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
Hi, Here is quite simple problem I didn't find solution yet. How to implement bidirectional tri-state buses inside FPGA design? I use XACT 6.0/ Viewlogic Schematic Interface. Any help or advises are wellcome. Thank You. Alex.Article: 3776
What is the efficiency of ASIC design versus FPGA design.I know that lots of silicon area is used for switches and connection lines in FPGA but is there any number that can show the efficiency based on silicon area ?eg. 10 or 15 . thanks HossainArticle: 3777
Alex Krynew <jmacow@dol.ru> wrote: >Hi, >Here is quite simple problem I didn't find solution yet. >How to implement bidirectional tri-state buses inside FPGA design? >I use XACT 6.0/ Viewlogic Schematic Interface. >Any help or advises are wellcome. >Thank You. >Alex. Just draw the busses as you would for a normal tri-state bus and use TBUF modules to gate the appropriate data onto the bus. MarkArticle: 3778
Lawrence Butcher wrote: > > AT&T flash serial proms observed to fail to configure reliably at powerup. > > I have had the exact same experience with AT&T serial proms. They work fine > if the power supply comes up fast, and don't configure if it comes up slow. > I am curious as to whether people with ATT SPROM config problems are using the active low RESET option, (RESET*), with RESET* driven by the lead FPGA INIT* output. According to the Xilinx databook ('94, p2-233) this removes the issue of the relative timing of FPGA/SPROM power-on reset and is generally recommended. Please inform me if the problem exists with this connection and I will be appropriately cautious about ATT SPROMS in future. Thanks - Tom Burgess tburgess@drao.nrc.caArticle: 3779
In article peter@xilinx.com says... > > is there any number > > that can show the efficiency based on silicon area ?eg. 10 or 15 . > > > Why do you care? He's learning, go easy on him! > Usually, FPGA chips are substantially bigger than equivalent capacity > ASICs, but there are several factors that help FPGAs: > FPGAs are mass-produced and move as fast as possible to a new, more > efficient process technology= smaller chip. ASICs usually stay with their > original design for cost reasons. FPGAs *AND* ASICS (both) move as fast as *practical* to a newer, whizzy process. You can't honestly say that 4000EX FPGAs are built on a process equivalent to Pentium Pros! I don't see how FPGAs have any inherent "edge" in migrating up the processing food chain. I *would* believe that standard cells can be had in faster/whizzier processes *sooner* than FPGAs. And sure, once the ASIC is designed in process X, it tends to stay in process X for its lifetime, and sooner or later FPGAs overtake the process X in which the ASIC was originally cast. But keep in mind that the ASIC (std cell or gate array) has a head start. Case in point: Xilinx' own HardWire product/service... lower cost per gate, better AC performance, lower power... -- yup, even with the exact same process, you can get better results in these three areas just by going semi-custom. > FPGAs are mass produced and travel down the manufacturing learning curve > faster than ASICs. Interesting thought. I don't believe it. Without getting too bogged down in the definition of "ASIC", most of the big guys (e.g. Moto, TI, Intel) do their humongous work in the same ASIC technology and process steps as the rest of us. They can't afford to do hand layout on every little via and metal run and transistor. They're doing cell-based design. > Finally, low-complexity but high pin-count devices might be "pad-limited", > which means their die size is dictated by the number of IO, not the logic. > In that case there may be no size difference between the two. > It's not a matter of square mm, it's a matter of cost. An that has far > more variables than just chip size. Agreed. > Peter Alfke, Xilinx Applications. Peter, we've known each other for a long time, and I respect your opinions greatly. Your comments just yanked my chain, on this topic. I've done lots of custom, semi-custom, and FPGAs. Having just told Peter to be "not quite so proud" of his FPGAs, most of the design work I do these days *is* in FPGAs (not ASICs). ASICs are still way ahead of FPGAs in some technical and cost areas, but nothing beats the low risk and engineering cost of FPGAs. I love you Peter, but you can't have my Bud Lite. Bob ************************************************************************** Bob Elkind mailto:eteam@aracnet.com CIS:72022,21 7118 SW Lee Road part-time fax number:503.357.9001 Gaston, OR 97119 cell:503.709.1985 home:503.359.4903 ******** Video processing, R&D, ASIC, FPGA design consulting *************Article: 3780
In article <DvC0EJ.5vv@news.uwindsor.ca>, hajimow@engn.uwindsor.ca wrote: > is there any number > that can show the efficiency based on silicon area ?eg. 10 or 15 . > Why do you care? Usually, FPGA chips are substantially bigger than equivalent capacity ASICs, but there are several factors that help FPGAs: FPGAs are mass-produced and move as fast as possible to a new, more efficient process technology= smaller chip. ASICs usually stay with their original design for cost reasons. FPGAs are mass produced and travel down the manufacturing learning curve faster than ASICs. Finally, low-complexity but high pin-count devices might be "pad-limited", which means their die size is dictated by the number of IO, not the logic. In that case there may be no size difference between the two. It's not a matter of square mm, it's a matter of cost. An that has far more variables than just chip size. Peter Alfke, Xilinx Applications.Article: 3781
In article <4tav6m$4lb@transfer.stratus.com>, Matt Cross <mcross@sw.stratus.com> writes >I purchased a copy of Lattice's ISP starter kit, and I am just toying with >programming teh 2032 part. I got a PLCC adapter that I've plugged into >my breadboard, and have wired up the programming cable. The programming >software ('Daisy Chain Download') that came with the kit recognizes the >chip, and says it succesfully programmed it, but that it cannot verify >it because the security fuse is set. But, when I generated the fuse map >(.JED file) I told it to not set the security fuse. My design does not >appear to be in the device from my simple tests. > >Any ideas? Am I just destroying the signals by running it on a breadboard? > >(In case you coudln't tell, I'm not a real hardware guy, I mostly do software. > I find digital logic fascinating, though, and I wanted to try it out...) > > -Matt Check out your power supply voltages. I had a similar problem verifying 3256 devices on low (or was it high) margins. That said the devices do function correctly, once programmed, over the specified supply voltage range. T.H.Article: 3782
Hello! I am currently working on a multi-FPGA project with XC4000 FPGAs. I have to face the problem of partition a design among several FPGAs. My questions are: - Which is the current state of the art in this problem? - Is there any comercial software for it? Does Xilinx plan to provide one? - Is there any free partitioner available? I will post an abstract of the answers that I get. Thank you. Regards. J. A. Herrera Camacho.Article: 3783
Hi, can someone tell me how I can assign LOCs to a 16 bit IOPAD in the Xilinx schematic entry ? I suppose I have to make a file with the LOCs. Or is the only solution using 16 times 1 IOPAD ? :( Thanks in advance ! Kim -- ============================================================================== ************************************ * Hofmans Kim * * * * tw38966@vub.ac.be * * khofmans@info.vub.ac.be * * * * Brouwerijstraat 62 * * 1630 Linkebeek * * Belgium * * * * 32-2-3771012 * * * ************************************Article: 3784
HighTech <hightech@mv.mv.com> wrote: > >Are we going to start posting job listings to these newsgroups? > Please don't. Job postings are for the job newsgroups. Once the headhunters start massively job posting in the technical newsgroups, it effectively kills the technical discussions in the newsgroups because it turns off too many engineers. It's all a matter of signal-to-noise; what does the fact that 8 different headhunters posting and reposting "HOT JOBS AT SGI!!!" have to do with my getting a solution to the EDA bug I'm currently fighting? Nothing. - John Cooley Part Time EDA Consumer Advocate Full Time ASIC, FPGA & EDA Design Consultant =========================================================================== Trapped trying to figure out a Synopsys bug? Want to hear how 4599 other users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, _] [_ Verilog, VHDL and numerous Design Methodologies. Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion."Article: 3785
> What is the efficiency of ASIC design versus FPGA design.I know that lots of >silicon area is used for switches and connection lines in FPGA but is there any number >that can show the efficiency based on silicon area ?eg. 10 or 15 . In general, when moving from an FPGA to an ASIC, you end up using a lot fewer gates than you expected. I have recently ported a 60%-full XC3090 design to an ASIC, and it uses only about 2000 gates. I would think the ASIC chip is much smaller than that of a 3090, but does this matter? You also gain a lot in dynamic power consumption. This is largely because you don't have the global clock nets, and you can gate clocks much more freely than one is able to in an FPGA. FPGA design tends to make use of the global clock nets, and you use the clock enables on individual D-types to decide which one gets clocked. In my above design, dynamic Icc dropped from 15-20 mA (and that was a XC3042, implementing a subset of the 3090 design) to about 1-2mA. However, this amount of power saving may not come if you just go to a straight netlist port like e.g. Orbit offer. But ASIC NREs and minimum quantities are much more demanding than they were in the ASIC heyday of some years ago. The only exceptions are the several firms around who can take your Xilinx netlist, but I have never used them for other reasons. Peter.Article: 3786
I think you'll have to assign 16 IOPADs - if you find an easier way, please let me know. -- Regards AndyGArticle: 3787
ANNOUNCEMENT: Immediate Release Model of the Month ==================== This month's FREE model is: Image Processing Cache Register Array You can find it at http://www.doulos.co.uk. You can also access previous FREE Models and Tips of the Month from the same site. ____________________________________________________________________ Also *** NEW *** for this month are: TIP of the MONTH: Encapsulation in VHDL Excerpts from the VHDL Golden Reference Guide Additions to our 'Hardware Engineers Guide to VHDL' series ____________________________________________________________________ Also visit the Doulos site 'THE WINNING EDGE' for: VHDL Quick Reference Card VHDL 93 Update Reference Card _____________________________________________________________________ Also details how to get a FREE online VHDL tutorial. DOULOS Church Hatch Tel: +44 1425 471 223 22 Market Place Fax: +44 1425 471 573 Ringwood BH24 1AW Email: info@doulos.co.uk UK _____________________________________________________________________Article: 3788
ANNOUNCEMENT: Immediate Release Tip of the Month ================== This month's FREE 'Tip' is: Encapsulation in VHDL You can find it at http://www.doulos.co.uk. You can also access previous FREE Models and Tips of the Month from the same site. ____________________________________________________________________ Also *** NEW *** for this month are: FREE MODEL of the MONTH: Image Processing Cache Register Array Excerpts from the VHDL Golden Reference Guide Additions to our 'Hardware Engineers Guide to VHDL' series ____________________________________________________________________ Also visit the Doulos site for: VHDL Quick Reference Card VHDL 93 Update Reference Card _____________________________________________________________________ Also details of how to get a FREE online VHDL tutorial. DOULOS Church Hatch Tel: +44 1425 471 223 22 Market Place Fax: +44 1425 471 573 Ringwood BH24 1AW Email: info@doulos.co.uk UK URL: http://www.doulos.co.uk _____________________________________________________________________Article: 3789
Hi all I'm looking for papers deal with the use of reconfigurable hardware in arithmetic circuit. In particular, the use of SRAM-based FPGA for reconfigurable arithmetic circuit is intersting for me; such as SRT based divide/square root circuit. Please, anyone has reference reply me. Pasquale Corsonello Dept. of Electronic Computer Science and System University of Calabria e-mail:pascor@ccusc1.unical.it corsone@nwdeis1.unical.itArticle: 3790
Reference Number: 1000417 Title: FPGA Design Engineer, Avionics Location: NH Type: Permanent Salary: Open, relative to experience Description: Design and develop electronic circuits that will be primarily digital with some custom components including FPGAs and/or ASICs. Designs may involve processors, data bus interfaces, combinatorial functions, and aircraft or spacecraft interfaces. The design process will include simulations using state of the art tools. Experience: Experience with simulation tools, FPGAs and/or ASICs is a plus. BS degree in Electronic Engineering from an accredited college/university or its equivalent. Must be a US citizen. If interested e-mail a TEXT file format of your resume to: khawes@cadstar.com or mail/FAX to: Kevin T. Hawes CADstar International 130 Middlesex Road, PO Box 704 Tyngsboro, MA 01879 Phone: (508)649-6828 FAX: (508)649-7856 e-mail: khawes@cadstar.comArticle: 3791
Reference Number: 1000420 Title: FPGA Computer Architecture Design Engineer Location: NH Type: Permanent Salary: Open, relative to experience Description: Demonstrated capability as a technical lead in advanced computer architecture, complex digital module and FPGA design, VHDL simulation, VHDL synthesis and working with emerging CAE technology a must. Processor design experience must include evaluating processing functions vs. architecture schemes for various signal and image processing applications. Experience: Experience in the innovative use of reconfigurable computing for image processing, and signal processing is a plus. The new ECAE technology to be used for this effort must become refined for broader use. BS in Electronic Engineering from an accredited college/university or its equivalent and 8 years work experience; or a MS and 7 years experience; or a Ph.D. with 4 years experience. If interested e-mail a TEXT file format of your resume to: khawes@cadstar.com or mail/FAX to: Kevin T. Hawes CADstar International 130 Middlesex Road, PO Box 704 Tyngsboro, MA 01879 Phone: (508)649-6828 FAX: (508)649-7856 e-mail: khawes@cadstar.comArticle: 3792
>Reference Number: 1000417 >Title: FPGA Design Engineer, Avionics >Location: NH >Type: Permanent >Salary: Open, relative to experience > >Description: > >Design and develop electronic circuits that will be primarily digital >with some custom components including FPGAs and/or ASICs. Designs may >involve processors, data bus interfaces, combinatorial functions, and >aircraft or spacecraft interfaces. The design process will include >simulations using state of the art tools. This one's easy; it's Lockheed-Sanders up in New Hampshire. I even know the engineers up there! If anyone's is looking for a job there, blow off this headhunter and send your resumes to me. I'll forward them to the engineers there and they can get the referal bonus! (Or if you want to send your resume up to their HR dept., call (603) 555-1212 and ask for "Lockheed-Sanders".) Why can't these headhunters post in the job newsgroups? - John Cooley Part Time EDA Consumer Advocate Full Time ASIC, FPGA & EDA Design Consultant =========================================================================== Trapped trying to figure out a Synopsys bug? Want to hear how 4599 other users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, _] [_ Verilog, VHDL and numerous Design Methodologies. Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion."Article: 3793
Rafiki Kim Hofmans (tw38966@vub.ac.be) wrote: : Hi, : : can someone tell me how I can assign LOCs to a 16 bit IOPAD in the Xilinx : schematic entry ? : I suppose I have to make a file with the LOCs. : : Or is the only solution using 16 times 1 IOPAD ? :( : : Thanks in advance ! : : Kim : -- u can also specify in a constranits file ( .cst ) for syntax look into the XACT manuals. -RaghuArticle: 3794
I second John's opinion. I can submit the resume for you and let an employee get the referral. Maybe the headhunters will get the message and stop posting in this newsgroup. In article <DvHwMG.AJK@world.std.com>, jcooley@world.std.com (John Cooley) writes: |> >Reference Number: 1000417 |> >Title: FPGA Design Engineer, Avionics |> >Location: NH |> >Type: Permanent |> >Salary: Open, relative to experience |> > |> >Description: |> > |> >Design and develop electronic circuits that will be primarily digital |> >with some custom components including FPGAs and/or ASICs. Designs may |> >involve processors, data bus interfaces, combinatorial functions, and |> >aircraft or spacecraft interfaces. The design process will include |> >simulations using state of the art tools. |> |> |> This one's easy; it's Lockheed-Sanders up in New Hampshire. I even |> know the engineers up there! If anyone's is looking for a job |> there, blow off this headhunter and send your resumes to me. I'll |> forward them to the engineers there and they can get the referal |> bonus! (Or if you want to send your resume up to their HR dept., |> call (603) 555-1212 and ask for "Lockheed-Sanders".) |> |> Why can't these headhunters post in the job newsgroups? |> |> - John Cooley |> Part Time EDA Consumer Advocate |> Full Time ASIC, FPGA & EDA Design Consultant |> |> =========================================================================== |> Trapped trying to figure out a Synopsys bug? Want to hear how 4599 other |> users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! |> |> !!! "It's not a BUG, jcooley@world.std.com |> /o o\ / it's a FEATURE!" (508) 429-4357 |> ( > ) |> \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, |> _] [_ Verilog, VHDL and numerous Design Methodologies. |> |> Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 |> Legal Disclaimer: "As always, anything said here is only opinion." -- Mark Christensen ascom-Nexion email: markc@nexen.com 289 Great Road ph: (508) 266-2315 Acton, MA 01720Article: 3795
In article <qKWR+DAAxi+xEwJ4@tennyson.demon.co.uk>, Nigel Burrows <nigel@tennyson.demon.co.uk> writes: >The first circuit I built didn't include the .01uF capacitor between >!ispEN and ground which caused the verification to fail. > >I've used the 2032 with a home made PLCC to DIL adaptor on a breadboard >OK. Thanks everybody for your help. I had a capactitor in the right place, but it was too small (silly software engineers who can't read capactitors :-) ). Once I put the right cap in there it worked great. -MattArticle: 3796
I'm sure I'm not the only one battling these problems, so I thought I'd put out a general call for help: I'm using Neocad (Xilinx) Foundry 7.0 software to place & route a synthesized design into 3195A and 40xx series parts. Neocad seems to be the weak link in the flow. Sometimes I have a design that works fine until I modify some other, unrelated part of the circuit. When the design is re-routed, it will cease to work. These are often fairly simple, synchronous designs. My usual strategy is to insert clock buffers on critical clock lines, use FREQUENCY NET and FREQUENCY constraints in the device.prf files, and make slight modifications to the top level HDL so the synthesized schematic will be slightly different (causing Neocad to route a slightly different design, hopefully with better results). This approach seems crude at best. I've thought about trying to use floorplanning, but I haven't attempted it yet for two reasons: 1. Synthesized schematics are usually very difficult to decipher. 2. When I can make sense of the signals (usually on smaller designs), the placement looks like it makes sense (all the flip-flops in a counter chain are located in the same area, for example). I do have a number of constraints on my system. Many of my I/O pins are constrained, which ends up limiting me to about 50-60% utilization. Anyone have any suggestions for how I can consistently get reliable timing results? Thanks, Jeff Hunsinger ---------------------------------------------------------------------- Jeff Hunsinger jeffh@oakhill-csic.sps.mot.comArticle: 3797
> > > is there any number > > > that can show the efficiency based on silicon area ?eg. 10 or 15 . > > > > > Why do you care? > > He's learning, go easy on him! > > > Usually, FPGA chips are substantially bigger than equivalent capacity ^^^^^^^^ > > ASICs, but there are several factors that help FPGAs: Thanks alot for the responses to my question.Yes I am learning .umm.. it is interesting, You say "Usually" , and my impression is it means "not always".So another question rises.In what designs FPGA chips are smaller than ASICS? Is it for the designs with higher number of inputs? HossainArticle: 3798
I have only once ever seen a case where a design does not work, and a straight recompilation "fixes" it, and that was recently with XACT6 place/route software. But I have more often come across a nastier problem, common to all FPGAs: say you have a 16-bit shift register. Obviously, for a SR to work, the clock skew stage-to-stage has to be less than the clock-to-Q propagation delay. IOW, you ideally want to clock all the stages together. The complex routing in an FPGA makes this hard to achieve. I generally put a SC=1 attribute on any such clock nets (requesting a skew of 1ns or less!) but this cannot be achieved when there are more than a few stages. The "proper" way to do this is to use one of the global clock nets to clock all the D-types, and use clock enables to decide which actually get clocked. Of course, in my above SR example, they all get clocked together anyway, but you may still need to gate the clocks, otherwise the SR will be shifting all the time! Unless I am oversimplifying your problem, it is possible that this is what you are seeing. Even if you do everything "synchronous", this is where you get caught. Incidentally, the problem with porting such an FPGA design to an ASIC is that you may get an ASIC with a lot higher power consumption than would be the case otherwise. This is one area where prototyping ASICs in FPGAs (a near-universal practice nowadays in ASIC design) is difficult. Peter.Article: 3799
In article <Pine.SUN.3.91.960729074337.16136A-100000@Bundy>, Rainer Scharnow <amigo@bintec.de> wrote: Of course, no > one describes the limitations of the products 8-). > Did it ever occur to you that the only thing the parameter values in the data sheet describearet device limitation? Propagation delays describe that the output doesnt change instantaneously Set-up and hold times tell you that the device may malfunction if you change data during that timing window Vcc describes that operation is only guaranteed between 4.75 and 5.25 V Temperature ranges describe that parameters are only guaranteed for a limited range of temperatures. Max storage temperature, max soldering temperature, max leakage current, and the list goes on... The data sheet parameters do not describe how good the devices are, they describe how "bad" they might be, in the extreme case. Our industry gives you hundreds ( thousands ? ) of pages of product limitations. I suppose you have never looked at it this way. :-) 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