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 42200

Article: 42200
Subject: Re: I2C Slave sampling edge
From: "Roland Zitzke" <zitzke@gmx.de>
Date: Thu, 18 Apr 2002 15:18:48 +0200
Links: << >>  << T >>  << A >>
Hi,
for designing a
> slave device. They say on pag 12 that everything (the state machine
> and associated counters and shift registers are clocked on the falling
> edge of SCL, 'cause on heavy loaded systems the rise time of the SCL
> line may be very slow and that is dangerous on a clock a signal,
> 'cause it can generate noise on it.
>
> However, on Philips I2C bus Specification v2.1 Jan-2000, pag 8 they
> say the data on the SDA line must be stable during the HIGH period of
> the clock.
>
And I strongly recommand to pedantically follow the Philips specification.
Philips has invented the i2c bus and it does restrict the capacitive load on
the bus to allow for better signal forms.
In fact, on high speed mode active driving is performed (boosting) to speed
up level change.
Note that for instance tracing tools expect the Philips standard to be
kept - otherwise they report an error see: www.tracii.de

Hope this helps
- Roland



Article: 42201
Subject: Re: 8051 Core for Motor Electronics
From: Russell <rjshaw@iprimus.com.au>
Date: Thu, 18 Apr 2002 13:24:17 GMT
Links: << >>  << T >>  << A >>
Steffen Thieringer wrote:
> 
> "Russell" <rjshaw@iprimus.com.au> wrote
> 
> > How about just using a smaller fpga as a dedicated pwm/timer peripheral
> > for a cpu. Better still, a motor control dsp.
> 
> sure, a motor control dsp solution is better, but not done in 10 minutes.
> i need a driver now, and for the changeover (8051-hardware to dsp-hardware)
> i want to go for the 8051 software.
> 
> the way to go for a speeded up 8051 ip-core is for using a program which has
> done the job very well on 8051 hardware, but became obsolete because of
> speed.

How about a Cygnal 8051 which does 50mips(?) allegedly. The 8051 seems
quite a complicated bit of hardware that i think would require quite a
big fpga, and the speed wouldn't be that high.

Article: 42202
Subject: Re: 1000 I/O Pins -- What is cheapest FPGA?
From: "UK Gary" <deton8@aol.com>
Date: Thu, 18 Apr 2002 13:31:53 +0000 (UTC)
Links: << >>  << T >>  << A >>
"Paul Baxter" <pauljnospambaxter@hotnospammail.com> wrote in message
news:3cbeabe7$0$8514$cc9e4d1f@news.dial.pipex.com...

> Secondly, why the single chip requirement? Is it a space issue? Because a
> two chip mounted on either side of a board would probably take less space
> and allow you to have far cheaper chips.

It's because I have a certain number of input channels to deal with, and
they all have to be dealt with together.  It's a big pain in the ass if I
have to go off chip as that is what my current design does.  When it's all
in one chip, I can perform certain optimizations which will make the overall
product superior.

Guerre



Article: 42203
Subject: Re: I2C Slave sampling edge
From: "Tauno Voipio" <tauno.voipio@iki.fi.SPAMBAIT_REMOVE.invalid>
Date: Thu, 18 Apr 2002 13:45:22 GMT
Links: << >>  << T >>  << A >>

"Roland Zitzke" <zitzke@gmx.de> wrote in message
news:a9mh3o$n0$00$1@news.t-online.com...
> Hi,
> for designing a
> > slave device. They say on pag 12 that everything (the state machine
> > and associated counters and shift registers are clocked on the falling
> > edge of SCL, 'cause on heavy loaded systems the rise time of the SCL
> > line may be very slow and that is dangerous on a clock a signal,
> > 'cause it can generate noise on it.
> >
> > However, on Philips I2C bus Specification v2.1 Jan-2000, pag 8 they
> > say the data on the SDA line must be stable during the HIGH period of
> > the clock.
> >

Take the data with a latch enabled by clock high - it'll pick the state at
falling edge (-setup time).

Tauno Voipio
tauno voipio @ iki fi



Article: 42204
Subject: Re: problems with Nios 2.0
From: "Matjaz Finc" <matjaz.finc@fe.uni-lj.si>
Date: Thu, 18 Apr 2002 15:55:40 +0200
Links: << >>  << T >>  << A >>
Thanx. Now it's working.

Matjaz

"Kerri Golden" <kgolden1@yahoo.com> wrote in message
news:3abccfd0.0204160745.2095aae7@posting.google.com...
> Hi Matjaz,
> I have the workaround for you.
> Nios 2.0 does not support Win98, but there are steps to make it work.
>
> First - install to a directory that is all lowercase.
>
> In Nios:
> 1. make sure that your germs mon has a good vecbase offset. - the end
> is best.
> 2. generate.
> 3. it will fail - it says that it made the sdk, but it didn't. So, you
> have to make it.
> 4. so, through the start menu, go to programs>Altera>Excalibur Nios
> 2.0> Nios SDK Shell
> navigate to your project dir and type mk_custom_sdk
> this will make your sdk.
> 5. when it is done, press generate again on your Nios design.
>
> This has solved all the problems we've had on Win98.
>
> Also note that Quartus runs MUCH faster on 2000 than on 98.
>
>
> Kerri
>
>
> "Paul Baxter" <pauljnospambaxter@hotnospammail.com> wrote in message
news:<3cb69b58$0$227$cc9e4d1f@news.dial.pipex.com>...
> > "Matjaz Finc" <matjaz.finc@fe.uni-lj.si> wrote in message
> > news:a961r4$o9$1@planja.arnes.si...
> > > Hello!
> > >
> > > Has anyone experienced any problems with Nios 2.0 (build 224) in
Quartus
> >  II
> > > 2.0? I can't generate it with SOPC builder - it always terminates with
an
> > > error. Even Altera representatives and their official support were of
no
> > > help (!?) until recently - they said it will be fixed in next update.
My
> >  OS:
> > > win98.
> >
> > Though I can't help with the specifics, I have found that all of the
tools
> > run far more smoothly under windows 2k. I used to dual-boot, but frankly
I
> > don't bother with win9x now.
> >
> > Another tip is to make sure all tools run from paths with no spaces. The
> > batch command line scripts etc have a habit of sometimes not working
with
> > spaces. (e.g. Leonardo integration etc)
> >
> > Paul



Article: 42205
Subject: Re: 8051 Core for Motor Electronics
From: "Steffen Thieringer" <steffen.thieringer@nmb.co.uk>
Date: Thu, 18 Apr 2002 15:58:28 +0200
Links: << >>  << T >>  << A >>


> > "Russell" <rjshaw@iprimus.com.au> wrote

> How about a Cygnal 8051 which does 50mips(?) allegedly. The 8051 seems
> quite a complicated bit of hardware that i think would require quite a
> big fpga, and the speed wouldn't be that high.

Very interesting, thank you for the link.

But despite the speed of the cynal 8051, going for a software solution will
better for me, because of implementing several parts of circuit(pll, pwm,
memory..) in one single piece of silicon. accuraccy will then be increased.
Steffen


Article: 42206
Subject: [Xilinx TRACE timing] two phase clock
From: Tsoi Kuen Hung <khtsoi@cse.cuhk.edu.hk>
Date: 18 Apr 2002 13:58:56 GMT
Links: << >>  << T >>  << A >>
Hi,
I have a design using the VirtexE DLL (HF). The Xilinx ISE 4.1i TRACE tools
report a critical path which count the same path twice. The Xilinx ans. for
this is that I use a two phase clock in the design which need to x2 the path
delay. My desing has only two clock (clk and clkdiv from DLL) and only rising
edge is used. What is the possible error. If this is by default, how can I get
the *real* timing report. The original design implementing under 3.1i does not
have this problem. I just add a small DFF in the original datapath to make it
faster and switch to 4.1i. What's wrong? Thanks in advance.
---- Brittle

Article: 42207
Subject: Re: 1000 I/O Pins -- What is cheapest FPGA?
From: "Sum" <anonymous@nospam.invalid>
Date: Thu, 18 Apr 2002 16:06:24 +0200
Links: << >>  << T >>  << A >>

"Steffen Thieringer" <steffen.thieringer@nmb.co.uk> a écrit dans le message
news: a9mf36$4g6ga$1@ID-41871.news.dfncis.de...
>
> "UK Gary" <deton8@aol.com> schrieb im Newsbeitrag
> news:a9m9g1$28d$1@helle.btinternet.com...
> > I need to design something where the FPGA must have about 1000 I/O pins.
> > What is my cheapest option?  The design must fit into a single FPGA, but
> > apart from that and the pin count I don't have any special requirements.
>
> Upcoming Altera Stratix devices do have more than 1000 I/O (up to 1310)!
>
> http://www.altera.com/products/devices/stratix/overview/stx-overview.html

Xilinx Virtex-II have more than 1000 I/O (up to 1108) and they are not
upcoming !
http://www.xilinx.com/xlnx/xil_prodcat_landingpage.jsp?title=Platform+FPGAs





Article: 42208
Subject: Re: 8051 Core for Motor Electronics
From: Russell <rjshaw@iprimus.com.au>
Date: Thu, 18 Apr 2002 14:07:57 GMT
Links: << >>  << T >>  << A >>
Steffen Thieringer wrote:
> 
> > > "Russell" <rjshaw@iprimus.com.au> wrote
> 
> > How about a Cygnal 8051 which does 50mips(?) allegedly. The 8051 seems
> > quite a complicated bit of hardware that i think would require quite a
> > big fpga, and the speed wouldn't be that high.
> 
> Very interesting, thank you for the link.
> 
> But despite the speed of the cynal 8051, going for a software solution will
> better for me, because of implementing several parts of circuit(pll, pwm,
> memory..) in one single piece of silicon. accuraccy will then be increased.
> Steffen

Could try asking in news:comp.arch.embedded

Article: 42209
Subject: Re: GND Outputs being optimized out using FPGA Express 3.6.1 in
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Thu, 18 Apr 2002 07:44:56 -0700
Links: << >>  << T >>  << A >>



William,

I am just pointing out that some synthesis tools seem to fight you
sometimes, and there are attributes that one can use to keep things in
the design that don't seem to agree with the synthesis tool that you
might be using.

The technical answer in the answers database came from a case that was
solved by this advice.

I agree that if you tie a logic '0' to an IOB, and drive it off chip,
that the tools should not get in your way:  this is a long known, and
often used method of improving the ground connection to the die, and
reduces ground bounce by a significant amount (sometimes as much as
cutting it in half).

Austin

William L Hunter Jr wrote:

> Austin I can understand this for inputs but don't agree on outputs.
> Inputs may not be required to implement a desired function and
> therefore be optimized out.  But setting a output to logic0 (i.e.
> output<='0') should also work. ThanksBill
>
>      "Austin Lesea" <austin.lesea@xilinx.com> wrote in message
>      news:3CBDA28C.8496D40F@xilinx.com...William,
>
>      See:
>
>       http://support.xilinx.com/xlnx/xil_ans_display.jsp?iLanguageID=1&iCountryID=1&getPagePath=9780
>
>      For how to "keep" and "don't touch" instantiations so the
>      tools don't rip them out....
>
>      Austin
>
>      William L Hunter Jr wrote:
>
>     > Hi
>     >
>     > I am doing a Virtex-II design using ISE 4.2.01 and FPGA
>     > Express 3.6.1.  I
>     > have a few pins on my top level entity configured as
>     > outputs which are used
>     > to set the address of some external devices.  In my
>     > architecture, I assign
>     > these outputs a '0'.  FPGA express is optimizing out this
>     > connection.  If I
>     > look at the synthesis optimized schematic viewer the
>     > connection to gnd and
>     > output pad are missing.  How do you connect outputs to GND
>     > in VHDL?  I have
>     > always assigned them a '0' in the VHDL and this always
>     > worked.  Any ideas on
>     > what could be wrong??
>     >
>     > Thanks in advance
>     > Bill
>



Article: 42210
Subject: Re: Programming for FPGA or ASIC
From: prashantj@usa.net (Prashant)
Date: 18 Apr 2002 07:51:39 -0700
Links: << >>  << T >>  << A >>
Thanks. I will have a look at it.

Prashant


Hari Devanath <harid@xilinx.com> wrote in message news:<3CBDE120.82B83123@xilinx.com>...
> Hi Prashanth,
> In particular, if you want to write HDL for Xilinx devices, this is a good
> place to start:
> http://toolbox.xilinx.com/docsan/xilinx4/data/docs/sim/preface.html
> Hari.
> 
> Prashant wrote:
> 
> > hi,
> > I read in a post previously that VHDL/Verilog programming style
> > differs when being written for an FPGA from being written for an ASIC.
> > Is there any website which mentions these differences ? I'm looking to
> > see what should I take care of when programming for an FPGA.
> >
> > Thanks,
> > Prashant

Article: 42211
Subject: Re: Telecom Bus info
From: "Kris Vorwerk" <no.spam.vorwerk@myself.com>
Date: Thu, 18 Apr 2002 11:24:45 -0400
Links: << >>  << T >>  << A >>
Unfortunately, the Telecom Bus is not a standardized bus.  It was =
originally written up in an IEEE draft and then cancelled.  Here is the =
bibtex reference for this draft (you will definitely have to be an IEEE =
member to find this file -- it isn't available electronically AFAIK):


@manual{bib:IEEE-P1396.1,
 title =3D "IEEE Draft Std. P1396.1, COMBUS: A Backplane Bus and Package =
for
 SONET Applications",
 address =3D "IEEE, New York, N.Y.",
        year =3D "1990",
        month =3D "November"}


PMC-Sierra's standardized bus is the SBI bus, which is very similar to =
the Telecom bus (i.e., can run in 77.76 MHz parallel mode, etc.) .... =
PMC sells a number of switches, serializers, and so forth that support =
both the Telecom bus and their own SBI bus.  

If you're using PMC parts, and have access to PMC's datasheets, I =
recommend looking at the SBI Bus Serializer (SBS) datasheet, or the =
actual SBI Bus info sheet (but they may not make the latter publically =
available, I dunno).

cheers,
-kris


> I am looking for spec/details on Telecom Bus (TelecomBus). I searched
> on the web and all I could get is a number of links to PMC Sierra
> which  offers Telecom bus bridges, converters, serializers etc but
> does not offer much details about Telecom bus protocol. Neither could
> I find anything on the IEEE standards website.
> 
> I am designing an ATM Cell Processor in FPGA which has Telecom bus
> interface. I would greatly appreciate if someone can point me to
> relevant info on Telecom Bus.


Article: 42212
Subject: Re: Looking for SpartanXL demo board
From: spam_hater_7@email.com (Spam Hater)
Date: Thu, 18 Apr 2002 15:26:29 GMT
Links: << >>  << T >>  << A >>

It looks like a nice board.

I'll contact you remotely about the software.



On Tue, 16 Apr 2002 12:06:15 +0200, Johann Glaser
<Johann.Glaser@gmx.at> wrote:

>Hi!
>
>> I'm looking for a SpartanXL demo board.  Any good ones out there?
>> 
>> Something I can download with my XChecker cable, and get at all the
>> pins.
>
>Look at http://www.digilent.cc/. They have a nice board. It is available
>with a Spartan or a SpartanXL FPGA (3.3V version).
>
>It is quite cheap, I think less than $100. The only problem is, that
>Spartan and SpartanXL are not supported by Xilinx WebPack.
>
>Downloading the configuration is done by a parallel cable. If you need
>instructions or souce code, you can have mine. I wrote a small Pascal
>program for Linux, but it is easily ported to whatever you need.
>
>Bye
>  Hansi


Article: 42213
Subject: how comes the clk suddenly pause?
From: lunaris_s@yahoo.com (mm)
Date: 18 Apr 2002 08:31:23 -0700
Links: << >>  << T >>  << A >>
Hi,all
   I'm now encountering a strange thing with fpga.
   CLKA is input to fpga through one of the GCLK pin.After IBUFG it
has the net name CLKA_IN.I watch CLKA and CLKA_IN on an oscillograph.
   The strange thing is that sometimes when there are still
transitions on CLKA,CLKA_IN will pause for a while either staying "1"
or "0".
   I also have another clk domain in that fpga CLKB and
CLKB_IN.CLKB_IN also pauses at the same time as CLKA_IN though CLKB
sitll has transitions as usual.
   Isn't it strange?I've never ever heard of this.What should I do
now?
thx in advance
mm

Article: 42214
Subject: Re: 8051 Core for Motor Electronics
From: "Felix Bertram" <f.bertram@trenz-electronic.de>
Date: Thu, 18 Apr 2002 17:58:47 +0200
Links: << >>  << T >>  << A >>
Steffen,

> I am looking for a 8051 Core to implement in an Xilinx Spartan2 FPGA.
> The task for it is a control of a sensorless brushless DC motor.
> The software for the processor is already done but needs to be speeded up.
> So i need pwm, capture/compare units, several timers ans so on...
> I have several in my mind but not yet found 'the one'.
> I am not looking for freeware, but if there are some out there, please let
> me know:-)

we developed an 8051 derivative which is planned to be released as a
commercial product later on. We are currently looking for people, who are
interested in cooperating with us, to debug and verify the core in "real
life applications". We will provide the core for free, to enable such a
cooperation.

Our core is cycle-compatible to the original 8051. It provides 4 ports, 128
bytes of internal RAM, supports the complete instruction set, and implements
timers and interrupts. With this setup, the core implements quite nicely in
Spartan2 FPGAs, as the following map report reveals. However, we are
assuming that there is still some room for optimizations.

Design Summary:
   Number of Slices:              1,245 out of  2,352   52%
   Number of Slices containing
      unrelated logic:                0 out of  1,245    0%
   Total Number Slice Registers:    495 out of  4,704   10%
      Number used as Flip Flops:                  451
      Number used as Latches:                      44
   Total Number 4 input LUTs:     2,297 out of  4,704   48%
      Number used as LUTs:                      2,211
      Number used as a route-thru:                 22
      Number used for 32x1 RAMs:                   64
      (Two LUTs used per 32x1 RAM)
   Number of bonded IOBs:           148 out of    140  105%
      IOB Flip Flops:                              94
   Number of GCLKs:                   1 out of      4   25%
   Number of GCLKIOBs:                1 out of      4   25%
Total equivalent gate count for design:  27,016
Additional JTAG gate count for IOBs:  7,152


Steffen, if you are interested in working with us, please drop me a line, so
that we may discuss further issues.

Best regards

Felix
_____
Dipl.-Ing. Felix Bertram
Trenz Electronic
Duenner Kirchweg 77
D - 32257 Buende
Tel.: +49 (0) 5223 49 39 755
Fax.: +49 (0) 5223 48 945
Mailto:f.bertram@trenz-electronic.de
http://www.trenz-electronic.de





Article: 42215
Subject: Re: Bidirectionnal bus...multiple sources driving the same signal...
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Thu, 18 Apr 2002 18:09:02 +0200
Links: << >>  << T >>  << A >>
"Eric Kral" <eric.kral@planetinternet.be> schrieb im Newsbeitrag
news:a9m34u$oaa$1@reader08.wxs.nl...
> Hi,
>
> I am trying to realize a bidirectionnal bus between two components. It
seems

Inside the FPGA, the is no need for a "real" bidirectional bus. It can be
made, but why? The are plenty of routing ressources, so use them and use
unidirectional busses.

> to work (I can synthesize the design but the simulation doesn't work
> properly, moreover I cannot implement the design and I can find the
> problem(s)).
> My synthesis tool is the latest version of the Webpack from Xilinx.
> The idea of the design is to enable a fast communication between several
> FPGA. The target FPGA is a Virtex-E from Xilinx. The TDM entity is to
enable

OK, this is another story, but then you have to be carefull. And ist not
easy to test the two modules inside one FPGA with INTERNAL connection, since
this will change the IO drivers. So if you want to connect multiple FPGAs,
create a test toplevel, drive the IOs outside (just as you would in the
final application) and connect then using physical wires.

--
MfG
Falk







Article: 42216
Subject: Understanding clock routing (or not)
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Thu, 18 Apr 2002 18:20:42 +0200
Links: << >>  << T >>  << A >>
Hello everyone,

Iam afraid Iam missing a major point in FPGA clock routing.
I have a Virtex-E (300), feed a clock from a Xtal (36 MHz) into a global
clock input, divide it down by 8 by means of a DLL and provide this clock to
a clock buffer. A big part of my design runs at this ~5Mhz clock. So far, so
good. But now comes the tricky part. I have some block which handles a
interface with another device (its a UTOPIA2). All output signals use IO
FFs. The clock is simply routed to an IO pin and from there to the other
chip. The interface is fully synchronous, everything works on the rising
edge.
I expected, that the datas are slightly delayed behing the clock, by the
clock-2-output time of the IOB FF, but they were NOT!!!!
The clock IS delayed by ~3 ns behind the data!!!!
First question, how does this happen? I thougt the clock nets are low skew,
means the clock arrives at all cell at the same time. But the P&R tools tell
me, that the clock is driving non-clock load, which can cause skew trouble
(and it does :-(
Second, what is the clean approach for clock distribution (without DLL
usage)
At the moment, I use a registers clocked on the other (falling) edge to
"simulate" the right phase relation between clock and data.

--
MfG
Falk




Article: 42217
Subject: Re: Bidirectionnal bus...multiple sources driving the same signal...
From: Eric Crabill <eric.crabill@xilinx.com>
Date: Thu, 18 Apr 2002 09:27:07 -0700
Links: << >>  << T >>  << A >>

Hi,

Your error message indicates that the synthesized design
has two TBUF (three-state buffers) named srampad_I4_0 and
tdmpad_I26 -- which have their outputs tied together, and
also share the same output enable signal.

So, in operation, you can have a situation where both of
them can turn on at the same time, which is an electrical
no-no.  I suspect you did not do this on purpose.  You need
to go back and re-check your code.  If you can't simulate
it correctly, you cannot possibly expect the synthesis
result to be meaningful.  I'd spend your time debugging
your code in the simulator.

Hope that helps,
Eric

Eric Kral wrote:
> Hi,
> 
> I am trying to realize a bidirectionnal bus between two components. It seems
> to work (I can synthesize the design but the simulation doesn't work
> properly, moreover I cannot implement the design and I can find the
> problem(s)).
> My synthesis tool is the latest version of the Webpack from Xilinx.
> The idea of the design is to enable a fast communication between several
> FPGA. The target FPGA is a Virtex-E from Xilinx. The TDM entity is to enable
> the fast communication (TDM stands for time division multiplexing).
> The SRAM is the interface for the SelectRam+ which is a blackbox in the
> virtex-E (RAM of 4096 bits)
> Finally the COMP entity should be the top design implemented in the FPGA.

Article: 42218
Subject: Re: 1000 I/O Pins -- What is cheapest FPGA?
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Thu, 18 Apr 2002 09:49:06 -0700
Links: << >>  << T >>  << A >>


Sum wrote:

> Xilinx Virtex-II have more than 1000 I/O (up to 1108) and they are not
> upcoming !
> http://www.xilinx.com/xlnx/xil_prodcat_landingpage.jsp?title=Platform+FPGAs

Yes, the XC2V6000 is in production, and it has 1,104 user I/O pins in the FF1517
package.  The small-quantity price is several thousand dollars. You can find it
listed at the distributors web sites. ( insight-electronics.com, avnet.com,
nuhorizons.com,)

Peter Alfke, Xilinx Applications



Article: 42219
Subject: Re: just bought Cohen's book...Real Chip Design and Verification using Verilog and VHDL
From: vhdlcohen@aol.com (VhdlCohen)
Date: 18 Apr 2002 17:04:04 GMT
Links: << >>  << T >>  << A >>
There were several comments about that topic, so I'll address some of the
issues here. 
1. Desired book: 
<What I would like to see in a book is simple:  Take five or six projects of
increasing complexity and have a chapter or set of chapters taking each one
from inception to completion.  This means from generating specifications, to
choosing the chip, writing the code, fitting it to the chip, making changes
and programming the chip.  That would be a very useful book. >
BEN: There are several issues in this proposal: Requirements definition,
architectural definition, testbench planning, design, verification, and
documentation. 
I have attempted to do that in my book "Component Design by Example ",  2001
isbn  0-9705394-0-1" where I used a UART of full complexity with error
detection, FIFO, and CPU interface with priority encoding.  That design is far
in complexity from the UART that is at my web site that merely consisits of a
parallel to serial, and serial to parallel register with clock detection.  The
UART in the book includes a reusable parser package.  The intent of that book
was really the process, and issues with design definition, implementation, and
verification.  

<I'm afraid that the ad-hoc style of presentation with poor and
inconsistently styled diagrams is the first thing to hit you.
Tables of text that look like they were converted to bitmap and then
stretched are really unacceptable in this day and age. Same applies to
certain equations (how long would that have taken to typeset
properly?)Probably an easy way of scanning source material though :( >
BEN: The goals for the book "Real Chip Design and Verification Using Verilog
and VHDL" were:
1. Address issues that I experienced with designers who thought that knowing an
HDL meant that you were ready to do designs.  That in why in the chapter on
Fundamentals, I covered with guidelines, the topics of reg, latch, 2 edges of
clk, and various types of counters (e.g., LFSR, Johnson, etc), memories,
primitives, and clockboost).  It is true that some material were extracted,
with permission, from various vendors and authors.  However, the object here
was to put relevent information in one place.  
2. Address the topic of asynchronous signals.  Again, you'll be surprised as to
the number of engineers who fail to understand how to deal with that issue. 
3. Verification: Here I used 2 examples: a counter and a memory with EDAC.
I was working on a project that spent many many (really many!) hours on the
design and verification of a memory with EDAC, anyway, far more thatn the $80
cost of the book.
4. Design process in defining a control machine: I have seen horrible code an
processes in the definition of such design.  Here I attempted to define a
usable, and hopefully practical approach  with a process, or a flow.  
5. In arithmetic machines, I was defining the usable packages, and issues with
Verilog. 
6. Mixed simulation is something that is common now because many vendors
require Verilog for the release sims, and many IPs come in Verilog (at first). 
7. I wanted to present Verilog as an HDL for several reasons: 
 a. Verilog is really gaining in popularity.
 b. Many IPs come in VErilog first.  
 c. Release sioms are often required in Verilog. 
By demonstrating designs in both HDLs, my intent was to facilitate the
transition from one HDL into the other, and to sho that the real design issues
are not necessarily the HDL, but the process and architectures.  I also
includes a tutorial of Verilog for VHDL users to explain the differences
between the two HDL.  

<Granted there's a CD with a lot of the files on, and the book does provide a
central repository, but I don't think that justifys the expenditure of your
effort> 
The book sells for $80.  If you can do better thatn than by searching things on
the web, extracting meaningful info, print the material needed for less thatn
$80, my hat goes to you!

<There are items that everyone from newcomers to moderately well-versed HDL
engineers will learn, but it isn't presented in a coherent style, instead my
impression is that its stitching articles together in the hope of getting a
book published quickly>
You are entitled to your opinions.  If you feel cheated after receiving the
book, you may return for a refund. However, the reviewers of the book,
including corporate VPs from Cadence and Synplicity really felt taht the book
had a lot of values as an investment.  
By the way, writing a book, and going the process of publishing (wheter thru a
major publisher or self-publishing) is not a "quick" manner and DOES take many,
many hours of work, particularly when the code has to also be verified and
explained.  It also does take research on the web, something that was brought
up as a "trivial manner, not worth the effort of the consolidation of the
necessary, and filtered information". 

I hope that this explanation provides some info as to why I wrote the book.  I
know that I did my best, and with a lot of effort.  Style is an issue own by
the beholder.  My style, like it or not, is by complete example and graphics. 
----------------------------------------------------------------------------
Ben Cohen     Publisher, Trainer, Consultant    (310) 721-4830  
http://www.vhdlcohen.com/                 vhdlcohen@aol.com  
Author of following textbooks: 
* Real Chip Design and Verification Using Verilog and VHDL, 2002 isbn
0-9705394-2-8 
* Component Design by Example ",  2001 isbn  0-9705394-0-1
* VHDL Coding Styles and Methodologies, 2nd Edition, 1999 isbn 0-7923-8474-1
* VHDL Answers to Frequently Asked Questions, 2nd Edition, isbn 0-7923-8115
------------------------------------------------------------------------------


Article: 42220
Subject: Re: Understanding clock routing (or not)
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Thu, 18 Apr 2002 10:08:48 -0700
Links: << >>  << T >>  << A >>
Falk, I do not really understand your design, but here is a brute-force
solution:
Use the opposite clock edge for your interface. This has beeenan old remedy,
applicable in any low-frequancy situation. And 36 MHz is low frequency these
days.

As I said, this is not an explanation, but rather a fix.
Gruß
Peter
=========================
Falk Brunner wrote:

> Hello everyone,
>
> Iam afraid Iam missing a major point in FPGA clock routing.
> I have a Virtex-E (300), feed a clock from a Xtal (36 MHz) into a global
> clock input, divide it down by 8 by means of a DLL and provide this clock to
> a clock buffer. A big part of my design runs at this ~5Mhz clock. So far, so
> good. But now comes the tricky part. I have some block which handles a
> interface with another device (its a UTOPIA2). All output signals use IO
> FFs. The clock is simply routed to an IO pin and from there to the other
> chip. The interface is fully synchronous, everything works on the rising
> edge.
> I expected, that the datas are slightly delayed behing the clock, by the
> clock-2-output time of the IOB FF, but they were NOT!!!!
> The clock IS delayed by ~3 ns behind the data!!!!
> First question, how does this happen? I thougt the clock nets are low skew,
> means the clock arrives at all cell at the same time. But the P&R tools tell
> me, that the clock is driving non-clock load, which can cause skew trouble
> (and it does :-(
> Second, what is the clean approach for clock distribution (without DLL
> usage)
> At the moment, I use a registers clocked on the other (falling) edge to
> "simulate" the right phase relation between clock and data.
>
> --
> MfG
> Falk


Article: 42221
Subject: Xilinx Programmable World 2002 - Review
From: Greg Neff <gregeneff@yahoo.com>
Date: Thu, 18 Apr 2002 13:17:24 -0400
Links: << >>  << T >>  << A >>
Yesterday I attended PW2002 in Toronto.  I have attended Xilinx
promotional seminars since the days of the XC2000.  Until now, I have
found that the Xilinx seminars provided enough technical information
to make the marketing content tolerable.  PW2002 was a radical shift
toward marketing glitz, at the expense of useful technical content.  

Heck, we weren't even supplied with VirtexII Pro data books, I had to
ask for one from our rep.  

During the IBM "technical training" track I actually got angry.  At
least 90% of the presentation was IBM marketing, describing various
IBM PowerPC processors and cores.  Now I can't remember which features
are in the core that is actually used in the PowerPC, since we were
bombarded with features spanning the entire IBM product line.

The keynote speakers talked about stuff that had nothing to do with
the VirtexII Pro, except in abstraction.

Apparently, this seminar was not targeted at engineers.  I'll keep
that in mind the next time I get an invite to a Xilinx seminar.


===================================
Greg Neff
VP Engineering
*Microsym* Computers Inc.
greg@guesswhichwordgoeshere.com

Article: 42222
(removed)


Article: 42223
Subject: Re: how comes the clk suddenly pause?
From: "Wayne" <whalcomb@lucent.com>
Date: Thu, 18 Apr 2002 12:23:49 -0500
Links: << >>  << T >>  << A >>
Is it feeding a DLL ?

"mm" <lunaris_s@yahoo.com> wrote in message
news:a2d0027.0204180731.3ca28548@posting.google.com...
> Hi,all
>    I'm now encountering a strange thing with fpga.
>    CLKA is input to fpga through one of the GCLK pin.After IBUFG it
> has the net name CLKA_IN.I watch CLKA and CLKA_IN on an oscillograph.
>    The strange thing is that sometimes when there are still
> transitions on CLKA,CLKA_IN will pause for a while either staying "1"
> or "0".
>    I also have another clk domain in that fpga CLKB and
> CLKB_IN.CLKB_IN also pauses at the same time as CLKA_IN though CLKB
> sitll has transitions as usual.
>    Isn't it strange?I've never ever heard of this.What should I do
> now?
> thx in advance
> mm



Article: 42224
Subject: Re: 8051 Core for Motor Electronics
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Thu, 18 Apr 2002 19:36:32 +0200
Links: << >>  << T >>  << A >>
"Felix Bertram" <f.bertram@trenz-electronic.de> schrieb im Newsbeitrag
news:a9mqfj$4l83o$1@ID-31589.news.dfncis.de...
> Steffen,
>
> > I am looking for a 8051 Core to implement in an Xilinx Spartan2 FPGA.

Hmm, maybe its worth a try to have a look at xapp213. The is a 8-Bit RISC,
very small, but very clever and fast.

16 8 bit register
256 instructions program size
>20 MIPS
8 bit IO port for additional hardware.
interrupt capability.
uses just 5% in a Spartan-II 100, even the smallest Spartan-II (XC2S15) can
hold 1 and has still 40% free.
all you have to add are the timers/PWM.
This is easy, but this will change the project from a pure software issue to
a FPGA/hardware issue.

--
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