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
Christos Dimitrakakis wrote: > > Anyone got any info on RNGs? > Since I'm only generating a 6-bit number with it > I could just use a simple counter scheme for it that runs @8Mhz > while using one of the other clock outputs provided in the chip, > say the 490Hz one, for the rest of the circuitry. > Will that reduce the decorrelate the timing of the RNG from the > rest of the chip? > > Is there any better way that does not take up a lot of chip area? > [I considered using a Fibonacci series, but it takes up too much space > if sufficient randomness is required] > Peter Alfke at Xilinx put together a good app note on compact long linear feedback shift registers for the 4000 family, available in pdf @ http://www.xilinx.com/apps/counter.htm Use a long enough one, and you probably don't have to worry about correlation. -- John L. Smith Univision Technologies, Inc. 6 Fortune Drive Billerica, MA 01821-3917 jsmith@univision.comArticle: 5376
Christos Dimitrakakis wrote: > > Anyone got any info on RNGs? > Since I'm only generating a 6-bit number with it > I could just use a simple counter scheme for it that runs @8Mhz > while using one of the other clock outputs provided in the chip, > say the 490Hz one, for the rest of the circuitry. > Will that reduce the decorrelate the timing of the RNG from the > rest of the chip? > > Is there any better way that does not take up a lot of chip area? > [I considered using a Fibonacci series, but it takes up too much space > if sufficient randomness is required] > Peter Alfke put together a good app note on compact long linear feedback shift registers for the 4000 family, available in pdf at: http://www.xilinx.com/apps/counter.htm With a long enough counter of this type , you don't need to worry about correlation, and it doesn't take up much space either. -- John L. Smith Univision Technologies, Inc. 6 Fortune Drive Billerica, MA 01821-3917 jsmith@univision.comArticle: 5377
Christos Dimitrakakis wrote: > > Anyone got any info on RNGs? > Since I'm only generating a 6-bit number with it > I could just use a simple counter scheme for it that runs @8Mhz > while using one of the other clock outputs provided in the chip, > say the 490Hz one, for the rest of the circuitry. > Will that reduce the decorrelate the timing of the RNG from the > rest of the chip? > > Is there any better way that does not take up a lot of chip area? > [I considered using a Fibonacci series, but it takes up too much space > if sufficient randomness is required] > If you are looking for a uniform distribution just use a linear feedback shift register counter (LFSR). The sequence out of the LFSR is pseudo-random and will repeat after 2^n -1 clocks (assuming the feedback chosen is for a maximal length sequence). The counter should have several more bits than your desired result, and the msbs of the output should be taken from the input end of the shift register to avoid a tendency toward an exponential function. Refer to the Xilinx app notes on LFSRs and on efficient shift registers using the CLB RAM for implementation details. BTW, the LFSR can be done in just a few CLBs. -Ray Andraka, P.E. Chairman, the Andraka Consulting Group 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://www.ids.net/~randrakaArticle: 5378
"John L. Smith" <jsmith@univision.com> wrote: >Christos Dimitrakakis wrote: >> >> Anyone got any info on RNGs? >> Since I'm only generating a 6-bit number with it >Peter Alfke at Xilinx put together a good app note on compact >long linear feedback shift registers for the 4000 family, >available in pdf @ > >http://www.xilinx.com/apps/counter.htm > >Use a long enough one, and you probably don't have to >worry about correlation. > >-- >John L. Smith >Univision Technologies, Inc. >6 Fortune Drive >Billerica, MA 01821-3917 >jsmith@univision.com Remember, you can use a long sequence RAM based LFSR as suggested by Peter Alfke/John L. Smith, or just a normal LFSR, if you don't need too many bits, Then just use a random process to permanently select 6, of the 'n' shift register bits. Now, you have a 6 bit output that will not repeat for an arbitrarily long time. Dave Decker ddecker@diabloresearch.com mush@jps.netArticle: 5379
"John L. Smith" <jsmith@univision.com> wrote: >Christos Dimitrakakis wrote: >> >> Anyone got any info on RNGs? >> Since I'm only generating a 6-bit number with it . . . > > >Peter Alfke at Xilinx put together a good app note on compact >long linear feedback shift registers for the 4000 family, >available in pdf @ > >http://www.xilinx.com/apps/counter.htm > >Use a long enough one, and you probably don't have to >worry about correlation. > >-- >John L. Smith >Univision Technologies, Inc. >6 Fortune Drive >Billerica, MA 01821-3917 >jsmith@univision.com Remember, you can use a long sequence RAM based LFSR as suggested by Peter Alfke/John L. Smith, or just a normal LFSR, if you don't need too many bits, Then just use a random process to permanently select 6, of the 'n' shift register bits. Now, you have a 6 bit output that will not repeat for an arbitrarily long time.Article: 5380
Knows anybody where you can or is willing to sell a 'simple' PCI prototyping board with a XC4013E or XC4013EX ? A scheme or PCB layout would also be great !!! Purpose : testing VHDL code for the PCI control. I haven't found it on the Xilinx website, even pci@xilinx.com never replied. (I don't need one with six or ten FPGA's , reconfigurable computing machine ..) No PCI prototyping board, PCI prototyping kit, PCI demo kit,... Thanks for any hints 8-]Article: 5381
I tried to e-mail you, but it got bounced. If you want info on this, e-mail me directly or provide a valid e-mail address. I have a board that has a 4013E PQ208 on it, and is attached to the PCI bus. I'm curious how you are going to do a PCI interface for a Xilinx in an HDL. Even doing it in schematics requires a lot of mapping, placement and logic massaging to make 33MHz. You can probably do a simple target implementation, but not a master, much less a burst master in an HDL unless you can map, place and control how the logic gets implemented... Austin Franklin ..darkroom@ix.netcom.com. Peeters <iolair@weirdness.com> wrote in article <01bc188a$b27fba80$8b0c07c2@mon00238>... > Knows anybody where you can or is willing to sell a 'simple' PCI > prototyping board with a XC4013E or XC4013EX ? > A scheme or PCB layout would also be great !!! > Purpose : testing VHDL code for the PCI control. > I haven't found it on the Xilinx website, even pci@xilinx.com never > replied. > (I don't need one with six or ten FPGA's , reconfigurable computing machine > ..) > > No PCI prototyping board, PCI prototyping kit, PCI demo kit,... > > Thanks for any hints 8-] > > > > >Article: 5382
Just a reminder to everyone. I have been compiling a semiconductor manufacturer website listing (and making it available to other engineers on the web) for about a year and a half now. I used to post it to sci.electronics every so often in the past, but it got way too big for that. There are about 350 companies on the list currently. These are chipmakers, so if you can't find the data sheet that you need because you don't know the manufacturer's website, try my listing. Bookmark it and tell your co-workers. I try to keep it the most up-to-date and useful listing anywhere on the web (by constantly searching for new URLs before most others find them) and it is now widely used by a lot of people. The following categories exist among my "engineering" subpages. - Semiconductor URLs (Brief listing) - Semiconductor URLs (Verbose listing; includes product categories) - New Semiconductor URLs - Missing Semiconductor URLs - Motivations - 25 Best Semiconductor Websites - 25 Worst Semiconductor Websites - Engineering Humor try it out, I think you'll find it to be quite useful. Let me know what you think. -- +---------------------------------------------+ | Gray Creager | | http://www.scruznet.com/~gcreager | +---------------------------------------------+ | "If you're not part of the solution, you're | | part of the precipitate." - Steven Wright | +---------------------------------------------+Article: 5383
Peter wrote: > I personally would like to see *schematic* (not HDL) designs for: Doing such designs in schematic is a waste of designer's time. This is so much easier accomplished in a HDL. > 1. The 16450 UART (with the FIFOs done in the schematic). > > 2. The 16550 UART (with the 16-byte FIFOs done using external RAM, and > thus obviously extensible). It would be a waste of an FPGA to do 32 > bytes of FIFO inside the FPGA. Why would you want the 16450/16550 with all the overhead (and limited) functionality? If you start with my compact UART, then you could easily add a FIFO buffer if you need it, and your design will consume very little FPGA space. By the way: If you use an Altera FLEX 10K, the dedicated RAM blocks are ideal for implementing a FIFO buffer. My Compact UART is located at: http://www.geocities.com/SiliconValley/Lakes/3656 _-_rune_-_Article: 5384
hi, does anybody know how the core of a XC4025 is connected to the pads of the package (HQ240). what i have to know is how the core-geometry looks like and which logic- cell is connected to which pad. is there a chance to get these informations from XACTstep? i played around a short time with the tool, so don't expect i really understand the software! thanksArticle: 5385
Juerg Haefliger (juerg@zaphod.ethz.ch) wrote: : hi, : does anybody know how the core of a XC4025 is connected to the pads : of the package (HQ240). : what i have to know is how the core-geometry looks like and which logic- : cell is connected to which pad. : is there a chance to get these informations from XACTstep? i played around : a short time with the tool, so don't expect i really understand the software! : thanks what i need is something like the 'CLB-to-PAD relationship' as described in the documentation for the XC50xx.Article: 5386
You can find out the CLB row/column relationship by using either the XDE chip editor, or the floorplanner. For XDE just open it and select a NEW chip, and set the type to the partnumber and package you will be using. For the floorplanner, you need to create a trivial dummy design so that the floorplanner has something to munch on. The info you want is NOT published in any of Xilinx's documentation for any XC 2000, XC3000, XC4000, XC4000E, XC4000A, XC4000H products. This has annoyed me for years, and I have repeatedly asked for Xilinx to include this info in their documentation. No luck yet ! Another alternative is LCA2PS.EXE at ftp.netcom.com : /pub/fl/fliptron/files_from_fliptron/lca2ps.exe This program needs an empty LCA file (you can create one with an editor) and in -template mode may be just what you need. You need a postscrip printer to view the output. Philip Freidin. In article <5dsic4$bf@elna.ethz.ch> juerg@zaphod.ethz.ch (Juerg Haefliger) writes: >hi, >does anybody know how the core of a XC4025 is connected to the pads >of the package (HQ240). >what i have to know is how the core-geometry looks like and which logic- >cell is connected to which pad. >is there a chance to get these informations from XACTstep? i played around >a short time with the tool, so don't expect i really understand the software! >thanksArticle: 5387
evjapps@inet.uni-c.dk wrote: >Hello all Altera designers! >I would like to promote the idea of a FreeCore Library. The FreeCore >Library consists of free, parameterizable building blocks for Altera >programmable logic that can be used in your design for absolutely free! >My first contribution is the parameterizable Compact UART. This is a >simple, complete and compact UART. An 8-bit UART may consume as little >as 63 logic cells in a FLEX 8000 or FLEX 10K architecture, including >transmitter, receiver and baud rate generator. >The documentation and download page is available at: >http://www.geocities.com/SiliconValley/Lakes/3656/ >Please give me your feedback on the FreeCore philosophy. I plan to set >up a FreeCore site soon, and I hope that you, the designers and >application guys, would like to contribute to the library. >What kind of functions would you like to see in the FreeCore library? >I'm planning my next release soon: A single master I2C (I squared C) >controller. >NOTE: >Although I work as a Field Application Engineer for one of the Altera >distributors, I would like to state that this is totally my personal >initiative. Altera is not involved in any way. >Uffe Tyrsted Toft >------------------------------------------- >ACTE NC Denmark A/S >E. V. J. Elektronik >Titangade 15 >DK 2200 Copenhagen N >Phone: +45 35 86 90 22 >Fax: +45 35 86 90 00 >E-mail: evjapps@inet.uni-c.dk >------------------------------------------- Just checked out and downloaded your UART design. This is GREAT!! I do Altera designs (ALOT nicer than that OTHER companies tools!) and have an application I'll be working on in the next couple of months. I will plan on using your design and see how it works. Keep in touch. Regards, Aaron Quantz Mgr Software Development, TCS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ HR Textron | Phone: (805) 253-5471 25200 W. Rye Canyon Rd. | Fax: (805) 253-5962 Valencia, CA USA 91355-1265 | Email: aquantz@ibm.net Visit the Textron web site: http://www.textron.com ----------------------------------------------------------------------Article: 5388
David Jacobson <jacobson@cello.hpl.hp.com> wrote: }Michael Koch <mkoch@earthlink.net> wrote: }>I have no data whatsoever }>that columbian cocaine dealers encrypt their pick-up schedules w/ DES, } }I don't have a good idea what the bandwidth needed by the columbian }drug system is, but it can't be all that high. A hundred kilobytes a }day would seem like plenty. So all they have to do is exclusive-OR }their message with the LSBs of some pre-agreed-on music CD, used as a }one-time pad. A 70 minute CD has 92 MB of just LSBs, good for }encrypting 920 days of traffic at 100KB per day. ... No, no, no -- the solution is for our government to ban the export of these one-time pads. One exception would be "In a Metal Mood" which we should carpet bomb all of our enemies with. John -- John Hascall, Software Engr. Shut up, be happy. The conveniences you ISU Computation Center demanded are now mandatory. -Jello Biafra mailto:john@iastate.edu http://www.cc.iastate.edu/staff/systems/john/welcome.html <-- the usual crudArticle: 5389
Call For Papers: FPGA IEEE Design and Test of Computers Special Issue Spring 1998, Volume 15, Number 1 Web Page: http://www.cs.tamu.edu/faculty/lombardi/datsi.htm Deadline: May 1st IEEE Design and Test of Computers seeks original manuscripts for a theme issue on Field Programmable Gate Arrays (FPGAs) scheduled to appear in the first issue of 1998. Articles concerning applied research and practical experience reports are solicited. The topics of interest include, but are not limited to: FPGA Fabrication and Technology. Manufacturing, process control, yield enhancement, and novel architectures for and device technology to support field programmable chips. Exploitation of Field Programmability. Configurable computer architectures, rapid prototyping, programmable interconnect architectures, field configurable memories, programmable I/O systems, and novel implementations. Reliable Online Implementations. Online testing, built-in self-test, concurrent testing, design for rapid testability, latency reduction, fault containment, verification of reprogramming, online reconfiguration approaches, and design for reconfigurability. Synthesis Approaches and Tools. Partitioning, logic minimization and technology mapping, placement and routing, test generation, verification of synthesis, and design for synthesizability. Submitted articles must not have been previously published or currently submitted for publication elsewhere. Authors should submit their original work to the guest editor by May 1, 1997, formatted according to the instructions below. Notification of acceptance will be sent September 1, 1997. Camera-ready copy for accepted papers will be due November 1, 1997. Submit articles by May 1, 1997 to: Prof. Fabrizio Lombardi, Guest Editor Department of Computer Science Phone: (409) 845--5464 Texas A & M University Fax: (409) 847--8578 College Station TX 77843-3112 E-Mail: lombardi@cs.tamu.edu Important dates: May 1, 1997: Submission deadline September 1, 1997: Authors notified of acceptance with requested revisions November 1, 1997: Final copy due to Design & Test Managing Editor Spring 1998: Publication in IEEE Design and Test of Computers Submission requirements: Send six (6) copies of the manuscript, in English, to the guest editor. Manuscripts are not to exceed 35 double-spaced pages, inclusive of figures and tables, in A4 or 8.5 by 11 inches. Type size must be at least 12 point. Each copy of the manuscript must contain a cover page with author contact information (name, postal address, telephone number, and e-mail address) and a 100-word abstract. Manuscripts must be cleared for publication. Accepted manuscripts will be edited for technical content, structure, style, clarity, and grammar. Detailed information for authors can be found at the Computer Society D&T website at http://www.computer.org/pubs/d&t/d&t.htm or in the Spring 1996 issue of Design & Test.Article: 5390
Nicholas C. Weaver wrote: > Back to comp.arch: Anyone have a circuit for GENERATING > random bits? Preferably something that can be built into a VLSI chip? > See my post or those from John Smith from yesterday regarding random number generators in Xilinx -Ray Andraka, P.E. Chairman, the Andraka Consulting Group 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://www.ids.net/~randrakaArticle: 5391
Unfortunately, the hack of Steve Schossow won't work for Max7000s chips... (no .ttf file). So if you don't want to pay 200$ for a piece of wire, you can make your own ByteB****** with the following schematic: Parallel port JTAG connector SubD25: HC244 buffer: HE10 10pins: 2 ______________4_|\_16_____________________1 |/ 3_______________2_|\_18_____________________5 |/ 8_______________6_|\_14_____________________9 |/ 7_______________8_|\_12_ |/ | 10______________________| 11______________9_/|_11____________________3 \| 13______________7_/|_13____________________7 \| 18-25___Gnd Vcc____4 Gnd____2,10 Gnd______1,10,19 of HC244 Vcc______20 of HC244 Enjoy !Article: 5392
Juerg Haefliger wrote: > > hi, > does anybody know how the core of a XC4025 is connected to the pads > of the package (HQ240). > what i have to know is how the core-geometry looks like and which logic- > cell is connected to which pad. > is there a chance to get these informations from XACTstep? i played around > a short time with the tool, so don't expect i really understand the software! > thanks I've run into the same thing several times. Xilinx doesn't have the pad to column/row info published anywhere for any of the 2K,3K or 4K parts. You can get the info by either going into the LCA editor or the floorplanner. For the floorplanner, you will need a trivial design. You also will need to get the package files for the 4025 from Xilinx, since they weren't included with the XACT6 release. -Ray Andraka, P.E. Chairman, the Andraka Consulting Group 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://www.ids.net/~randrakaArticle: 5393
Duncan Davis wrote: > > I hoping someone can help me. > > I am convinced I read a posting from someone regarding the topic of > simulating at the gate level on Quicksim, having syntheised from Galileo. > I seem to remember it involved EDIF. I think it appeared about a month or > two ago. You're right. I've not tried it personally because we do not have the EDIF reader option on our quicksim; however, we were able to simulate at gate level with QuickHDL and vital libraries. Regards, Kayvon IraniArticle: 5394
First of all, I'd like to thank all those that replied to my previous question, I really appreciate it. To the problem: We are using MG to synthesize VHDL here at the Uni, with Xilinx-supplied software to actually route and upload the designs to the chips. The problem is that if the no-flatten option is used in Autologic the component names associated with x-blocks are in capital letters. That wouldn't be a problem if it weren't for a bug in the xilinx software that convers everything to lower-case. This makes the xilinx software unable to find the x-blocks components. If the flatten option is used then the x-blocks seem to be splitted into gates and the xilinx software works - but does it actually re-create the x-blocks from the collection of gates, or does it leave the gates as they were, wasting a lot of valuable space?? I tried a number of work-arounds, including: A)renaming the components and correcting the broken references created by the renaming B)making an xnf netlist and skipping some of the stages that you normally go through to place and route a design but neither worked. As there is no available Xilinx documentation apart from the crude online help, I am at a loss. Any ideas? Thanks in advance, -- Christos Dimitrakakis --------------------- mailto:mbge4cd1@fs4.eng.man.ac.uk mailto:mbge4cd1@afs.mcc.ac.uk http://www.man.ac.uk/~mbge4cd1Article: 5395
I am currently designing 106 Bytes(848 bits)FIFO in XILINX 4000E. Is there any other more efficient way rather than using RAM cell macros? Thank you! -------------------==== Posted via Deja News ====----------------------- http://www.dejanews.com/ Search, Read, Post to UsenetArticle: 5396
Hi, I've just seen a LOG/iC2 design system supporting all CPLDs and XILINX / ACTEL FPGAs offered on the ISDATA homepage for 4.999 DM including simulation. Any experiences with LOG/iC2? ThanksArticle: 5397
I work for a Lucent Disti, just so you know. On Thu, 06 Feb 1997 05:46:45 GMT, kenk@teleport.com (Ken Krolikoski) wrote: >Most of Lucents sales comes from captive AT&T accounts. ORCA is not a >major player in the general marketplace. So shipping millions of dollars of product in volume to large accounts that are price/performance sensitive doesn't make you a major player in the "general marketplace"? I believe Xilinx likes to count AT&T as one of its customers anyway, so I don't think they could exactly be described as 'captive'. When a company's programmable logic offering is FPGA's, and high density ones at that (3000 series excluded), I would think that a market turnover of around $100M would be pretty good going. I'm sure that if Xilinx had a 'better deal' then some of the major companies using ORCA would switch. But they haven't.BTW What would you suggest they replace a 2T40A-4 with? or even a 2C40 from way back in 1995? >The new routing sw that xilinx is releasing now for workstations and >in april for windows NT will significantly improve your route times. Aah. Do you mean the NeoCAD engine? I recall Lucent shipped their own version_way_before Xilinx after the buy-out. Lucent's Foundry 9.0 has been running on NT for about a year now. XACT doesn't (5.x in a DOS box?) Good speed too. Well usable for large parts so long as you have enough memory etc. >the new 4Kex and 4kxl devices have double the routing resources of the >4ke. Can't get a 4010EX though, can you? and Lois Cartier says that ORCA has more resource than 4000E. (XBRF005.pdf). How gracious! >the next three parts available from xilinx will be the 4062(now), >4085(2q97) and 40125(3q97). I found the press release on their web >site. (www.xilinx.com) I could blether on about the Lucent 3C series and the 200K+ gate parts etc. but I don't consider this kind of posturing relevant to designers doing work today. Actel say 500K gates, Xilinx says a million. Come on guys, it's boring. <snip of Xilinx vapour-ware press release> Xilinx, Altera, Lucent, Actel (apologies for missed ones) all have user SRAM implementations of one sort or another. Some fit well in some areas, some don't. Please don't slag off your competitors' business models. Customers will spot it for the smoke screen it really is. Lucent is a $20B+ dollar company so I guess they must be getting something right: #1 Worldwide in Digital Cellular DSP's #1 Worldwide in Standard Cell ASIC #1 Worldwide in videoconferencing IC's #1 Worldwide in Telecom Power Systems Compare FPGA's for specific applications. Sometimes Xilinx will have a winning feature, other times Lucent, Altera, or anyone else. Just my two cents StuartArticle: 5398
On Thu, 6 Feb 1997 09:06:59 -0000, eteam@aracnet.com (bob elkind) wrote: >I don't see any reason why Lucent is any less of an FPGA >vendor than Xilinx or Altera, for any size project or >company. Agreed >On the subject of "whose xxx is bigger": If you're going >to use *lots* of gates (list of huge Xilinx FPGAs deleted), >and use them at a high enough bandwidth, then power can become >a limiting factor before you are bound by gates, routing >channels, etc. Last time I looked, the Lucent 2cA series >had a speed/power advantage of X 4KE, but that may have >changed. Both the L and the X company have made much-needed >improvements in the thermal characteristics of their packages, >but last time I looked the Lucent 2C40 with its cavity-down >packaging had a definite advantage in the field. Take a look at the Lucent OR2T40A. It's a 0.35 micron device which runs a 3.3V core (5V tolerant I/O's) It will run the same speed as the five volt parts, but gives a very satisfactory reduction in power consumption. It is also in full production. Correct me if I'm wrong, but Lucent appear to be the only FPGA vendor with same speed 5V and pure 3.3V parts (available today). StuartArticle: 5399
In article <01bc17f5$32984e80$745c34c1@ensbr.bretagne.ens-cachan.fr>, "bergeon" <bergeon@bretagne.ens-cachan.fr> wrote: > Can someone tell me where I can find a structure of a system which can do a > 1/T > (T is in BCD) with registers. > Yves, there are several methods, and your choice depends on: € required accuracy € available time for converson € available resorces, registers, counters etc. In other words: acceptable cost € acceptable design effort. You can kludge together something that is slow and crude and wasteful, or you can dig deeper and design something far more elegant. I would be tempted fo first convert your decimal numbers to binary, then get successive reciprocal bits by successive approximation multiplication, one bit per clock or per two clocks. But you can also emulate the human longhand method: keep it in BCD, normalize, subtract from 1.000000 as often as you can ( that gives you the first digit ) then shift one decimal position, and repeat... Needless to say, all of this can easily be done in an FPGA at >20 MHz clock rate. Peter Alfke, Xilinx Applications
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