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 apologise, will be careful next time. Andy Peters wrote: > Dalip K. Singh wrote in message <393C0819.DDDED147@hns.com>... > >I have used Quickswitch for 3.3/5V xlation and it works great. > >P/n: QS34X2245. I had to interface altera max7032aelc 3.3v part to 5v > >ttl logic. > > > >-Dalip. > >[Image] > > Please don't post data sheets here. It's annoying. Post a link to a web > site instead. > -- > ----------------------------------------- > Andy Peters > Sr Electrical Engineer > National Optical Astronomy Observatories > 950 N Cherry Ave > Tucson, AZ 85719 > apeters (at) noao \dot\ edu > > "A sufficiently advanced technology is indistinguishable from magic" > --Arthur C. ClarkeArticle: 22976
--------------7607B5902295183738F8A6DC Content-Type: text/plain; charset=us-ascii; x-mac-type="54455854"; x-mac-creator="4D4F5353" Content-Transfer-Encoding: 7bit Jim Granville wrote: > Peter Alfke wrote: > > > >Q: What about the reverse PFET CMOS diode - does this not clamp to > Vccio+0.6v ? > No, that diode was "eliminated" when the output was designed to be "5-V tolerant". You can pull the output to 6 V with only microamps of current. Even when the 3.3-V supply is at zero volt! That took some effort and circuit trickery, but it works. Peter Alfke, Xilinx Applications --------------7607B5902295183738F8A6DC Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit <!doctype html public "-//w3c//dtd html 4.0 transitional//en"> <html> <p>Jim Granville wrote: <blockquote TYPE=CITE>Peter Alfke wrote: <br>> <br>>Q: What about the reverse PFET CMOS diode - does this not clamp to <br>Vccio+0.6v ? <br><a href="http://www.DesignTools.co.nz"></a> </blockquote> No, that diode was "eliminated" when the output was designed to be "5-V tolerant". You can pull the output to 6 V with only microamps of current. Even when the 3.3-V supply is at zero volt! <br>That took some effort and circuit trickery, but it works. <p>Peter Alfke, Xilinx Applications <br> </html> --------------7607B5902295183738F8A6DC--Article: 22977
This is a multi-part message in MIME format. --------------7CECC5631362FF15E6498E58 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hi, I've recently upgraded to the 2.1 version of the Xilinx foundation tools, and I'm having some problems with a very simple design. This VHDL code works when compiled using the version 1.5 tools, but does not work when using the 2.1 tools. Has anybody else experienced this? I have got the latest patches for the tools, and all the licence stuff is working OK. The design is a simple 8 bit register mapped onto the I/O space on a PC/104 bus (ISA bus). I've attatched the code for anybody interested. If anyone could help me with this it would be greatly appreciated. Cheers, James -- James Kennedy Electronics/Computer Design Engineer IntelliDesign Brisbane, Australia. james at intellidesign dot com dot au Tel: +61 7 3366 6478 Fax: +61 7 3366 6471 --------------7CECC5631362FF15E6498E58 Content-Type: application/x-unknown-content-type-ACTIVE.HDE; name="vhdltest.vhd" Content-Transfer-Encoding: base64 Content-Disposition: inline; filename="vhdltest.vhd" LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLQ0KLS0N Ci0tIFNpbXBsZSBWSERMIHRlc3QgZmlsZQ0KLS0NCi0tIEouS2VubmVkeSwgNiBKdW5lIDIw MDANCi0tDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tDQoNCmxpYnJhcnkgSUVFRTsNCnVzZSBJRUVFLnN0ZF9sb2dpY18xMTY0LmFsbDsNCg0K ZW50aXR5IFZIRExURVNUIGlzDQogICAgcG9ydCAoDQoJLS0gUEMvMTA0IHBpbnMNCglTQToJ CQlpbgkJU1REX0xPR0lDX1ZFQ1RPUiAoMTUgZG93bnRvIDApOw0KCVNEOgkJCWlub3V0CVNU RF9MT0dJQ19WRUNUT1IgKDcgZG93bnRvIDApOw0KCUlPUjoJCWluCQlTVERfTE9HSUM7DQoJ SU9XOgkJaW4gCQlTVERfTE9HSUM7DQoJQUVOOgkJaW4JCVNURF9MT0dJQzsNCglSU1REUlY6 CQlpbgkJU1REX0xPR0lDOw0KCQkNCgktLSBMRUQgcGlucw0KCUxFRDE6CQlvdXQJCVNURF9M T0dJQzsNCglMRUQyOgkJb3V0CQlTVERfTE9HSUM7DQoJTEVEMzoJCW91dAkJU1REX0xPR0lD DQoNCgkpOw0KZW5kIFZIRExURVNUOw0KDQphcmNoaXRlY3R1cmUgVkhETFRFU1RfYXJjaCBv ZiBWSERMVEVTVCBpcw0KDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tDQotLSBDb21wb25lbnQgRGVjbGFyYXRpb24gU1RBUlRVUCBNT0RV TEUNCi0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0NCmNvbXBvbmVudCBTVEFSVFVQDQoJUE9SVCggR1NSOiBJTiBzdGRfbG9naWMpOw0KZW5k IGNvbXBvbmVudDsNCg0KLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLQ0KLS0gSU5URVJOQUwgU0lHTkFMUw0KLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLQ0Kc2lnbmFsIENTX1NDUkFUQ0hQ QUQ6IFNURF9MT0dJQzsNCnNpZ25hbCBTQ1JBVENIUEFEOiBTVERfTE9HSUNfVkVDVE9SKDcg ZG93bnRvIDApOw0KDQoNCmJlZ2luDQoNCi0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0NCi0tIFN0YXJ0dXAgY29tcG9uZW50IGluc3RhbnRp YXRpb24gDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tDQpTVEFSVENPTVBPTkVOVCA6IFNUQVJUVVAgcG9ydCBtYXAgKEdTUiA9PiBSU1RE UlYpOw0KDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tDQotLSBQZXJtYW5lbnQgY29ubmVjdGlvbnMNCi0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0NCkxFRDEgPD0gU0NSQVRDSFBBRCgw KTsNCkxFRDIgPD0gU0NSQVRDSFBBRCgxKTsNCkxFRDMgPD0gU0NSQVRDSFBBRCgyKTsNCg0K LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLQ0K LS0gUkVHSVNURVIgQ0hJUCBTRUxFQ1RTDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tDQpDSElQX1NFTEVDVFM6ICBwcm9jZXNzIChBRU4s IFNBKQ0KYmVnaW4NCglDU19TQ1JBVENIUEFEIDw9ICcwJzsNCgkNCglpZiAoQUVOID0gJzAn KSB0aGVuDQoJCWlmICggU0EgPSAiMDAwMDAxMDAwMDEwMDAwMCIpIHRoZW4JCS0tIDQyMGgN CgkJCUNTX1NDUkFUQ0hQQUQgPD0gJzEnOw0KCQllbmQgaWY7DQoJZW5kIGlmOw0KZW5kIHBy b2Nlc3MgQ0hJUF9TRUxFQ1RTOw0KDQotLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tDQotLSBJU0EgQlVTIFRSQU5TRkVSUzogIFNDNDAwIC0+ IEZQR0ENCi0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0NCldSSVRFOiBwcm9jZXNzKENTX1NDUkFUQ0hQQUQsIElPVykNCmJlZ2luDQoJaWYg KCgobm90IElPVykgYW5kIENTX1NDUkFUQ0hQQUQpID0gJzEnKSB0aGVuDQoJCVNDUkFUQ0hQ QUQgPD0gU0Q7DQoJZW5kIGlmOw0KZW5kIHByb2Nlc3MgV1JJVEU7DQoNCi0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0NCi0tIElTQSBCVVMg VFJBTlNGRVJTOiAgRlBHQSAtPiBTQzQwMA0KLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0t LS0tLS0tLS0tLS0tLS0tLS0tLS0tLS0tLQ0KUkVBRDogcHJvY2VzcyhDU19TQ1JBVENIUEFE LCBJT1IpDQpiZWdpbg0KCVNEIDw9ICJaWlpaWlpaWiI7DQoJaWYgKCgobm90IElPUikgYW5k IENTX1NDUkFUQ0hQQUQpID0gJzEnKSB0aGVuDQoJCVNEIDw9IFNDUkFUQ0hQQUQ7DQoJZW5k IGlmOw0KZW5kIHByb2Nlc3MgUkVBRDsNCg0KZW5kIFZIRExURVNUX2FyY2g7DQo= --------------7CECC5631362FF15E6498E58--Article: 22978
Hi, I have try to designed the FPGA for digital picture processing (a very simple one) using schematic editor. I have problem that count of used CLB's is too high. I tried to simplify the design, but nothing seems to affect to the used CLB count. Does anyone know what 'compnents' I shouldn't use for not wasting CLB resources ? Of course,a bigger Spartan would solve this problem, but is it really necessary ? Now I have used XCS10-version of Spartan. Jukka Popponen FinlandArticle: 22979
Hi all. Since we are going to buy a Virtex development board, we would like to know what do you think of a Ballynuey 2 from Nallatech. Is there anybody using it? Is it worth the price? We intend to use it for developping software radio algorithms (downconverter). Is there a link (beyond the official Nallatech site) where we can get more information? Thanks, Giuseppe. -- __________________________________________________________________ | | |Ing. Giuseppe Baruffa, | | | |c/o Dipartimento di Ingegneria Elettronica e dell'Informazione | |Università degli Studi di Perugia, | |via G. Duranti 93, 06125 Perugia (Italy). | |__________________________________________________________________| | | | |Tel.: +39 75 5853626 |E-mail:baruffa@diei.unipg.it | |Fax: +39 75 5853654 | | | |Home Page: | | | http://dante.diei.unipg.it/~baruffa | |_______________________|__________________________________________|Article: 22980
iglasner@my-deja.com wrote: >> Even tho' I'm not looking for any free tool/hw/sw etc I was still > wondering why should a CD be made which for sure will mean a certain > cost. > > If the intension was to make those kind of thing free the way I would > think should be used is simple putting all the "staff" on the web maybe > on several server and than it will really be free. > From the distributors' point of view: why is disk space and web server maintainance free? From the downloader's point of view: which ISP is free? And who wants to spend ages downloading 650Mb (a CD's worth) over the net. Even at 30Kbytes a second (the fastest I've ever got anything over the net, even with a fixed connection) that would take 6 hours. And for us poor ppl with 56Kbit modems, its at least 26 hours. Since I rarely get better than 28.8, it would be quicker and cheaper for somebody to post me a CD. -- Andrew MacCormack email: andrewm@cadence.com -- Senior Design Engineer -- Cadence Design Systems, Alba Campus, Livingston EH54 7HH, Scotland -- Phone: +44 1506 595360 Fax: +44 1506 595959Article: 22981
Hi, I think that you can use floor planner editor, you can move blocs et generate a place constraint file. I think, It will be better just moving element that required short delay. Put as near as possible elements that communicate together... And good luck. For information, be carefull, timing simulator is really bad and tri-state timing are false, so you will prefer using something like model-sim to have good timing results... See you Paul <jthioude@my-deja.com> a écrit dans le message : 8h8jei$pfa$1@nnrp1.deja.com... > First of all, exuse me for my poor english.I've got a disign on a XC4044XLA > , implemented with Foundation 1.5i. This design is "critiqal in delay".I'd > like to make few modifications on this design if possible, without modifying > the first routing. What is the best method?? > > > Sent via Deja.com http://www.deja.com/ > Before you buy.Article: 22982
I had the honor of talking to Mr. Optimagic today, and asked if he was aware that his web presence was sub-optimal. He told me that he was aware of the problem (incompetence at network solutions), and that he hoped that things would be restored in a day or two. Ping early, and Ping often: www.optimagic.com Philip FreidinArticle: 22983
It appears that FPAG compiler is ignoring timing specs of the select rams in the vertex fpgas. Xilinx design manager properly points out those path in the timing analysis tool. I looked at all the obvoius places for answers but could not find anything. Anyone else experienced that, or knows what I'm doing wrong ? I tryed both, directly insantiating RAM block (in my verilog) code, and use memories generated with core gen. Both times, there appear to be notiming constrains on clock to output (and I presume on input). Thanks for any tips suggestion ! bkk Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22984
James Kennedy wrote: > I've recently upgraded to the 2.1 version of the Xilinx foundation > tools, and I'm having some problems with a very simple design. This > VHDL code works when compiled using the version 1.5 tools, but does not > work when using the 2.1 tools. Has anybody else experienced this? > > I have got the latest patches for the tools, and all the licence stuff > is working OK. > > The design is a simple 8 bit register mapped onto the I/O space on a > PC/104 bus (ISA bus). I've attatched the code for anybody interested. I have a fairly large project that builds and works perfectly on my target hardware with F1.5i, however everything goes to hell after I upgraded to F2.1i. However, my problem is that the design does not build at all. I am doing an HDL flow, but with a top-level schematic generated VHDL netlist. The bus handling seems different between the two versions, and I don't understand how to connect a bus to a pin with a different number of lines. The method for 1.5i doesn't seem to work in 2.1i. I also have a problem with some RLOC_ORIGIN requirement that has magically appeared because of a tri-state buffer. I currently have two cases open at Xilinx support, and for about a week or two for both, and they are not moving at all. I can't find any decent documentation about upgrading projects from 1.5i to 2.1i. I've re-installed 1.5i until these things are sorted out, but if any others have experience with this, I'd also appreciate it. When my cases are resolved, I'll make sure to post the results to this group. thanks in advance Joshua Lamorie Systems Designer Xiphos Technologies Inc.Article: 22985
Hey, I bought a copy of Xilinx Foundation Student Ed. and went through the registration process, received the stuff for license.dat. I then plugged it into the path it should be by the command line parameter, and I am not able to run anything other than Xilinx Design manager. (Even then, when it starts to init, it gives errors) Generally the error is that there are no "server" lines in my license.dat however, this means very little to me other than "no im not going to work." Any ideas on what I might have missed? (I've done the web-support, but have yet to get a response) _seiya_ @ nergal.org (just remove spaces) for the spam bots, try abuse@fbi.govArticle: 22986
Try to check --> the path for license.dat file in autoexec.bat --> if you HD id or Network ID is correct into the license file I haven't more idea !! PaulArticle: 22987
Hi, I've noticed a strange relation in Virtex prices. XCV50E has more system gates, more available IO pins, and more blockram, and it's still cheaper than XCV50 (the same is for bigger chips) !! Why ? How ? Maybe it's just a promotional trick. What do you think, how will the prices (and price relations) behave in the next two years ? regards, ------------------------------------------- - Domagoj - - Domagoj@engineer.com - -------------------------------------------Article: 22988
In article <8hlor8$b6l$1@bagan.srce.hr>, Domagoj <domagoj@engineer.com> wrote: >Hi, > I've noticed a strange relation in Virtex prices. XCV50E >has more system gates, more available IO pins, and more blockram, >and it's still cheaper than XCV50 (the same is for bigger chips) !! >Why ? How ? The E series is fabricated on a .18 instead of a .25uM process, uses 1.8 instead of 2.5 volt power supply, and does not support 5V tolerant I/O. Since this is largely a process shrink, the die size is smaller for the Virtex E. > Maybe it's just a promotional trick. What do you think, >how will the prices (and price relations) behave in the next two >years ? It's more a process trick. -- Nicholas C. Weaver nweaver@cs.berkeley.eduArticle: 22989
Help, I need somebody Badly need exact format/examples for Synplicity design time constrains. Pls Help!Article: 22990
Philip Freidin wrote: > > I had the honor of talking to Mr. Optimagic today, and asked if he was > aware that his web presence was sub-optimal. He told me that he was aware > of the problem (incompetence at network solutions), and that he hoped > that things would be restored in a day or two. > > Ping early, and Ping often: www.optimagic.com > > Philip Freidin Why didn't you post the resurrection of optimagic? I think thats more efficient than polling ;-) Patrick -- Patrick Schulz (schulz@rumms.uni-mannheim.de, pschulz@ieee.org) University of Mannheim - Dep. of Computer Architecture 68161 Mannheim - GERMANY / http://mufasa.informatik.uni-mannheim.de Phone: +49-621-181-2720 Fax: +49-621-181-2713Article: 22991
Hi, > The E series is fabricated on a .18 instead of a .25uM > process, uses 1.8 instead of 2.5 volt power supply, and does not > support 5V tolerant I/O. > Since this is largely a process shrink, the die size is > smaller for the Virtex E. I've assumed that , but it still doesn't make sense to me. Xilinx probably wants to squeeze every last drop of juice from plain old Virtex , so why would they sell better chips cheaper. This doesn't help them selling plain Virtex , except if someone needs 5V tolerant IOs. ------------------------------------------- - Domagoj - - Domagoj@engineer.com - -------------------------------------------Article: 22992
What is the nature of your "digital picture processing"? What kind of data rates do you need to handle? Answers to these questions will dictate the design architecture. If the processing is relatively complex and you need to work at a video frame rate you may not have enough logic resources in an XCS10 no matter how you slice it. If it is less demanding, you might consider reconfiguration to do the processing in steps and/or using bit serial techniques to reduce the size of the hardware. In any event, it is not so much what components not to use, as much as designing to the requirements and to the architecture. "Jukka Pöppönen" wrote: > Hi, > > I have try to designed the FPGA for digital picture processing (a very > simple one) using schematic editor. I have problem that count of used CLB's > is too high. I tried to simplify the design, but nothing seems to affect to > the used CLB count. Does anyone know what 'compnents' I shouldn't use for > not wasting CLB resources ? > > Of course,a bigger Spartan would solve this problem, but is it really > necessary ? > Now I have used XCS10-version of Spartan. > > Jukka Popponen > Finland -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com or http://www.fpga-guru.comArticle: 22993
Hello, I am looking for TTL device libraries (macros), ideally in CUPL, but other HDL's would do.. There must be dusty archives of these things somewhere.. CUPL comes with only about 30 TTL equiv macros, other systems may come with more.. TIA - Jim G.Article: 22994
Hi What are the differences (if any) between an fpga and an epld?Article: 22995
In article <OYVt$EJ0$GA.311@cpmsnbbsa07>, "Seiya" <valid@email.address.com> wrote: > Hey, I bought a copy of Xilinx Foundation Student Ed. > and went through the registration process, received the stuff > for license.dat. I then plugged it into the path it should be by the command > line parameter, and I am not able to run anything other than Xilinx Design > manager. (Even then, when it starts to init, it gives errors) Generally the > error is that there are no "server" lines in my license.dat however, this > means very little to me other than "no im not going to work." > Any ideas on what I might have missed? > > (I've done the web-support, but have yet to get a response) > _seiya_ @ nergal.org (just remove spaces) > > for the spam bots, try abuse@fbi.gov > > Which command line parameters? You need to set up an environment variable LM_LICENSE_FILE, which points to the license file. Example : set LM_LICENSE_FILE=C:\flexlm\license.dat -- Klaus Falser Durst Phototechnik AG I-39042 Brixen Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22996
"Domagoj" <domagoj@engineer.com> wrote in message news:8hmqcd$gq0$1@bagan.srce.hr... > Hi, > > > The E series is fabricated on a .18 instead of a .25uM > > process, uses 1.8 instead of 2.5 volt power supply, and does not > > support 5V tolerant I/O. > > Since this is largely a process shrink, the die size is > > smaller for the Virtex E. > > I've assumed that , but it still doesn't make sense to me. Xilinx probably > wants to squeeze every last drop of juice from plain old Virtex , so why > would they sell better chips cheaper. This doesn't help them selling > plain Virtex , except if someone needs 5V tolerant IOs. > When looking at the entire range of Xilinx parts, including package/speed/temperature variations, I'm pretty sure they would have preferred to have far fewer types. So they probably want to discourage use of XCV by pricing XCV-E lower. And if someone complaints about XCV-E not being 5V tolerant, they could always argue that you could spend some of the savings on external level translators..... ;-) Regards, - OlafArticle: 22997
I hope that this situation seems familiar to someone with more vhdl/fpga experience than I. If it does, I would be interesting in seeing your opinion or advice. I have just completed the design of a data acquisition system that incorporates two Actel parts: a 40mx04 and a 42mx09. The development tools that I used are the "Actel DeskTop". They include a subset of Synplicity. I have found that Synplicity makes only limited use of the library macros in 40mx compilation as compared to 42mx compilations. A specific example is that 42mx outputs will include and2, and2a, and and2b while the 40mx output will contain only and2. and2a and and2b are both available in the 40mx macro library, and they are implemented in a single cell. An and2a is a 2-input and gate with one of the inputs inverted; the and2b has both inputs inverted. To implement the equivalent of an and2a in 40mx, Synplicity creates an inverter and an and2. The problem here is, of course, that 2 cells instead of 1 are needed, and there is a greater propogation delay. The back-end Actel "place and route" tools evidently take care of some, but not all, of this problem in post-systhesis optimization. I have verified this by synthesizing the same behavioral VHDL code using both 40mx and 42mx compilers. In instances where the 42mx compiler has output macros that are not available in the 40mx library, I edited the structural output by instantiating equivalent 40mx macros or macro combinations. I then processed both structural outputs through the Actel back-end tools. The code that was processed with the 42mx Synplicity compiler produced smaller and faster outputs. The performance was verified under simulation (in ModelSim) and in actual hardware implementation. Being relatively new to fpga design, I initially suspected that deficiencies in the size and performance of my designs were due to my limited experience with and knowledge of the process. I scrutinized my work thoroughly before looking at the perfomance of my tools. Does my assesment seem to be valid? Have any of you had these experiences? And, if so, what approach enabled you to overcome the problems? Thanks for looking this over, Frank MadisonArticle: 22998
Frank Madison wrote: > I hope that this situation seems familiar to someone with more vhdl/fpga > experience than I. If it does, I would be interesting in seeing your opinion > or advice. Not too familiar with this exact problem but interesting trouble you have. :-) ================================= > I have just completed the design of a data acquisition system that > incorporates two Actel parts: a 40mx04 and a 42mx09. The development tools > that I used are the "Actel DeskTop". They include a subset of Synplicity. I > have found that Synplicity makes only limited use of the library macros in > 40mx compilation as compared to 42mx compilations. Remember that the 40mx is really an Act 1 architecture and a 42mx is an Act 2 architecture. The 40mx, for combinational logic, has a slightly different logic element than the 42mx. (the two muxes up front have separate selects in Act 1; in Act 2, they give you an extra AND gate on one of the mux selects, making it a bit better for up and down counters. Often, the Act 1 element seems to be more useful and powerful). ====================================== > A specific example is that 42mx outputs will include and2, and2a, and and2b > while the 40mx output will contain only and2. and2a and and2b are both > available in the 40mx macro library, and they are implemented in a single > cell. An and2a is a 2-input and gate with one of the inputs inverted; the > and2b has both inputs inverted. To implement the equivalent of an and2a > in 40mx, Synplicity creates an inverter and an and2. The problem here is, > of course, that 2 cells instead of 1 are needed, and there is a greater > propogation delay. > > The back-end Actel "place and route" tools evidently take care of some, but > not all, of this problem in post-systhesis optimization. I have verified this > by synthesizing the same behavioral VHDL code using both 40mx and 42mx > compilers. In instances where the 42mx compiler has output macros that are > not available in the 40mx library, I edited the structural output by > instantiating equivalent 40mx macros or macro combinations. I then > processed both structural outputs through the Actel back-end tools. The code > that was processed with the 42mx Synplicity compiler produced smaller and > faster outputs. The performance was verified under simulation > (in ModelSim) and in actual hardware implementation. There is an optimizer in the Combiner which tries to eliminate logic that is not necessary and will handle a bunch of these things. Another trick is to take the output of the VHDL synthesizer and then run it through Actmap, in netlist optimizer mode. It has been shown for some VHDL synthesizers that this will produce very large performance gains; for others, it has almost no effect. It depends on the structure of the logic generated by the synthesizer. ============================================== > Being relatively new to fpga design, I initially suspected that deficiencies > in the size and performance of my designs were due to my limited experience > with and knowledge of the process. I scrutinized my work thoroughly before > looking at the perfomance of my tools. Does my assesment seem to be valid? > Have any of you had these experiences? And, if so, what approach enabled you > to overcome the problems? It seems that you have looked at this carefully and it does seem to be valid. If you have some free time on your hands, you can try running the code using Act 3 and SX technologies. Act 3 is the same as Act 2 with the addition of another input to the S-module so that the clear signal isn't shared, helping combinability. The SX does have a more powerful logic module. In one test that was run with a particular version of software, it was found that setting the target to Act 3, with the less powerful logic module, got better results, for a purely combinational design, than SX. Why was this? SX was new and all the algorithms were not yet implemented in that version of the software. I suspect, but have no evidence, that this may be the source of your problem. Act 1/40MX are relatively small devices and as such have the least suitability for HDL synthesis. Also, the Act 1 module is different, by a bunch, from Act 2 and Act 3. So, given limited resources, one can suspect that Synplicity, at least for the version that you have, simply did not put much time in developing and optimizing their synthesis engine for Act 1 technology. Did the Synplicity folks have anything else to say about this? Here are some options that may enable you to eliminate your problem. 1. Like the old joke with the doctor, how to fix the problem when your arm hurts when you lift it funny, don't do that. How about using the A1225A (or whatever the 42MX equivalent is) instead of the 40MX device, as these are comparable device densities. This will actually help you out in board level simulations since they will both reference the same Actel models. 2. Design with schematics. 3. If you did not update to the latest version of software, use the freebie Actel VHDL synthesizer, Actmap. For combinational logic, this seems to be the best tool for synthesis (although I most use it for Act 2/3/SX). > Thanks for looking this over, No problem, interesting story. Have a good day, ---------------------------------------------------------------------- rk But Mother Nature, unlike Congress stellar engineering, ltd. and the press and even the space stellare@erols.com.NOSPAM workers, can't be bluffed. Hi-Rel Digital Systems Design -- James Oberg, 2000Article: 22999
POSTE : Ingénieur Électronicien Design Numérique DATE D'EMISSION : 8 Juin 2000 Référence : PP_IEN1 POSTE Description : Design Cartes ou Modules Numériques Lieu de travail : Bretagne, Grand-Ouest Date de disponibilité : Immédiate à 3 mois Type de contrat : CDI Rémunération : 200 KF à 300 KF selon expérience Création de poste : oui CONTACT ------------- SOCIÉTÉ : IPSIS Secteur : Ingénierie spécialisée Personne à contacter : Patrick PENEAU Adresse : 3 square du Chêne-Germain 35510 CESSON-SÉVIGNÉ Téléphone : 02.99.27.53.24 Fax : 02.99.27.53.28 E-mail : commercial@ipsis.com AUTRES INFORMATIONS -------------------------------- Connaissance : Microcontrôleurs, FPGA/EPLD, . Expérience programmation/simulation VHDL, . Expérience de la CAO (Mentor, .) IPSIS,filiale rennaise du groupe ITLINK, est une société d'ingénierie spécialisée sur les domaines de l'Automatique, des Télécommunications et des Hyperfréquences. Pour faire face à la croissance rapide de ses activités, IPSIS recrute des ingénieurs expérimentés ou débutants dans ses domaines d'expertise. -- Patrick PENEAU _________________________________ IPSIS - Ingénierie Pour Signaux & Systèmes ITLINK Group Phone : +33 2 99 27 53 24 Fax : + 33 2 99 27 53 28 Email : commercial@ipsis.com Web : http://www.ipsis.com http://www.itlink.fr
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