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
I've been discussing this topic via email with a person who does not wish to be indentified, but who has provided the following comment with permission for me to post it: Xilinx hasn't changed the structure of their core tools for many years as you can tell by watching what the programs do, especially if you are familiar with PPR's intermediate files from the hard macro days. Porting this struct to the Windows DOS box seems to us advanced users a bad technical decision as far as robustness and performance, but given the architecture of their existing tools it makes some sense that perhaps they couldn't have done anything different to ship a Windows product in the near term. My solution is just not to use the Windows interface. XACT 5.2 under DOS works just fine. It will be interesting to see how the new Neocad based tools work. Could it be that Xilinx bought Neocad strictly because they know how to write software better, as opposed to how to write better software? -- Tom Liehe - tliehe@rmi.net - toml@metrum.comArticle: 2726
Try Needham Electronics (usually advertise in the back of EE Times)... they have something for under $200. Lattice has an In Circuit Programmable kit for 'round $90, that includes there spin of a 22V10. Doug Shade rxjf20@email.sps.mot.comArticle: 2727
In article <mark.stephens-2901961009190001@mstephens.gsfc.nasa.gov>, mark.stephens@gsfc.nasa.gov says... >Do I need to get another programmer or can one home brew for a specific chip? Elektor magazine did a project article for a GAL programmer(other magazines may have also). I built it, and they had a later article with an add-on for the 22V10 GAL. Elektor is no longer printed in this country, but Old Colony Sound Labs carries back issues, and PC boards, and the software. Old Colony's phone number is (603)924-6371. The issue with the original article was May 1992. The PC board was $19 and the software was also $19. I forget how much the 22V10 upgrade was. There were no really exotic parts in the original design. The upgrade has an A/D converter that's about $10 from Farnell Components. Email me if you have any questions. bblick@execpc.comArticle: 2728
In article <hersman.821993240@aplcomm.jhuapl.edu>, hersman@aplcomm.jhuapl.eduļ says... > >Does anyone know of a way to program an Actel FPGA in the circuit? > >I am aware that even if it is possible, it's not straight forward. > >Chris Hersman If Actels are specifically required check out the Lattice ISP family for extremely simple in ciruit programing.Article: 2729
In article <4eji8k$n4i@newsbf02.news.aol.com>, Jim Drew writes: > Dump the GALs, they are absolutely worthless and expensive. PEELs are only moderately more capable than GALs. PEEL Arrays are nicer but vertually unobtainable from hobbyist sources. JDR and Jameco price GAL22V10-25 and PEEL22V10-25's exactly the same: $4.95 ICT's software is free but it is also buggy and PEEL's are not well suported by third parties. That said, PEEL's are usefull and really don't even have to choose. PEEL's are JDEC compatible with GAL's. Most low cost device programmers will handle both. So get you PEELs. Use them as PEEL's or GAL's whichever way suites. I use a EE Tools Allmax with Intel/Altera's PLDshell and/or ICT's PLACE. I prefer PLDshell but I need PLACE to take advantage of the extra PEEL features (like bi-directional I/O) > PEELs provide a wide range of different Macro Cells, making them much more > versatile than GALs. They are about .75 cents for an 18CV8 and about > $1.75 for a 22CV10 (although I pay less than this for 1000+ quantities). Where are you finding 22V10's for $1.75? Are you still talking large quantities? ---- "Not many fishes, in the sea. Not many fishes, just Londo and me" Remember the home hobbyist computer: Born 1975, died April 29, 1994Article: 2730
Sigh... Most of 'em want a committment, not just for hobby use. If you find a couple of them that will put up with fooling around on weekends, would you post addresses? <gdr> Jim Jim Weir VP Engineering | You bet your sweet patootie I speak RST Engineering | for the company. I OWN the cotton- Grass Valley CA 95945 | pickin' company. voice/fax 916/272-1432 | AR Adv. WB6BHI Cessna 182A N73CQ rst-engr@oro.net | Comm'l Pilot & CFI Airplane/Glider A&P MechanicArticle: 2731
You may take a look at http://www.aaeon.com. There is a univeral programmer good for EPROM, PEEL, GAL, FPGA etc................. Several pages of the device list that it can program. Its price is very very low comparing to its performance. I got one in the lab and it works perfect.Article: 2732
In article <4e8ed3$boc@agate.berkeley.edu>, Steve Pope <spp@bob.EECS.Berkeley.EDU> wrote: >verschue@eb.ele.tue.nl (Ad Verschueren) replise to my post, > >>>This is because whenever a design shop designs a chip for one >>>customer, its elements then become potential "off the shelf cores" for >>>the next customer. Problem is, the actual level of circuit >>>re-use is always lower than one hopes. Rule of thumb -- if you >>>can get by with less than half of a design being newly-designed >>>circuitry, you're doing really well. > >| In my experience, the amount of curcuit re-use increases >| tremendously when you are allowed to make (small) modifications >| to a previous design - and I don't mean setting a few parameters >| differently here... We do have large groups of students working >| and learning here using our own tools, and they are copying and >| pasting all over the place. Everything they made over the last >| couple of years is placed in a central depository, we do not >| allow anything in there which is not thoroughly documented (both >| on paper and within the designs themselves). Not only complete >| designs are open for re-use, but also any sub-part of a design >| can be copied in separation. > >I agree that you are describing techniques that serve to increase >re-use. But I'll stand by my estimate that anything better than >50% re-use -- for REQUIREMENTS DRIVEN projects -- is really >pretty good, even with the techniques you mention. > I agree, 50% re-use is very good, but we should try to get it even much higher - for any kind of project... >Of course, for RESOURCE DRIVEN projects, you will get more re-use >more readily. (e.g., "What cores do I already have, and what >can I design with them" -- i.e. a typical university situation.) > I do give a course in industry on system synthesis. One of the main motto's is 'do not design anything new unless it is really necessary'. The method derives a system hardware architecture from an existing functional descrip- tion by moving pieces of functionality around, combining them in modules for which 'preliminary implementation choices' must be made in the process. This way, the requirements in the functional description are matched against the resources as much as possible. I always ask the audience to generate a list of implementations (for data processing/storage *and* communication) they know they can use - it is surprising to see how quickly a (sometimes huge) list they can produce. Even though I state in the course that a method like this should be supported by appropriate design entry/modification/analysis/simulation tools (which are, sadly, not yet available...), feedback I get indicates the philosophy works. The main problem is not only that a (if possible, large) number of existing designs must be available in an easy to access and parametrise/ modify form (the 'resources') - at least as large a problem is the fact that it becomes difficult to choose the 'optimal' implementation given an amount of functionality to implement. This matching process gets more difficult when the list of available implementations grows, especially if all kinds of modifications/extensions are allowed. If the list of resources is treated like a 'company asset', open for everybody within the company for re-use (like it should be...), you can get the situation that not all the designers know everything about all the available implementations within the company (I do know some horror stories about people working in the same building in a large company 'nearby' who were designing almost the same stuff without knowing from eachother). They should be able to decide, though, that a specific implementation is usable for their problem. This decision should be easy to take, too - if you have a list of a few hundred implementation methods, you don't wanna spend an hour on each of them figuring out if they are suitable. Our proposed solution to this problem is to abstract the requirements and implementation capabilities into a number of key figures and match on those figures. We intend to do the abstraction process automatically with analysis tools and have already shown that this is -in principle- possible. >Steve Ad -- --(dr.ir.) Ad (A.C.) Verschueren-----------------------VERSCHUE@EB.ELE.TUE.NL-- Eindhoven University of Technology Digital Information Processing Systems Smail: Room EH 10.26 ---- P.O. Box 513 ---- 5600 MB Eindhoven, Netherlands Voice: +31-40-2473404 FAX: +31-40-2448375 [corner for rent, apply within]Article: 2733
In a previous article mark.stephens@gsfc.nasa.gov (Mark Stephens) writes: > <I'd really, really like to get rid of the NOR gate glue logic and replace >it with some species of GAL. I have PLDasm and it's manual and can now <create the JDEC files. Unfortunately, my PROM programmer is just that... >PROMs only. < >Do I need to get another programmer or can one home brew for a specific chip? Lattice makes a 22V10 equivalent that is in-circuit reprogrammable via serial down load cable. No special voltage is required, and you can make the down load cable yourself and hook it up to the printer port of your PC.Article: 2734
Thanks for all the great responces! I've been avoiding this but looks like the "best" path is to get an ol' 286/386 box and a card based programmer. If the Mac only had a parallel port...sigh. mark -- mark stephens "In constraint, NASA GSFC Code 521 is freedom" Greenbelt, MD (301) 286-4269 mark.stephens@gsfc.nasa.govArticle: 2735
Tom Liehe (tliehe@rainbow.rmii.com) wrote: : I've been discussing this topic via email with a person who does not : wish to be indentified, but who has provided the following comment : with permission for me to post it: : Xilinx hasn't changed the structure of their core tools for many years : as you can tell by watching what the programs do, especially if you : are familiar with PPR's intermediate files from the hard macro : days. Porting this struct to the Windows DOS box seems to us : advanced users a bad technical decision as far as robustness and I agree 100% : performance, but given the architecture of their existing tools it makes : some sense that perhaps they couldn't have done anything different to : ship a Windows product in the near term. My solution is just not to use : the Windows interface. XACT 5.2 under DOS works just fine. I am doing exactly this. : It will be interesting to see how the new Neocad based tools : work. Could it be that Xilinx bought Neocad strictly because they : know how to write software better, as opposed to how to write better : software? : -- : Tom Liehe - tliehe@rmi.net - toml@metrum.comArticle: 2736
I wonder if we'll see more "object oriented" high-level design languages. I don't see any problem with automating design of >10M chips, given the rise of computational power. But the high-level object descriptions might be re-used. -ThomasArticle: 2737
MMartin950 (mmartin950@aol.com) wrote: > You may take a look at http://www.aaeon.com. There is a univeral > programmer good for EPROM, PEEL, GAL, FPGA etc................. Several > pages of the device list that it can program. Its price is very very low > comparing to its performance. I got one in the lab and it works perfect. If you actually saw my response to this before it died (because I killed my article) please ignore my complaints about their home page. It works fine, as long as you use Netscape. You can't get product listings if you use mosaic. RogerArticle: 2738
I would be grateful for anyone's views on which they prefer; the Xilinx or Altera range of FPGA's. I am at present using the Xilinx XACT software however the sales rep made quite a case for Altera (obviously!) and I must admit the software sounded quite impressive. I would like to know has anyone made the move from xilinx to Altera and if so do they think it was the right move? I would appreciate it if I could also have email replies, I will post a summary back to the group, my service provider is quite slow on the news access at the moment so sometimes I do not get all the threads. Thanks, -- Stephen BellArticle: 2739
I'm upgrading a 6-pal Atmel 2500 CUPL design to a FPGA type device. I'm thinking of using Xilinx XC3164 or an Atmel 8636 (need 300-400 ff). I intend to implement using mostly Boolean/state machine text descriptions but I have orcad sw; and I'll want to have some explicit control over resource usage. I think my two greatest concerns should be chip availability & especially, design software performance/usability/bugs (maybe cost if above 2-3k). Any feedback will no doubt add weeks of quality time to my life. THANKS!Article: 2740
>> Dump the GALs, they are absolutely worthless and expensive. > >PEELs are only moderately more capable than GALs. PEEL Arrays are nicer >but vertually unobtainable from hobbyist sources. PEEL are far more versatile than GALS, as each output can be a different MACROCELL, unlike GALs. PEELs also use far less current than GALs, and have a much better source/sink ability. The max clock rate of PEELs is also much higher than GALs. Bi-directional I/O, internal feedback (which can be latched), etc... lots of goodies that GALs are missing. >JDR and Jameco price GAL22V10-25 and PEEL22V10-25's exactly the same: >$4.95 What do you expect from places like these? >ICT's software is free but it is also buggy and PEEL's are not well >suported by third parties. Huh? I have been using the PLACE software for 4 years now, making dozens of different 18CV8 and 22CV10 designs, and I have yet to see a single bug (I am an assembly language programmer, so I do know what bugs are!). >That said, PEEL's are usefull and really don't even have to choose. >PEEL's are JDEC compatible with GAL's. Most low cost device programmers >will handle both. So get you PEELs. Use them as PEEL's or GAL's >whichever way suites. I use a EE Tools Allmax with Intel/Altera's >PLDshell and/or ICT's PLACE. I prefer PLDshell but I need PLACE to take >advantage of the extra PEEL features (like bi-directional I/O) They are JDEC format, but not pin or MACROCELL compatible. > PEELs provide a wide range of different Macro Cells, making them much more > versatile than GALs. They are about .75 cents for an 18CV8 and about > $1.75 for a 22CV10 (although I pay less than this for 1000+ quantities). Where are you finding 22V10's for $1.75? Are you still talking large quantities? I pay about $1.45 for 22CV10s and $.65 for 18CV8s in 1000 lots. I can order as little as 100 for the $1.75 price, but I do buy factory direct. Jim Drew, Utilities Unlimited International, Inc.Article: 2741
>I've been avoiding this but looks like the "best" path is to get an ol' >286/386 box and a card based programmer. If the Mac only had a parallel >port...sigh. It can. There are several programmers for the MAC that use a parallel interface. You use the PowerPrint cable (which converts the 8 pin mini-din to parallel) with these devices. Jim Drew, Utilities Unlimited International, Inc.Article: 2742
Jim Drew (jimdrew@aol.com) wrote: : Dump the GALs, they are absolutely worthless and expensive. : Go with PEELs. Contact ICT for literature. And contact your distributor to see if you as a new customer can get any. We have had several hunderd 18cv8s on order since last July. We are being promised April, maybe. ICT is fabless, and at the mercy of their fabs. Last year, their Singapore fab, Chartered, cut their wafer allocation. Instant shortage. They are currently working with a Rohm fab in San Jose. Things should improve soon, but I'll believe it when parts show up on my dock. Never again. -- Gerry Belanger, WA1HOZ wa1hoz@a3bbak.nai.net Newtown, CT g.belanger@ieee.orgArticle: 2743
Capable of routing 3090 series devices (or better) offers ? MArticle: 2744
In article <4em0uo$klc@newsgate.sps.mot.com>, Doug Shade writes: > Lattice has an In Circuit Programmable kit for 'round $90, that > includes > there spin of a 22V10. The trouble with Lattice ISP is the same as with almost everything except generic 22V10's. Where do you get parts? Digikey no longer caries them. AFAIK, no other low volume supplier has ever carried them. ---- "Not many fishes, in the sea. Not many fishes, just Londo and me" Remember the home hobbyist computer: Born 1975, died April 29, 1994Article: 2745
>And contact your distributor to see if you as a new customer can get any. >We have had several hunderd 18cv8s on order since last July. We are >being promised April, maybe. Yeah, I guess being on allocation prevents these types of problems. I get thousands drop shipped monthly, so I don't see any shortage. I am having this problem with MOSFETs though. I wish I was on allocation for them. I won't see any until Jan '97. :-\ Jim Drew, Utilities Unlimited International, Inc.Article: 2746
I'm aware of their isp line. At $12 a pop (apx.), it's a tad overkill compaired to a GAL. However, the ability to alter the circuit board configuration in the field may come in handy someday! mark In article <4env0o$bjg@neptune.myri.com>, wen-king@myri.com (Wen-King Su) wrote: >In a previous article mark.stephens@gsfc.nasa.gov (Mark Stephens) writes: >> ><I'd really, really like to get rid of the NOR gate glue logic and replace >>it with some species of GAL. I have PLDasm and it's manual and can now ><create the JDEC files. Unfortunately, my PROM programmer is just that... >>PROMs only. >< >>Do I need to get another programmer or can one home brew for a specific chip? > >Lattice makes a 22V10 equivalent that is in-circuit reprogrammable via >serial down load cable. No special voltage is required, and you can make >the down load cable yourself and hook it up to the printer port of your PC. -- mark stephens "In constraint, NASA GSFC Code 521 is freedom" Greenbelt, MD (301) 286-4269 mark.stephens@gsfc.nasa.govArticle: 2747
Hi, I need advice from experts, I have to build some networkcards for parallel computing using PVM. The operating system is going to be LINUX. The networkcards will be build in Pentium PC's. For the PCI interface I'm using XILINX FPGA's. The link adapters are STC101's and the link switch is a STC104. I have some questions regarding the PCI interface. 1) Will a X4000E be fast enough ? Or is it recommended to use a 3000 or 7000 ? 2) In the PCI-examples that Xilinx provided, they are using two X7000's One for the businterface, and one for the parity check. Thus for maximizing the logic capicity. Is this necessary ? And will this be necessary if a X4000E will be used? 3) And do you guys think I'll need the optional interrupt signal INTA# for my design ? I'd also like to know if there is some newsgroup or website where to get a lot of info about STC101 link adapters. Thanx in advance !!!! Kim Hofmans tw38966@vub.ac.be khofmans@info.vub.ac.beArticle: 2748
Stephen Bell <sbell@stephenb.demon.co.uk> wrote: >I would be grateful for anyone's views on which they prefer; the Xilinx >or Altera range of FPGA's. I am at present using the Xilinx XACT >software however the sales rep made quite a case for Altera (obviously!) >and I must admit the software sounded quite impressive. I would like to >know has anyone made the move from xilinx to Altera and if so do they >think it was the right move? >I would appreciate it if I could also have email replies, I will post a >summary back to the group, my service provider is quite slow on the news >access at the moment so sometimes I do not get all the threads. Are you interested in just P&R tools? Quality/ease of synthesis is a big issue as well, if you're doing a large design.Article: 2749
: I've met too many people that have made the mistake of hiring on : "buzzwords". They're not what makes the engineer. They may make : it easier to make a decision, but they mean nothing to a good manager. Without the proper buzzwords, one may not able to pass through the HR. For big company, the HR do not even read your resume. In order to reduce cost. They simply scan you resume using scanner and run a problem to match the keywords. Also the program will assign a mark to your resume by number of keywords match. They will look for not only VHDL .... etc., but also things like "team player" ,.....etc. Also if your resume use some unusual fonts. You might be out of luck. The program might give you a poor marks. -- Edwin Tsang, Email:edwint@bnr.ca , NORTEL Opinion is mine only and I reserved the right to change it
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