ( for every IC manufacturer. Ask Intel! )

>
>
> * Ability to switch off power to non-volatile hardware (everything
> except for          configuration memory, RAM and flip/flops).

All storage in our FPGAs is inherently volatile, but CoolRunner CPLDs
uses non-volatile storage for their configuration. And they consume
extremely low power, down to microamps. Powered by grapefruits.  :-)

>
>
> * Remove all the unnecessary circuitry - like all the different I/O
> standard stuff. ("Unnecessary" in this specific high volume
> application)

Removing it would separate us from very attractive markets. Although it
costs area, it costs hardly any extra power, and if, then only in
high-frequency outputs.

>
>
> * Ability to prune the clock tree, hence only deliver the clock to the
> parts of the device in use.

That's what we do. We distribute the clocks on a horizontal spine, but
activate the vertical "ribs" only when needed. Thanks for the suggestion,
we anticipated your idea.  :-)

>
>
> * Use very low power SRAM for configuration memory (since this cannot
> be switched off without incurring a huge latency/energy penalty next
> time the FPGA is required).

The configuration store used to consume microamps at most. It is only
recently that deep sub-micron technology has increased this current to
milliamps. There is little we can do about subthreshold leakage.

>
>
> any comments or opinions on how far energy wastage could be reduced
> and the consequences in terms of die area/speed?

For the present generation FPGAs, speed, flexibility (versatility), and
cost have been the main driving forces.
Our CoolRunner CPLDs address the issue of low power brilliantly. They are
by far the least power-hungry programmable devices in the universe. And
they are easy to design with. And the software is free. But they do not
reach the complexity level of the bigger Virtex and Spartan device.

Everything is a trade-off. Believe us that we try to keep the static and
dynamic power as low as possible, but for FPGAs, speed and versatility
come first.
You can still buy XC3000L, but you would not be impressed by the
functionality and performance.
You might also look at SpartanXL, a 3.3-V, XC4000-derived family with
fairly low power and more modern features.  But CoolRunner is the
all-time champion of low power in programmable logic.

Peter Alfke, Xilinx Applications



Article: 36139
Subject: BRAM usage reduction in FIFO design: First Scenario
From: kahhean@hotmail.com (Chua Kah Hean)
Date: 30 Oct 2001 22:29:02 -0800
Links: << >>  << T >>  << A >>
Hi all experts out there,

I would like to seek help in ways to reduce BRAM usage in the
following two scenarios.  This post contains only the first scenario. 
I will post the second scenario in another post.

(1)

I need a 2600 deep 16 bit wide FIFO (independent read/write clocks).

Using Xilinx's 4Kbit DPRAM, I realize I will need 16 blocks of them,
giving me a 4096 by 16 bit FIFO instead.  (In fact, I am using the
gray-coded FIFO as supplied by Xilinx in XAPP131, after some
modification).

I am wasting too many block rams.  So I am thinking of cascading a
2048x16 and a 1024x16 FIFOs to give me a 3072x16 FIFO (I stuck to
depth of power 2 to avoid muxing control/output and also to keep it
easier for the gray codes).  There will be a simple data relaying
between the 2048x16 and the 1024x16 FIFOs: whenever the second FIFO is
not full and the first FIFO is not empty, read out the first FIFO and
write into the second FIFO (taking into account the delayed response
of the empty/full flags).

I would like to seek out opinions on any pitfall in this approach.

If there are others ways of reducing block ram usage, I would be glad
to hear them.

Thanks in advance.  TA TA.

Best Regards,
kahhean

Article: 36140
Subject: Second Scenario: BRAM usage reduction in FIFO design
From: kahhean@hotmail.com (Chua Kah Hean)
Date: 30 Oct 2001 22:39:17 -0800
Links: << >>  << T >>  << A >>
Hi all experts out there,

I would like to seek help in ways to reduce BRAM usage under two
scenarios.  This post contains only the second scenario.  I have
posted the first scenario in an earlier post.

(2)
I need a 1024 by 9 FIFO (independent read/write clocks).  If I
implement a 1024 by 16, I will be needing four 4Kbit BRAMs and
actually wasting 7/16 of the capacity.  Suppose I can afford only two
block RAMs.  Can I implement a 1024 by 8 (two blocks) BRAM-based FIFO
and then parallel it with a 1024 by 1 FIFO implemented with
distributed SRAM?

I am personally not uncomfortable with this approach, because the
full/empty flags of the two FIFOs may not be in-sync (especially now
that one is BRAM based, and another distributed SRAM based).  And if
they do go out of sync, the only way back is to reset both FIFOs.  :-P

Supposing I can solve the above problem, how do I code a 1024 by 1
FIFO for Xilinx Virtex-E FPGA?  I noticed that Coregen only allows
depth of 256 for both async FIFO and DPRAM.  If I have to build
SRAM-based from scratch, is there a primitive for SRAM-based Dual
Port?

Thanks in advance.  TA TA.

Regards,
kahhean

Article: 36141
Subject: Re: Clock attribute problem
From: Phillip <p.basters@hotmail.com>
Date: Wed, 31 Oct 2001 00:00:34 -0700
Links: << >>  << T >>  << A >>
Hello banana,

Unfortunaltely you cannot use both the rising as the falling edge in a single process. What you can do is 1. double the clock (by using a dll)
2. use the rising edge of that clock signal to generate a 'switch bit' (0 - 1 - 0 - 1 - 0)
3. yhen use the double clock in a process together with the 'switch bit' (if double_clk'event and double_clk = '1' then

if switchbit is 1 then 
   ...code...
else 
   ...code...
end if;

good luck, philip

Article: 36142
Subject: Clock attribute problem
From: cappellainfuocata@yahoo.it (Banana)
Date: 30 Oct 2001 23:35:06 -0800
Links: << >>  << T >>  << A >>
Good Morning,
I sinthesize a counter by 3 that I also use to interpolate the clock
by one,
this code is working with Active HDL but It give me some problem with
Xilinx , in particular during the synthesis I've the error :


Analyzing Entity <counter_divider_3> (Architecture
<counter_divider_3_arch>).
ERROR :   (VHDL_0047).
C:\Documenti\XilinxDesign\QPSKModulator\counter_divider_3.vhd (Line
23). Bad Form of Clock Condition.
--> 


Following there is the code, can you explain me why it doesn't work on
Xilinx and how I can modify it to work properly ?? The line that is
not good is
          "if clk'event then "

Thanks ,
          Bananone


library IEEE;
use IEEE.std_logic_1164.all;	  
use IEEE.std_logic_unsigned.all;

entity counter_divider_3 is
	port (
		clk			: in  STD_LOGIC;
		reset		: in  STD_LOGIC;
		count_3		: out STD_LOGIC_VECTOR (2 downto 0);
		clk_div_3 	: out STD_LOGIC 
		);
end counter_divider_3;


architecture counter_divider_3_arch of counter_divider_3 is
begin
	process (clk, reset)
		variable count_3_interno : STD_LOGIC_VECTOR (2 downto 0);
	begin
		if reset = '1' then  
			count_3_interno := "000"; 
		else
			if clk'event then 
				if falling_edge(clk) then
					if count_3_interno < 2 then
						count_3_interno := count_3_interno + 1;
					else  -- si deve riazzerare il contatore
						count_3_interno := "000";
						clk_div_3	<= '0' ;
					end if;
				else -- vuol dire che siamo sul fronte positivo
					if count_3_interno = 1 then
						clk_div_3 <= '1' ;
					else null;
					end if ;
				end if ;
			else null ;
			end if;
		end if ;
		count_3 <= count_3_interno;
	end process;
	
end counter_divider_3_arch;

Article: 36143
Subject: one SPROM for 2 XCS30XLs?
From: Utku Ozcan <ozcan@netas.com.tr>
Date: Wed, 31 Oct 2001 09:51:59 +0200
Links: << >>  << T >>  << A >>

In ds030.pdf, v1.8, I haven't found any PROM that can hold config data
of 2 separate XCS30XLs. Is there any such big PROM? Or shall I use 2
separate XC17S30XLs?

Utku



Article: 36144
Subject: Re: Field Programmable Logic in energy poor environments
From: Kolja Sulimma <kolja@sulimma.de>
Date: Wed, 31 Oct 2001 10:07:49 +0100
Links: << >>  << T >>  << A >>


Peter Alfke wrote:

> Irwin Kennedy wrote:
>
> > Hi,
> > I have a been thinking about the feasibility of using FPGA's in
> > wireless computing devices like the much talked about "3rd Generation"
> > cellular terminals. <snip>

Have a look at
http://bwrc.eecs.berkeley.edu/Research/Old/Infopad/

For the same netlist, an ASIC requires a lot less power than an FPGA, but
with FPGAs the netlist can be specializes to the current application. This
results in a smaller netlist and less switching.

<snip>

> > * Use less "leaky" transistors! (?)
>
> Until recently, there was extremely little junction leakage or
> sub-threshold leakage current. XC3000L could run on 50 microamps.
> Unfortunately, as we approach 100 nm technology, subthreshold leakage
> current becomes significant ( for every IC manufacturer. Ask Intel! )
>
> > * Use very low power SRAM for configuration memory (since this cannot
> > be switched off without incurring a huge latency/energy penalty next
> > time the FPGA is required).
>
> The configuration store used to consume microamps at most. It is only
> recently that deep sub-micron technology has increased this current to
> milliamps. There is little we can do about subthreshold leakage.

You can. You do not need fast transistors in your configuration memory.
As you are using an SOI technology for Virtex-II you could easily use a
higher
threshold voltage for your configuration SRAM transistors as for your CLB
logic transistors. (Hmmm. You probably do allready.)

You could do the same for your logic: A configuration bit could select
between fast, high leakage transistors and slow, low leakage transistors on a
per CLB basis.
The software tools could automatically set this bit for all CLBs that are not
on the critical path.
A power down mode could even dymically controll the threshold voltage.

Kolja Sulimma


Article: 36145
Subject: Re: Cloning someone else's IP core
From: Kolja Sulimma <kolja@sulimma.de>
Date: Wed, 31 Oct 2001 10:14:00 +0100
Links: << >>  << T >>  << A >>


Eric Smith wrote:

> Kevin Brace wrote:
> > Xilinx IP Evaluation License Agreement
> >
> > Xilinx Evaluation IP is owned and controlled by Xilinx and
> > must be used solely for design, simulation, implementation and
> > creation of design files limited to Xilinx devices or technologies.
> > Use with non-Xilinx devices or technologies is expressly prohibited
> > and immediately terminates your license.
>
> Kolja Sulimma <kolja@sulimma.de> writes:
> > This violates the "principle of first sale"
>
> Many things in the US violate the principle of first sale.  In general,
> U.S. copyright law does.  For instance, when you buy a VHS tape of a
> movie, you aren't allowed to use it for public exhibition.

Yes, But  Sony Home Entertainment video can not terminate your license when
you play it on a Grundig VCR.

> License agreements may well violate the principle of first sale.  The
> question is whether the license is binding and enforceable.

Kolja Sulimma


Article: 36146
Subject: Re: Cloning someone else's IP core
From: Kolja Sulimma <kolja@sulimma.de>
Date: Wed, 31 Oct 2001 10:27:40 +0100
Links: << >>  << T >>  << A >>


Eric Crabill wrote:

> Hi Kolja,
>
> I'm not a lawyer, so I can't say for sure...  But I think
> the Xilinx IP Evaluation License Agreement means that Xilinx
> retains ownership of the IP and is licensing it, not selling
> it, to whoever happens to download the evaluation.  I believe
> the same is true of the full-featured product.

A contract to transfer a good for an unlimited amount of time without paying
a recurrent fee is a purchase, no matter what it is called in the contract.
I am pretty sure that after purchasing the core package I own the
documentation, constraints files and the core generator software.

To be able to sell copies of the core to customers I need of course a
license as this can not be covered by the purchase alone. In a b2b context
this license can be more or less contain arbitrary agreements between me and
the vendor.

However, a license that states that Xilinx remains the owner of the IP is
worthless, as this would mean that Xilinx would own part of any product that
contains the core.
At least the end customer of my product must become owner of the netlist in
the FPGA.
And this of course allows me to become owner of the netlist myself.

(Would you buy a network adapter that comes with a license agreement that
states that part of the adapter is owned by Xilinx?)

If I use part of the netlist or source code in my own PCI core I of course
violate theire copyright. I can however not violate any Xilinx patents
because I licensed them from SIGPCI.

Kolja Sulimma

(DISCLAIMER: I want to state that I designed my own PCI core _before_ I
downloaded the Xilinx PCI core)


Article: 36147
Subject: Re: Cloning someone else's IP core
From: Kolja Sulimma <kolja@sulimma.de>
Date: Wed, 31 Oct 2001 10:31:38 +0100
Links: << >>  << T >>  << A >>
> I am sure that it depends on the skill of the designer, but from my
> experience trying to meeting timings with my own PCI IP core, I think
> meeting 33MHz PCI timings with a synthesizable (HDL based) PCI IP core
> with only automatic P&R is very challenging to say the least.

I found that most constraints violations were false paths.
For a PCI slave only TRDY and the output enable generation is a real problem.
In Virtex architectures there is a special logic block for this. The functionality was documented earlier
in this newsgroup.

Kolja Sulimma



Article: 36148
Subject: searchin for High density non bga packages something like PGA.
From: Richard Meester <rme@quest-innovations.com>
Date: Wed, 31 Oct 2001 10:42:41 +0100
Links: << >>  << T >>  << A >>
Hello all,

I am searching for a high density non bga package FPGA. I don't want to
use BGA since the high prototyping costs involved. My estimation is that
PGA packages are cheaper to produce. Are there any SPartanII Virtex/II/E
devices in this package?

Hope you can help.

Richard

--


Quest Innovations
tel: +31 (0) 227 604046
fax: +31 (0) 227 604053
http://www.quest-innovations.com



Article: 36149
Subject: Re: BRAM usage reduction in FIFO design: First Scenario
From: Falk Brunner <Falk.Brunner@gmx.de>
Date: Wed, 31 Oct 2001 10:48:08 +0100
Links: << >>  << T >>  << A >>
Chua Kah Hean schrieb:
> 
> Hi all experts out there,
> 
> I would like to seek help in ways to reduce BRAM usage in the
> following two scenarios.  This post contains only the first scenario.
> I will post the second scenario in another post.
> 
> (1)
> 
> I need a 2600 deep 16 bit wide FIFO (independent read/write clocks).
> 
> Using Xilinx's 4Kbit DPRAM, I realize I will need 16 blocks of them,
> giving me a 4096 by 16 bit FIFO instead.  (In fact, I am using the
> gray-coded FIFO as supplied by Xilinx in XAPP131, after some
> modification).
> 
> I am wasting too many block rams.  So I am thinking of cascading a
> 2048x16 and a 1024x16 FIFOs to give me a 3072x16 FIFO (I stuck to
> depth of power 2 to avoid muxing control/output and also to keep it
> easier for the gray codes).  There will be a simple data relaying
> between the 2048x16 and the 1024x16 FIFOs: whenever the second FIFO is
> not full and the first FIFO is not empty, read out the first FIFO and
> write into the second FIFO (taking into account the delayed response
> of the empty/full flags).
> 
> I would like to seek out opinions on any pitfall in this approach.

Hmm, possible BUT,

- this will increase latency

- using two FIFO from Xapp131 will take 3 clock nets (write clock,
in-between clock, read clock)

> 
> If there are others ways of reducing block ram usage, I would be glad
> to hear them.

Why dont you rearange the RAMS? Just use 12 256x16 BRAMs and create a
chip select from the address counter[11:8]. Since this is just a 4 bit
decoder it take only 1 LUT, so only 1 level of locic -> very fast

-- 
MFG
Falk





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