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

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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

Custom Search

Messages from 68200

Article: 68200
Subject: Re: Question : Serial PROM
From: Eric Smith <eric-no-spam-for-me@brouhaha.com>
Date: 29 Mar 2004 16:57:22 -0800
Links: << >>  << T >>  << A >>
someone2003@gawab.com (someone2003@gawab.com) writes:
> I was wondering if there is a way to "copy" a serial prom device

Yes.

Article: 68201
Subject: Re: study verilog or vhdl?
From: johnjakson@yahoo.com (john jakson)
Date: 29 Mar 2004 17:57:49 -0800
Links: << >>  << T >>  << A >>
Jim Lewis <Jim@SynthWorks.com> wrote in message news:<106gm8h2p2vf8a5@corp.supernews.com>...
> Jim Lewis wrote:
> >>What is the user base really at?  Based on newsgroup
> >>traffic (comp.lang.vhdl vs. comp.lang.verilog),
> >>VHDL usage is 2X that of Verilog.
> >>
> 
> john jakson wrote:
> > Aha, about as useless as counting the no of HDL textbooks in the local
> > EE store. But the book writing industry is remote from HW design, they
> > write what the Universities request. Infact its a wonder Verilog
> > survived the almost 100% teaching of VHDL in most schools. I rarely
> > hear of Verilog being tought in any school.
> 
> How can you question this measure and not realize that
> what Cooley has published is nothing more than twisting
> EDA sales numbers trying to make the user base look like
> it is mostly Verilog.  In Cooley's numbers, I did not
> see any normalization for tool prices.
> 
> So why do you let Cooley pull the wool over your eyes.
> 

You jest surely!

Going back a few yrs long before Virtex and any interest in FPGAs I
was immersed only in ASIC design, where ever I worked, and everyone I
knew was using Verilog. I had no contact with edu, dod, eec, fpga EEs
so I never saw VHDL. I did see a dearth of Verilog books in the
stores. I very much got the impression that the professors or whoever
were out to save the ASIC industry from itself, I could be mistaken.
Imagine if all the C++ books got replaced with ADA or Java both of
which have avid users that claim them to be better than C++. Would
that force people to switch, nope don't think so (I respect all
languages, but I don't have time to use them all). I am also sure the
VHDL books were not aimed at FPGA users either (FPGAs were too puny
back then), they were aimed at any HW engineers, but I felt they were
mostly replicating each other. I only like or use the Douglas Smith
book for V to V comparison.

The ASIC industry is a very low level detail place where lack of libs
for VHDL harmed it, that didn't get better till too late. The VHDL
users are obviously working at higher levels than the typical Verilog
user since the language allows so much more abstraction. Sometimes
better features turn off potential users as some HW EEs are less
familiar with CS and more powerfull langs. The improvements to Verilog
ie SV are almost entirely aimed at the testbench crowd who often come
from CS and not so much EE backgrounds. I would say that most of the
Verilog people I know are also averse to programming and anti C if it
should suggest any usefulness to HW design.

As for Cooley, I think he's ok, he has his pulse on the industry,
mostly ASICs though, and lots of people write him with inside info. As
for the EDA industry and Cooley, it is Cooley who is left to
straighten out the sales figures from EDA by measuring tapeouts, info
that these companies are very secretive about. He gets his reports
from credible sources with the plusses and lots of minusses. The usual
thing is to do with layout SW, Magma v Cadence v Avanti etc. The lang
issue was not top of list.

If you don't believe the ASIC industry stats, go ask him yourself or
see his site or visit DAC.

I don't know how tool prices can be normalized, I do see high end FPGA
tools going up in price, just look at all the Symplicity goodies, far
out of my reach for sure. I suspect the ASIC EDA companies will come
to FPGAs for future growth, if FPGAs really could handle 1M gates
equiv, they are going to need far better tools than we have now. The
layout SW is a prime example.

Since there never was a comp.arch.asic (or vlsi or cmos whatever),
Cooley site serves a similar purpose to this NG. One should expect a
huge influx of new ASIC guys into FPGAs, some may even show up here.

regards

johnjakson_usa_com

Article: 68202
Subject: Re: CLB usage: Xilinx XCS20 and Foundation 3.1
From: Ken McElvain <ken@synplicity.com>
Date: Tue, 30 Mar 2004 03:58:31 GMT
Links: << >>  << T >>  << A >>
Try another synthesis tool.  Synplify can usually get a substantial
area reduction which could give you the room for your new functions.

John Braun wrote:

> I have a design that come very close to filling a Xilinx XCS20 - it
> has 400
> complex logic blocks, and I'm using 388 (97% usage).  Under Foundation
> 3.1,
> the design needs multiple reentrant routes to work properly, and
> sometimes
> bringing out debug information to a spare pin causes the whole thing
> to fail to route.
> 
> Yes, I know this is cutting it close, but I'm trying to make a new set
> of functions fit on an old card.  We're stuck with this design for now
> (we could go to the larger XCS40, but
> that would require getting new parts and replacing the existing ones -
> time
> consuming).  My questions are: what is a good rule of thumb for CLB
> usage %,
> and are there particular issues with Foundation 3.1 that I should be
> aware
> of?
> 
> (reply in thread so others can benefit from your expertise too...)
> 
> Thanks in advance,
> 
> John


Article: 68203
Subject: Re: FPGA Engineer w/clearance - where do you look for a job?
From: johnjakson@yahoo.com (john jakson)
Date: 29 Mar 2004 21:06:11 -0800
Links: << >>  << T >>  << A >>
kellydingee@adelphia.net (Kelly) wrote in message news:<62b176ec.0403291155.425d4272@posting.google.com>...
> I'm trying to find an experienced FPGA Engineer (currently holding a
> clearance) - where do people like this look for jobs typically?  The
> clearance is critical to this opening.
> 
> Thank you for your help - 
> 
> Kelly
> kellydingee@adelphia.net

How about careerbuilder or monster, they both have lots of "clearance
only" positions. Somebody must be looking.

Article: 68204
Subject: Re: ISE and EDK Incompatible?
From: "Frank van Eijkelenburg" <someone@work.com>
Date: Tue, 30 Mar 2004 08:36:55 +0200
Links: << >>  << T >>  << A >>
You can have multiple installations of ISE on your pc (choose a seperate
directory for each installation). Just make sure the %XILINX% environment
variable is pointing to the desired installation for that moment.

Frank

"George" <george_mercury@hotmail.com> wrote in message
news:6d167a0a.0403290615.3e97ec3f@posting.google.com...
> Hello,
> Here In our company we have just installed the new ISE 6.2 plus the
> SP1. Previoudsly we have been using the ISE 6.1 SP3 and the EDK 6.1
> SP2. Now that we have installed the ISE6.2 the EDK ( Xilinx Platform
> Studio ) won't start. We get the following error: $XILINX does not
> point to and iSE 6.1 installation. So does this mean that we have to
> install the ISE6.1 back?
>
> Best Regards
> George Mercury



Article: 68205
Subject: Re: study verilog or vhdl?
From: hmurray@suespammers.org (Hal Murray)
Date: Tue, 30 Mar 2004 08:45:05 -0000
Links: << >>  << T >>  << A >>
>PLAIN TEXT IS GOOD.

I'm happy with binary files as long as there is enough info available to
parse and manipulate them.  One good way to do that is to provide
programs to convert between binary and text format.  Extra credit
if they provide the source code so I can use their .h files rather
than inventing my own and look at the way the process things for
some hints when getting started.  The vendors probably need that sort of
program anyway for their own internal use.


I remembered another thing for the list...

You need to be able to figure all of the library files that are needed
to compile/build a design.  And which tools are used and what files
that includes.  The basic idea is that when you are happy with a design,
you want to archive a copy of "everything" needed to reconstruct
your working environment so you can put it back together again if
you ever need to make a simple change.  A note with the version of
the tools might be enough if you have the CDs handy, but then you
have to keep track of patches and such.

If you are slightly paranoid, you need to know the version of the OS
and such.  Setting the hard disk aside might be a reasonable approach.

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.


Article: 68206
Subject: Re: study verilog or vhdl?
From: hmurray@suespammers.org (Hal Murray)
Date: Tue, 30 Mar 2004 08:51:45 -0000
Links: << >>  << T >>  << A >>
>Now, the people who try to use C as a hardware description language
>are the ones I really don't understand.   Two that I know about
>(though not much about) are Handel C and Transmogrifier C.

What's wrong with c for "describing" hardware?  You just need
a good library.  The actual code is just subroutine calls and
argument shuffeling.  Using c might be slightly less dense than
better approaches, but it works fine.

-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.


Article: 68207
Subject: Re: CLB usage: Xilinx XCS20 and Foundation 3.1
From: Bedrich <>
Date: Tue, 30 Mar 2004 00:53:14 -0800
Links: << >>  << T >>  << A >>
Learn the architecture and use schematic design entry. This gives you almost 
full control over FPGA resources. Think how to make the design denser (for 
example converting FFs to RAM and working in time multiplex - for example 
recently I have made this way 16 UART receivers with almost equal FPGA resources 
as the single one) - human brain is still much better than any expert system. 


Article: 68208
Subject: Re: Schematic Edition Tool : Suggestions
From: Bedrich <>
Date: Tue, 30 Mar 2004 01:19:57 -0800
Links: << >>  << T >>  << A >>
You may wonder but we make almost all our designs in OrCAD SDT 386+. For us (no 
flame, please! ;-) it is the best design entry tool ever. We drawn compact 
libraries for Spartan2+ and Virtex FPGAs, we modified the EDIF 2 0 0 netlister 
and we created an interface for embedding VHDL blocks into the schematics. 
Although this old program has some disadvantages in today's Windows envirinment, 
it allows for almost direct "brain-to-PC" connection than other, not so user's 
friendly tools before we will create our own one. And the 2D schematics is 
significantly more suitable for parallel tasks in FPGA than text files 
description. Last note - even large projects are composed from small block, so I 
do not agree the argument that schematics is not well suited for large 
designs... 



Article: 68209
Subject: Virtex2 partial reconfiguration
From: "Jaan Sirp" <jaan.sirp@liewenthal.ee>
Date: Tue, 30 Mar 2004 01:27:14 -0800
Links: << >>  << T >>  << A >>
When using modular design and partial reconfiguration options, the ISE 
translator replaces all FD-s with FDC-s, CLR-pin (SR) connected to constant GND. 
As SR is shared pin of the slice, SRL (or RAM) and flip-flop next to it can't 
reside in the same slice and if clock is 300+MHz, timings aren't meet. If 
partial configuration option isn't used, the implementation is correct. The FD 
replacing with FDC which has constantly grounded CLR is meanless. How to avoid 
this? 

Jaan. 



Article: 68210
Subject: Is there any Sync separator IP(Intellectual property) exists?
From: "coreDEVIL" <coredev@korea.com>
Date: Tue, 30 Mar 2004 18:58:36 +0900
Links: << >>  << T >>  << A >>
I need sync separator IP which works like Gennum GS4982.

Can I get any information about that?

Or is there any web site that introduces existing IPs?



Thank you for your answer.


Article: 68211
Subject: Re: Is there any Sync separator IP(Intellectual property) exists?
From: Jonathan Bromley <jonathan.bromley@doulos.com>
Date: Tue, 30 Mar 2004 12:05:34 +0100
Links: << >>  << T >>  << A >>
On Tue, 30 Mar 2004 18:58:36 +0900, "coreDEVIL" 
<coredev@korea.com> wrote:

>I need sync separator IP which works like Gennum GS4982.
>Can I get any information about that?

Gennum's website :-)

But this is an analog function.  You could not implement
it in any FPGA I can think of.

You could digitise the incoming video, remembering that
your A/D conversion needs enough dynamic range to digitise
sync tips as well as the full video span.  That means you
need one more bit of precision in the A/D than you would
otherwise require.  Having digitised it, you would then
need some fairly simple DSP to identify the sync tips
and black level, slice the sync signal at a threshold
halfway between these two values, then black-level-clamp
your video by subtracting the black level from every 
sample.  Beware interactions between these DC-restoring 
operations in your DSP, and the AC coupling circuitry that
is sure to exist at the input of your A/D.

Note also that the GS4982 also performs various sanity
checks on the incoming signal, such as checking that its
line scan rate is reasonable.  These are fairly easy to
do in DSP, but need some careful thought to be sure that
you are implementing the checks in a sufficiently 
general way.

This is a fairly straightforward project and we would be
happy to quote you for the development.

>Or is there any web site that introduces existing IPs?

www.opencores.org and www.free-ip.com are good places to
start, although I wasn't hugely impressed with the free-ip
CORDIC module when I looked at it.  Caveat raptor.
-- 
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223          mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573                Web: http://www.doulos.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

Article: 68212
Subject: Re: Multiple DCM ? (Virtex II)
From: Rudolf Usselmann <russelmann@hotmail.com>
Date: Tue, 30 Mar 2004 19:00:34 +0700
Links: << >>  << T >>  << A >>
Rudolf Usselmann wrote:
> 
> So, I'm trying to use these fancy DCMs that everybody
> is talking about ! Looking at the data sheet, these are
> some damn impressive pieces of work.
> 
> So I have a design that uses two clocks. Neither one is
> available on my development PCB. Lets throw in some numbers
> to make it a bit simpler: I need 60 and 75 MHz, my development
> board only has a 100MHz oscillator.
> 
> Hmm, sounds like a simple problem to solve. Pull up Arch.
> Gen., configure two DCMs, one fo 60 the other for 75 MHz.
> edit the files to pull out the IBUFG to serve both of the
> DCMs, done, right ?! wrong !
> 
> Now the fight with the tools begins. First Synplify, had to
> learn I had to put the clock in to separate groups so it
> will constrain them independently. So far so good.
> 
> No I want for ISE to treat them separate as well. But it
> just refuses. It will accept the input clock as 100MHz,
> and one of the other once (either 60 or 75) as an independent
> clock, but the other one it will make dependent on the input
> clock, and come up with cycle times of 22++ nS.
> 
> Now I did do a trial synthesis and P&R without the DCMs
> and know that theoretically there should be no problems
> whatsoever to meet the required timing. But ISE starts of
> the timing report with some 10 ns as the first entry:
> Slack:                  -9.417ns (requirement - (data path - clock path
> skew + uncertainty))
> 
> I have no idea where it is taking this from !
> 
> How do I tell ISE to treat ALL clocks separate, even though
> some of them are used in the same module ?
> 
> Thanks,
> rudi


I'm following up to my own message, I made some additional
discoveries:

1) It appears the architecture wizard is instantiating
BUFGs and IBUFG. Both of them are modeled as a simple
buffer, with out the "synthesis black box" directive,
which makes the syntheis took remove them. Adding "synthesis
black_box" to both modules in the xilinx libraries, leaves
them.

2) If I use different clock inputs for both DCMs all my
problems seem to go away. How can I use a single clock
input for both DCMs ???

Thanks,
rudi               
========================================================
   ASICS.ws   ::: Solutions for your ASIC/FPGA needs :::
..............::: FPGAs * Full Custom ICs * IP Cores :::
FREE IP Cores -> http://www.asics.ws/  <- FREE EDA Tools


Article: 68213
Subject: FIFO Depth(Length) Calculation
From: Anand P Paralkar <anandp@sasken.nospam.com>
Date: Tue, 30 Mar 2004 18:01:23 +0530
Links: << >>  << T >>  << A >>
Hi,

I have often seen the question "How do you calculate the
required FIFO depth (length) ?" been asked.  How do you
answer this question?

I wonder if this is what people mean:

  Consider two asynchronous blocks each generating data
  at rates R1 and R2 such that R1 > R2.  Calculate the depth
  of a FIFO required between these two blocks so that there
  is no data dropped.  No feedback/handshake mechnism should
  be assumed between the two blocks.

In such a case, do we have a general solution?

Thanks,
Anand


Article: 68214
Subject: Re: Is there any Sync separator IP(Intellectual property) exists?
From: "coreDEVIL" <coredev@korea.com>
Date: Tue, 30 Mar 2004 21:53:07 +0900
Links: << >>  << T >>  << A >>
"Jonathan Bromley" <jonathan.bromley@doulos.com> wrote in message
news:9fki60l45bt7p7eu8l0ojttb9ju30klue0@4ax.com...
> On Tue, 30 Mar 2004 18:58:36 +0900, "coreDEVIL"
> <coredev@korea.com> wrote:
>
> >I need sync separator IP which works like Gennum GS4982.
> >Can I get any information about that?
>
> Gennum's website :-)
>
> But this is an analog function.  You could not implement
> it in any FPGA I can think of.
>
> You could digitise the incoming video, remembering that
> your A/D conversion needs enough dynamic range to digitise
> sync tips as well as the full video span.  That means you
> need one more bit of precision in the A/D than you would
> otherwise require.  Having digitised it, you would then
> need some fairly simple DSP to identify the sync tips
> and black level, slice the sync signal at a threshold
> halfway between these two values, then black-level-clamp
> your video by subtracting the black level from every
> sample.  Beware interactions between these DC-restoring
> operations in your DSP, and the AC coupling circuitry that
> is sure to exist at the input of your A/D.
>
> Note also that the GS4982 also performs various sanity
> checks on the incoming signal, such as checking that its
> line scan rate is reasonable.  These are fairly easy to
> do in DSP, but need some careful thought to be sure that
> you are implementing the checks in a sufficiently
> general way.
>
> This is a fairly straightforward project and we would be
> happy to quote you for the development.
>
> >Or is there any web site that introduces existing IPs?
>
> www.opencores.org and www.free-ip.com are good places to
> start, although I wasn't hugely impressed with the free-ip
> CORDIC module when I looked at it.  Caveat raptor.
> -- 
> Jonathan Bromley, Consultant
>
> DOULOS - Developing Design Know-how
> VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services
>
> Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
> Tel: +44 (0)1425 471223          mail:jonathan.bromley@doulos.com
> Fax: +44 (0)1425 471573                Web: c
>
> The contents of this message may contain personal views which
> are not the views of Doulos Ltd., unless specifically stated.

I forgot to say "I use that IP on asic work".
I know this is FPGA News group, but there's no asic news group that I
know,
so I asking in this group. Sorry.
I visited Gennum web site but I can't get any information on that.
I'm not searching for free IPs. I can pay for some GOOD IPs.
Can anybody suggest me for some good commersial IP company?

Sorry for my poor English, and Mr.Bromley, Thanks for your answer.


Article: 68215
Subject: Re: FIFO Depth(Length) Calculation
From: Jonathan Bromley <jonathan.bromley@doulos.com>
Date: Tue, 30 Mar 2004 14:13:23 +0100
Links: << >>  << T >>  << A >>
On Tue, 30 Mar 2004 18:01:23 +0530, Anand P Paralkar
<anandp@sasken.nospam.com> wrote:

>  Consider two asynchronous blocks each generating data
>  at rates R1 and R2 such that R1 > R2.  Calculate the depth
>  of a FIFO required between these two blocks so that there
>  is no data dropped.  No feedback/handshake mechnism should
>  be assumed between the two blocks.
>
>In such a case, do we have a general solution?

If you insist on no feedback, how can you ever possibly 
solve the problem of data overrun in the R1->R2 direction
when R1 > R2 ?

FIFOs are good when you have "bursty" data transfer,
so that the long-term data rate is slow enough for the
slower receiver to catch up during idle times.  If 
that's what you mean, it seems to me that the calculation
is fairly simple.  Given a burst of N data items transmitted
at the higher rate R1, and received at the slower rate R2:

  burst duration = N/R1
  items received in that time = (N.R2)/R1
  backlog that must be stored in FIFO = N - (N.R2)/R1
                                      = N.(R1-R2)/R1
Any sensible designer will then make the FIFO somewhat
longer than this, to accommodate any latency or response
time in the receiver.  Given a response time T, you need
an additional T.R1 locations in the FIFO.

It's also clear that the receiver will need some time
to consume this backlog, so the idle time between bursts
must be long enough;  the minimum time is of course
  mop-up time = backlog/R2 = N.(R1-R2)/(R1.R2)

Going in the other direction, from a slow transmitter to
a fast receiver, the FIFO needs to be only long enough
to hold any data items that might be received during
the receiver's response time, T.R2

Have I missed something, or is it really that easy?
-- 
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223          mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573                Web: http://www.doulos.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

Article: 68216
Subject: Re: maybe a stupid question
From: "Jaan Sirp" <jaan.sirp@liewenthal.ee>
Date: Tue, 30 Mar 2004 05:27:12 -0800
Links: << >>  << T >>  << A >>
Hi Marija, 

The delay of a path includes delays of the components and delays of the routing. 
Synthesis knows nothing about routing delays. Don't pay attention to the 
frequency, synthesis shows, it just a estimated number. You have to run PAR, 
then Timing Analyser and then go back to the source. 

Jaan. 



Article: 68217
Subject: Re: Is there any Sync separator IP(Intellectual property) exists?
From: Jonathan Bromley <jonathan.bromley@doulos.com>
Date: Tue, 30 Mar 2004 14:29:43 +0100
Links: << >>  << T >>  << A >>
On Tue, 30 Mar 2004 21:53:07 +0900, "coreDEVIL" 
<coredev@korea.com> wrote:

[...]
>> >I need sync separator IP which works like Gennum GS4982.
[...]
>I forgot to say "I use that IP on asic work".

OK - so you have analog cells in your process, I guess.

>I know this is FPGA News group, but there's no asic news
> group that I know,
>so I asking in this group. Sorry.

No problem.

>I visited Gennum web site but I can't get any information on that.
>I'm not searching for free IPs. I can pay for some GOOD IPs.
>Can anybody suggest me for some good commersial IP company?

Did you ask Gennum?  They also do custom ICs,
and they have very good expertise in mixed-signal.

In the UK, you could try Wolfson Microelectronics:
  http://www.wolfsonmicro.com/

>Sorry for my poor English

It is **MUCH** better than my Korean :-)
-- 
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223          mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573                Web: http://www.doulos.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

Article: 68218
Subject: Re: Is there any Sync separator IP(Intellectual property) exists?
From: Jan Panteltje <pNaonSptaemltje@yahoo.com>
Date: Tue, 30 Mar 2004 14:22:09 GMT
Links: << >>  << T >>  << A >>
On a sunny day (Tue, 30 Mar 2004 12:05:34 +0100) it happened Jonathan Bromley
<jonathan.bromley@doulos.com> wrote in
<9fki60l45bt7p7eu8l0ojttb9ju30klue0@4ax.com>:

>On Tue, 30 Mar 2004 18:58:36 +0900, "coreDEVIL" 
><coredev@korea.com> wrote:
>
>>I need sync separator IP which works like Gennum GS4982.
>>Can I get any information about that?
>
>Gennum's website :-)
>
>But this is an analog function.  You could not implement
>it in any FPGA I can think of.
>
>You could digitise the incoming video, remembering that
>your A/D conversion needs enough dynamic range to digitise
>sync tips as well as the full video span.  That means you
>need one more bit of precision in the A/D than you would
>otherwise require.  Having digitised it, you would then
>need some fairly simple DSP to identify the sync tips
>and black level, slice the sync signal at a threshold
>halfway between these two values, then black-level-clamp
>your video by subtracting the black level from every 
>sample.  Beware interactions between these DC-restoring 
>operations in your DSP, and the AC coupling circuitry that
>is sure to exist at the input of your A/D.
>
>Note also that the GS4982 also performs various sanity
>checks on the incoming signal, such as checking that its
>line scan rate is reasonable.  These are fairly easy to
>do in DSP, but need some careful thought to be sure that
>you are implementing the checks in a sufficiently 
>general way.
>
>This is a fairly straightforward project and we would be
>happy to quote you for the development.
>
>>Or is there any web site that introduces existing IPs?
>
>www.opencores.org and www.free-ip.com are good places to
>start, although I wasn't hugely impressed with the free-ip
>CORDIC module when I looked at it.  Caveat raptor.
>-- 
>Jonathan Bromley, Consultant
Him, jou mention DSP.
I did this without, but there are some catches.
First of cause there is thr DC level, it is not that you need ONE
more bit (although theoretically you are right), you just need to
dedicate the ful 256 steps to the composite, and sync will be 30%
of that.
So you would likely want to slice at 15 %.
But this requires a GOOD DC sync tip clamp, and that clamp needs a
clamp pulse, which wont't be there until you slice sync right :-)
So, to 'lock in' I just made (without DSP) a register that kept
the minimum (bottom sync) everytime the signal was lowest, and then
each frame (or was it several lines) increased, so slowly increased.
Also some limit against overflow :-)
So now, the 'syn tip register' slowly follows the sync tip.
Add some counts to that and slice there.
This works very well.
Once the H PLL is locked (to the sync), you can do more..
But main point no DSP whatsover I used.
I am a bit reluctant to release the Verilog code, as this project
is one I hope to actually sell, but this should help one on the way.
JP 

Article: 68219
Subject: Re: rs232 interface on nios
From: Derek Young <LL_mit_edu@dereky.nospam>
Date: Tue, 30 Mar 2004 09:43:07 -0500
Links: << >>  << T >>  << A >>
 >>I have pic (12f675) which is conected to a sensor. PIC will digitise the
 >>signal from sensor and sends it over one of the on of IO pins. This pin
 >>is connected to Rx wire of the rs232 cable. the other end of rs232(male)
 >>is connected to console port of Altera nios board. the PIC continuously
 >>sends 8 bit data over the Rx line of the console. My problem is, when I
 >>am trying to read a byte from console using the default function
 >>provided ie nr_uart_rxchar(). Its always returning -1.  I have tested
 >>the my program on PIC by connecting it to serial line of a pc and
 >>reading the serial line using a program. It just works fine. I did not
 >>change any parameters on nios board as I am using the default nios

 > Gabor Szakacs wrote:

 > If you're really connecting the I/O pin of the PIC directly
 > to the RS232 line, you must be aware that you're not meeting
 > the standard voltage levels.  It probably works with the PC
 > because they don't use standard circuits, either.  True RS232
 > requires voltages to swing below ground to guarantee switching
 > the receiver.  Some receivers are modified with positive
 > threshold offset to accommodate TTL input signals.  I don't
 > know what's used on the nios, but you should probably look at
 > the output of the receiver to see if your signal gets through.

I was actually just using the serial port on the Nios Development board...

The development board uses a Maxim RS232 transceiver (MAX3237).  I 
looked really quickly at the datasheet, and it seems like it should be 
okay with a TTL input--input threshold low and high are 1.2/1.5 V.  You 
should probably check yourself to make sure.

One other thing that you can check once you're happy with the levels is 
the type of cable you're using to connect.  For a normal PC serial port, 
pin 2 is receive, and pin 3 is transmit.

http://www.interfacebus.com/Design_Connector_RS232.html

But on the Nios development board, Pin 3 is receive and 2 is transmit. 
As a result, you can connect the Nios board to the PC using a regular 
serial cable (without a null-modem adaptor).  I'm thinking that since 
you with no problem connecting the PIC to the PC, the serial connection 
probably needs to be switched in order to connect to the Nios.

Hope that helps...

Derek


Article: 68220
Subject: More Chipscope JTAG Blues...
From: Sean Durkin <smd@despammed.com>
Date: Tue, 30 Mar 2004 16:46:19 +0200
Links: << >>  << T >>  << A >>
Hi,

I've been happily using Chipscope Pro 5.2i (Release 2) together with 
ISE5.2.03i for quite awhile.

Now suddenly it won't connect to the target anymore:

COMMAND: open_cable
ERROR: <P4JTAGCable> Error during Parallel 4 TAP Navigation
ERROR: Failed detecting JTAG device chain
ERROR: Opened Xilinx Parallel Cable but failed to detect JTAG Chain.

This is exactly what is talked about in Xilinx Answer Record #17079, but 
the patch mentioned there a) does not help and b) should not be 
necessary for ChipScope Pro 5.2i (Release 2).

iMPACT runs fine on the same machine, so the cable and the connection 
are perfectly ok, it is just ChipScope that won't connect anymore. Of 
course I tried reinstalling and such, doesn't help...

I guess I must've done something to mess it up, but I haven't the 
slightest idea what could possibly have caused this.

Any ideas?

Best regards,
Sean Durkin

Article: 68221
Subject: Athlon FX vs Pentium 4 benchmarks for xilinx's par
From: tmlo@networks.nera.no (Tomas)
Date: 30 Mar 2004 07:23:32 -0800
Links: << >>  << T >>  << A >>
Hi,

we are planning to upgrade our Linux cluster with new nodes. These
machines will be maily dedicated to run Place&Route and, as we are
mainly targetting Xilinx devices, the main performance criteria will
be P&R times.

has anyone benchmarks showing run times for both these processors?
Does anyone know if Xilinx is coming with a 64 bits native code? I
think that this is happening with Altera's Quartus 4.1.

Thanks in advance for the answers!

Regards
Tomas

Article: 68222
Subject: Re: AHDL, VERILOG or VHDL??
From: petersommerfeld@hotmail.com (Peter Sommerfeld)
Date: 30 Mar 2004 07:30:05 -0800
Links: << >>  << T >>  << A >>
Hi Hendra,

Yes Maxplus II/Quartus only support synthesizable code. You can
simulate using waveforms but no behavioural testbenching.

-- Pete

"Hendra Gunawan" <u1000393@email.sjsu.edu> wrote in message news:<c4a458$6gkqf$1@hades.csu.net>...
> "Peter Sommerfeld" <petersommerfeld@hotmail.com> wrote in message
> news:5c4d983.0403271458.43ee7ad9@posting.google.com...
> > Unfortunately the AHDL
> > stuff could not be testbenched though
> 
> Is that why Altera MaxPlus II does not support testbench for Verilog either?
> Other tools support testbench, MaxPlus II doesn't!
> 
> Hendra

Article: 68223
Subject: Re: Is there any Sync separator IP(Intellectual property) exists?
From: Jonathan Bromley <jonathan.bromley@doulos.com>
Date: Tue, 30 Mar 2004 16:47:52 +0100
Links: << >>  << T >>  << A >>
On Tue, 30 Mar 2004 14:22:09 GMT, Jan Panteltje
<pNaonSptaemltje@yahoo.com> wrote:

>
>Him, jou mention DSP.
>I did this without, but there are some catches.
>First of cause there is thr DC level, it is not that you need ONE
>more bit (although theoretically you are right), you just need to
>dedicate the ful 256 steps to the composite, and sync will be 30%
>of that.

Leaving only about 200 grey levels.  That's approaching poor
quality, I suggest.

>So you would likely want to slice at 15 %.
>But this requires a GOOD DC sync tip clamp, and that clamp needs a
>clamp pulse, which wont't be there until you slice sync right :-)

Yes.  But you can DC-clamp to the negative peaks (sync tips),
then reference everything off that - just as you describe below.

>So, to 'lock in' I just made (without DSP) a register that kept
>the minimum (bottom sync) everytime the signal was lowest, and then
>each frame (or was it several lines) increased, so slowly increased.

I hate to say this, Jan, but that sounds an awful lot like DSP
to me.  Not very hard DSP, but you're processing a signal, and
it's been digitised...  fits my definition of DSP.

>Also some limit against overflow :-)

I would suggest that it would be better to use 9-bit or even 10-bit
data, and allow yourself plenty of dynamic range.  That way, it is
no longer essential for your analog sync-tip tracking to be 
excellent, and you can still get 8 bits or even 9 bits of
precision in digitising the true video signal.

8 bits is IMHO the absolute minimum that you need for good 
visual quality on a good monitor.  7 bits looks imperfect,
6 bits is rubbish.

>So now, the 'syn tip register' slowly follows the sync tip.
>Add some counts to that and slice there.
>This works very well.

Agreed.

>Once the H PLL is locked (to the sync), you can do more..
>But main point no DSP whatsover I used.

I've disagreed with you, and you can make your own
judgment.  OK, it has no large FIR filter and no FFT;
but why are those things needed to qualify as DSP?

>I am a bit reluctant to release the Verilog code, as this project
>is one I hope to actually sell, but this should help one on the way.

It seems to me that the Verilog is pretty simple stuff.  The
clever part is getting the sync-tip tracking right, in
all possible combinations of noisy sync signals, occasional
spikes on the signal, a DC level that drifts around, lots
of jitter on the sync (thank you very much for nothing, VCRs!),
colour signals that can push the video level below black, 
out-of-spec slew rate on the sync edges, non-monotonic edges
on the sync,  ...

OK, I'll stop now.  You get the idea.  My point to Jan is 
that he already knows about all this terrible stuff, and 
knows the tricks (analogue or digital) to get around them.
It's that special application knowledge that will make 
or break your product - the Verilog code is easy once
you have the design sorted out.  Good luck with it.
  
Good to know that you are making interesting progress with
your Verilog-driven video processing.  It's something that
has been close to my heart for some time :-)  Did you get 
anywhere with your R-2R ladder successive-approximation 
converter in a Xilinx FPGA, using an LVDS receiver as 
a comparator???
-- 
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223          mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573                Web: http://www.doulos.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

Article: 68224
Subject: incremental design flow question (PACE)
From: "Yttrium" <Yttrium@pandora.be>
Date: Tue, 30 Mar 2004 15:51:03 GMT
Links: << >>  << T >>  << A >>
hey, i'm just trying out the incremental design flow (and modular) and when
i was reading the xapp from xilinx it said that the slice utilization of the
area groups have to have a similar percentage. how come? and i was just
wondering, do a lot of designers use this type of design flow (or the
modular)?

thanx for your reply's in advance,

kind regards,

Yttrium





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

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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

Custom Search