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
try to check the opencores organization you ma find some one interested in this idea and to work with Jamil Khatib http://www.opencores.org Gary Watson wrote: > I've noticed that there are now a couple of hardware TCP/IP implementations > around, like the Seiko S7600, and there is a VHDL TCP/IP stack in some IP at > www.iReady.net. Does there exist a free or low cost attempt at a TCP/IP > stack which would fit in a Xilinx FPGA? I'm thinking in terms of a > system-on-a-chip web appliance, and/or a high speed NAS processor. > > -- > > Gary Watson > gary@nexsan.sex (Change dot sex to dot com to reply!!!) > Nexsan Technologies Ltd. > Derby DE21 7BF ENGLAND > http://www.nexsan.comArticle: 22501
Hi. I made a simple ATMEL EEPROM Programmer.. you can find it at: http://members.xoom.com/microele/paralelo/aplicaciones.html I hope this will help you. -- Hernán Sánchez Professor Universidad Pontificia Bolivariana Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22502
needham electronics EMP20 is supposed to do serial EEPROMs. Before you buy get a money back guarantee if it doesn't work out. I have an older version of the software for it. It programs the data in a 1701L fine, but doesn't handle the reset polarity correctly. It took them a very long time (several months) to get a patch and that patch didn't work right either. Perhaps by now they have it fixed. I haven't turned it on in over a year now. Emmanuel Lecomte wrote: > Hello, > > You could have a look to http://www.mvd-fpga.com/ -> products -> tiny prog > > Greetings, > > Emmanuel > > Rick Filipkiewicz a écrit : > > > Dan wrote: > > > > > The Data IO programmers cost $1500ish to program Xilinx SPROM. Does anyone > > > know of a more reasonably priced one ? > > > > > > Thanks Dan > > > > If you're in smallish volume why not consider the reprogrammable 1800 series > > devices from Xilinx. They are JTAG programmable. Once you've got the design > > stable you could get your SPROM distributor to program one-time ones for you. > > > > Alternatively avoid SPROMs altogether & use a standard EPROM+small CPLD to > > generate the bit stream. -- P.S. Please note the new email address and website url -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: 22503
Hello there , I am facing some problem in ASIC packaging . I could not find the appropriate forum to ask So I am asking here . I have a die 5mmx5mm after fab. of 256 pads. But the ASIC packaging house can not package this due to bonding wire length violations.It can be done in 208 pin package. So I need to cut down some pads ( keeping them unconnected and not carrying out them to pins via bonds ) Now while doing so , I need to cut some power as well as some functional pads. The result of this is uneven distribution of power pins on IC ( PQFP package ) This package has no substrate ( as with BGA ) so that I can have ring there . 1) What is general ratio of number of power pins to number of functional pins. 2) Can uneven distribution of power pins be done and still the IC make work. What I assume is any way inside there is a power ring in die (ASIC). 2) If I left some of the input PADS open , what will be the effect of this (The techonology is 0.25 micron CMOS ) 3) If I keep some power pads open , what will be the effect. 4) Can the pads be bonded to VDD , VSS inside the package. Can anyone help Waiting for your valueable reply. Please mail a cc to my email address also yang_li1@yahoo.co.uk Thanks Yang Li Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22504
Quartus ( maxplus ) does not understand the Leonardo-Spectrum's way for dealing with generics In order to get it work you must blackboxing using attribute instead of generics entity RAM_INST is port (clock, we: in bit; data : in bit_vector (7 downto 0); address: in bit_vector (7 downto 0); q: out bit_vector (7 downto 0)); end myram; architecture arch1 of RAM_INST is component RAM_INST_8X8 port (data: in bit_vector (7 downto 0); address: in bit_vector (7 downto 0); inclock, outclock, we: in bit; q: out bit_vector (7 downto 0) ); end component; -- Declare the black_box and parameters and their types attribute black_box: boolean; attribute LPM_WIDTH: integer; attribute LPM_WIDTHAD: integer; attribute LPM_TYPE: string; -- Assign values to each attribute attribute black_box of RAM_INST_8X8: component is true; attribute LPM_WIDTH of RAM_INST_8X8: component is 8; attribute LPM_WIDTHAD of RAM_INST_8X8: component is 8; -- Specify the name of the LPM function as the value of the -- LPM_TYPE attribute attribute LPM_TYPE of RAM_INST_8X8: component is "LPM_RAM_DQ"; begin -- Instantiate the LPM component u_ram: RAM_INST_8X8 port map(data, address, clock, clock, we, q); end arch1; For simulation You need to use generics --use translate off /on directive to comment out the generics for synthesis. On Tue, 9 May 2000 09:37:49 +0200, "Thorsten Bunte" <t.bunte@beckhoff.de> wrote: >Hi, > >I want to use the Megafunction SCFIFO created with Maxplus 2. >The properties are: >- width: 1Bit >- deep: 8 >- both sides are synchronized by one clock >- outputs: empty >- asynchronous clear >- show-ahead synchronous FIFO mode > >I am targeting a FLEX6k. > >After synthesis in Leonardo I get the message: >Warning, component scfifo has no visible entity binding. > >During place&route the following message appears: >Warning: Illegal memory cell name "scfifo_1_8_on_on" >Error: can't find design file "scfifo_1_8_on_on" > >If I use only Maxplus 2 everything works fine. > >Is it possible to use megafunctions created with MaxPlus 2 any how? >How can I use these functions with Leonardo? > >I hope someone can help me. > >Thanks, >Thorsten >Article: 22505
Long wrote in message ... >Anyone have any experience with the EETools Topmax universal device >programmer? I have the TopMax "Universal Programmer" w/the TM-A48 pod sitting here on my desk. >- frequency of software updates: EMP-20 and EMP-30 software last updated >3/27/200. Topmax last updated 5/4/200 with at least 3 updates within the >last two weeks or so when I started researching device programmers. I think the updates are mostly to add new device support. >- DOS and Windows support: EMP-20 supports DOS and Windows 95/98, EMP-30 >supports DOS and Windows 9X/NT4/NT2000, Topmax supports DOS and Windows >9X/NT4/NT2000. Didn't have any problems running either in a DOS box >under Windows 98 SE with ECP/EPP printer port mode setting. EMP-30 has >Windows version of software but is listed as beta with fewer devices >support than DOS version. Topmax also has Windows software version with >same revision number as DOS software (haven't tried). I'm using the Windows software on an NT box. >- software ease of use: EMP-20 requires manual look-up of devices. >Topmax can search by typing in a few letters of manufacturer and/or part >number. Topmax also supports auto device ID on some devices and testing >of memory and logic ICs. Edge goes to Topmax software. Except the software does not have online help, so figuring out what some of the features are is a bit of a challenge. Good luck trying to figure out how to load a hex-formatted file. And there's a "config option" dialog which lets you set the port speed, and it's not obvious why you'd want to change that option (but see below). And there are spelling and grammar errors EVERYWHERE. "Device signature dose not match." "The polarity of the Reset/OE pin may be toggle. Press ok to make RESET to toggle polarity." OK, that's picking nits, but I think that if they can't spell right, can they code correctly? >- support: Both Needham's and EETools are local Northern California based >companies. Did a search of the newgroups with good things said about >both companies' support. Both companies answer with a live person when >calling by phone. I've sent emails to EETools and got a response within >a few days. I haven't requested for a new device support but they both >seem willing to help. EEtools willing to help? Sure, maybe, until you have an actual problem. It turns out that we couldn't get the thing to program our parts (Atmel AT17LV010), and calling their tech support was less than useless. The woman who answers the phone must be having a bad YEAR, because another engineer called there and had the same problem. This person is arrogant and nasty and immediately assumes that the problems are the user's fault, regardless of the details. So I asked to speak to the "president of the company" and eventually spoke to someone who still couldn't answer my question. It turns out that the port speed had to be reduced to "slow," which I discovered by accident. Do they test their software? I don't really think so. Clearly, they don't understand how the end-user would actually use the product. Did I mention the lack of a help file? Oh, yeah, the 20-pin PLCC-to-DIP deadbug programming adapter shipped with Pin 1 in the wrong position. What's funny/sad is that this was the SECOND Topmax box with the same problem. They gave us an RMA for the first one, which we sent back, only to have the second one have the same problems. It costs them money. I should have sent the goddamn thing back, just on general principles. -- andy ----------------------------------------- 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: 22506
Rick Filipkiewicz wrote in message <3918666E.A9D2149B@algor.co.uk>... >For Virtex parts an LUT can be configured as a Shift register in much the same >way as it can be configured as a sync. RAM. See the SRL16xx components in the >lib guide. ah, didn't know that. neat. sorry for shootin' my mouth off. -- a ----------------------------------------- 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: 22507
For virtex, there is no need to specify a startup clock to Synplify. This is done when running bitgen. This is unlike the XC4000 series devices. I know this to be true when using the cclk or jtag tck for your startup, but I'm unsure about that when using the user clock. chad In article <V6pR4.1395$Qf.18285@news2.bora.net>, "Ben" <ejhong@future.co.kr> wrote: > I have a question on using synplify for virtex configuration. > When I implement and configure virtex design within synplify, there seems no > way of choosing a "Startup Clock". To the contrary, in Foundation, we can > select a startup clock among "User Clock", "CCLK", and "JTAG Clock". > Certainly there'll be a way of making a choice, but I cannot find it. > Will somebody tell me how? > > TIA, > > Ben > > Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22508
Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input without any problems, external pullups, etc? Mark.Article: 22509
I am interested in it too!But I can't find any information about it!Please give me some hint. cyrilw@sina.com In article <391964C5.CFCCA47D@yahoo.com>, khatib@ieee.org wrote: > try to check the opencores organization you ma find some one interested in this > idea and to work with > > Jamil Khatib > http://www.opencores.org > > Gary Watson wrote: > > > I've noticed that there are now a couple of hardware TCP/IP implementations > > around, like the Seiko S7600, and there is a VHDL TCP/IP stack in some IP at > > www.iReady.net. Does there exist a free or low cost attempt at a TCP/IP > > stack which would fit in a Xilinx FPGA? I'm thinking in terms of a > > system-on-a-chip web appliance, and/or a high speed NAS processor. > > > > -- > > > > Gary Watson > > gary@nexsan.sex (Change dot sex to dot com to reply!!!) > > Nexsan Technologies Ltd. > > Derby DE21 7BF ENGLAND > > http://www.nexsan.com > > Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22510
On Wed, 10 May 2000 16:51:25 GMT, "Mark Harvey" <mark.harvey@iol.it> wrote: >Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input >without any problems, external pullups, etc? The logic HIGH input threshold of a true CMOS receiver is ~ 0.7*Vcc, or about 3.5V if the device is run from a 5V supply. Driving such an input from a 3.3V CMOS output is unreliable. You'll need a pullup or a level translator. Bob Perlman ----------------------------------------------------- Bob Perlman Cambrian Design Works Digital Design, Signal Integrity http://www.best.com/~bobperl/cdw.htm Send e-mail replies to best<dot>com, username bobperl -----------------------------------------------------Article: 22511
This is a multi-part message in MIME format. --------------19AB8D450D392D40E173840D Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit > Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input > without any problems, external pullups, etc? > > Mark. 5V CMOS has a logic threshold around 2.5V. The Spartan XL data sheet lists the minimum output level for a logic 1 as 2.4V at a current of 4 mA. The minimum output level goes up to 90% of Vcc (3.0V) if the current is limited to 500 uA. The second case seems more like what you will encounter with CMOS input circuits, so the answer seems to be yes and you have about 0.5V of margin. But I have never actually done this. -- || Dr. Dave Van den Bout XESS Corp. (919) 387-0076 || || devb@xess.com 2608 Sweetgum Dr. (800) 549-9377 || || http://www.xess.com Apex, NC 27502 USA FAX:(919) 387-1302 || --------------19AB8D450D392D40E173840D Content-Type: text/x-vcard; charset=us-ascii; name="devb.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Dave Vanden Bout Content-Disposition: attachment; filename="devb.vcf" begin:vcard n:Vanden Bout;Dave tel;fax:(919) 387-1302 tel;work:(919) 387-0076 x-mozilla-html:FALSE url:http://www.xess.com org:XESS Corp. adr:;;2608 Sweetgum Drive;Apex;NC;27502;USA version:2.1 email;internet:devb@xess.com title:FPGA Product Manager x-mozilla-cpt:;-16464 fn:Dave Vanden Bout end:vcard --------------19AB8D450D392D40E173840D--Article: 22512
In article <hsgS4.7106$uk2.82101@news.infostrada.it>, "Mark Harvey" <mark.harvey@iol.it> wrote: > Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input > without any problems, external pullups, etc? > > Mark. > > No. For example, the minimum high level input voltage for 74AC logic (Vcc = 5.5V) is 3.85V. If you don't need speed, then use pseudo open collector outputs with pullups. If you need speed, then use a 74ACT buffer. If you need very high speed, then you will have to hunt around for some other solution. -- Greg Neff VP Engineering *Microsym* Computers Inc. greg@guesswhichwordgoeshere.com Sent via Deja.com http://www.deja.com/ Before you buy.Article: 22513
Ok, I'll bite - I am missing something here. While all of this information is quite useful (Thank you Ray), what is the real issue? You said that they need to act as pipeline registers to help break up the long travel time across the chip. Doesn't the LUT implementation do that? The signals are being registered ... correct? So obviously I am missing something. Please clue me in ... Cheers, Gary spivey@ieee.org "Ray Andraka" <ray@andraka.com> wrote in message news:39192201.728B434A@andraka.com... > Synplify infers the SRL16 or SRL16e if there is no reset term. These, as > you know, use the LUT as a 16 bit shift register in the VIrtex devices. > Put a reset term on the shift register to keep it out of the LUT. If you > don't use the reset term, connect it to a dummy signal and keep that from > getting optimized out in synthesis by putting a syn_keep attribute on the > dummy signal. > > > > William LenihanIii wrote: > > > I have some shift registers in a Xilinx Virtex design and between the > > synthesis tool (FPGA Compiler II) and the Xilinx Alliance P&R (2.1i), it > > is placing these shift registers inside the Look-up tables, not in the > > 'regular' registers in the slices/CLBs -- which is where I need them > > since they are acting as pipeline registers to help break up the long > > travel time from one side of the chip to the other -- and forcing them > > inside a "SRL16" of one CLB isn't going to do that. > > > > Is there a way of coaxing the synthesis and/or P&R tool to put shift > > registers in a resource of the designers' choosing (without manual > > instantiation of SRL16's vs. FDCE's)? > > > > -- > > ======================== > > William Lenihan > > lenihan3we@earthlink.net > > ======================== >Article: 22514
Mark Harvey wrote: > Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input > without any problems, external pullups, etc? The answer is: Not reliably, not worst-case. If you try it in the lab at nominal supply voltages, it will work, but you don't want to ship it in production. The output High voltage, when driving a CMOS load is exactly at the 3.3-V supply, which of course might be as low as 3.0 V. And the 5-V supply might be at 5.25 V, and the threshold at 70% of that. It is unlikely that all these conditions come together, but we are designers, not gamblers. :-) So you need a pull-up resistor to Vcc, and you also have to make the output "open collector", otherwise the upper output transistor (~20 Ohm) will clamp the Voh to slightly above the 3.3-V Vcc. This purely resistive pull-up creates a very slow risetime, but there is a remedy: Control the Spartan tristate with an AND gate that has the internal data and the output pin as inputs. This results in an active ( low impedance = fast ) rise time, until the pin voltage reaches about 1.5 V, when it turns off the output pull-up transistor, and lets the resistor ( 470 Ohm or higher) pull all the way to 5 V. Adding some internal delay to the AND gate helps even more. So you can maintain reasonable speed, but you must use an external resistor. Peter Alfke, Xilinx Applications > >Article: 22515
On Wed, 10 May 2000 13:20:27 -0400, Dave Vanden Bout <devb@xess.com> wrote: >> Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input >> without any problems, external pullups, etc? >> >> Mark. > >5V CMOS has a logic threshold around 2.5V. The Spartan XL data sheet >lists the minimum output level for a logic 1 as 2.4V at a current of >4 mA. The minimum output level goes up to 90% of Vcc (3.0V) if the >current is limited to 500 uA. The second case seems more like what >you will encounter with CMOS input circuits, so the answer seems to >be yes and you have about 0.5V of margin. But I have never actually >done this. There's a difference between switching threshold (the 2.5V you mentioned) and the DC thresholds for logic HIGH and LOW. For true CMOS devices, the logic HIGH and LOW DC thresholds are ~0.7*Vcc and ~0.3*Vcc, respectively. The driving device must be able to drive higher than the logic HIGH threshold and lower than the logic LOW threshold to guarantee DC noise margin. Bob Perlman ----------------------------------------------------- Bob Perlman Cambrian Design Works Digital Design, Signal Integrity http://www.best.com/~bobperl/cdw.htm Send e-mail replies to best<dot>com, username bobperl -----------------------------------------------------Article: 22516
Okay, what device programmer would you recommend? Andy Peters <apeters.Nospam@nospam.noao.edu.nospam> wrote in message news:8fc25b$g0q$1@noao.edu... [SNIP of long Vent against Topmax]Article: 22517
William is trying to use multiple registers across the chip so that the delay between each of them is very small. This will allow him to use a very high clock rate. The LUT implementation puts all of the registers in one spot so that it only cuts the travel time in half at best. This is one of the many problems with using an HDL to synthesize a design. You spend less time describing a design at the logical level, but you then spend more time getting the implementation you really want (assuming you need tight control). In many designs you really don't need to specify the exact implementation since it is not important or the tools do a good job. In other cases, the tools have a very hard time giving you what you really need. In this case William has been told that if he specifies the reset behaviour, he will get the separate registers which he can then place where he wants them. But the reset code then needs to be documented as an implementation control so that a future designer does not take it out and break the design. Gary Spivey wrote: > > Ok, > I'll bite - I am missing something here. While all of this information is > quite useful (Thank you Ray), what is the real issue? You said that they > need to act as pipeline registers to help break up the long travel time > across the chip. Doesn't the LUT implementation do that? The signals are > being registered ... correct? So obviously I am missing something. Please > clue me in ... > > Cheers, > Gary > spivey@ieee.org > > "Ray Andraka" <ray@andraka.com> wrote in message > news:39192201.728B434A@andraka.com... > > Synplify infers the SRL16 or SRL16e if there is no reset term. These, as > > you know, use the LUT as a 16 bit shift register in the VIrtex devices. > > Put a reset term on the shift register to keep it out of the LUT. If you > > don't use the reset term, connect it to a dummy signal and keep that from > > getting optimized out in synthesis by putting a syn_keep attribute on the > > dummy signal. > > > > > > > > William LenihanIii wrote: > > > > > I have some shift registers in a Xilinx Virtex design and between the > > > synthesis tool (FPGA Compiler II) and the Xilinx Alliance P&R (2.1i), it > > > is placing these shift registers inside the Look-up tables, not in the > > > 'regular' registers in the slices/CLBs -- which is where I need them > > > since they are acting as pipeline registers to help break up the long > > > travel time from one side of the chip to the other -- and forcing them > > > inside a "SRL16" of one CLB isn't going to do that. > > > > > > Is there a way of coaxing the synthesis and/or P&R tool to put shift > > > registers in a resource of the designers' choosing (without manual > > > instantiation of SRL16's vs. FDCE's)? > > > > > > -- > > > ======================== > > > William Lenihan > > > lenihan3we@earthlink.net > > > ======================== > > -- Rick Collins rick.collins@XYarius.com remove the XY to email me. Arius - A Signal Processing Solutions Company Specializing in DSP and FPGA design Arius 4 King Ave Frederick, MD 21701-3110 301-682-7772 Voice 301-682-7666 FAX Internet URL http://www.arius.comArticle: 22518
This is a multi-part message in MIME format. --------------AEDE1D642C4628788CEABBAB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Bob Perlman wrote: > On Wed, 10 May 2000 13:20:27 -0400, Dave Vanden Bout <devb@xess.com> > wrote: > > >> Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input > >> without any problems, external pullups, etc? > >> > >> Mark. > > > >5V CMOS has a logic threshold around 2.5V. The Spartan XL data sheet > >lists the minimum output level for a logic 1 as 2.4V at a current of > >4 mA. The minimum output level goes up to 90% of Vcc (3.0V) if the > >current is limited to 500 uA. The second case seems more like what > >you will encounter with CMOS input circuits, so the answer seems to > >be yes and you have about 0.5V of margin. But I have never actually > >done this. > > There's a difference between switching threshold (the 2.5V you > mentioned) and the DC thresholds for logic HIGH and LOW. For true > CMOS devices, the logic HIGH and LOW DC thresholds are ~0.7*Vcc and > ~0.3*Vcc, respectively. The driving device must be able to drive > higher than the logic HIGH threshold and lower than the logic LOW > threshold to guarantee DC noise margin. Yes, I agree that you are correct. I re-read the original question and I see he asks whether it can be done "without any problems". You can probably get away with it in a lab, but not in production. > > > Bob Perlman > > ----------------------------------------------------- > Bob Perlman > Cambrian Design Works > Digital Design, Signal Integrity > http://www.best.com/~bobperl/cdw.htm > Send e-mail replies to best<dot>com, username bobperl > ----------------------------------------------------- -- || Dr. Dave Van den Bout XESS Corp. (919) 387-0076 || || devb@xess.com 2608 Sweetgum Dr. (800) 549-9377 || || http://www.xess.com Apex, NC 27502 USA FAX:(919) 387-1302 || --------------AEDE1D642C4628788CEABBAB Content-Type: text/x-vcard; charset=us-ascii; name="devb.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Dave Vanden Bout Content-Disposition: attachment; filename="devb.vcf" begin:vcard n:Vanden Bout;Dave tel;fax:(919) 387-1302 tel;work:(919) 387-0076 x-mozilla-html:FALSE url:http://www.xess.com org:XESS Corp. adr:;;2608 Sweetgum Drive;Apex;NC;27502;USA version:2.1 email;internet:devb@xess.com title:FPGA Product Manager x-mozilla-cpt:;-16464 fn:Dave Vanden Bout end:vcard --------------AEDE1D642C4628788CEABBAB--Article: 22519
I get the following error from Quartus when targetting an APEX20K. --- Error: The 7 path(s) clocked by clock |clk have clock skew larger than the data delay -- Circuit will not operate. See the fmax report panel for details. Info: Hold Time Violation between |a and |b for clock |clk (Hold Time is 29 ps) Info: + Largest Clock skew is 0.000 ns Info: + Longest Clock path from |clk to destination register is 6.500 ns Info: 1: + IC(0.000 ns) + CELL(1.400 ns) = 1.400 ns; Loc. = Unassigned; CLK Node = '|clk' Info: 2: + IC(5.000 ns) + CELL(0.100 ns) = 6.500 ns; Loc. = Unassigned; REG Node = '..' Info: - Shortest Clock path from |clk to source register is 6.500 ns Info: 1: + IC(0.000 ns) + CELL(1.400 ns) = 1.400 ns; Loc. = Unassigned; CLK Node = '|clk' Info: 2: + IC(5.000 ns) + CELL(0.100 ns) = 6.500 ns; Loc. = Unassigned; REG Node = '..' Info: - Micro clock to output delay of source is 0.100 ns Info: - Shortest register to register delay is 0.321 ns Info: 1: + IC(0.000 ns) + CELL(0.000 ns) = 0.000 ns; Loc. = Unassigned; REG Node = '..' Info: 2: + IC(0.180 ns) + CELL(0.141 ns) = 0.321 ns; Loc. = Unassigned; REG Node = '..' Info: + Micro hold delay of destination is 0.450 ns Warning: Timing requirements were not met. See Report window for details. Info: test_design: Netlist extraction and synthesis were unsuccessful. 1 error, 671 warnings. --- Now, how can the clock skew (0.000 ns) be larger then the data delay (0.321 ns)? I can't add LCELLs because the design is coming from VHDL though a 3rd party synthesis tool. I have cliqued blocks with problem paths but this hasn't helped. Any help is greatly appreciated. DonArticle: 22520
Take a look at the following: http://www.embedded-solutions.ltd.uk/tech_info_3.htm http://www.annapmicro.com/ http://www.associatedpro.com/ http://www.vcc.com/prod1.html Patrick Schulz wrote: > Hi all, > > we are targeting an PCI-based network interface ASIC prototype on VIRTEX1000(E), > but I found only very few boards which meet our requirements (PCI_64, VIRTEX1000(E)). > One of this is the DN2000k10 from the Dini Group (http://www.dinigroup.com). The problem > with this board is, that it is not compatible with the PCI LogicCore from Xilinx and I'm > not willing to design a PCI-Interface. > > Does anyone has experience with this board? > Does anyone has an idea for a appropriate board?? > > Thanks 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: +49-621-181-2713 -- ***************************** Anna M. Acevedo Xilinx University Program 2100 Logic Drive San Jose, CA 95124 PH: (408) 879-5338 FAX: (408) 879-4780 Email: anna.acevedo@xilinx.com http://www.xilinx.com/programs/univ.htm *****************************Article: 22521
Where can I get some info on using the partial reconfiguration (while powered up) features of the Xilinx Virtex family. I can't find much on their site. I have read that floorplanning is key as the parts only do a column at a time. I have never used the floorplanner. Is it tough? Design is in Verilog, using many RPM cores from coregen, including block RAM. Using the Foundation 2.1i. Thanks. -- Jim Patterson jpatters@stny.rr.comArticle: 22522
Hi Patrick, OK, I'll bite: If you are doing a PCI based network interface in an ASIC, wouldn't you want control over the PCI interface design too? The xilinx core is designed to run in the FPGA, not as an ASIC core. In fact, I think the license for it says something about retargeting it doesn't it? As far as the Dini Board goes, I'm using one for a current project. Its pretty good for what it is supposed to be: an ASIC prototyping platform. If your ASIC is to include the PCI interface, then you'll most likely want to be doing your own design or buying a suitable ASIC core. Besides which, a design targetted at an ASIC is probably not going to run at full speed in an FPGA. If your design doesn't include a PCI interface, then Dini Group has a simple PCI target that works just fine for interfacing to your logic in most cases. Patrick Schulz wrote: > Hi all, > > we are targeting an PCI-based network interface ASIC prototype on VIRTEX1000(E), > but I found only very few boards which meet our requirements (PCI_64, VIRTEX1000(E)). > One of this is the DN2000k10 from the Dini Group (http://www.dinigroup.com). The problem > with this board is, that it is not compatible with the PCI LogicCore from Xilinx and I'm > not willing to design a PCI-Interface. > > Does anyone has experience with this board? > Does anyone has an idea for a appropriate board?? > > Thanks 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: +49-621-181-2713 -- P.S. Please note the new email address and website url -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: 22523
Peter Alfke wrote: > Mark Harvey wrote: > > > Can a SpartanXL (3.3V) device drive a 5V CMOS (i.e. not ACT type) input > > without any problems, external pullups, etc? > > The answer is: Not reliably, not worst-case. > If you try it in the lab at nominal supply voltages, it will work, but you > don't want to ship it in production. > The output High voltage, when driving a CMOS load is exactly at the 3.3-V > supply, which of course might be as low as 3.0 V. And the 5-V supply might > be at 5.25 V, and the threshold at 70% of that. > It is unlikely that all these conditions come together, but we are > designers, not gamblers. :-) <rk applauds> ---------------------------------------------------------------------- rk The Soviets no longer were a threat stellar engineering, ltd. in space, and in the terms that stellare@erols.com.NOSPAM became commonplace among the veteran Hi-Rel Digital Systems Design ground crews, as well as the astronauts, the dreamers and builders were replaced by a new wave of NASA teams, bureaucrats who swayed with the political winds, sadly short of dreams, drive, and determination to keep forging outward beyond earth. -- Shepard and Slayton.Article: 22524
I stand corrected then. I simply was not aware of any Lucent based boards that are commercially available. I guess the marketing on those is not that great :-) As far as the gate counts go, I think tailoring the design to the architecture and floorplanning play a bigger part than the amount of memory. For example, I've got a virtex design in a 1000 that uses none of the block RAMs and is still over a million gates. There's no way that design would fit without designing it to the Virtex architecture. Advertised gate count on the devices is what I call "marketing gates" which has little bearing on the actual attained gate count. You are much better off counting LUTs to determine the size of your design, and by thinking LUTs you are likely to get to a better design anyway. Whether you are counting LUTs or gates, you need a pretty detailed concept of the design to get anywhere close to an accurate estimate. I do think the vendors would be more forthcoming by sizing the parts by number of LUTs than by "gates". At least that is not design dependent. Rickman wrote: > Hey Ray! What about the PC104C31 which uses not only one, but four > Lucent FPGA chips? I realize that our price does not fit the $300 limit > imposed in this case, but don't say there aren't any commercial boards > around that use the Lucent chip! :) > > Of course our board is not really intended to be an FPGA evaluation > board, but rather a DSP board with FPGA capability. But it is so loaded > with FPGA capability that anyone needing a lot of FPGA in multiple chips > would do well to look at it and just consider the DSP to be a fancy boot > loader. In fact the OR3T30 or OR3T55 that can be used on our board would > be a very good choice for an attached coprocessor for the TMS320C31 that > is on the board. > > I believe that APS also has a Lucent board much like their APS-X84 board > although they may not be marketing it. They sent me a board at one point > as a beta test which I never got around to using. > > > > Andy Holt wrote: > > > > > This is the sort of thing I would expect to be an FAQ, but there doesn't > > > seem to be one for this group. > > > > > > I have been thinking about "playing" with an FPGA both from the view of > > > learning about an interesting-looking technology and with the hope of > > > constructing an emulation of a '60s mainframe (more about this later). > > > > > > I am looking for advice on low-cost ways of doing this (Let's say price > > > ceiling of about £200 [$300]). It seems that I am going to need two main > > > things: > > > * A package of software. > > > * A prototyping/evaluation board. > > > > > > Taking the second of these first there seem to be few choices available > > > (without paying lots of $$$) - > > Our board is a lot of money by comparison. The single piece price I have > been quoting is between $2000 and $3000 depending on features and speed. > But this will be changing to something a little more palatable at the > end of the month. And of course the price is much better at quantity. > > ...snip... > > > As for software, the Kanda and Atmel packages come with some, for the > > > Xess one I would also have to spend another $100 for the Foundation > > > student edition. > > The PC104C31 comes with a boot monitor program to allow you to program > the Flash, boot the FPGAs and perform a lot of testing of the various > parts of the board. Full source is provided along with the designs for > the FPGAs. A low end package for the Lucent FPGAs is available for $150 > very much like the introductory package from Xilinx. This package > includes Viewlogic schematic capture and simulation tools. I also > believe it includes VHDL synthesis. > > > > ** so, first question: any known "gotcha's" with the above alternatives? > > > [ISTR a recent hint that the Atmel software was weak in one respect - > > > it is noticeable that their web site seems to say almost nothing about > > > its functionality - and the low cost version of Foundation doesn't > > > include VHDL?] > > > Are there other reasonable options? > > > > > > The other main question I have concerns estimating how big an FPGA I > > > would need for the mainframe emulation. I assume that the "usable gate" > > > counts for all devices tend to be as much marketing as technical > > > statements. I have detailed (but only "almost complete") descriptions of > > > the logic design for the mainframe that I am interested in (ICT 1905 - > > > aka FP6000) and I can be reasonably confident that it has less than 6000 > > > gates including FPU ... probably less than 4000 without. Is this likely > > > to fit in a "10000 gate" FPGA?, a "20000 gate" one, or whatever? > > Keep in mind that your gate count also needs to include the IO system > for accessing the memory and any peripherals that will be attached. This > can easily double the size of the design. > > Yes, you are right about the gate counts for the FPGAs. But the largest > part of the gate inflation comes from the inclusion of memory in the > gate counts. They make certain assumptions about the percentage of the > memory used. Your design may use more or less of that memory. This will > change the used gate count dramatically. > > -- > > Rick Collins > > rick.collins@XYarius.com > > remove the XY to email me. > > Arius - A Signal Processing Solutions Company > Specializing in DSP and FPGA design > > Arius > 4 King Ave > Frederick, MD 21701-3110 > 301-682-7772 Voice > 301-682-7666 FAX > > Internet URL http://www.arius.com -- P.S. Please note the new email address and website url -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.com
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