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
A moment of your time please, Tom... Tom Kean (tom@algotronix.com) wrote: : Mike McCarty wrote: : > )->In reconfigurable hardware, the behaviours and interconnections of the : > )->constituent electronic primitives can be repeatedly changed. Artificial : > )->evolution can automatically derive a configuration causing the system to : > )->exhibit a pre-specified desired behaviour. A circuit's evolutionary [snip] : > Let me translate this little bit into English. : > : > Hardware which can be changed can be changed more than once. It : > is sometimes possible to make the hardware configure itself. [snip] : <FLAME> : Actually, its not 'obfuscated bullshit' its precise scientific english: you : just have to turn off the TV, put down your beer and engage your brain : before reading it. I enjoy drinking beer. I also enjoy thinking. Are the two mutually exclusive? Most of my friends are philosophical drunks. ;) Actually, the summary is obfuscated. That's not necessarily bad, that just happens to be a side-effect of the type of summary being written: a dense, high information-per-sentence paragraph that attempts to explain a general concept (which takes an entire book to fully describe, or so one infers) in a short amount of space -- hence, a summary of the concept. In order to explain such a wide concept in precise terms in a short summary, dense language is used which must be read carefully. Obfuscation (as webster says, "to make obscure") is not the purpose of such language, rather it is a side effect of the concept-condensing process. This is not the only kind of summary which could have been written. It was written thoughtfully and deliberately. Alternatively it could have been a topical overview of the specific discussions within the text, etc. However, the author of the summary believed that this sort of dense conceptual specification would engage people who would be interested in reading the book. The summary is obviously not bullshit, since it imparts very precise and accurate information to the people who extract it. I sincerely hope that your flame against anyone not willing to extract that information was just a knee-jerk reaction (careful with those, now). There was a pointed comment that the summary could have be written in a more accessible fashion: you disagree (maybe the book itself is not accessible, and such a summary would be misleading?). Fine, disagree, but don't automatically light up a blowtorch (care with those, too!). : For example: : : 'Hardware that can be changed can be changed more than once' FALSE : what about fuse technologies? The original text does not make this claim. The original text (as quoted by my newsreader, tin) says: "In reconfigurable hardware, the behaviours and interconnections of the constituent electronic primitives can be repeatedly changed." ^^^^^^^^^^ The use of the word "repeatedly" implies that the hardware can be changed more than once. It may be false to infer that the *same* behavior or connection can be changed more than once, but it is very clear that the hardware in general can be. Do you disagree with this interpretation? : 'It is sometimes possible to make the hardware configure itself' : The text does not say this either - what it says is that a configuration : can be derived from artificial evolution. From the quoted portion of the message I saw, I would agree. It is interesting to note that the summary uses the phrase "a circuit's evolutionary..." If this refers to the evolutionary capabilities of the entire system in redesigning a specific circuit, you are right. If this refers to the *physical* circuit, stating that the *physical* hardware circuit(ry) has evolutionary capabilities (outside of the context of a surrounding system), then one would assume that the circuit has the ability to modify itself (since it can evolve outside of a system, when there is nothing else to modify it!). : </FLAME> </rebuttal> : Tom. With respect, Gordon. ------------------------------------------------------------------ Common sense and a sense of humor are the same thing, moving at different speeds. A sense of humor is just common sense, dancing. -- Clive JamesArticle: 11601
Jeff Berryhill <berryhill@uchicago.edu> wrote in article <Ey9E1B.4Bx@midway.uchicago.edu>... > In article <6rt9de$j01$1@nnrp1.dejanews.com>, > jcvilleneuve@hotmail.com writes: > > > > Is anybody has used something else than a PROM for the > > configuration cycle of > > their FPGA (example direct CPU to fpga)? > > > We are currently designing a board, upon which resides upwards of a dozen > Altera FLEX10K FPGAs. The board resides in a crate with VMEbus access from > a PC. The board itself contains an Altera MAX7000 chip which coordinates > VME I/O, a large (16 Mb) FLASH memory, and an additional MAX9000 > 'boot chip' which emulates the configuration cycle of a FLEX 10K. > I did something very similar for a VXI based A/D converter board. An Altera MAX7128S decodes the VXI A16 space registers and contains a register that controls downloading 5 EPF10K20s and 2 EPF10K10s. I first program the MAX7128S using a Byteblaster and then the rest of the board is configured via the board's device drivers. This allows reconfiguration of the board for different applications. Daniel Lang dblx@tyrvos.caltech.edu (but remove the x)Article: 11602
In article <35E3A88B.7FEBCF88@studm.hrz.uni-siegen.de>, Justen <justen@studm.hrz.uni-siegen.de> wrote: >Hi, > >I would like to build a digital PLL with a XILINX 4010 FPGA. The >generated frequency should be 9.5MHz to 10MHz in steps of 1kHz. So I >have to use a 1kHz Referenz (this is also the frequenzy of the >PhaseDetector) and multiply it with 9.500 to 10.000. At the moment, I >use a PhaseDetector von Xilinx AppNote and an Integrator to close the >loop. But this isn't a good design. The PhaseJitter is to big. Could >anybody help me to improve this design? If you look at the design equations for PLLs you discover that jitter is usually specified as a phase angle relative to the reference (1KHz) frequence... so your PLL is probably working fine, but not at the very high performace you had hoped for. Here are some things to try: - The reference must be very low jitter. It _must_ be xtal oscillator derived. - The VCO should also be very low jitter. Varactor tuned LC oscillators tend to be much better than relaxation oscillators. The tuning range should be as narrow as possible (9.5 to 10 sounds good). - If you can get away with it, increase your reference frequency. A multiplier of 950 is much easier to deal with than 9500- you're getting phase corrections 10 times more frequently. - Increase the bandwidth of the loop (reduce the value of the integrating capacitor). This gives you more reference frequency feed-through, but less jitter. Also make sure that you add the extra zero in the loop filter or the loop will not be stable: If you consider VCO-rate phase as the controlled variable, then you have two integrators (the VCO is an integrator of phase) in series. When you close the loop, you'll get a second order system with no damping and lots of gain (the division ratio ends up being part of the gain from this point of view). - Use a low noise op-amp for the integrator. - Add a little droop to the integrator (pull the input up or down with a large resistor). Any sequential phase detector not made with equal delays (as in random Xilinx place and route) will have a dead band where the loop is essentially open and has no control over the jitter. The droop will force the loop to have a correction on each reference clock edge, thereby eliminating this problem. - Get the Motorola PLL and MECL databooks and read the application notes therein. -- /* jhallen@world.std.com (192.74.137.5) */ /* Joseph H. Allen */ int a[1817];main(z,p,q,r){for(p=80;q+p-80;p-=2*a[p])for(z=9;z--;)q=3&(r=time(0) +r*57)/7,q=q?q-1?q-2?1-p%79?-1:0:p%79-77?1:0:p<1659?79:0:p>158?-79:0,q?!a[p+q*2 ]?a[p+=a[p+=q]=q]=q:0:0;for(;q++-1817;)printf(q%79?"%c":"%c\n"," #"[!a[q-1]]);}Article: 11603
I managed to get this stopped - you've probably noticed that they haven't posted for the last few days. I won't go into the details of how i did it, since the original offenders have probably found this thread on dejanews. One of the people I listed admitted that he hosted the sites, and that he had now asked them to move elsewhere. However, if you don't have any legal backing, then this procedure will probably work in future cases: Spammers (of this sort) will generally have a web site. To get some contact names for their domain name, do a WHOIS search at Internic (http://internic.net/cgi-bin/whois). These are real names and email addresses - Internic needs them to contact the domain, to collect their registration fees. There's a section at the bottom of the search page, titled 'domain servers in listed order'. For many (not all) sites, this will list the upstream service provider. Email any addresses you can find on the 'whois' output, although it isn't necessary to include the domain servers at this stage. Tell them, politely, that your newsgroup has been spammed, and that you will take steps to contact their upstream providers if the abuse isn't stopped. Use a false return address if you want, but there shouldn't be any problem using a real address with this sort of message. This should do the trick. If it doesn't, copy the email to the domain servers listed on the 'whois' output. These are just domain names, so you'll need to find a real address - postmaster@domain and abuse@domain should be Ok. Some sites list themselves, or a related site, as their own domain server, in which case you can use 'traceroute' to find any upstream connections. There are lots of traceroute servers - just do a search for 'traceroute'.Article: 11604
If you read Joseph's post, he specifically states he is building a digital PLL. Your hints, while correct for analog PLLs are not all that applicable for a digital one. To improve the jitter specification on a digital PLL, Joseph needs to raise the frequency of his master clock (not the reference). Reading Joseph's post again, I think what he really wants is digital frequency synthesis. If he is not attempting to lock to an external reference, then he should use an accumulator with a programmable increment and a high frequency master clock. If he is attempting to lock to an external signal (the 1KHz reference), then a PLL or DPLL is needed. If there is a VCO, then it is a PLL not a DPLL. The DPLL needs a higher frequency master clock. The amount this is higher than the target frequencies determines the amount of jitter. -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 11605
Hello Reza, Different vendors have different characteristics and different price ranges for different perfromance levels.. I guess the best starting point is to estimate your design gate-count/speed-requirement.. I have experience w/ Altera and Xilinx(Two leading FPGA vendors, in my opinion).. Lots of literature is avalibale at altera.com and xilinx.com I feel the Altera FPGAs are cheaper compared to Xilinx FPGAs.. The Altera EPLDs have EABs which can be used as big chunks of memory blocks unlike distributed memort element slike in XIlinx.. So, if your design has lots of memory, Altera might be better.. The Altera place&rooute tools are alsoeasy to learn as I recently did while designing w/ 10K100-1 which is kind of the largest and fastest FPGA available in 240-pin package.. With the new Xilinx versions, I guess it is also easy to learn the Xilinx tools.. Reza Bohrani wrote: > I am new to to world of FPGA and would like to now which vendor to choose. > We have Actel, Altera, Xilinx, Lucent ... and many more. Can anyone please > tell me which one is the cheapest, the "best"? > Sincerely > RezaArticle: 11606
This is a multi-part message in MIME format. --------------566787CF4671EAA22D6AFE34 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit benyamin@my-dejanews.com wrote: Any thoughts? (on constant vector dot variable vector product logic minimization)I don't think my posts have been much help, you'veput a lot more thought into the topic than I have time to. Thanks for an interesting topic though, its been fun thinking about. You might consider extending your investigations into (constant matrix)(variable vector). Depending on the nature of the constants, there may be many redundancies in the calculations that can be exploited to produce simpler circuits. May I ask if the S/W you're putting together is for your own use, a commercial product, or are you willing to share? - John --------------566787CF4671EAA22D6AFE34 Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for John L. Smith Content-Disposition: attachment; filename="vcard.vcf" begin: vcard fn: John L. Smith n: Smith;John L. org: Visicom Imaging Products adr: 1 Burlington Woods;;;Burlington;MA;01803;USA email;internet: jsmith@visicom.com title: Principal Engineer tel;work: 781-221-6700 tel;fax: 781-221-6777 x-mozilla-cpt: ;0 x-mozilla-html: FALSE version: 2.1 end: vcard --------------566787CF4671EAA22D6AFE34--Article: 11607
Hello every one ! I am a new user to Altera and fpga technology and I am looking for how to define the port in Altera board Maxp+II(7xxx) series, as in shown in my code (below to see) but whenever I am going to assign the port signal to fpga it is always giving me the error that pin # is not defined in VHDL code , also I dont know how to transfer the compiled code to Altera fpga as I am unable to browse the assigned fpga, please help me. Thanking all of you in anticipation Rizwan ------------------------------------------------------------------ entity shifter is port (shftin :in bit_vector (0 to 3); shftout :out bit_vector(0 to 3); shftctl :in bit_vector (0 to 1)); end shifter; architecture behav of shifter is --architecture body begin f2: process(shftin, shftctl) variable shifted:bit_vector (0 to 3); --proc. decl. part begin case shftctl is when "00" =>shifted :=shftin; when "01" =>shifted :=shftin(1 to 3) & '0'; when "10" =>shifted :='0' & shftin(0 to 2); when "11" =>shifted :=shftin(0) & shftin(0 to 2); end case; shftout<=shifted after 10 ns; end process f2; end behav;Article: 11608
THE FACTS about the New Evolutionary Electronics Book. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ BACK-COVER: `Hardware Evolution' describes the use of computer simulations of Darwinian Evolution as a way of designing electronic circuits automatically. Amidst increasing interest from academic and industrial communities, researchers are just beginning to map out the exciting possibilities and practicalities of the idea. In this book, Adrian Thompson concentrates on the fundamentals - what sort of thing can an evolved circuit be? Through a sequence of arguments and experiments, it is shown that evolution can produce `bizarre but useful' circuits, beyond the scope of conventional design. The experiments (the first of their kind) apply evolution to a Field-Programmable Gate Array (FPGA) chip, and show the promise of significant real-world engineering applications. COMMENT: This was my PhD thesis. It was selected by the British Computer Society to be published unchanged by Springer. The summary I originally posted is a scientific abstract, so yes it is more formal than a book would normally be. Thanks to those who've pointed out that the attempts given by responders to `translate into plain English' end up saying things that are wrong. A precise statement in plain-speaking English without loss of information would have been longer than the regulations allow a thesis summary to be. In my original posting, I was attempting to be as brief and informative as possible for a science/R&D audience. This has backfired somewhat in provoking a discussion across several newsgroups! I'm pleased that people care, and for those who don't want to read the summary (which does indeed take some consideration), the back-cover text is provided above. The book itself, I think, is quite readable: that's one of the BCS's selection criteria. Ordering Information ^^^^^^^^^^^^^^^^^^^^ Springer-Verlag, 1988. Distinguished Dissertations Series ISBN 3-540-76253-1 http://www.springer.co.uk/comp/books/distinguished.html email: postmaster@svl.co.ukArticle: 11609
> Now tabulate the additions (as Rick suggests below): > > 8 H > 4 D E F G > 2 B C F G > 1 A C E G > Such representations is exactly what I have been working on! That's why I posted the question in the first place...I wanted to see if stuff like this existed. I have come up with a few rules about how one can minimize expressions when the data is formatted in such a way, and I have also written code to implement it. The problem is that the search space can become quite large, so I am trying now different approaches of keeping the problem managable. Dan benyamin@ucla.edu -----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member ForumArticle: 11610
Hello, I would like to learn about CPLD/FPGA programing. I found that there are two low cost tool from Cypress and Xilinx for $95. Can someone give me suggestion on which one to purchase and why? Any other low cost alternative? Thanks in advance for feedback. Tiger -----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member ForumArticle: 11611
In article <6s1rpl$3p4$1@nnrp1.dejanews.com>, tigerz@my-dejanews.com writes >Hello, > >I would like to learn about CPLD/FPGA programing. I found that there are two >low cost tool from Cypress and Xilinx for $95. Can someone give me suggestion >on which one to purchase and why? Any other low cost alternative? Thanks in >advance for feedback. > >Tiger > >-----== Posted via Deja News, The Leader in Internet Discussion ==----- >http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum Machxl2.1 is free from vantis - www.vantis.com -- Kindest Regards | gerry@devantech | We manufacture Pic programmers, 8031, Gerald Coe | .demon.co.uk | 68302, 64180, 80C188EB cpu modules. http://www.devantech.demon.co.uk | Full custom uP control systems designed.Article: 11612
Ray Andraka <no_spam_randraka@ids.net> wrote in article <35E40FC2.4DB3B88@ids.net>... > If you read Joseph's post, he specifically states he is building a digital PLL. > Your hints, while correct for analog PLLs are not all that applicable for a digital > one. To improve the jitter specification on a digital PLL, Joseph needs to raise > the frequency of his master clock (not the reference). Reading Joseph's post > again, I think what he really wants is digital frequency synthesis. If he is not > attempting to lock to an external reference, then he should use an accumulator with > a programmable increment and a high frequency master clock. If he is attempting to > lock to an external signal (the 1KHz reference), then a PLL or DPLL is needed. If > there is a VCO, then it is a PLL not a DPLL. The DPLL needs a higher frequency > master clock. The amount this is higher than the target frequencies determines the > amount of jitter. > Given the original posters specs (9.5 MHz to 10.0 MHz in 1 KHz steps), it is unlikely that he is using a completely digital approach. I pulled the app note off the Xilinx website and it is indeed a hybrid analog/digital PLL with analog low-pass filter and VCO. Joseph Allens tips are good. I would like to mention one additional item. When the reference leads the signal you get a pump down output and when the reference lags the signal you get a pump up output (or vice versa depending on the VCO). You must adjust the timings so that when the phase error is very small, there is a small overlap where you get both the pump up and pump down signals. Use separate resistors from the pump up and pump down outputs to the capacitor to avoid simultaneous conduction. This avoids a dead zone in the comparator's phase response. Daniel Lang dblx@tyrvos.caltech.edu (but remove the x)Article: 11613
PALASM from AMD is free. I also like the Xilinx Software - available from discount book stores. I haven't tried any other stuff yet. Simon tigerz@my-dejanews.com wrote: >Hello, > >I would like to learn about CPLD/FPGA programing. I found that there are two >low cost tool from Cypress and Xilinx for $95. Can someone give me suggestion >on which one to purchase and why? Any other low cost alternative? Thanks in >advance for feedback. > >Tiger > >-----== Posted via Deja News, The Leader in Internet Discussion ==----- >http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum Opinions expressed herein are solely my own and may or may not reflect my opinion at this particular time or any other.Article: 11614
I am in search of information about the digital phase lock loop, can any one send me information that can help me in designing a digital phase lock loop, I want to design it by using FPGA Altera chip, or give me hint about the material finding about digital phase lock loop. -----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member ForumArticle: 11615
hi, has anyone worked with the OLE automation interface to viewdraw or other portions of WorkViewOffice ? Does anyone have a simple script to create a new schematic and instantiate a component, save, netlist etc? thanks muzo WDM & NT Kernel Driver Development Consulting <muzok@pacbell.net>Article: 11616
Hi, Try Lattice's free ISP Synario. You can download it from http://www.latticesemi.com. balabala tigerz@my-dejanews.com wrote in article <6s1rpl$3p4$1@nnrp1.dejanews.com>... > Hello, > > I would like to learn about CPLD/FPGA programing. I found that there are two > low cost tool from Cypress and Xilinx for $95. Can someone give me suggestion > on which one to purchase and why? Any other low cost alternative? Thanks in > advance for feedback. > > TigerArticle: 11617
Hello everyone .. I am synthesizing my 8-bit uC using FPGA-compiler from Synopsys, targeting to XC4025EPG299-2. I follow all procedures for this Xilinx(XC4000)-Synopsys compilation from their docs/app notes. First, compile using : compile -map_effort med The result : report_timing : slack = +9.32(MET) report_fpga : Total num. of CLBs = 1119 Total num. of cells = 1143 After PPR : 100% connections routed (after quite a lot iterations and still some nets failed to meet the constraint) Second, compile using : compile -map_effort med -ungroup_all (as new Xilinx app notes/v.1995 suggested) The result : report_timing : slack = +32.17(MET) -- definitely better report_fpga : Total num. of CLBs = 696 Total num. of cells = 721 -- definitely smaller After PPR : Only 86% connections routed (it gives up !) Can somebody explain me this ? Any pointer to yield a better result for second effort is appeciated. FYI all designs before this are successfully 100% routed using the second method, since it will guarantee more compact and smaller area for cells (definitely better yield compare to first method). But why this failed ? TIA -- -iman =============================================== = Watch the transformation or be a part of it = ===============================================Article: 11618
We recently compared the Foundation / Synopsys sythesiser against Galileo Extreme 4.2.2, for a number of existing VHDL projects in the 5k - 20k gate range which had been synthesised with Galileo in the past. For the designs that the Synopsys tools were able to compile, it produced results that were about 20% bigger. There were a number of designs that required simple modifications to compile, and some that (would have) required major modifications. [ Things like generic parameters being 2 diminsional arrays, vhdl'93, etc, caused problems. ] I guess you get what you pay for. I hear that Synplify is about the same as Galileo, but I have not independently confirmed this. -- Allan. On Tue, 25 Aug 1998 21:17:09 -0700, Scott Campbell <sjcampbe@earthlink.net> wrote: >I am not sure what tool is the BEST on the market, but I have Xilinx's >Foundation series 1.4 software. It includes Synopsys' FPGA express for synthesis >as well as a HDL entry ( VHDL/Verilog/ABEL ), and schematic entry tools. I >believe that the student version ( schematic / ABEL entry ) is only $100, and >will soon be upgradable to VHDL/Verilog for free in their 1.5 release coming >soon. Of course the student version is limited to synthesis of Xilinx parts less >than 10,000 gates, but I believe the (1.5) upgrade will include larger device >size and support for Altera and Lucent parts. Check out their web page >http://www.xilinx.com Hope this helps. > >Scott Campbell >sjcampbe@ece.ucdavis.edu >University of California Davis > >Reza Bohrani wrote: > >> Which pc-based synthesis tool is the best on the market. I have heard of >> Synplify; is that a good tool or should I go for Leonardo or are there other >> good tools?Article: 11619
tigerz@my-dejanews.com wrote in message <6s1rpl$3p4$1@nnrp1.dejanews.com>... >Hello, > >I would like to learn about CPLD/FPGA programing. I found that there are two >low cost tool from Cypress and Xilinx for $95. Can someone give me suggestion >on which one to purchase and why? Any other low cost alternative? Thanks in >advance for feedback. The replies you've gotten so far suggest free software. This is certainly worth a look, given the price, but as far as I'm aware the only thing you'll get for free is schematic entry, ABEL entry, and JEDEC-style simulation. This stuff is fun to play with, but if you have ambitions of doing serious FPGA designs, you'd be much better off spending that $100. The $100 you send to Xilinx gets you the most compehensive package. You get full simulation capabilities, schematic entry, state machine entry, and logic synthesis for all of Xilinx's CPLDs and the more modest FPGAs. The Xilinx university software is version 1.3, which is unfortunate since version 1.4 and above include VHDL synthesis as well. It is legal to install 1.4 with your 1.3 license but you need to find someone with the CDs. Xilinx is going to upgrade all those student copies to 1.5 later this year. They're probably holding off on upgrading people to 1.4 because the VHDL synthesis (provided by Synopsys' FPGA Express) isn't very well integrated into the design flow yet. The $100 you send to Cypress will get you a VHDL synthesis package today. And a decent book. And, um... that's about it. There is a JEDEC-style simulator, but it's not really all that useful. The book is useful for learning VHDL, whereas the book that comes with the Xilinx stuff is more an introduction to digital design techniques that happens use ABEL (primarily) as strictly a means to an end. The guy who wrote that book has translated all of the ABEL into VHDL lately and has it on his web site, however. Note that the Cypress $100 package includes software updates for awhile; you can get the same package without the update service for <$50 at a bookstore; the software is bundled with the book. If you have to get just one, in general I'd go for the Xilinx software. You'll get VHDL no later than the end of this year. If you really want to learn VHDL, the Cypress package is better to start with. One other thing I can think to mention is that Xilinx doesn't support the student editions of the software; you're supposed to go bug your instructor (which is unfortunate in a way, since not all instructors know much more about the software than the students do...). On the other hand, Cypress doesn't designate their software as a student version, so while you're pretty much stuck e-mailing them, you can get them to answer any question you have. ---Joel KolstadArticle: 11620
I've been trying to use busses in Foundation's schematic capture program (Active CAD "lite" by Aldec, I believe), and I have a question. If I instantiate some component for the Xilinx unified component libraries, say, a 16 bit counter, and I only want to look at the lower 12 bits of that output, how can I do it using a bus? If I just declare a bus with 12 members and connect it to the counter's output, it'll connect Output[15:4] to my Bus[11:0] given the "left to right" bus pin to bus terminal mapping rule that the on-line help discusses. (I wish their on-line help gave a few more examples... they seem to give examples of exactly the kind of connection I'm trying NOT to make...) I'm thinking I need to label the bus connection as something like garbage[3:0],Bus[11:0] so that the upper four bits just head off into hyperspace and I get the 12 lowermost bits that I wanted. However, I'm sure the software is going to complain if I start creating these orphan nets with names like "garbade". Any ideas? The Xilinx help line guy was trying to tell me that you have to take that bus and draw 12 bus taps to get what I want, but I had a real hard time believing that. He's still checking into it however. Thanks a lot. ---Joel Kolstad Joel.Kolstad@USA.NetArticle: 11621
Hi Detlef, Your main problem is the big dividing ratio (10,000). For these dividing ratios you need reference clock with very, very low jitter. Your output jitter will be crystal oscillator jitter multiplied with 10,000 plus additional jitter of the VCO itself. Try to keep the VCO power supply as clean as possible. Regards, Mirko Justen wrote in message <35E3A88B.7FEBCF88@studm.hrz.uni-siegen.de>... >Hi, > >I would like to build a digital PLL with a XILINX 4010 FPGA. The >generated frequency should be 9.5MHz to 10MHz in steps of 1kHz. So I >have to use a 1kHz Referenz (this is also the frequenzy of the >PhaseDetector) and multiply it with 9.500 to 10.000. At the moment, I >use a PhaseDetector von Xilinx AppNote and an Integrator to close the >loop. But this isn't a good design. The PhaseJitter is to big. Could >anybody help me to improve this design? > >Thanks Detlef > >------------------------------------------------------------------ >Dipl.-Ing. D.Justen ______| _____| ___| ___| >Center for Sensor Systems (ZESS) __| __| __| __| >Paul-Bonatz-Str. 9-11 __| ____| __| __| >57074 Siegen __| __| __| __| >Germany ______| _____| ____| ____| >Tel.: ++49271/ 740-2432 >Fax.: ++49271/ 740-2336 >E-Mail: justen@zess.uni-siegen.de >Homepage: http://www.zess.uni-siegen.de >---------------------------------------------------------------- > >Article: 11622
> (on constant vector dot variable vector product logic minimization)I don't think my > posts have been much help, you'veput a lot more thought into the topic than I have > time to. > Thanks for an interesting topic though, its been fun > thinking about. > > You might consider extending your investigations > into (constant matrix)(variable vector). Depending > on the nature of the constants, there may be many > redundancies in the calculations that can be exploited > to produce simpler circuits. > > May I ask if the S/W you're putting together is for > your own use, a commercial product, or are you > willing to share? > > - John Everyone posting have helped me out a lot, and the discussions have been great as well. I think there is a lot more one can do with these ideas, and as John noted, matrix products can be minimized quite well -- the more coefficients, the more constant subexpressions, the more minimization. As shown by Feig and Winograd ("Fast Algorithms for the Discrete Cosine Transform"), the 8x8 matrix of DCT coefficients can be reduced to two 4x4 matricies. This was exploited by Woods (working w/ the 6200) and Bergmann (w/ the 4000). Both authors implemented the dual 4x4 transforms using Distributed Arithmetic. Feig used modulo addition/multiplications to illustrate how pairs of variable terms {x(i) + x(k-1-i)} and {x(i) - x(k-1-i)} can be grouped into subexpressions. The minimization approach discussed here does the same thing, but discovers the subexpressions in a "brute force" manner. Moreover, the truncation caused by fixed point arithmetic introduces further subexpressions that did not exist in the original cosine terms. I am submitting a paper discussing this topic to FPGA'99, so you're welcome to that, which outlines the algorithms used. I will have to check with those in power (my advisor(s), DARPA, etc.) about making the source available! Thanks again, Dan Benyamin benyamin@ucla.edu -----== Posted via Deja News, The Leader in Internet Discussion ==----- http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member ForumArticle: 11623
One way to lower the jitter is to increase the 1KHz reference. If you still want the 1KHz resolution your options are: 1. Fractional-N techniques (this requires a DAC). Basicaly, a digital phase accumulator is used to inject a saw-tooth like waveform into the PLL loop. A technique introduced by HP (I think) about twenty years ago. 2. Divide both the VCO output and the reference with variable (and much smaller) numbers. For example (I am assuming a 10MHz reference frequency): - for 9.5MHz use 19/20 instead of 9500/10000 - for 9.673MHz use 148/153 instead of 9673/10000 and you will get 9.673203MHz, etc. Of course there will be frequencies that cannot be approximated by a fraction of small numbers, for example 9.501MHz. For these just use a second (or even a third) reference frequency which is not in a small harmonic ratio to 10MHz. The ratios can be easily computed on the fly with a modified Euclid algorithm (which can also be implemented in an FPGA) or they can be precomputed and stored in a ROM table (which might also fit in an FPGA). 3. Look at the Cypress part ICD2061A (www.cypress.com). It might do exactly what you want. It is a programmable dual clock PLL generator used in graphics applications. It has a very wide range and I believe low jitter. Doesn't require any external components and is programmed trough a simple serial interface. Catalin Baetoniu Mirko Kovacevic wrote: > Hi Detlef, > > Your main problem is the big dividing ratio (10,000). For these dividing > ratios you need reference clock with very, very low jitter. Your output > jitter will be crystal oscillator jitter multiplied with 10,000 plus > additional jitter of the VCO itself. Try to keep the VCO power supply as > clean as possible. > > Regards, > MirkoArticle: 11624
Try http://miles.winlink.no/~tom/hwb/co_ISA_Tech.html APS <resp@associatedpro.com> wrote in article <35DECDFF.FA39A3D0@associatedpro.com>... > check at http://www.associatedpro.com > > G. R. Jeffrey wrote: > > > I am new to the fpga world. I have some Altera flex10k devices that I > > want to use to implement a routing protocol. > > > > Does anyone know where I can get some (simple) example codes that use > > the ISA bus to interact with fpga/cpld devices. I spent all of > > yesterday, unsuccessfully, trying to write/read to/from ram on a device. > > > > Also, is there a web site that explains the ISA bus signals? > > > > Thanks, > > > > Gersham Jeffrey. > > > > -- > __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ > > Richard Schwarz, President EDA & Engineering Tools > Associated Professional Systems (APS) http://www.associatedpro.com > 3003 Latrobe Court richard@associatedpro.com > Abingdon, Maryland 21009 > Phone: 410.569.5897 Fax:410.661.2760 > > __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ > > >
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