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 28625

Article: 28625
Subject: Re: revision control tools ??
From: Christof Abt <Christof.Abt@marconi.com>
Date: Thu, 18 Jan 2001 17:50:21 +0100
Links: << >>  << T >>  << A >>
strshn99@my-deja.com wrote:
> 
> Dear Gurus,
> 
> I have a newbie question to ask you.  What kind of Revision control
> tools do you use when you are working on a design?  Any recommendation
> as to which is better?
I used RCS and SCCS and was quite happy with both of them
> 
> What tools are recommended for coverage analysis?
VHDLCOVER is easy to use but the big gun called specman does more.
> 
> What tools are recommended for formal verification
The new Synopsys tool should be fine
> 
> What tools are recommended for timing verification,
I like Synopsys' Primetime 
> is it acceptable to
> conduct timing verification by embedding timing delays in a VHDL test
> bench? you know X <== ( Y AND Z ) AFTER 12 NS
No, it is isn't

> 
> thanks in advance
> 
> Sent via Deja.com
> http://www.deja.com/


-- 


Mit freundlichen Grüßen 

Christof Abt      

+---------------------------------------------------------------+
|  Christof Abt            Marconi Communications MC-AC/ASD     |
|  P.O.Box 30 00           Phone: +(7191) 13-2820               |
|  D-71520 Backnang        Fax  : +(7191) 13-4321               |
|  Germany                 email: Christof.Abt@marconi.com      |
+---------------------------------------------------------------+

Article: 28626
Subject: Re: revision control tools ??
From: "Jamie Sanderson" <jamie@nortelnetworks.com>
Date: Thu, 18 Jan 2001 11:50:44 -0500
Links: << >>  << T >>  << A >>
<eml@riverside-machines.com.NOSPAM> wrote in message
news:3a66ce37.8952109@news.dial.pipex.com...
> 1)
> the obvious one - it's got a GUI. Ok, I'm not in favour of GUIs for
> the sake of it, but if any tool needs a GUI then it's a VCS (or any
> file manager, come to think of it). You get an immediate view of the
> whole project, you can see which files are modified, you can sort on
> tags, you can sort on modified files (irrespective of directory) so
> that you can check in all modded files at once, you can checkout only
> the required files without having to painfully remember and type each
> one, or creating a new module name for each set you're likely to want,
> you can see who's got what file checked out, and a dozen other things.

WinCVS can also you a flat view of the project. You can also save several
project settings (the local directory where you've checked it out, and the
location/type of the repository, and some other options). One of the
less-used features "watch" can also let you know who's got a file checked
out.

> 2)
> The other obvious one - CVS doesn't have reserved checkouts. In fact,
> I did a quick survey of VCS tools recently, and opinion was divided on
> whether a tool should support reserved checkouts or not. RCS, SCCS,
> MKS, Clearcase, WebDAV, Sourcesafe, and Perforce all thought that it
> should, CVS thought that it shouldn't.

Isn't this what "cvs admin -l" does?

> I had an argument with someone recently who had used CVS on a large
> ASIC project, and their reply was that "this doesn't happen in
> practice". IMO, it will happen eventually, and the whole point of a
> management tool is to enforce systems to prevent this sort of thing
> happening.

There's a reason I'm not sure how the locking features works, it hasn't been
necessary "in practice". :)

> 3)
> MKS implements a security mechanism in which you can set up projects
> and groups, and assign permissions to people in groups, within a given
> project. For instance, you may want to give a couple of people write
> access to a given project, and give everyone else only read access; or
> you may want to make sure that only some people can create or delete
> archives. IMO, this is important when you have development groups with
> a large range of abilities in them (and in which only a few people, if
> any, have previously used a VCS).

CVS can also do this using its password, readers, etc. files in the CVSROOT
folder. This doesn't seem to get used very often (other than creating a
single read-only account). Still, it's there for the case you describe
above.

> >CVS. However, I had some problems with the Windows port of CVS. Today
> >I tried to check in a netlist and I got:
> >
> >cvs server: Up-to-date check failed for `PSB63.edf'
> >cvs [server aborted]: correct above errors first!
> >
> >I had to transfer the files over to Solaris and check them in from
> >there...
>
> Unfortunately, I occasionally get errors as well

The main problem seems to be with CVS repositories that are mounted as a
local directory under Windows, instead of using pserver (or ntserver). There
are problems overriding the read-only file permissions in the repository. I
don't know if this applies to MKS or not.

> Here's an example from a couple of weeks ago. I wanted to import a
> module using WinCVS. The "import module" selection is the next one
> down on the menu from "create repository". I clicked "create
> repository" by mistake, and then quickly clicked "ok" by mistake,
> instead of "cancel". The new repository defaults to the same place as
> the old repository - it took several hours to find out what had
> happened and to restore from tape afterwards. Ok, you can't make this
> mistake from the command line interface, but my point is that project
> security is more complex than making sure that you have a secure
> connection to the server, and that's one thing that commercial
> products do better than CVS.

Sounds like you should have read-only access... Seriously, though, I would
much rather see someone take it on and improve it, rather than buying into a
proprietary product like ClearCase or Synchronicity (no one's mentioned this
one yet). Cyclic was trying this, but seemed to be afraid to break
compatibility. At one point, I was trying to set up a CVS pserver, and was
having all sorts of problems with it trying to find local preferences files.
Browsing through the CVS source, it was quite amusing to see that they put
into comments that they wanted to do away with this feature altogether. How
about CVS2 anyone, all the things that are good in CVS, minus the junk no
one needs, and some new features (such as locking?) thrown in.

Cheers,
Jamie






Article: 28627
Subject: Re: revision control tools ??
From: Doug Hillmer <doug.hillmer@innovisiontv.com>
Date: Thu, 18 Jan 2001 10:00:44 -0800
Links: << >>  << T >>  << A >>

When I was at HP we used to use an internal tool
called HMS, worked great for multisite HDL projects
and binaries.

It then became productized by HP under the name of
Softbench, but the HP dropped it and the developer
formed his own company for it, but I forgot what
the name of the company is.

Any pointers on where I could get it would be 
welcomed.

I like it better than CVS (which we are now using).
And it was not as complicated as Clearcase (which
I have used for software projects in the past too).
And it handled binaries, which RCS does not, and worked 
very well for both geographically disperse and close projects.

Doug Hillmer.

Article: 28628
Subject: Tool to partially, dynamically reconfigure Virtex?
From: Michael Boehnel <boehnel@iti.tu-graz.ac.at>
Date: Thu, 18 Jan 2001 19:58:10 +0100
Links: << >>  << T >>  << A >>
Is there a tool to partially dynamical reconfigure a Virtex FPGA via
JTAG-Interface? I think with JTAG-Programmer it isn't possible.

The Virtex FPGA is configured with bitstream 1. With JBits I manipulate
this bitstream 1 to get a new bitstream 2. What tool can I use to
download the bitstream 2 without generating a device reset. I'd like the
unaffected parts to continue their work without interruption. Is this
possible?

Thanks,

Michael

PS: My board doesn't support the JBits XHWIF so I think somehow I have
to use the JTAG-Interface directly for reconfiguration.



Article: 28629
Subject: Re: FPGA for radar digital downconversion
From: Ray Andraka <ray@andraka.com>
Date: Thu, 18 Jan 2001 19:27:48 GMT
Links: << >>  << T >>  << A >>
Ken, 

I didn't realize you were a lurker here.  Thanks for chiming in.  You are, of
course,
absolutely correct.  DSP processors can't hope to keep up in software
radio/radar 
applications where the sampling is being done at IF at tens of MHz.  

In the case of downconversion, there are a number of hardware tricks that can be
used
including polyphase decomposition filters, distributed arithmetic, and CORDIC. 
It is quite 
possible using these techniques to get 100x the performance of a DSP
microprocessor for
close to the same cost.  For some details on downconversion, see my article in
the 
winter 2000 issue of the Xilinx XCell Journal (sorry, it is not on-line at this
time, 
you can order your subscription by searchning for XCell on the Xilinx website). 
I've got
some radar and radar simulator examples in papers available on my website.  One
is doing
the equivalent of 11 billion multiplies per second in a Virtex-4 (slow speed
grade) part.

The catch is there are not all that many people versed in hardware DSP, so it
can be challenging
to get this kind of performance.  Xilinx has closed the gap somewhat with their
system generator
tool, but it is far from a complete solution (It basically lets you intantiate
cores from within
Matlab simulink).  THe tool knows nothing about algorithms, and in hardware DSP
the big thing is
tailoring the algorithm to a hardware-friendly implementation.  The new virtex
II is also exciting 
because it offers some more "software DSP like" features.

My advice is, if it will run at the desired sample rate in a single DSP
processor, use the DSP:
It is alot easier and cheaper to find software DSP expertise than hardware DSP
expertise.  If 
a DSP doesn't cut it by just a little, you might still be better off going with
a few DSPs, depending
on the available expertise.  For much more than that, you probably want to bring
someone in who
can either do the design or architecture, or teach you people how to do it (the
learning is not an
over night process, but pointing out some of the hardware techniques goes along
way toward breaching 
some of the hurdles.)


Ken Chapman wrote:
> 
> Dear Tom,
> 
> In simple terms, this is about the required horsepower for the application.
> If you can really perform all the processing that you need to in a single
> DSP processor chip then it does provide you with a very easy way to
> prototype your system.
> 
> However, at 50MHz sample rate you will soon burn up the available number of
> MIPS available to you in a DSP processor. If we consider the
> TMS320VC5410-10 device, it has a single 17x17 multiplier clocked at 100MHz.
> This would only allow you to perform 2 operations at 50MHz. Obviously you
> will be performing down-conversion to carry out most of your work at a
> lower sample rate, but this down-conversion in itself will almost certainly
> require more than this 100MIPS. These processors are really excellent when
> you have lots of time between samples such as in audio. At At 48KHz the
> 100MIPS provides over 2000 operations per sample.
> 
> In contrast, a device such as the Xilinx Spartan-II XC2S150 is a 'blank
> sheet' onto which a DSP design can be implemented. It contains 864
> Configurable Logic Blocks which could implement 10 full 16x16 multipliers
> operating at over 100MHz. That provides over 1000 MIPS. Unlike the fixed
> architecture of a processor where flexability is derived via software code
> that utilises 'spare' clock cycles (if you have any), the Xilinx route
> provides you the flexability to produce any architecture containing
> multipliers, adders, comaparitors etc all working in parallel. It is highly
> unlikely you will need full 16x16 multipliers and this means that the
> effective MIP count can be increase as smaller modules are constructed.
> 
> The Spartan devices are in the $5 to $20 range providing a lot of
> processing potential per $ compared with a DSP processor. If you have not
> looked at Xilinx DSP before then you will need to take some time to look
> into it and how designs can be implemented. There are tools that generate
> complete FIR filters and NCOs for you as well as the simple blocks.
> 
> I have worked on RADAR systems before joining Xilinx, and many times with
> Xinlinx customers during my 10 years with the company. Some of the most
> successful systems are the ones that combine the merits of both the Xilinx
> devices and the DSP processors. The Xilinx device is used to perform the
> intensive processing of the data path flow. The DSP processor acts as a
> controller and less time critical calculator. In your case the actual beam
> forming would probably be done by the FPGA, but the calculation of
> coefficients to set the beam pattern would be made by the DSP processor.
> 
> There are DSP seminars by Xilinx on the 25th January all over North
> America. Try to go to one!
> 
> Yours sincerely,
> 
> Ken Chapman
> 
> Xilinx Applications (UK)

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

Article: 28630
Subject: Best design for asyn. interface DSP <-> FPGA?
From: "Chris G. Schneider" <chris@cgschneider.com>
Date: Thu, 18 Jan 2001 20:40:19 +0100
Links: << >>  << T >>  << A >>
I want to design an interface between a DSP and a FPGA. The DSP should
be able to access FPGA registers, via the DSP parallel port. The
DSP parallel port has address, data, chip select, read and write lines.
Just the usual stuff.

I just designed this interface, but I am not happy with the results in
terms of speed and design stability. So I  want to ask this audience for
some design hints.

1) Write access: I just synchronize the DSP write signal and generate an
internal write strobe to fetch the data. Is it a better solution to use
the write signal to store the data in a register and then synchronize
the register contents to the internal clock? This makes it possible to
detect very short write strobes of the DSP and to reduce the number of
WS, in the application.

2) Read access: I use a async. read access at the moment. Is it possible
to make this work synchronously? The problem is that synchronizing just
eats up a lot of time, and timing constraints of modern DSP are pretty
tight. There will be a problem if the FPGA does not release the DSP data
bus fast enough, because read and write accesses can appear very soon
after each other.

Any suggestions, or literature tips?

--
chris



Article: 28631
Subject: Re: Virtex-II officially launched
From: Andy Peters <"apeters <"@> n o a o [.] e d u>
Date: Thu, 18 Jan 2001 12:49:17 -0700
Links: << >>  << T >>  << A >>
Austin Lesea wrote:
> 
> Bob,
> 
> The impedance updates are asynchronous, and are done in such a manner that
> the updates trim values are all extremely small.  The resultant dR/dt is
> undetectable in the signal integrity of the output.

How will you handle the variable impedances in the IBIS models?

Please tell me you'll have IBIS models ...

-- a
----------------------------
Andy Peters
Sr. Electrical Engineer
National Optical Astronomy Observatory
950 N Cherry Ave
Tucson, AZ 85719
apeters (at) n o a o [dot] e d u

"It is better to be silent and thought a fool, 
 than to send an e-mail to the entire company
 and remove all doubt."

Article: 28632
Subject: Re: VHDL question
From: Andy Peters <"apeters <"@> n o a o [.] e d u>
Date: Thu, 18 Jan 2001 12:51:10 -0700
Links: << >>  << T >>  << A >>
#ROBERTUS WAHENDRO ADI# wrote:
> 
> Hi all,
> Can variable be shared between one process to another?
> if yes, How to do it?

You can use a shared variable.

You might want to ask your question in comp.lang.vhdl.

-- a
----------------------------
Andy Peters
Sr. Electrical Engineer
National Optical Astronomy Observatory
950 N Cherry Ave
Tucson, AZ 85719
apeters (at) n o a o [dot] e d u

"It is better to be silent and thought a fool, 
 than to send an e-mail to the entire company
 and remove all doubt."

Article: 28633
Subject: Re: revision control tools ??
From: eml@riverside-machines.com.NOSPAM
Date: Thu, 18 Jan 2001 21:12:39 GMT
Links: << >>  << T >>  << A >>
On Thu, 18 Jan 2001 11:50:44 -0500, "Jamie Sanderson"
<jamie@nortelnetworks.com> wrote:

>WinCVS can also you a flat view of the project.

I haven't found this (I'm on 1.1b16) - can you tell me how do it?

> You can also save several
>project settings (the local directory where you've checked it out, and the
>location/type of the repository, and some other options). One of the
>less-used features "watch" can also let you know who's got a file checked
>out.

I've never used it - email notification seems a very clunky way to
find out what's going on. It's nice to able to run up the GUI and
immediately see who's doing what.

>> 2)
>> The other obvious one - CVS doesn't have reserved checkouts. In fact,
>> I did a quick survey of VCS tools recently, and opinion was divided on
>> whether a tool should support reserved checkouts or not. RCS, SCCS,
>> MKS, Clearcase, WebDAV, Sourcesafe, and Perforce all thought that it
>> should, CVS thought that it shouldn't.
>
>Isn't this what "cvs admin -l" does?

Hmmm... not really sure. The docs say that you can use it in
conjunction with the Perl script I mentioned, to give reserved
checkouts. I read somewhere else that it didn't work particularly
well. So, I found the script (which I understand only marginally)
after reading your reply, and it tells you to lock the file with rcs,
and the script can then prevent a commit if someone else has the lock.
Maybe admin -l is actually rcs -l, the rcs lock. It all seems very
kludgey - the CVS docs say you're on your own if you use it.

>> 3)
>> MKS implements a security mechanism in which you can set up projects
>> and groups, and assign permissions to people in groups, within a given
>> project. For instance, you may want to give a couple of people write
>> access to a given project, and give everyone else only read access; or
>> you may want to make sure that only some people can create or delete
>> archives. IMO, this is important when you have development groups with
>> a large range of abilities in them (and in which only a few people, if
>> any, have previously used a VCS).
>
>CVS can also do this using its password, readers, etc. files in the CVSROOT
>folder. This doesn't seem to get used very often (other than creating a
>single read-only account). Still, it's there for the case you describe
>above.

Very interesting - I haven't heard of this; I'll have a look.
 
>Sounds like you should have read-only access... Seriously, though, I would
>much rather see someone take it on and improve it, rather than buying into a
>proprietary product like ClearCase or Synchronicity (no one's mentioned this
>one yet). Cyclic was trying this, but seemed to be afraid to break
>compatibility. At one point, I was trying to set up a CVS pserver, and was
>having all sorts of problems with it trying to find local preferences files.
>Browsing through the CVS source, it was quite amusing to see that they put
>into comments that they wanted to do away with this feature altogether. How
>about CVS2 anyone, all the things that are good in CVS, minus the junk no
>one needs, and some new features (such as locking?) thrown in.

Sounds good, but it's back to the old 'free' v. commercial argument -
if I'm paying someone, then their job depends on it, and they're
hopefully going to be more likely to get it right than a volunteer.
Anyone doing a chip has to invest a minimum of several K$, and maybe
several hundred K$ - I personally am certainly willing to pay a small
fraction of that for a commercial tool. To each his own.

Evan

Article: 28634
Subject: Re: revision control tools ??
From: "Chris G. Schneider" <chris@cgschneider.com>
Date: Thu, 18 Jan 2001 22:33:37 +0100
Links: << >>  << T >>  << A >>
Did somebody try to archive to tools as well?

--

chris




Article: 28635
Subject: Re: revision control tools ??
From: Martin Darwin <martin.darwin@alcatel.com>
Date: Thu, 18 Jan 2001 17:31:28 -0500
Links: << >>  << T >>  << A >>
eml@riverside-machines.com.NOSPAM wrote:

> >Isn't this what "cvs admin -l" does?
> 
> Hmmm... not really sure. The docs say that you can use it in
> conjunction with the Perl script I mentioned, to give reserved
> checkouts. I read somewhere else that it didn't work particularly
> well. So, I found the script (which I understand only marginally)
> after reading your reply, and it tells you to lock the file with rcs,
> and the script can then prevent a commit if someone else has the lock.
> Maybe admin -l is actually rcs -l, the rcs lock. It all seems very
> kludgey - the CVS docs say you're on your own if you use it.

It works fine. It has been used on several projects. It is kludgey
though.

We have scripts called lock and unlock.
All files are checked-out read-only (a CVS setting). 
Lock, locks the file (cvs admin -l), updates it to the newest version
and then gives it write permissions.
Unlock, unlocks the file (cvs admin -u), removes write permissions.
Unlock is rarely used as a checkin will unlock the file anyway.

The annoying thing about this is that when you checkin the file it
unlocks it automatically. If you are just doing an incremental checkin
and still want the file locked you have to re-lock it again.

The perl script is there to prevent checkins of files that are not
locked. The perl script is called in CVS (add a line in the commitinfo
file to call the script).

I personally would rather not use locks because it gets annoying during
top level edits. You have to run over and tell the person that has it
locked to either: unlock the file or do the edit for you. Everyone seems
leery of unreserved checkins, I like them.

Once you are using file locking, you can write a script to parse "cvs
log" and determine who has which files are locked.

CVS works fine. It has its problems, but it is simple. Very short
learning curve (a few simple commands to remember). No problems with
repository corruption.

MD

--
Martin Darwin	        
ASIC Design Engineer       Tel +1 (613) 784-8873
Alcatel Networks Corporation

Article: 28636
Subject: Re: Best design for asyn. interface DSP <-> FPGA?
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Thu, 18 Jan 2001 16:53:30 -0800
Links: << >>  << T >>  << A >>

--------------25706E7DA74EADA6AEACA0EE
Content-Type: text/plain; charset=us-ascii; x-mac-type="54455854"; x-mac-creator="4D4F5353"
Content-Transfer-Encoding: 7bit

Chris, you may want to look at an article about crossing asynchronous clock
boundaries that I published about a year ago. Mayeb it helps you.

http://www.isdmag.com/editorial/2000/design0003.html

Peter Alfke, Xilinx Applications
================================
"Chris G. Schneider" wrote:

> I want to design an interface between a DSP and a FPGA. The DSP should
> be able to access FPGA registers, via the DSP parallel port. The
> DSP parallel port has address, data, chip select, read and write lines.
> Just the usual stuff.
>
> I just designed this interface, but I am not happy with the results in
> terms of speed and design stability. So I  want to ask this audience for
> some design hints.
>
> 1) Write access: I just synchronize the DSP write signal and generate an
> internal write strobe to fetch the data. Is it a better solution to use
> the write signal to store the data in a register and then synchronize
> the register contents to the internal clock? This makes it possible to
> detect very short write strobes of the DSP and to reduce the number of
> WS, in the application.
>
> 2) Read access: I use a async. read access at the moment. Is it possible
> to make this work synchronously? The problem is that synchronizing just
> eats up a lot of time, and timing constraints of modern DSP are pretty
> tight. There will be a problem if the FPGA does not release the DSP data
> bus fast enough, because read and write accesses can appear very soon
> after each other.
>
> Any suggestions, or literature tips?
>
> --
> chris

--------------25706E7DA74EADA6AEACA0EE
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit

<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
Chris, you may want to look at an article about crossing asynchronous clock
boundaries that I published about a year ago. Mayeb it helps you.<u></u>
<p><u><A HREF="http://www.isdmag.com/editorial/2000/design0003.html">http://www.isdmag.com/editorial/2000/design0003.html</A></u>
<p>Peter Alfke, Xilinx Applications
<br>================================
<br>"Chris G. Schneider" wrote:
<blockquote TYPE=CITE>I want to design an interface between a DSP and a
FPGA. The DSP should
<br>be able to access FPGA registers, via the DSP parallel port. The
<br>DSP parallel port has address, data, chip select, read and write lines.
<br>Just the usual stuff.
<p>I just designed this interface, but I am not happy with the results
in
<br>terms of speed and design stability. So I&nbsp; want to ask this audience
for
<br>some design hints.
<p>1) Write access: I just synchronize the DSP write signal and generate
an
<br>internal write strobe to fetch the data. Is it a better solution to
use
<br>the write signal to store the data in a register and then synchronize
<br>the register contents to the internal clock? This makes it possible
to
<br>detect very short write strobes of the DSP and to reduce the number
of
<br>WS, in the application.
<p>2) Read access: I use a async. read access at the moment. Is it possible
<br>to make this work synchronously? The problem is that synchronizing
just
<br>eats up a lot of time, and timing constraints of modern DSP are pretty
<br>tight. There will be a problem if the FPGA does not release the DSP
data
<br>bus fast enough, because read and write accesses can appear very soon
<br>after each other.
<p>Any suggestions, or literature tips?
<p>--
<br>chris</blockquote>
</html>

--------------25706E7DA74EADA6AEACA0EE--


Article: 28637
Subject: Re: CMOS or TTL
From: rk <stellare@nospamplease.erols.com>
Date: Thu, 18 Jan 2001 20:55:16 -0500
Links: << >>  << T >>  << A >>
erika_uk@my-deja.com wrote:
> 
> hello,
> 
> which provides better performance CMOS or TTL based devices?
> I believe that CMOS gives better performance, and it's the CMOS based
> devices which are more expensive...but my friend told me the inverse ?
> 
> but what let the better to be better ?
> are all fpga based CMOS devise ?
> 
> any input  ?

The one recent FPGA that I know of that wasn't CMOS was the Dynachip
DL5000 series, which was produced in a biCMOS process at IBM - 0.8 um -
as usual, if I remember correctly to this whole note.  The modules were
power-switched and took on the order of 50mW per module; configuration
memory was CMOS.

Dynachip is now gone and I think Xilinx bought up their intellectual
property.

Bipolar devices for digital stuff is still used sometimes, but not very
often, with biCMOS devices used on interfaces (high impedance when
powered off, line drivers/receivers).  Some of the newer CMOS devices
are now tolerant to I/O voltages higher than the logic supply and are
high impedance with power removed (cold sparing) or are capable of hot
insertion.

I remember "fast" TTL buffer chips taking 0.500 watt each, just idling,
with pin to pin delays of like 8 ns!  Ouch!

----------------------------------------------------------------------
rk                               How the hell do I know? I'm just a
stellar engineering, ltd.        common, ordinary, simple savior of
stellare@erols.com.NOSPAM        America's destiny.
Hi-Rel Digital Systems Design    -- Pat Paulsen

Article: 28638
Subject: Re: CMOS or TTL
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Thu, 18 Jan 2001 18:35:12 -0800
Links: << >>  << T >>  << A >>


rk wrote:

> The one recent FPGA that I know of that wasn't CMOS was the Dynachip
> DL5000 series, which was produced in a biCMOS process at IBM - 0.8 um -
> as usual, if I remember correctly to this whole note.  The modules were
> power-switched and took on the order of 50mW per module; configuration
> memory was CMOS.
>
> Dynachip is now gone and I think Xilinx bought up their intellectual
> property.
>

And more importantly, we got their experienced designers, who are now
giving us super-fast and versatile I/O designs.
Everybody came out a winner: They, we, and you, our Virtex-II customers.

Peter Alfke, Xilinx Applications


Article: 28639
Subject: Re: FPGA driving clock line
From: murray@pa.dec.com (Hal Murray)
Date: 19 Jan 2001 04:34:07 GMT
Links: << >>  << T >>  << A >>

> Better yet, let the clock run freely, and use the FPGA to create a
> one-clock-period Enable signal when you want to single step. Driven by a simple
> state machine in less than one CLB, this is the solution that will give you the
> least trouble ( none at all)

The disadvantage that I see with a traditional Enable signal is that
you have to route it all over the place and the enable logic is often
very handy for other things.

Here is a suggestion that might avoid that.


Start with a 2x clock.  Generate the working clock with a FF.
Put the enable on that FF.

You can do the same thing with a 1x clock and appropriate gates
but that depends upon prop times so you will have to work
a bit to get it right and convince yourself it is correct.


-- 
These are my opinions, not necessarily my employers.  I hate spam.

Article: 28640
Subject: Re: Best design for asyn. interface DSP <-> FPGA?
From: "Manfred Kraus" <newsreply@cesys.com>
Date: Fri, 19 Jan 2001 08:18:05 +0100
Links: << >>  << T >>  << A >>
Hi chris,

You dont mention the FPGA vendor for your project.
If it was Xilinx, think about the dual-clock FIFO that
is available free of charge (CoreLib).
It has totaly independent clocks for both sides.

Manfred




Article: 28641
Subject: Reconfiguration of Armel fpga
From: xgeorg@my-deja.com
Date: Fri, 19 Jan 2001 08:21:36 GMT
Links: << >>  << T >>  << A >>
Hi folks,

has anybody expieriences with the dynamic reconfiguration of Atmel FPGAs
?
We have signed a NDA, but information is difficult to get.

So does somebody have any C / ASSEMBLER - ROUTINES for MODE 4
Reconfiguration ?
Or some hints how to start that ?
Any Comments...
regards
Georg Heinrich


Sent via Deja.com
http://www.deja.com/

Article: 28642
Subject: Re: spartanII chip availability
From: "Ulf Samuelsson" <ulf@atmel.dot.com>
Date: Fri, 19 Jan 2001 13:32:18 +0100
Links: << >>  << T >>  << A >>
:     I am looking at building a small number 10-20 pcbs that will use =
one of
: Xilinx's FPGAs.  The logic required is small but may be expanded =
later.
: Essentially it is a FPDP i/f feeding some fifos that are feeding a =
D/A.
:=20
:     I am looking initially at a Spartan II for a fifo or DPRAM to hold =
some
: data for the D/A.
: I like the idea of using the block ram for the small DPRAM or fifo. =
Speed is
: slow by most people's standards on this group. D/A will be run at =
120KSa/s
: or so but I am doing
: 8 channels + on each FPGA.
:=20
You may want to try the Atmel AT40K FPGA family.
They have "distribued" 10 ns SRAM suitable for FIFOs.
There is a 32 x 4 DPRAM for every 16 macrocell,
and they are hardwired, do not use configuration RAM.
The AT94Kxx FPSLIC has 12kB of SRAM useable for the FPGA
on  top of the distributed SRAM.=20
They are pin compatible with the XC4000 series
Tools are for free at Atmels homepage.
The FPSLIC demoboard will set you back $500 for a 4 months
license but then you get really good tools.
Mentor Modelsim, Leonardo,Seamless.

--=20
Best Regards
Ulf at atmel dot com
These comment are intended to be my own personal view
and may or may not be shared by my Employer Atmel Sweden.





Article: 28643
Subject: Re: CMOS or TTL
From: krw@btv.ibm.com (Keith R. Williams)
Date: Fri, 19 Jan 2001 13:19:19 GMT
Links: << >>  << T >>  << A >>
On Thu, 18 Jan 2001 18:35:12 -0800, Peter Alfke
<peter.alfke@xilinx.com> wrote:


>> Dynachip is now gone and I think Xilinx bought up their intellectual
>> property.
>>
>
>And more importantly, we got their experienced designers, who are now
>giving us super-fast and versatile I/O designs.
>Everybody came out a winner: They, we, and you, our Virtex-II customers.

Everyone except DynaChip's customers, possibly. :-(  I had a DY8000
design done, ready for fab and poof!

----
   Keith


Article: 28644
Subject: Re: About programming cables
From: "Tomasz Nakielski" <tnakiels@poland.com>
Date: Fri, 19 Jan 2001 05:38:29 -0800
Links: << >>  << T >>  << A >>
Hello.

I have a problem with programming Virtex (XVC200PQ240) via 
JTAG Parallel Cable III Model DLC5 and Programmer 3.3.06i. from one of my PC.
I receive message: "Communication with the cable could not be established". I have tried with all settings of Parallel Port (lpt 1).
On the other computer everything is OK (There is older version of Foundation). On both PC is Windows 98 SE so I think the Parallel Cable is OK. 

If someone knows what's wrong just write to me.

Tomasz

Article: 28645
Subject: Re: revision control tools ??
From: "Jamie Sanderson" <jamie@nortelnetworks.com>
Date: Fri, 19 Jan 2001 09:50:54 -0500
Links: << >>  << T >>  << A >>
<eml@riverside-machines.com.NOSPAM> wrote in message
news:3a675bb8.45182544@news.dial.pipex.com...
> On Thu, 18 Jan 2001 11:50:44 -0500, "Jamie Sanderson"
> <jamie@nortelnetworks.com> wrote:
>
> >WinCVS can also you a flat view of the project.
>
> I haven't found this (I'm on 1.1b16) - can you tell me how do it?

I'm using beta 17, but I don't recall if this was a new feature or not.
Under the "View" menu, there is an item called "Flat Mode". Selecting that
will show you every file in the project in a single listing.

> I've never used it - email notification seems a very clunky way to
> find out what's going on. It's nice to able to run up the GUI and
> immediately see who's doing what.

I agree, but it doesn't have to be email. You can execute any command and
pass the user name as an argument. All sorts of interesting solutions come
to mind, but they are admittedly more work than something built-in to the
tool. On the other hand, the fact that you can choose your method of
notification is also nice. One could easily send messages to a newsgroup,
instead of mail, or even update a web page or database.

> >Isn't this what "cvs admin -l" does?
>
> Maybe admin -l is actually rcs -l, the rcs lock. It all seems very
> kludgey - the CVS docs say you're on your own if you use it.

Kludgey is the word that comes to everyone's mind. I personally haven't had
to find out because I've been able to live without reserved checkouts. For
the most part, the admin commands are there for when the proper process
fails (like people wanting to correct log messages/descriptions, or deleting
revisions).

> To each his own.

Absolutely! My only goal in replying to the group(s) was to clear up any
inaccuracies on what CVS/WinCVS can do, not to say that it's better.

Cheers,
Jamie



Article: 28646
Subject: problem with Xilinx CORE Generator
From: Lars Rzymianowicz <larsrzy@ti.uni-mannheim.de>
Date: Fri, 19 Jan 2001 16:23:07 +0100
Links: << >>  << T >>  << A >>
Hi,

got Xilinx Alliance 2.1i running on Solaris.
I'd like to use the CORE Generator, but when i start 'coregen',
no cores are listed in the dirs. Not a single core cell is listed
under the categories "Basic Elements", etc. But the resources.lib
file is full of core refs, i also fetched the latest update from the
Xilinx site.

Read through the User Guide, sec. troubleshooting, but no hint about
this. But when i start coregen with the -d debug flag, i get the
following msgs:


Debug: reading preference file /opt6/xilinx/alliance2.1i/coregen/preferences/coregen_lr.prf
Debug: i18N resource file not found for basename message, defaulting to english text
Debug: Load LibraryManager resources from /opt6/xilinx/alliance2.1i/coregen/ip/resources.lib
Debug: Changing File mode to world write: /opt6/xilinx/alliance2.1i/coregen/preferences/known.prj
Debug: Loaded known Project file /opt6/xilinx/alliance2.1i/coregen/preferences/known.prj
Debug: Parsing Plug-in file: /opt6/xilinx/alliance2.1i/coregen/lib/com/xilinx/encore/util/plugin/fixed.cgd
...[more of this]...
Debug: Parsing Plug-in file: /opt6/xilinx/alliance2.1i/coregen/lib/com/xilinx/encore/util/plugin/sim.eld
Debug: i18N resource file not found for basename gui, defaulting to english text
Debug: Loaded Project file /home/lr/tmp/coregen/coregen.prj
Debug: Loading IP library /opt6/xilinx/alliance2.1i/coregen/ip/xilinx/varmult_v1_0/com/xilinx/ip
Debug: Loading IP library /opt6/xilinx/alliance2.1i/coregen/ip/xilinx/catalog_v1_0/com/xilinx/ip
...[more of this]...
Debug: Loading IP library /opt6/xilinx/alliance2.1i/coregen/ip/xilinx/sincos_v1_0/com/xilinx/ip
Set current Project to /home/lr/tmp/coregen


So the tool loads the resources.lib file and all the IP libraries! But
they don't show up in the GUI. Puuhh, what's goin' on here?

Any help welcome!
Lars
-- 
Address:  University of Mannheim; B6, 26; 68159 Mannheim, Germany
Tel:      +(49) 621 181-2716, Fax: -2713
email:    larsrzy@{ti.uni-mannheim.de, atoll-net.de, computer.org}
Homepage: http://mufasa.informatik.uni-mannheim.de/lsra/persons/lars/

Article: 28647
Subject: About JTAG
From: "Juan M. Rivas" <jmrivas@media.mit.edu>
Date: Fri, 19 Jan 2001 10:55:55 -0500
Links: << >>  << T >>  << A >>
This is a multi-part message in MIME format.
--------------5334DB3F45F5CD39B92C0BD0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hi all!

Does anyone know about a good book or article to understand JTAG and
boundary scan?  What I've found so far is not really clear...

Thanks

Attn. Juan
--------------5334DB3F45F5CD39B92C0BD0
Content-Type: text/x-vcard; charset=us-ascii;
 name="jmrivas.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Juan M. Rivas
Content-Disposition: attachment;
 filename="jmrivas.vcf"

begin:vcard 
n:Rivas;Juan M.
tel;home:(617)255-1268
tel;work:(617) 253-5097
x-mozilla-html:FALSE
org:Media Lab, M.I.T.;Object Based Media
version:2.1
email;internet:jmrivas@media.mit.edu
title:Research Assistant
adr;quoted-printable:;;20 Ames Street=0D=0A;Cambridge;Massachusetts;02139;U.S.A.
fn:Juan M. Rivas
end:vcard

--------------5334DB3F45F5CD39B92C0BD0--


Article: 28648
Subject: FPGAs with a partial reconfiguration
From: pawel5732@my-deja.com
Date: Fri, 19 Jan 2001 16:21:02 GMT
Links: << >>  << T >>  << A >>
Hi,
What FPGAs do support a partial reconfiguration?
What FPGAs can be reconfigured "on-the-fly"?
Is this possible to use this feature with a tool?
Thank you,
Pawel


Sent via Deja.com
http://www.deja.com/

Article: 28649
Subject: Re: Virtex-II officially launched
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Fri, 19 Jan 2001 08:44:40 -0800
Links: << >>  << T >>  << A >>
Andy,

We have IBIS models.  See the web page.

For those real power users who want to use DCI at 65 ohms, or 35 ohms, we have
IBIS models for those as well.

Since the IBIS models for DCI's at any impedance is just silly, we picked about
6 different system impedances to cover the cases.  If you are concerned about 68
ohms, and we have a model for 65 ohms, then I think you need to seriously
consider going to an SI class (to learn what is important, and what is not).

If you need impedances for systems other than 50 ohms, contact your FAE to
obtain the IBIS models.  We may also place these on the web if there is a lot of
demand for them.  It really is a tremendous amount of work that we had to do to
get ready for this release.  I hope someone somewhere appreciates that we
thought of these things.

By the way, good catch.  Thought you had us, didn't you?  We are not perfect
(!!!! especially me), so if we do miss something, tell us right away,

On a practical note, making impedances higher than 50 ohms is almost impossible
today (lines too skinny to fab), and lower than 50 ohms takes up too much room.
The ULVDS specification just got changed from 60 ohms to 50 ohms because of
that.

Austin

Andy Peters wrote:

> Austin Lesea wrote:
> >
> > Bob,
> >
> > The impedance updates are asynchronous, and are done in such a manner that
> > the updates trim values are all extremely small.  The resultant dR/dt is
> > undetectable in the signal integrity of the output.
>
> How will you handle the variable impedances in the IBIS models?
>
> Please tell me you'll have IBIS models ...
>
> -- a
> ----------------------------
> Andy Peters
> Sr. Electrical Engineer
> National Optical Astronomy Observatory
> 950 N Cherry Ave
> Tucson, AZ 85719
> apeters (at) n o a o [dot] e d u
>
> "It is better to be silent and thought a fool,
>  than to send an e-mail to the entire company
>  and remove all doubt."




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