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
In article <9j6ukm$73u$1@nh.pace.co.uk>, Brendan Lynskey <brendan.lynskey@pace.co.uk> writes >I thought that SystemC was intended only as a descriptive language - not as >a language for synthesis. I suspect that Synopsys (CoCentric SystemC Compiler), Frontier Design (A|RT Designer) and others would disagree. OTOH maybe one should ask the customers :-) -- Jonathan Bromley DOULOS Ltd. Church Hatch, 22 Market Place, Ringwood, Hampshire BH24 1AW, United Kingdom Tel: +44 1425 471223 Email: jonathan.bromley@doulos.com Fax: +44 1425 471573 Web: http://www.doulos.com ********************************** ** Developing design know-how ** ********************************** This e-mail and any attachments are confidential and Doulos Ltd. reserves all rights of privilege in respect thereof. It is intended for the use of the addressee only. If you are not the intended recipient please delete it from your system, any use, disclosure, or copying of this document is unauthorised. The contents of this message may contain personal views which are not the views of Doulos Ltd., unless specifically stated.Article: 33226
"Stéphane" <sjulhes@free.fr> wrote in message news:Irw57.158$Q34.447547@nnrp3.proxad.net... > My question is : > Are the APEX's LVDS I/O directly compatible with such signals ? > The main problem is the common mode which seems to be to small. I haven't used the Apex LVDS yet, but I will offer some advice on LVDS in general. Look at the differential signal when evaluating the quality of the overall signal. The common-mode signal will be nearly invisible to the receiver. This is the advantage of using differential signals. As someone else suggested, you may also want to check the cable. In the past, I've used a cable from Gore called "twin-ax" for carrying LVDS signals. The cable is like coax, but with two conductors carried within a single elliptical cross-section shield, instead of a single conductor with a circular cross-section shield. Differential signals are really only useful if they are treated as such. If the individual signals see very different skew and noise, the benefits of using differential signalling go away. Cheers, JamieArticle: 33228
Hmmm, whatdoyaknow.... constant seems to work just fine. Thanks. The assignmet I used does work in the silicon using MAX+PLUS II, Quartus, or Leonardo and targetting Apex, Acex, MAX, or FLEX devices. That's the limit of my experience so that's all I can say. Thanks again, Tim cyber_spook <pjc@cyberspook.freeserve.co.uk> wrote in message news:<3B4B70A6.3D4F3FE1@cyberspook.freeserve.co.uk>... > Tim O'Connell wrote: > > > I would like to know also. I've never been able to get a constant > > signal that way. If I were to code it I would use the lines > > Use the CONSTANT keyword. > > > > > > > SA <= "11110000"; > > SB <= "00001111"; > > > > instead of the initialization in the signal declaration. This will > > work. I don't know why the initialization doesn't work though. > > Correct me if I'm wrong (and they will). When this is in slicon the initialisation dose not happen - > this is for simulation. You should always code your logic with constants like this. > > >Article: 33229
"Stéphane" wrote: > The fact is that signal at my side of the cable are very perturbed, I don't > know exactly what they look like but I was told there is a lot of, skew, > jitter, and voltages are disturbed. The clock is also transmitted on the > cable in LVDS protocol and I have to get the FPGA's clock out of it. Skew for a 40MHz signal shouldn't be a problem since skew should not amount to a significant part of the 25nS cycle time. What *is* a problem is if the LVDS pairs have skew within the two wires! LVDS is a differential signalling standard and MUST use differential pairs (two wires twisted together over their length before being packaged with other pairs and jacketed). Twinax probably isn't required for 40MHz but is helpful in getting cleaner signals. The twisted pairs should have 100 ohms differential impedance - mismatches at the transmitter and/or receiver can cause problems. For 30m of Cat5 cable, the tables in the National Semiconductor "LVDS Owner's Manual, 2nd Edition" it appears you'll have 20% jitter at 100MHz. 40MHz should be cleaner. As far as I'm aware, there is no "protocol" to the LVDS signalling standard. The clock may be embedded in an 8b/10b format, for instance, but this is design dependent. If the clock is coming over as an LVDS clock, this should be able to run natively in the Altera devices.Article: 33230
Hi, It's me again. I am using Xilinx WebPACK, and I get this error when trying to 'implement' my design: Checking expanded design ... ERROR:NgdBuild:455 - logical net 'buildclock9600_cnt_counterno0_count_net_sig_inv_updown' has multiple drivers My questions are: 1) Does it mean that I have something like this? z <= x; z <= y; 2) How may I know wich are theese multiple drivers? Thank you verry much in advance, GonzaloArticle: 33232
I am working on a board with a virtex xcv100. I am hoping to locate a student edition of foundation series that is capable of compiling for a virtex xcv100. If anyone has used the student version for a virtex xcv100 please reply.Article: 33234
Wolfgang Loewer wrote: > > In the MAX+plus II simulator you can display a waveform for every single > word of all memories used. So in your case you would just select the first > memory word that you expect the statemachine to write to and watch it during > simulation... Thanks WL, that worked too:) BTW, i can't find a library function for a simple RS (reset-set) flip-flop (to use as a sticky flag). Do these need to be made manually from a D flip-flop or AND gates? --RussellArticle: 33235
In AHDL, cast the SR (or RS) flop (in VARIABLES section) as SRFF or SRFFE. -- Bob Elkind Russell Shaw wrote: > ... > BTW, i can't find a library function for a simple RS (reset-set) > flip-flop (to use as a sticky flag). Do these need to be made > manually from a D flip-flop or AND gates? > > --RussellArticle: 33236
ramshankar <rramshankar@excite.com> writes: > I am working on a board with a virtex xcv100. I am hoping to locate a > student edition of foundation series that is capable of compiling for > a virtex xcv100. If anyone has used the student version for a virtex > xcv100 please reply. If you don't need exact timing, use WebPACK, and target the Spartan II XC2S100. It uses exactly the same configuration bitstream as the XCV100.Article: 33237
"Brendan Lynskey" <brendan.lynskey@pace.co.uk> writes: >I thought that SystemC was intended only as a descriptive language - not as >a language for synthesis. I recently attended a presentation where it was claimed that it is possible to synthesise the HW parts of the system from SystemC. Unfortunately, the presentation was quite boring, causing me to doze off in the middle, which means that I have no idea how mature the support for synthesis is. My reaction to SystemC was slightly negative, probably because I find C++ ugly, but also because the language/library/whatever you like to call it, seems to be "unfinished". I got the impression that the language/library/whatever you like to call it was "unfinished", in a way similar to when the Java hype started. (Libraries that changed, functions that were deprecated between releases, and so on.) I'm very interested in hearing about some real-life experience with SystemC. I know that there are people that are very enthusiastic about it, so it could be that I'm to stubborn to realize that it in fact is a good thing. /w. -- Rädda viltet, förbjud O-ringen!Article: 33238
I started using Altera 7032S and 7064S in fire safety systems (for tanks and related weapons) about 3 years ago with huge success. The electronics gets exposed to high temperatures, enormous shock and vibration and the box is situated next to a cable sometime carriying 700Ampere through it. For hazardous environments I would not use a FPGA for fear that the device might not be configured correctly. The one thing you can do is to use a couple of the lines as test outputs. E.g. 8 wires in a 1 0 1 0 1 0 1 0 output after config. Use something like a 74LS688 (yeah I know, ancient) to compare the pattern after config or during operation. If you get something else then reconfig. ...but you can still get the right value even after the device goes corrupt. 2nd option is to implement a watch dog timer (MAX690/1/2/3). The FPGA must contain a counter plus output. When the device is configured correctly the FPGA will strobe the the WDT. When the stobe output stops the WDT will reset the FPGA (reconfig). 3rd option. Use many of the same circuits and then check if they output the same values. A client of mine uses 4 CPLDs calculating the same outputs. The final 4 outputs are fed into a 4 input hardcoded AND circuit (relays or qualified solid state). If the any one of the outputs differ the output is not implemented. Remember to check for how long your system can tolerate a faulty FPGA. In medical equipment no errors at all is preferred. When switching a relay output the time is less than the reaction time of the relay. In you airborne app. how much "faulty" time must pass before the aircraft start behaving incorrectly? I can write a whole book about what to look for. You must just ask the right questions and then you might even find a place for a buggy PIC microcontroller. Victor "Jon Harrison" <jon.harrison@gecm.com> wrote in message news:3b5575d0@pull.gecm.com... > Hi, > > We are starting out on a new design for an airborne application, which will > employ a large numer of Virtex-E/II FPGAs. The issue of safety > classification has raised it's head, and whether the volatility of SRAM > FPGAs constitutes an intrinsic hazard. Conventionaly with SRAM devices which > hold program information in microprocessor based systems this is addressed > by the use of continuous memory testing and checksumming. > > Has anyone else crossed this thorny issue yet ? Guidance as to any > procedures / processes followed would be useful. > > We would propose to address this issue by using device readback to allow > checksumming of the configuration data - has this been found to be > acceptable eleswhere ? > > Cheers, > > Jon > > > >Article: 33239
Hi, currently, synthesis support for SystemC such as Synopsys CoCentric is available. However, the SystemC subset supported is basically equivalent to that of VHDL accepted by Synopsys Behavioural Compiler . So I don't see any advantage to code in SystemC for synthesis, esp. if there is no standardized SystemC synthesis subset. You can stick to VHDL for synthesizable models and wait for model converter tools to SystemC which will surely appear as soon as the SystemC language got standardized. -- robert Vivian schrieb: > Hi, > Does anyone have experience with systemC synthesis tools for FPGAs. > > What works best ? > > Viv -- ------------------------------------------------------------------------------ Dipl.-Ing. Robert Siegmund email: rsie@infotech.tu-chemnitz.de Chemnitz University of Technology Dpt. of Systems and Circuit Design www: http://www.tu-chemnitz.de/~rsie Chemnitz,Germany ------------------------------------------------------------------------------Article: 33240
Is there a posibility to watch all signals (watch *) in the Foundation Timing Simulator? I haven't found such a command yet? My intended workaround is to extract the signals from the design files and place a watch for every single signal. But in what design files can I find the signals? Do I have to analyze the netlist? That would be very laborious for me. MichaelArticle: 33241
Hi, I assume you use VHDL... Gonzalo Arana wrote: > Hi, > > It's me again. > I am using Xilinx WebPACK, and I get this error when trying to > 'implement' my design: > > Checking expanded design ... > ERROR:NgdBuild:455 - logical net > 'buildclock9600_cnt_counterno0_count_net_sig_inv_updown' has multiple > drivers > > My questions are: > > 1) Does it mean that I have something like this? > z <= x; > z <= y; The answer is yes, if these are "concurrent signal assignments" (= signal assignments which are not located within a process or subprogram). Roughly spoken: each process (or concurrent signal assignment) which includes at least one assignment statement that targets a specific signal forms a driver for this signal. Instantiated components may also form a source if the corresponding signal it is connected to a component port of mode "out", "inout" or "buffer". > > > 2) How may I know wich are theese multiple drivers? If your tool does not give you this information then you have to check out the source code by yourself. Try to find all assignments that target the corresponding signal and also search for all instantiated component ports that build a source for this signal. See http://www.vhdl.org/comp.lang.vhdl/FAQ1.html#drivers for further information. -- EdwinArticle: 33242
In article <3B549BE4.5AD1A371@xess.com>, Dave Vanden Bout <devb@xess.com> writes >You might try looking at a few chapters of our online text at >http://www.xess.com/pragmatic-2_1.html. Excellent work, Dave, I just wish you'd published this three years ago :-) Cheers -- Keith WoottenArticle: 33243
Hi, I found the SRFF library component. Unfortunately, its useless because it needs a clock. I have a state machine controlled case statement that sets a flag: srff.s=vcc. However, srff.q only changes at the next clock edge which is also when the transition into the next state case-statement happens. But this is too late for another flip-flop which clocks in that flag to do something else. A clockless RS flip-flop wouldn't have that problem. Are asynchronous things like that bad for some reason? I could fix it by inserting extra states, but that just makes things slower. Maybe things like flags should work on a x2 or x4 clock? bob elkind wrote: > > In AHDL, cast the SR (or RS) flop (in VARIABLES section) as SRFF or SRFFE. > > -- Bob Elkind > > Russell Shaw wrote: > > > ... > > BTW, i can't find a library function for a simple RS (reset-set) > > flip-flop (to use as a sticky flag). Do these need to be made > > manually from a D flip-flop or AND gates? > > > > --Russell -- ___ ___ / /\ / /\ / /__\ / /\/\ /__/ / Russell Shaw, B.Eng, M.Eng(Research) /__/\/\/ \ \ / Victoria, Australia, Down-Under \ \/\/ \__\/ \__\/Article: 33244
On Thu, 19 Jul 2001 22:36:42 -0300, Gonzalo Arana <gonzaloa@sinectis.com.ar> wrote: >Hi, > >It's me again. >I am using Xilinx WebPACK, and I get this error when trying to >'implement' my design: > >Checking expanded design ... >ERROR:NgdBuild:455 - logical net > 'buildclock9600_cnt_counterno0_count_net_sig_inv_updown' has multiple >drivers > >My questions are: > >1) Does it mean that I have something like this? > z <= x; > z <= y; Probably... >2) How may I know wich are theese multiple drivers? In ModelSim (Xilinx offer a freebie version of this tool) there is a useful command to list all drivers on any particular signal. - BrianArticle: 33245
Hello. The problem I am facing is that when I try to readback Spartan XL device using Multilinx cable, Hardware Debugger informs: Total of 1299 bits mismatched but device is working properly. Number of mismatched bits is not constant . I have no such problem when I'm performing readback on Spartan device. Thanks in advance. Piotr ForytArticle: 33246
Good Morning, I'm Producing a QPSK modulator using an NCO, I use the classical schema including 2 branches, each with a Polyphase SquareRootRaisedCosine Filter whose output go to a multiplier having on the other input the output of the NCO , the two branches then come to a adder, my question is about the number of bits in the different part of the circuit, for example : 1) How much bits I've to use for the coefficients of the polyphase filter ?? 2) the data arriving to the polyphase are 1 and -1 so the multiply is reduced to invert the coefficient or not, how much bits I've to consider at the output providing the coefficient are expressed in 2'complement and are -0.0103914189205138260 -0.0090424537413504452 -0.0015462472598948768 0.0104572775382219900 0.0225120416117039310 0.0285826337727227570 0.0233188590342006150 0.0046304157926415250 -0.0244042584244363110 -0.0551805195177435540 -0.0751980048558142520 -0.0714382299222202640 -0.0345746562804935130 0.0373165930609706130 0.1368592502236164500 0.2481225405617600900 0.3500278381428516000 0.4215598203650190400 0.4472906424227021100 0.4215598203650190400 0.3500278381428516000 0.2481225405617600900 0.1368592502236164500 0.0373165930609706130 -0.0345746562804935130 -0.0714382299222202640 -0.0751980048558142520 -0.0551805195177435540 -0.0244042584244363110 0.0046304157926415250 0.0233188590342006150 0.0285826337727227570 0.0225120416117039310 0.0104572775382219900 -0.0015462472598948768 -0.0090424537413504452 -0.0103914189205138260 by the way how you suggest me to implement these coefficients ?? 3) If for example one input of the multiplier is 10 bits, also the other must be ten bits ?? and how much for the output ??? 4) By the way , it is important that the inputs of the multiplier have the same rate ?? 5) if for example the input of the following adder are both 10 bits, how much bits I've to provide for the output. I know many of these question could be stupid , but, how I can say, I've not the answer so if you have some answer also only at some of them I'll be really happy if you tell it to me or also if you redirect me to some resource speaking strictly about these thinghs ... Antonio D'OttavioArticle: 33247
Hi My Modelsim XE 5.3d claims Incompatible modes for port abc with port abc being an 'inout' port. Is there a special trick I do not know about? How can I simulate bidirectional ports? I am using XILINX ISE 3.1i Klaus HiltropArticle: 33248
Our client has an outstanding opportunity for a Senior Integrated Circuit Engineer. The ideal candidate will have: *5+ years of digital ASIC design experience *Experience with verilog coding for modelling and synthesis *Experience with digital simulation *Knowledge of ASIC verification methodologies ************************************************************** Opportunity is available in the Ottawa area. To email resume for this opportunity directly and in confidence, or for more information, contact: Ellen@ingeniuspeople.com ************************************** You may also submit your resume in our database at www.ingeniuspeople.com ******************************************* Ellen Ann Nichol - Staffing Consultant - Ingenius People...formerly InGenius Engineering 30 Rosemount Ave.Ottawa, Ontario K1Y 1P4 Phone: (613) 729-6400 x 241 Fax: (613) 729-6770 www.ingeniuspeople.comArticle: 33249
Brian Drummond <brian@shapes.demon.co.uk> wrote in message news:<ok3gltgatf6qcii7pcnc1ce9nctk14fhda@4ax.com>... > On Thu, 19 Jul 2001 22:36:42 -0300, Gonzalo Arana > <gonzaloa@sinectis.com.ar> wrote: > > >Hi, > > > >It's me again. > >I am using Xilinx WebPACK, and I get this error when trying to > >'implement' my design: > > > >Checking expanded design ... > >ERROR:NgdBuild:455 - logical net > > 'buildclock9600_cnt_counterno0_count_net_sig_inv_updown' has multiple > >drivers > > > >My questions are: > > > >1) Does it mean that I have something like this? > > z <= x; > > z <= y; > Probably... > > >2) How may I know wich are theese multiple drivers? > > In ModelSim (Xilinx offer a freebie version of this tool) there is a > useful command to list all drivers on any particular signal. > > - Brian This is why it is best to use std_ulogic and std_ulogic_vector where possible. That way, your simulations will catch any multiply driven signals. There are issues with some simulation tools not support ulogic though. A good rule of thumb is to use std_ulogic and std_logic_vector. Synopsys, Synplify and Leonardo certainly have no problems with this strategy.
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