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
"H.L" <alphaboran@yahoo.com> schrieb im Newsbeitrag news:a6pkdl$2fh6$1@ulysses.noc.ntua.gr... > Hello all, > I use a BUFGDLL in a Virtex-E FPGA to succeed a proper distribution of a > clock (155MHz). I instantiate BUFGDLL in my code and I did the port map. In > this way I have not access to the RST pin of the DLL (I want to set it '0' > as xilinx suggests). I checked that DLL's CLKIN,CLKFB and CLK0 pins are OK > but I cant check if the RST pin is '0' by default. How can I check if the > RST pin is grounded or not? Have a look at the design in FPGA editor. OR just instanciate the DLL and BUFG seperately, which is saver, since it gives you access to the reset (just in case) -- MfG FalkArticle: 40751
"Markus Meng" <meng.engineering@bluewin.ch> schrieb im Newsbeitrag news:aaaee51b.0203132355.60183b6b@posting.google.com... > Hi all, > > on the Xilinx documentation I do find, that the bitstream > format has a minimum of 8 '1' at the very beginning of > the serial bitstream. Is it possible to extend the minimum > 8 of '1' to any higher number before the start-pattern > is being applied? I dont know at all. Why do you want to do this? The SpartanXL are known to be a little bit more sensitive on configuration data clocking, since the use a length counter encoded into the first bits of the datastream. Spartan-II(E) are much easier to use, since the configuration is much simplified. -- MfG FalkArticle: 40752
"Martin Sauer" <msauer@gmx.net> schrieb im Newsbeitrag news:a6pqk7$cpa$06$1@news.t-online.com... > Hello, > > one question: Is it possible to connect a 5V IO Device direct to the Xilinx > Virtex FPGA series? Yes. The Virtex inputs (as well as Spartan-II) are 5V tolerant. You can also drive a 5V device with the 3.3V IO of the Virtex, when the 5V device has TTL inputs (1.4V switching level) Virtex-E/Spartan-IIE/Virtex-II are NOT 5V tolerant. But you can add tolerance using a 100Ohm series resistor. But this will influence timing. -- MfG FalkArticle: 40753
A first order observation is that the tool results are pretty close; you know they try each others sythesizers ou and see what they produce so the battle goes back and forth. For your benchmarking, one thing to make sure of, is that you run the FPGA vendor P&R tools on the EDIF output from the sythesizers and compare the final P&R numbers. I've found HUGE differences pre and post place and route circuit speeds. I can understand the marketing reasons for doing this but still think is an underhanded method of wooing potential customers. Then report your results anonymously on this newgroup for everyone to see. This non-compete stuff is highly questionable in my view; of course I've never done this, but occasionally an anonymous poster comes on here and does. Regards "Arash Salarian" <arash.salarian@epfl.ch> wrote in message news:<3c907f23$1@epflnews.epfl.ch>... > hmmmmmmmm, quite strange! It can explain why I could not find any useful > benchmark on the web for the recent verstions of these tools.... > > But the question remains, and it's that if there is no benchmark > comparision, how would one choose one tool over another one? Just looking at > the feature list and vendor claims does not look like the best way... > > A year ago, I used one of my big DSP designs as benchmark for these tools > and at that time Synplify outperformed both Leonardo and FPGA compiler for a > Flex10K target. But today? Who knows how these tools compete against each > other? > > Should I try to downlaod the trial version of the tools, and run all of them > on my latest designs to choose one? For each new design? :) I don't think > so... > > Regards > Arash > > > "S. Ramirez" <sramirez@cfl.rr.com> wrote in message > news:H3Rj8.118337$Dl4.13204394@typhoon.tampabay.rr.com... > > "Arash Salarian" <arash.salarian@epfl.ch> wrote in message > > news:3c8f4cb7$1@epflnews.epfl.ch... > > > Hello, > > > > > > Do you know of any recent review/benchmark comparision of major > synthesis > > > tools for the FPGA? I'm interested in such a comparision between > Synplify, > > > LeonardoSpectrum and FPGA compiler. And is there any major perfromance > > > difference between these tools when targetting different FPGAs from > Xilinx > > > and Altera? > > > > > > Best Regards > > > Arash > > > > You will be hard pressed to find these comparisons, since the license > > agreements of these vendors say something close to "Licensee shall not .. > > disclose the results of any benchmarking of the SOFTWARE, or use such > > results for its own competing software development activities, without the > > prior written permission of Blah Blah." > > Don't blame me, I'm just the messenger! > > Simon Ramirez, Consultant > > Synchronous Design, Inc. > > Oviedo, FL USA > > > >Article: 40754
Vertex 2 came after Vertex E. In general, FPGA families follow steps in CMOS fab process technology. Each new process seems to spawn a new family. The previous process technology becomes the "value" product line (Spartan). To the end user, the Vertex 2 line gives (over the Vertex E) you higher densities, higher speeds (20% in my application), lots more on chip ram, hardwired multipliers and new tools to debug. Regards Martin Sauer <msauer@gmx.net> wrote in message news:<3C904A7E.5010304@gmx.net>... > Hi, > > can you tell me the difference between the Xilinix Virtex-II and the > Virtex-E Series? > Thanks for your answer. > > bye > > martinArticle: 40755
Hi Nahum, please tell more about the "hidden features" of the Virtex-II chip. -Manfred Kraus >.......... Like most > designs, there are a lot of hidden features that we placed in there for > test, but may be useful for some applications. > .........Article: 40756
XILINX Answer # 10835 says: 175 Ohm -Manfred Kraus Falk Brunner <Falk.Brunner@gmx.de> schrieb in im Newsbeitrag: a6qsl1$gikto$4@ID-84877.news.dfncis.de... > "Martin Sauer" <msauer@gmx.net> schrieb im Newsbeitrag > news:a6pqk7$cpa$06$1@news.t-online.com... > > Hello, > > > > one question: Is it possible to connect a 5V IO Device direct to the > Xilinx > > Virtex FPGA series? > > Yes. > The Virtex inputs (as well as Spartan-II) are 5V tolerant. You can also > drive a 5V device with the 3.3V IO of the Virtex, when the 5V device has TTL > inputs (1.4V switching level) > > Virtex-E/Spartan-IIE/Virtex-II are NOT 5V tolerant. But you can add > tolerance using a 100Ohm series resistor. But this will influence timing. > > -- > MfG > Falk > > > > >Article: 40757
Have you checked the Cesys FPGA boards ? Click on "Products" at www.cesys.com There is a USB and a PCI Version available. If you dont like to use the interfaces to transmit data between a PC software and the FPGA you can simply treat them as a "configuration download channel". Doubleclick on the Bit or EXO file that is generated by the Xilinx WebPack and it will be downloaded. There is an discount for educational institutions, just ask. Sorry for making advertisment in this NG. I couldnt resist. Manfred Kraus CESYS GmbHArticle: 40758
"Manfred Kraus" <newsreply@cesys.com> schrieb im Newsbeitrag news:a6qtja$gg5ad$1@ID-22088.news.dfncis.de... > XILINX Answer # 10835 says: 175 Ohm Hmm, not sure. I remember to read somewhere about 100 Ohm. -- MfG FalkArticle: 40759
--------------E64C3234107AE28A006C100A Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit David Brown wrote: > Just when I think I am getting somewhere about the languages, I find out > more and end up changing my mind yet again... > > Am I right in thinking that ABEL is only supported by a few vendors? Yes - Xilinx and Lattice. ABEL used to be owned by Data I/O. They were acquired by Minc Synario. Minc Synario was acquired by Xilinx. (If my memory serves me correctly). At the time of the Minc Synario acquisition Lattice bought the source code for ABEL and the Xilinx bought the source code along with the rights. Xilinx has since improved the language and brought many new features to it such as bus support and HDL simulation. > I have > had a brief look at information on tools from Altera, Atmel and Xilinx, and > none of them appear to support ABEL as far as I can see. While this is > going to be a Lattice/Mach project, it would be silly to spend a lot of > effort learning a language that I can't otherwise use for other designs. I have a fondness for ABEL because of its simplicity and direct mapping to the device architecture. However, I do understand the quandary you are in. Just providing the facts though, VHDL and Verilog do seem to be the dominant HDL languages of the day. It does allow easy migration between device families, vendors, and architectures. From a language standpoint my impressions are that VHDL leads with about 50% of the designs being done in this format and 35% in Verilog and about 15% in ABEL. All of these languages are supported by the ISE WebPACK available for free from Xilinx. The ISE WebPACK does support HDL conversion as well. If you have an ABEL source and would like to convert it to either VHDL or Verilog this can be done easily enough. Since you have an existing design, this conversion process may allow you to get a preview how the various languages appear. This should be taken as a rough preview since it is machine generated code. While it may be functional, it probably is not as elegant as the code could be written from scratch. My thoughts on this topic. Regards, LarryArticle: 40760
Yes, I am already aware that a Xilinx Spartan-II IOB tri-state buffer is active low (Took a lot of time to figure out the IOB merging rules.), so my design already handles that part fine. Following your suggestion, I attached "keep" attribute to tell synthesis tool not to tinker with the only OE FF for AD[31:0] and the only OE FF for C/BE#[3:0], but XST ignored the "keep" attribute anyway, and duplicated them. Just as an experiment, I attached "keep" attribute to the output FFs of AD[31:0] and C/BE#[3:0] which I don't normally do (Because that way the PCI IP core won't meet Tco (Tval) of 11ns.), but XST overdid that, too, probably because Pack I/O Registers into IOBs option has some kind of precedence over the "keep" attribute. Kevin Brace (Don't respond to me directly, respond within the newsgroup.) Brian Drummond wrote: > > I had the opposite problem for a long time, until I realised the OE > signal _into_ the ENBFFs has to be active low (mine was active high). > Maybe changing the polarity of your OE signal would prevent this > "optimisation"? > > Or, check if your tool chain supports "dont_touch" or "preserve_signal" > attributes on specific signals? > > - BrianArticle: 40761
Larry McKeogh wrote: > > > > David Brown wrote: > > Just when I think I am getting somewhere about the > languages, I find out > more and end up changing my mind yet again... > > Am I right in thinking that ABEL is only supported by a few > vendors? > > Yes - Xilinx and Lattice. ABEL used to be owned by Data I/O. They > were acquired by Minc Synario. Minc Synario was acquired by Xilinx. > (If my memory serves me correctly). At the time of the Minc Synario > acquisition Lattice bought the source code for ABEL and the Xilinx > bought the source code along with the rights. Xilinx has since > improved the language and brought many new features to it such as bus > support and HDL simulation. The PHDL used in XPLA Coolrunner tools, bought by Xilinx, was also an ABEL-variant. Altera's AHDL language is also an ABEL derivative. Atmel offers ABEL and CUPL flows, tho ABEL is being phased out. I would call ABEL and CUPL Tier1 HDLs, and most vendors do have a Tier1 offering. Code porting across these Tier1 languages is not transparent, but a lot easier than porting Assembler :) Many of the report files, from the Fitters reading EDIF files, are in Tier1 language formats. > > I have had a brief look at information on tools from Altera, Atmel > and Xilinx, and > none of them appear to support ABEL as far as I can see. > While this is > going to be a Lattice/Mach project, it would be silly to > spend a lot of > effort learning a language that I can't otherwise use for > other designs. > > I have a fondness for ABEL because of its simplicity and direct > mapping to the device architecture. <snip> Check the devices you expect to target in future, as well as this project: The higher HDLs like VHDL and Verilog do not map universally onto the SPLDs ( mostly, they expect a EDIF import fitter ). They also have detail control issues that become more imporant on the smaller devices. Also, ABEL and CUPL can create Test vectors, allowing hardware test in programmers, expecially usefull on the smaller packages viz 20-44 pins. -jgArticle: 40762
Peter Waldeck wrote: > Hi, > I'm trying to do a design for a VirtexII using ISE 4.1i and I can't > understand the synthesis errors I'm getting. (I am fairly new to FPGA's as > well...) I have two .vhd files - both of which are included in the work > library. The first, mygpio.vhd, tries to instantiate an entity from the > second - pselect.vhd. Within pselect, an architecture imp is declared. But > the synthesis tool doesn't like it - > > ERROR:HDLParsers:3281 - C:\Peripheral\peripheral/mygpio.vhd Line 81. imp is > not an architecture body for pselect in library work. > > I don't understand it - are there some path settings that are wrong? Or do > I have to do things in a specific order? Those slashes going both ways sure looks like something is configured wrong, or a path was entered wrong. If this is on a Windows PC, then all slashes should be the "\" kind. JonArticle: 40763
Utku Ozcan wrote: > Peter Ormsby wrote: > > > newman <newman5382@aol.com> wrote in message > > news:e6038423.0203071338.2e220f61@posting.google.com... > > > I saw on the Xilinx Web site that ISE4.2 is now supported on RedHat7.2 > > > > > > Newman > > > > Newman, > > > > This is slightly misleading. See this post: > > > > http://groups.google.com/groups?hl=en&scoring=d&selm=D9Bh8.13448%24Or3.14935 > > 91%40typhoon.mn.ipsvc.net > > > > -Pete- > > Google: Unable to retrieve article. (I also tried to concatenate wrapped lines, > doesn't go either) Worked for me. But the gist is : Just to be clear here, the Xilinx Linux solution is currently limited to command-line only on WINE. Yuck! command line only? How do you do schematics, floorplanner, FPGA editor, etc.? I have VmWare installed with Windows 2000 under Mandrake 8.1, and it is a dream. VmWare is commercial software, but it is very cheap to educational institutions. I think it is affordable, even to everyone else. They do have a demo version. My Linux/Win 2000 system has been up 97 days! Vmware is www.vmware.com JonArticle: 40764
ssy wrote: > Hi everyone > > in my design, some module use high active reset, and other use low > active reset, and the global reset is low active, so all high active > reset is generate by pass global reset through an inverter, if this > will cause any problem? At least in most Xilinx parts, although you may SPECIFY an inverter, the reset inputs to the FFs actually have a mux to select either polarity of the reset signal, so it is actually done at each FF separately, and the time should be the same through either path. JonArticle: 40765
Austin Lesea wrote: > Jay, > > It is intended to be exactly equivalent to a 1N914 diode. But, it is not a separate diode floating in space. It is built on the same substrate as the rest of the chip, and may well have protection diodes to the Vss and Vdd, as well, so you'd need to keep those reverse biased to properly sense the diode. JonArticle: 40766
Hi Another option is to look at the Altera FIR Compiler after downloading the free eval to evaluate real FIR filters. http://www.altera.com/products/ip/altera/m-alt-fir-compiler.html or http://www.altera.com/products/ip/altera/m-alt-iircompiler.html --Philippe Jacky Renaux <renaux.jacky@wanadoo.fr> wrote in message news:<2002312-184818-183279@foorum.com>... > Hi > I would recommand you to look into www.andraka.com there is a > great tutorial on fir filter using distributed arithmetic > > I am sur It will help you > jackyArticle: 40767
Jon, Exactly correct. There are parasitics present that require the voltages on the diode pins stay with the limits of the power supply to the chip. Austin Jon Elson wrote: > Austin Lesea wrote: > > > Jay, > > > > It is intended to be exactly equivalent to a 1N914 diode. > > But, it is not a separate diode floating in space. It is built on the > same substrate as the rest of the chip, and may well have protection > diodes to the Vss and Vdd, as well, so you'd need to keep those > reverse biased to properly sense the diode. > > JonArticle: 40768
thanks, Ray. I'll do that. Has anyone got a good, beginner-level resource for the best way to go about assigning groups / constraining paths? a note: I'm observing speeds over 120% faster than reported, on a well-used device and with a really lousy clock signal; surely more accurate numbers are possible from the tools. JP "Ray Andraka" <ray@andraka.com> wrote in message news:3C90D449.992B9266@andraka.com... > The old 4000 series stuff seemed to be pretty conservatively timed. It was not > unusual to be able to clock something at 50% faster than the timing report > stated in the lab under ideal conditions, however I would not put such an > overclocked design into production unless you like spending your career fixing > production problems. The numbers reported by the timing analysis are worst > case over voltage, temperature and process. Chances are in the lab you will > not hit the worst case on any of the three much less all of the three. > > For the timing, run the static timing analyzer and set it to report paths > failing timing. Those are sorted so that the worst paths show up first. From > that report, you'll be able to see what CLBs the worst paths go through, so > you'll have an idea where to look. > > Josh Pfrimmer wrote: > > > Hi all, > > For a class (calmer down.. I'm not looking for answers to my homework) > > I've designed an 8-bit CPU. Fifteen instructions, pre-fetching, branch > > predicting, load-store, etc. etc. So far, it cruises along at ~50MHz on an > > XC4010XL-3. This is better than what's necessary for the class, but I'd > > like to push it further. > > > > Now that I have a ballpark speed, what's the best way to get started > > specifying constraints in a .ucf? How do I get the (old, old, old) > > Foundation 3.1 software in the labs to tell me what the worst paths are, so > > that I can start picking out false paths, etc. > > > > Incidentally, the post-layout timing report gives a figure of ~20MHz fo > > the maximum clock speed, but on the board, I'm well over 45.. is this > > typical? How can I get more accurate numbers out of the reports? > > > > Thanks for your attention > > JoshP >Article: 40769
"Austin Lesea" <austin.lesea@xilinx.com> wrote in message news:3C90C799.E64C72C2@xilinx.com... > Virtex E was a shrink to 0.18u of the classic Virtex architecture and > circuitry. Virtex E added LVDS input buffers to the original Virtex > design, but little else was changed. > > Virtex II was a complete redesign in 0.15u at 1.5V for the core, that > extended the reach of the interconnect, and buffered virtually all paths > to reduce loading effects. The CLB had more features added (more LUT > RAM modes, SRL modes), as well as the horizontal carry (useful for > p-terms). The block RAMS got bigger, and there are more of them (4K vs > 18K), as well as having three read/write modes instead of one in > addition to a 18X18 multiplier in each BRAM block. Hi Austin. The multipliers were hella fast when the spec first came out, and have gradually crept down in speed as new "improved" speed files have been issued. Are you working on getting the multipliers back to their original hoped-for speed, or will we have to wait for the next family?Article: 40770
Nitin, If you are a member of the university staff (meaning professor, associate professor, a.k.a non-student) I would suggest you contact the Xilinx University Program. Let them know your course objectives and how you want to use the demo boards. Under the right circumstances you may be able to get your demo boards and software at no cost, or at least significantly reduced cost. (This has been my experience with two universities in my area). Regarding the development board you definitely want to stay away from the SpartanXL and get into the SpartanII for your application. Regards, JasonArticle: 40771
Nahum Barnea wrote: > Hi. > Xilinx tradionally publish only max values for the clock to out > delays. > This isn't quite correct. Although quite often the early datasheets don't publish min delays when a part family matures a little they do tend to appear. Its even possible to do a min timing simulation. > > I remember that there is an application note that say that the min clock > to out value can be taken as 1/3 'rd of the max value (or 1/4 'th if > you are a conservative person). > Check out the `SPEEDPRINT' utility. There's a min timing flag and it will tell you either the min timings or a message saying they are not available. Its some time since I dumped the min & max timings and did a comparison but IIRC the ratio for the 1000001 timing values for original Virtex was, overall, about ~ 1/3.Article: 40772
Markus Meng wrote: > Hi all, > > on the Xilinx documentation I do find, that the bitstream > format has a minimum of 8 '1' at the very beginning of > the serial bitstream. Is it possible to extend the minimum > 8 of '1' to any higher number before the start-pattern > is being applied? > > markus I don't know why you'd want to do this but the answer, for Virtex/Virtex-E and Spartan2 parts is yes. In fact if you look at the bit stream produced by the Xilinx s/w for these devices you'll probably see an initial 32 1's before the sync word. If you want to delay configuration its probably better to hold INIT~ active until you're ready.Article: 40773
Falk Brunner wrote: > "Martin Sauer" <msauer@gmx.net> schrieb im Newsbeitrag > news:a6pqk7$cpa$06$1@news.t-online.com... > > Hello, > > > > one question: Is it possible to connect a 5V IO Device direct to the > Xilinx > > Virtex FPGA series? > > Yes. > The Virtex inputs (as well as Spartan-II) are 5V tolerant. You can also > drive a 5V device with the 3.3V IO of the Virtex, when the 5V device has TTL > inputs (1.4V switching level) > > Virtex-E/Spartan-IIE/Virtex-II are NOT 5V tolerant. But you can add > tolerance using a 100Ohm series resistor. But this will influence timing. > Or, if board area and cost permit, use some QuickSwitch style parts to do the level translation. These are just a bunch of FETs whose on resistance is very low (5-10R) until the driving signal gets to about 0.7V below its power pin, when it starts increasing rapidly. I think, for something like PCI, this is much less intrusive than a bunch of 100Rs. We tend to still use the QS3245 or Pericom equivalent but with the VCC Zener'ed down to 3.9V, or even connected to the 3.3V supply.Article: 40774
Josh Pfrimmer wrote: > thanks, Ray. I'll do that. Has anyone got a good, beginner-level resource > for the best way to go about assigning groups / constraining paths? > > a note: I'm observing speeds over 120% faster than reported, on a well-used > device and with a really lousy clock signal; surely more accurate numbers > are possible from the tools. > The first and most basic thing to do is very simple. Just put a `PERIOD' time spec on your clock(s). If you have only one clock then that's almost good enough on its own. From then on you are basically into relaxation constraints [or tightening ones if you've got any clock domain crossings or async input sampling]. I can remember asking a similar question to yours some years ago and getting an answer - read the manuals. I thought that was a bit unhelpful but for once this was absolutely the correct reply. The problem with Xil contraints is that beyond the basic `PERIOD' it rapidly starts getting murky with issues like contraint priority rearing up. In contrast with most of the Xilinx tools the contraints documentation is so poorly written [probably because `Hey we've got a constraints GUI, who needs all that antique Gutenberg stuff!'] that the even more correct answer is to read it N times with N >= 4. I actually allocated a whole, otherwise free, Sunday for this process and got to about 50% understanding of the most basic timing stuff at the end of it. Note that 4.X now has a single Contraints manual instead of dispersing the info between the Development System Reference Guide, the Libraries Guide, and a few other places as well but .... in many ways its worse, whoever wrote it had just been told about hypertext ... Somebody, somewhere, needs to do the world a service and write `Xilinx Timing Constraints for Dummies' but AFAIK this magic volume doesn't exist.
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