Custom Search
Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
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
Utku Ozcan wrote: > Are the lines separated by newline characters? Maybe > Ngdbuild UCF parser might not handle newline characters. > Can it be the case? There is only one line with all the FFs names. I even tried to replace the separating spaces with dots (FFS(copier_cout(41):copier_cout(42):...) but it gave the same error. I'm puzzled that a Xilinx tool generates things that are not recognized by other Xilinx tools... -- Nicolas MATRINGE IPricot European Headquarters Conception electronique 10-12 Avenue de Verdun Tel +33 1 46 52 53 11 F-92250 LA GARENNE-COLOMBES - FRANCE Fax +33 1 46 52 53 01 http://www.IPricot.com/Article: 28526
Hi all, The Xilinx documentation says that a CLB can implement any 5-variable function. However, this is not the case when I compile the design implementing a 5-variable function. Usually two or more CLBs are required. Please help how to fix this problem. Kwong,Article: 28527
Nicolas Matringe wrote: > Utku Ozcan wrote: > > > Are the lines separated by newline characters? Maybe > > Ngdbuild UCF parser might not handle newline characters. > > Can it be the case? > > There is only one line with all the FFs names. I even tried to replace > the separating spaces with dots > (FFS(copier_cout(41):copier_cout(42):...) but it gave the same error. > > I'm puzzled that a Xilinx tool generates things that are not recognized > by other Xilinx tools... Have you tried: TIMEGRP "your_group"= FFS (net1) FFS (net2) FFS (net3) ... ; neti, i=1,2,.. might include wildcharts. That's what I have used to group 30 nets, and I didn't give any error. Tool was M1.5 with latest patch. UtkuArticle: 28528
What happened to Synplicity's newsgroup? The newsgroup browser tells that the server is down... UtkuArticle: 28529
Kwong Chan wrote: > Hi all, > > The Xilinx documentation says that a CLB > can implement any 5-variable function. However, this is not > the case when I compile the design implementing a 5-variable > function. Usually two or more CLBs are required. > > Please help how to fix this problem. > > Kwong, In the worst case, have you not used map with -k option? UtkuArticle: 28530
Utku Ozcan wrote: > Have you tried: > > TIMEGRP "your_group"= FFS (net1) FFS (net2) FFS (net3) ... ; > > That's what I have used to group 30 nets, and I didn't give any > error. Tool was M1.5 with latest patch. I haven't tried to separate each FF. I made 4 groups but each one gave an error. I'll try it. -- Nicolas MATRINGE IPricot European Headquarters Conception electronique 10-12 Avenue de Verdun Tel +33 1 46 52 53 11 F-92250 LA GARENNE-COLOMBES - FRANCE Fax +33 1 46 52 53 01 http://www.IPricot.com/Article: 28531
Article: 28532
> The thing is that there would be no reason (except power consumption) to > implement correlation based approach in an FGPA since you can already > get real-time in sowtware with recent CPUs or DSPs... This sounds not possible in my eyes for large resolutions and complete-image-disparities with more sophisticated filters, as in http://www.ai.sri.com/~konolige/papers/svm.pdf only relatively small images have been taken (320x240, 32 disparities). I believe that this problem maps far better to an FPGA design than to a DSP.Article: 28533
Hi Juan, on http://www.x2e.de/ you'll find a wide range of so-called "Spyder-Virtex" PCI based FPGA boards (including the XCV2000E;-), especially designed for prototyping. Just send me an email, if you are interested in further details! --Sven Fleck University of Tuebingen Germany "Juan Antonio Gómez Pulido" <jangomez@unex.es> wrote in message news:3A62C11A.57D215C8@unex.es... > Hi! > > I am looking for a prototyping board with a great Virtex FPGA. This > board should have PCI communication, RAM modules and interfacing > software. > > Thanks a lot, > > -- Juan > University of Extremadura. Spain > >Article: 28534
> 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 use QVCS from http://www.qumasoft.com/ . It's almost free at $25. Small, easy to install & use. Runs under windoze ... you can't always choose. Berni.Article: 28535
Ray Andraka wrote: > NO! > > There are a few things the automatic placement is exceptionally poor at, with > pin placement leading the list, followed by tbuf placement, and data-path > placement. You should always specify the pin placement rather than letting the > tools do it. Even a bad guess is likely to be better than the random placement > generated by the tools, especially if you start itrating a design. See earlier > posts in this thread for some guidelines on assigning pins. > > As for number of pins used, 100% is not a problem...most of the designs I touch > have 100% of the pins defined. The only time 100% pin utilization becomes a > problem is when you let the tools do the assigning! > > Jakab Tanko wrote: > > > > NO, pins are better left to be picked by the place&route tool. > > At minimum I think you should put together a dummy design, > > if you don't have time for a detailed one, do a quick place and route > > and go with that. As for the pins 100% is 20% to many used pins, I > > would select a larger device or different package to get more I/O pins. > > This is of course just my opinion and I could be wrong, > > > > jakab > > Another consideration is that leaving the pin assignment to the P&R tools can lead to a pinout that makes the PCB autorouter give up & die. Esp the case with the full square FG devices e.g. FG676.Article: 28536
Michael Boehnel wrote: > Hi! > > When I start a new design with partially coded (not yet fully > implemented) black boxes and location constraints for a fixed pin > assignment I often have the problem that the unused pins are optimized > away by the synthesis tool. This results in an error of the place&root > tools since the location constraint can't be met. > > Is it really necessary that the P&R tool results in an error if a pin in > the constraint file isn't existent? I think a warning would be enough so > that I don't have to set extra attributes in the VHDL/Verilog code!? > > Whats' the easiest way to prevent that (yet) unused pins are optimized > away? > > Thanks, > > Michael The easiest way is to temporarily change all the pins to outputs & drive them either to 0 or - better - to the value of one of the already defined inputs.Article: 28537
Hello, I know this is a little off topic ... FPGA that I'm using have CMOS threshold on CLK input. Vtrh=3.1v and Vtrl=1,3v ( 5V supply ). Standard Quartz oscillator build with TTL gates inside can reach these thresholds on lower frequencies <10MHz. On higher frequencies amplitude drops and threshold are not reached or, in some cases, FPGA is randomly triggered. Also CLK input have very low impendance. I have build my own quartz oscillator with BF246 transistor but it requires 30V to archive 3Vpp output when loaded with FPGA. Can anybody give me an advice how to implement crystal oscillator with: - output amplitude >2.5 v - work with crystals from 1MHz to ~50 MHz ( fundamental frequency ) - high fanout - without expensive special components. Thankful for any design experience. -- Veselic Mladen Sigma Laboratorija ECO InzenjeringArticle: 28538
Peter Alfke wrote:Sorry for the blatant propaganda. Got carried away by my enthusiasm. > > I had been impatiently waiting for the Ides of January, for a long > time ! > > Peter Alfke, Xilinx Applications > ===================================== Peter, Sorry to sound a dampening note here among all the enthusiasm but a real world question occurs: Are Xilinx going to avoid the SpartanII mistake of letting a few major customers soak up all the Virtex-II production leaving those of us reliant on the disti's completely out of luck ? [lead time > 20 weeks => no design in] [Anybody remember Ray's joke about the SpartanII T-shirt ?]Article: 28539
Nicolas Matringe wrote: > Utku Ozcan wrote: > > > Have you tried: > > > > TIMEGRP "your_group"= FFS (net1) FFS (net2) FFS (net3) ... ; > > > > That's what I have used to group 30 nets, and I didn't give any > > error. Tool was M1.5 with latest patch. > > I haven't tried to separate each FF. I made 4 groups but each one gave > an error. I'll try it. > > -- > Nicolas MATRINGE IPricot European Headquarters > Conception electronique 10-12 Avenue de Verdun > Tel +33 1 46 52 53 11 F-92250 LA GARENNE-COLOMBES - FRANCE > Fax +33 1 46 52 53 01 http://www.IPricot.com/ Hi There Nicholas, One possibility that has happened to me before is that the real error is in the line *before* the one that NGDBUILD complains about.Article: 28540
In article <93v2rr$f97$1@nnrp1.deja.com>, bjorn_lindegren@my-deja.com wrote: > Hi > > I use a Spartan device, and some thing is wrong when programming it. > > I know that I can check on an out pin for a check sum. > This check sum tells me when the fpga is right programed. > > But whitch pin do I have to check on? I use an LED on the DONE pin. Leon -- Leon Heller, G1HSM Tel: (Mobile) 079 9098 1221 (Work) +44 1327 357824 Email: leon_heller@hotmail.com Web: http://www.geocities.com/SiliconValley/Code/1835 Sent via Deja.com http://www.deja.com/Article: 28541
Utku Ozcan wrote: > What happened to Synplicity's newsgroup? > The newsgroup browser tells that the server is down... > > Utku I've never been able to get to it either - I contacted Synplicity about this a long time ago but never got any result.Article: 28542
Mladen If you lived in the US, the answer would be very simple: Go to the nearest parts distributor and buy a canned crystal oscillator for about $1 to $2 in small quantity. ( It runs on 3.3 or 5 V and drives your FPGA easily) The FPGA input has a 10 pF max impedance, and no dc current ! There are also a number of oscillator manufacturers in Europe ( I remember some Swiss manufacturers from the recent Electronica fair). Building your own with a transistor, and ending up with an unacceptable supply voltage is not the right solution. Regards, Peter Alfke ========================== Mladen Veselic wrote: > Hello, > > I know this is a little off topic ... > > FPGA that I'm using have CMOS threshold on CLK input. Vtrh=3.1v and > Vtrl=1,3v ( 5V supply ). > Standard Quartz oscillator build with TTL gates inside can reach these > thresholds on lower frequencies <10MHz. On higher frequencies amplitude > drops and threshold are not reached or, in some cases, FPGA is randomly > triggered. Also CLK input have very low impendance. > I have build my own quartz oscillator with BF246 transistor but it requires > 30V to archive 3Vpp output when loaded with FPGA. > Can anybody give me an advice how to implement crystal oscillator with: > - output amplitude >2.5 v > - work with crystals from 1MHz to ~50 MHz ( fundamental frequency ) > - high fanout > - without expensive special components. > > Thankful for any design experience. > > -- > Veselic Mladen > Sigma Laboratorija > ECO InzenjeringArticle: 28543
Hi All, Is it possible to drive a clock line for a Spartan II device and three Xilinx XC9500XL CPLD devices from a user IO pin on another Spartan II. The application is one where I want logic within the main Spartan to select between two different input clocks. To be used for itself and the other devices. Thanks Dean ArmstrongArticle: 28544
Sorry, I also forgot to mention that this will only be operating at low frequencies (~2MHz). Dean Armstrong wrote: > Hi All, > > Is it possible to drive a clock line for a Spartan II device and three > Xilinx XC9500XL CPLD devices from a user IO pin on another Spartan II. > The application is one where I want logic within the main Spartan to > select between two different input clocks. To be used for itself and the > other devices. > > Thanks > Dean ArmstrongArticle: 28545
"Dean Armstrong" <daa1@cs.waikato.ac.nz> wrote in message
news:3A64DFBC.A895C809@cs.waikato.ac.nz...
> Sorry, I also forgot to mention that this will only be operating at low
> frequencies (~2MHz).
>
> Dean Armstrong wrote:
>
> > Hi All,
> >
> > Is it possible to drive a clock line for a Spartan II device and three
> > Xilinx XC9500XL CPLD devices from a user IO pin on another Spartan II.
> > The application is one where I want logic within the main Spartan to
> > select between two different input clocks. To be used for itself and the
> > other devices.
> >
> > Thanks
> > Dean Armstrong
Dean,
It' hard to determine exactly what you are talking about. Let me see if I
have this right. You have two Spartan II devices and three CPLDs. You call
one Spartan II device the "main" device, and you want logic on this device
to select from two different input clocks for itself as well as the other
Spartan II device and three CPLDs.
This means that the main Spartan II will have two or more input clocks, and
it is going to select the clock and drive itself as well as the other
Spartan II and the three CPLDs. I thnk that the following are applicable:
1. You won't have a clock out of main Spartain II
until after configuration. It could glitch, too.
Can you handle this?
2. The logic selecting the clock is totally combinatorial
and not dependent on any of the clocks.
3. You drive the selected clock to a IO pad and distribute
the clock to both Spartan IIs and CPLDs.
4. The IO pad will be tristated and weakly pulled high.
You might want to pull it high with an external pull up
resistor.
5. The clock network will probably have no termination
or a mild termination (pullup resistor?) due to the
Spartan II 24ma limit. The clock may be 2 MHz, max,
but you still have to contend with those edges.
6. You're not too worried about duty cycle. At 2 MHz,
I think this would be true.
Without having more information to work with, this is what I came up with.
It would be nice to know how you are originally generating the clock, as the
clock selection can be made there, too.
With what you have, though, I would be tempted to select the clocks in the
CPLDs, though, because they come right up after power on.
Simon Ramirez, Consultant
Synchronous Design, Inc.
Oviedo, FL USA
Article: 28546Hi Simon, Thanks for your reply. You understood me correctly. The intention is that one clock source be a crystal oscillator at about 2MHz, the other will be a push button (which will be debounced in programmable logic somewhere). A third input (an on/off switch) will select between the two. This will allow the devices to run on the clock or be switched into a single-step mode. Basically, what I am designing is a computer system that can free-run, or be switched into a mode where the CPU and bus operations are single stepping. Some of these devices (eg UART, programmable timer) will still need the high speed clock to perform their operations correctly, but all bus transactions with them will be single stepped. Thanks, Dean "S. Ramirez" wrote: > "Dean Armstrong" <daa1@cs.waikato.ac.nz> wrote in message > news:3A64DFBC.A895C809@cs.waikato.ac.nz... > > Sorry, I also forgot to mention that this will only be operating at low > > frequencies (~2MHz). > > > > Dean Armstrong wrote: > > > > > Hi All, > > > > > > Is it possible to drive a clock line for a Spartan II device and three > > > Xilinx XC9500XL CPLD devices from a user IO pin on another Spartan II. > > > The application is one where I want logic within the main Spartan to > > > select between two different input clocks. To be used for itself and the > > > other devices. > > > > > > Thanks > > > Dean Armstrong > > Dean, > It' hard to determine exactly what you are talking about. Let me see if I > have this right. You have two Spartan II devices and three CPLDs. You call > one Spartan II device the "main" device, and you want logic on this device > to select from two different input clocks for itself as well as the other > Spartan II device and three CPLDs. > > This means that the main Spartan II will have two or more input clocks, and > it is going to select the clock and drive itself as well as the other > Spartan II and the three CPLDs. I thnk that the following are applicable: > 1. You won't have a clock out of main Spartain II > until after configuration. It could glitch, too. > Can you handle this? > 2. The logic selecting the clock is totally combinatorial > and not dependent on any of the clocks. > 3. You drive the selected clock to a IO pad and distribute > the clock to both Spartan IIs and CPLDs. > 4. The IO pad will be tristated and weakly pulled high. > You might want to pull it high with an external pull up > resistor. > 5. The clock network will probably have no termination > or a mild termination (pullup resistor?) due to the > Spartan II 24ma limit. The clock may be 2 MHz, max, > but you still have to contend with those edges. > 6. You're not too worried about duty cycle. At 2 MHz, > I think this would be true. > > Without having more information to work with, this is what I came up with. > It would be nice to know how you are originally generating the clock, as the > clock selection can be made there, too. > > With what you have, though, I would be tempted to select the clocks in the > CPLDs, though, because they come right up after power on. > Simon Ramirez, Consultant > Synchronous Design, Inc. > Oviedo, FL USAArticle: 28547
"Juan Antonio Gómez Pulido" <jangomez@unex.es> wrote in message
news:3A62C11A.57D215C8@unex.es...
> Hi!
>
> I am looking for a prototyping board with a great Virtex FPGA. This
> board should have PCI communication, RAM modules and interfacing
> software.
>
> Thanks a lot,
>
> -- Juan
> University of Extremadura. Spain
Juan,
Have you looked at PLX's prototyping board? Check out:
http://www.plxtech.com/tools/rdk/9054rdk-lite.htm
This kit is a PCI prototype adaptor board with a PLX PCI9054 32-bit/33MHz
PCI interface chip. It also has surface mount footprints on the board that
support industry standard packages. It comes with a CD-ROM which includes
complete hardware design info and software tools, inclduing a reference
manual, OrCAD schematics, Gerber files, BOM, software tools, etc.
In short, this kit has everything you need to put some devices on the
board, wire it together to the PCI chip, plug it into a PCI slot and start
working on the software. This kit is great if you have to do a one of a
kind product. The PCI interface chip is an industry standard, too.
Simon Ramirez, Consultant
Synchronous Design, Inc.
Oviedo, FL USA
Article: 28548hamish@cloud.net.au (Hamish Moffatt VK3SB) writes: > Sure, but ClearCase is much better. For one thing, because it > has a custom virtual file system on the client (MVFS -- multi version > file system) it can do a lot of things which CVS can't do. On the other hand, depending on a special file system is also a disadvantage. It's not portable, and it prevents you from using some tools that are available for normal file systems. > It's integrated quite well into NT, as well as Unix. Too well, some of us claim. CVS has some deficiencies, but for most projects it works quite well.Article: 28549
"Mladen Veselic" <veselic@eunet.yu> writes: > FPGA that I'm using have CMOS threshold on CLK input. Vtrh=3.1v and > Vtrl=1,3v ( 5V supply ). > Standard Quartz oscillator build with TTL gates inside can reach these > thresholds on lower frequencies <10MHz. On higher frequencies amplitude > drops and threshold are not reached or, in some cases, FPGA is randomly > triggered. Why not buy CMOS-output oscillators? I've had no trouble with them.
Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
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