Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search

Messages from 40700

Article: 40700
Subject: Any data about SFI 4 interface ?
From: "Zak smith" <zakhama@sympatico.ca>
Date: Wed, 13 Mar 2002 00:25:04 -0800
Links: << >>  << T >>  << A >>
can we implement SFI-4 interface in an FPGA ???
How many gates it will take ?
Is there application notes ?

Thanx !



Article: 40701
Subject: Re: Mystery two wire interface, or am I being dense?
From: "Tauno Voipio" <tauno.voipio@iki.fi.SPAMBAIT_REMOVE>
Date: Wed, 13 Mar 2002 09:24:12 GMT
Links: << >>  << T >>  << A >>

"michael brown" <n5qmg@nospam.earthlink.net> wrote in message
news:zGwj8.94983$dj3.3473465@typhoon.austin.rr.com...
> > One of the reasons that "two wire" interfaces that are nominally IIC
> > are given a different name is that they aren't fully compatible with
> > the spec, even though they interwork in every practical way.
> >
> > Regards,
> > Allan.
>
> That may be partially true, but the primary reason that other
manufacturers
> don't call their parts IIC compatible is that Philips would make them pay
> royalties to do so.  Pretty much whenever you hear about two-wire
> interfaces, it's probably IIC compatible (more or less).
>

More or less - that's the key.

Pay special attention to start and stop conditions - they seem pretty often
messed up. It's a pity that most of the changes lose something of the
original, pretty clever protocol.

Tauno Voipio
tauno voipio @ iki fi



Article: 40702
Subject: Re: How would I know somebody has copied my files in Unix?
From: Timmestein <3218@SPAM.thrijswijk.nl>
Date: Wed, 13 Mar 2002 10:40:48 +0100
Links: << >>  << T >>  << A >>
Kelvin Hsu wrote:

> Hi,
> 
> How would I know somebody has copied my files in Unix?
> 
> --
> Best Regards,
> -----------------------------------------------------------------
> Xu Qijun
> Engineer
> OKI Techno Centre (S) Pte Ltd
> Tel: 770-7049 Fax: 779-1621
> Email: qijun@okigrp.com.sg

I guess this isn't the right group for these type of questions.
But, to answer your question: as a normal user you can't. Maybe su to root 
and view everyones history (if they didn't clear it). Or look at the logs 
of your ftp-daemon. Or maybe the system is compromised. Or ...

-- 
To reach me by email, remove the obvious from my email-address.

Article: 40703
Subject: Re: Mutual Clock Synchronization
From: allan_herriman.hates.spam@agilent.com (Allan Herriman)
Date: Wed, 13 Mar 2002 10:06:26 GMT
Links: << >>  << T >>  << A >>
On 12 Mar 2002 22:39:35 -0800, kindalin2002@yahoo.com (showbiz) wrote:

>allan_herriman.hates.spam@agilent.com (Allan Herriman) wrote in message news:<3c870132.9726435@netnews.agilent.com>...
>> On Wed, 06 Mar 2002 18:19:46 -0500, Greg Neff <gregeneff@yahoo.com>
>> wrote:
>> 
>> >On Wed, 06 Mar 2002 21:08:39 GMT, John_H <johnhandwork@mail.com>
>> >wrote:
>> >
>> >I have not looked at the CEPT standard.  We are using E1 physical
>> >layer stuff for a non-telecom application.  Does the CEPT standard
>> >talk about mutual clock synchronization?  If so, I would appreciate a
>> >pointer to this standard.  I though that E1/T1 systems were typically
>> >clocked in a master/slave arrangement, or lived with slips.
>> >
>> > With regard to stability, I was thinking about the stability of the
>> >democratic clock, due to multiple feedback paths to the multiple phase
>> >error detectors on each PLL.
>> >
>> >>The references for stability...  shouldn't they be contained in the
>> >>ITU-T specifications regarding E1 reference timing?  The clock
>> >>distribution schemes, switchover conditions on signal loss, acceptable
>> >>jitter transfer and tolerance are all part of the CEPT standard.  Very
>> >>constrained.
>> >(snip)
>> 
>> As the other posters have mentioned, a "democratic" approach to timing
>> is unusual, and (to me) doesn't make sense.
>> 
>> Regarding the ITU-T specifications, I think you should look at the
>> SONET/SDH specs instead of the E1 ones.
>> Try: ITU-T G.707 (SDH) and the G.810 series (SDH timing).
>> Also GR-253 (SONET).
>> 
>> Even the Network Time Protocol (RFC1305) that is used to synchronise
>> time of day clocks on computers has a hierarchy of timing sources.
>> But it might give you some ideas that will help you with your problem.
>> http://www.ietf.org/rfc/rfc1305.txt
>> 
>> Oh, you should probably look at the datasheets for chips that are
>> designed for phone network timing, e.g. Zarlink (nee Mitel) MT90401
>> http://products.zarlink.com/partfinder/prodprofile.cgi?device=1127
>> or maybe this one from Semtech:
>> http://www.semtech.com/products/sets.html
>> 
>> Regards,
>> Allan.
>
>I have two questions related to the timing soloutions you have
>mentioned:
>
>1)I've found a similar chip from Agere systems which is TSWC01622. But
>there is nothing more than a product brief on their website.
>http://www.agere.com/netcom/docs/PB00144.pdf
>I'll be grateful if someone provide me with more information (like a
>complete datasheet)about it.
> 
>2)According to MT90401 datasheet the two input references should have
>the same frequency. In my application I want to switch between two
>different clock rates so I can't feed both of them to the chip. After
>any failure of references MT90401 enters a holdover mode(suppose the
>failed clock frequency X). I WANNA KNOW if returning from holdover
>mode to a diffrent clock rate Y is safe and slip free. I know the
>manufacture can answer the question best but I thought you may know
>the structure of such devices.

If both inputs are a multiple of 8kHz, you could divide them both to
this frequency and then apply them to the MT90401.

If you can't manage that, you could always roll your own clock
management circuit in an FPGA.  (You may need an external analog PLL
to get the jitter within limits, but that's dependent on your
application.)

Regards,
Allan.

Article: 40704
Subject: Single-event upsets in ROM
From: "Michael Boehnel" <boehnel@iti.tu-graz.ac.at>
Date: Wed, 13 Mar 2002 02:26:47 -0800
Links: << >>  << T >>  << A >>
Hello!

I'd like to implement a ROM in a Xilinx Virtex either by using a block SelectRAM or LUTs. Is there any difference concerning SEU (bit-flip due to heavy-ions etc.) resistance? Are LUTs more resistant or is the same technologie (area, speed) used as for SelectRAM. 

Thank you for your comments.

Michael

Please send any comments to my email, too.

Article: 40705
Subject: DES implementation in Handel C
From: saupal@indiatimes.com (Saurabh Pal)
Date: 13 Mar 2002 02:29:25 -0800
Links: << >>  << T >>  << A >>
Hi !

Can anyone tell whether the source code of DES implementation in
Handel C is available publically or not.

Thanks in advance.

Saurabh

Article: 40706
Subject: EDA tools(from front to end)
From: sfjg <eftjh@alkfhv.zlskdv>
Date: Wed, 13 Mar 2002 03:41:47 -0800
Links: << >>  << T >>  << A >>
when i design a project,i always use many,many eda tools(simulation/synthesis/P&R/check...).
so i have to switch  in eda tools,and copy some files(source file/edif/edn/sdf....etc).
who have had some scripts(tcl/perl/batch/unix shell....) run from front to end so that i only run it ?that is,some switch will be finished by scripts automatic.it don't need me to intervene.in fact ,it is the same as batch(.bat) on Windows.can you share it ?

Article: 40707
Subject: Re: Pointer Processor for OC192
From: hamish@cloud.net.au
Date: 13 Mar 2002 11:56:04 GMT
Links: << >>  << T >>  << A >>
Zak smith <zakhama@sympatico.ca> wrote:
> Do you have any data how many gates it take to implement the pointer
> processor in FPGA ?
> Do you have data about this design ?

Can't you just run your OC48 design 4x faster?

Hamish
-- 
Hamish Moffatt VK3SB <hamish@debian.org> <hamish@cloud.net.au>

Article: 40708
Subject: Synthesis tools comparison?
From: "Arash Salarian" <arash.salarian@epfl.ch>
Date: Wed, 13 Mar 2002 13:52:37 +0100
Links: << >>  << T >>  << A >>
Hello,

Do you know of any recent review/benchmark comparision of major synthesis
tools for the FPGA? I'm interested in such a comparision between Synplify,
LeonardoSpectrum and FPGA compiler. And is there any major perfromance
difference between these tools when targetting different FPGAs from Xilinx
and Altera?

Best Regards
Arash



Article: 40709
Subject: Re: IBIS simulation (was Re: max frequency of obuf_lvdci_dv2_18)
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Wed, 13 Mar 2002 07:36:50 -0800
Links: << >>  << T >>  << A >>
Eric,

Try downloading the free demo from Hyperlynx.  Try before you buy.

Austin

Eric Smith wrote:

> Austin Lesea <austin.lesea@xilinx.com> writes:
> > Run an IBIS simulation.
>
> Is there any not-too-insanely-expensive software that
> can be used for this?


Article: 40710
Subject: Re: IBIS simulation (was Re: max frequency of obuf_lvdci_dv2_18)
From: Austin Lesea <austin.lesea@xilinx.com>
Date: Wed, 13 Mar 2002 07:50:30 -0800
Links: << >>  << T >>  << A >>
Eric,

To make it easier, their site has changed around:

 http://www.hyperlynx.com/democontent.html

Austin

Eric Smith wrote:

> Austin Lesea <austin.lesea@xilinx.com> writes:
> > Run an IBIS simulation.
>
> Is there any not-too-insanely-expensive software that
> can be used for this?


Article: 40711
Subject: IP cores availability
From: prashantj@usa.net (Prashant)
Date: 13 Mar 2002 08:23:44 -0800
Links: << >>  << T >>  << A >>
hi,

Would anyone know if there are IP cores available for Baseline CDMA
2000 / IS-95 handset receiver ?

Thanks,
Prashant

Article: 40712
Subject: minimum value for clock to output
From: nahum_barnea@yahoo.com (Nahum Barnea)
Date: 13 Mar 2002 08:28:41 -0800
Links: << >>  << T >>  << A >>
Hi.
Xilinx tradionally publish only max values for the clock to out
delays.

I remember that there is an application note that say that the min clock 
to out value can be taken as 1/3 'rd of the max value (or 1/4 'th if
you are a conservative person).

Can someone point me to that application note ?
I want to verify it.

ThankX,
Nahum.

Article: 40713
Subject: Re: How do I infer a carry-chain parity generator in XST?
From: John_H <johnhandwork@mail.com>
Date: Wed, 13 Mar 2002 16:37:08 GMT
Links: << >>  << T >>  << A >>
As far as I know - you can't implement an XOR carry chain in the Xilinx
devices (that's the scope of XST, right?  Xilinx only) though Altera carry
structers allow the greater flexibility.  To do parity, you need to XOR the
result from the previous stages' xor chain;  this requires the ability to
select either the carry or the invert of the carry.  The carry invert isn't
available without going external to the carry chain and totally blowing all
performance gains.

Maybe it's a good thing XST didn't infer the parity generator with carry
chains!


Kevin Brace wrote:

>         I will like to know if there is a way to infer a parity
> generator that uses Virtex's carry-chains rather than LUTs?
> The reason I am asking this question is because when I synthesize a
> parity generator, XST uses LUTs rather than carry-chains even if "XOR
> Collapsing" option is checked (enabled).
> I will like to conserve valuable LUTs, and use carry-chains because
> carry-chains are rarely used in my design.
> Will I have to use Virtex specific primitives to do so, or is there a
> way to do so without using Virtex specific primitives?
> IF I have to use Virtex specific primitives, how do I do it?
> Or is this problem XST's fault, and do other synthesis tools handle XOR
> stuff differently than XST?
> The parity generator here I am talking about is a one that generates the
> parity of 36-inputs for PCI bus.
>
> Thanks,
>
> Kevin Brace (Don't respond to me directly, respond within the
> newsgroup.)


Article: 40714
Subject: Re: How to Align 7x DDR Data Input to a XC2V6000-5?
From: John_H <johnhandwork@mail.com>
Date: Wed, 13 Mar 2002 16:58:53 GMT
Links: << >>  << T >>  << A >>
                 ______        ______        ______
clk10x         _/      \______/      \______/
               __ _____________ _____________ _____
toggle         __X_____________X_____________X_____
                _   _   _   _   _   _   _   _   _
clk35x_shift  _/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_/ \_
              ______ _______________ ___________ __
togp          ______X_______________X___________X__
              ________ ___________ _______________
togn          ________X___________X_______________X
                     _             _             _
tog_diff      ______/ \___________/ \___________/ \
                                     ___
samplep       ______________________/   \__________
                       ___
samplen       ________/   \_______________________/
               _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
ddr_data      X6X0X1X2X3X4X5X6X0X1X2X3X4X5X6X0X1X2X

You're right about the toggle period.
Though it's tough to show prop delays (the combinatorial tog_diff doesn't show a delay in
the diagram) it seems that the pulse that covers timeslot 3 and 4 in this phase shift
version of clk35x is independent of which clk35x transition you use to sample the
difference.  You can apply the same timing diagram to the clk10x90deg signal with an
unshifted clk35x and get similar results but the timing analysis is up to you.  Fun stuff!


Mark wrote:

> I think that with the toggling signals, they have a period of twice the clock period.
> It seems when using clk35x_p and clk35x_n, toggle35x_p and toggle35x_n can only resolve
> down to one clk35x resolution.  The 90 degree phase shift sounds like one way to find a
> difference.  I believe that I once saw an appnote that described using 90 deg phases to
> lock(?) onto an input signal.
>
> Now that you mention it, I do remember reading messages about clock skew and jitter in
> FPGAs.  Also, several weeks ago, I looked into the Virtex-II skew/jitter when
> multiplying up the clock by 7x to the rate of 420 MHz.


Article: 40715
Subject: Universal FPGA Programmer
From: "Dionissis Efstathiou" <eyden@mhl.tuc.gr>
Date: Wed, 13 Mar 2002 20:46:34 +0200
Links: << >>  << T >>  << A >>
Hi,

I'm trying to find if there is a low cost Universal programmer for FPGAs.
(Not exclusively using JTAG).
If someone could help me i would be much obliged.

Thanks in advance.



Article: 40716
Subject: Re: Universal FPGA Programmer
From: "Falk Brunner" <Falk.Brunner@gmx.de>
Date: Wed, 13 Mar 2002 21:08:30 +0100
Links: << >>  << T >>  << A >>
"Dionissis Efstathiou" <eyden@mhl.tuc.gr> schrieb im Newsbeitrag
news:a6o6p7$q5f$1@ulysses.noc.ntua.gr...
> Hi,
>
> I'm trying to find if there is a low cost Universal programmer for FPGAs.
> (Not exclusively using JTAG).
> If someone could help me i would be much obliged.

Look at the Xilinx homepage->support->hardware.

There you can find a schematic for the parallel-III cable, which allows you
to do JTAG and Serial Slave downloads.

--
MfG
Falk





Article: 40717
Subject: Re: digital video PLL
From: Frank Vorstenbosch <frank-spam@kingswood-consulting-spam.co.uk>
Date: Wed, 13 Mar 2002 20:26:24 +0000
Links: << >>  << T >>  << A >>
John_H wrote:

> There are techniques to develop a stable average frequency from a high
> frequency reference using the 64us pulse as a correction.  I'd suspect,
> however, that the jitter information that would be inherent in the DPLL
> could form visual patterns that detract from your true needes.
> 
> In my opinion, good video needs good analog elements in key places like
> this:  timing.

Nonsense.  Look at video front-end (digitizer) chips for examples of
all-digital video decoding.  The Philips SAA7110 has been around for
aeons, and the Bt848/878 are very common in PCI cards.  Both use all
digital circuitry to lock to the syncs and demodulate the colour info
etc.

Frank

-- 
------------------------------------------------------------------------
Frank A. Vorstenbosch                               <SPAM_ACCEPT="NONE">
Wimbledon, London SW19        frank-spam@kingswood-consulting-spam.co.uk


Article: 40718
Subject: Re: digital video PLL
From: John_H <johnhandwork@mail.com>
Date: Wed, 13 Mar 2002 21:12:18 GMT
Links: << >>  << T >>  << A >>
>From the SAA7110 datasheet:

15 CLOCK SYSTEM
15.1 Clock generation circuit
The internal CGC generates the system clocks LLC, LLC2
and the clock reference signal CREF. The internally
generated LFCO (triangular waveform) is multiplied by
four via the analog PLL (including phase detector, loop
filter, VCO and frequency divider). The rectangular output
signals have a 50% duty factor.


Thanks for giving a fellow engineer the benefit of the doubt.  I could swear
that words like "triangular" and "analog" aren't quite digital elements.


Frank Vorstenbosch wrote:

> John_H wrote:
>
> > There are techniques to develop a stable average frequency from a high
> > frequency reference using the 64us pulse as a correction.  I'd suspect,
> > however, that the jitter information that would be inherent in the DPLL
> > could form visual patterns that detract from your true needes.
> >
> > In my opinion, good video needs good analog elements in key places like
> > this:  timing.
>
> Nonsense.  Look at video front-end (digitizer) chips for examples of
> all-digital video decoding.  The Philips SAA7110 has been around for
> aeons, and the Bt848/878 are very common in PCI cards.  Both use all
> digital circuitry to lock to the syncs and demodulate the colour info
> etc.
>
> Frank
>
> --
> ------------------------------------------------------------------------
> Frank A. Vorstenbosch                               <SPAM_ACCEPT="NONE">
> Wimbledon, London SW19        frank-spam@kingswood-consulting-spam.co.uk


Article: 40719
Subject: Re: digital video PLL
From: Frank Vorstenbosch <frank-spam@kingswood-consulting-spam.co.uk>
Date: Wed, 13 Mar 2002 21:32:16 +0000
Links: << >>  << T >>  << A >>
John_H wrote:

> From the SAA7110 datasheet:
> 
> 15 CLOCK SYSTEM
> 15.1 Clock generation circuit
> The internal CGC generates the system clocks LLC, LLC2
> and the clock reference signal CREF. The internally
> generated LFCO (triangular waveform) is multiplied by
> four via the analog PLL (including phase detector, loop
> filter, VCO and frequency divider). The rectangular output
> signals have a 50% duty factor.
> 
> 
> Thanks for giving a fellow engineer the benefit of the doubt.  I could swear
> that words like "triangular" and "analog" aren't quite digital elements.


Oops!  The Bt8x8's "UltraLock" is alleged to be all digital, though.


Frank

-- 
------------------------------------------------------------------------
Frank A. Vorstenbosch                               <SPAM_ACCEPT="NONE">
Wimbledon, London SW19        frank-spam@kingswood-consulting-spam.co.uk


Article: 40720
Subject: Is XST's Keep Hierarchy option broken?
From: Kevin Brace <ihatespam99kevinbraceusenet@ihatespam99hotmail.com>
Date: Wed, 13 Mar 2002 16:08:28 -0600
Links: << >>  << T >>  << A >>
        I have seen other people using XST saying when Keep Hierarchy
option is checked, and if a tri-state buffer that ultimately connects to
a pin is located anywhere other than the top module (like in
instantiated submodules), things go wrong during MAP.
To verify if this happens to me, I synthesized my design (a PCI IP core
) which has tri-state buffers two submodules below the top module with
this Keep Hierarchy checked.
The result is that the synthesized design which normally has lots of
bi-directional ports (inout in Verilog) now has output ports instead.
Yes, I do understand that this problem can be worked around if I uncheck
Keep Hierarchy option, but since this Keep Hierarchy option exists in
XST, shouldn't my design synthesize correctly even if the Keep Hierarchy
option is checked?
From what I see this problem seems like a bug of XST, so I hope it gets
fixed in the next revision.




Kevin Brace



P.S. A Xilinx employee reading this posting, please forward this posting
to the XST development team.

Article: 40721
Subject: XST duplicates unnecessary IOB OE FFs
From: Kevin Brace <ihatespam99kevinbraceusenet@ihatespam99hotmail.com>
Date: Wed, 13 Mar 2002 17:04:37 -0600
Links: << >>  << T >>  << A >>
        I am currently experiencing a problem with XST's Pack I/O
Registers into IOBs option.
The problem I am having is that XST duplicates OE FFs when I don't want
it to do so.
In my design (a PCI IP core), to control the Output Enable (OE) of
AD[31:0] and C/BE#[3:0], I use one FF for AD[31:0] and one FF for
C/BE#[3:0].
Some people might say the way I am handling the OE will prevent me from
merging those two OE FFs into IOB, and I understand that, and that is
what I want.
Yes, if I wanted the OE FFs to be merged into IOBs, I will manually
duplicate the OE FFs myself in my design (I used to do that.), but the
design decision I made needs only one OE FF for AD[31:0], and one OE FF
for C/BE#[3:0].
However, I do still need those output FFs for AD[31:0] and C/BE#[3:0] to
be merged into IOBs, and since IOBs don't have a feedback path, those
output FFs have to be duplicated, and must reside inside CLB.
So, when I synthesize my design with Pack I/O Registers into IOBs option
set to "True," XST duplicates the output FFs for AD[31:0] and C/BE#[3:0]
correctly so that it can be merged into IOBs.
But XST also duplicates AD[31:0] OE FF 32 times, and C/BE#[3:0] OE FF 4
times.
I don't appreciate XST overriding the design trade off I made in the
design, and do I have a way to prevent XST from duplicating the OE FF?
I am using ISE WebPACK 4.1WP3.0's XST (XST E.33), and Spartan-II XC2S150
is the target device.
I feel like this OE FF duplication thing should not happen, and hope
that the future version of XST will give its users an option to disable
OE FF duplication if the user doesn't want it.



Kevin Brace



P.S. A Xilinx employee reading this posting, please forward this posting
to the XST development team.

Article: 40722
Subject: Re: Synthesis tools comparison?
From: "S. Ramirez" <sramirez@cfl.rr.com>
Date: Wed, 13 Mar 2002 23:14:47 GMT
Links: << >>  << T >>  << A >>
"Arash Salarian" <arash.salarian@epfl.ch> wrote in message
news:3c8f4cb7$1@epflnews.epfl.ch...
> Hello,
>
> Do you know of any recent review/benchmark comparision of major synthesis
> tools for the FPGA? I'm interested in such a comparision between Synplify,
> LeonardoSpectrum and FPGA compiler. And is there any major perfromance
> difference between these tools when targetting different FPGAs from Xilinx
> and Altera?
>
> Best Regards
> Arash

   You will be hard pressed to find these comparisons, since the license
agreements of these vendors say something close to "Licensee shall not ..
disclose the results of any benchmarking of the SOFTWARE, or use such
results for its own competing software development activities, without the
prior written permission of Blah Blah."
   Don't blame me, I'm just the messenger!
Simon Ramirez, Consultant
Synchronous Design, Inc.
Oviedo, FL  USA



Article: 40723
Subject: Re: nOOb: wants to start using an fpga
From: Kevin Brace <ihatespam99kevinbraceusenet@ihatespam99hotmail.com>
Date: Wed, 13 Mar 2002 17:20:23 -0600
Links: << >>  << T >>  << A >>
Timmestein wrote:
> 
> Hi,
> 
> I'm fairly new to this kind of desinging circuitry. I've used VHDL at
> school to program some simple things, but what I want to do is the
> following.
> 
> I want to connect my own circuitry (like a microcontroller or some other
> logic) to the PCI-bus of my computer. There are PCI interface ic's, but
> this also can be done using a fpga and an ip core from
> http://www.opencores.org . It's obvious I want to use the PCI-bridge core.
> This way my device keeps it's flexibility.
> 

        When I started developing my own PCI IP core, I had a similar
motivation like the one you have of wanting to connect something to my
computer, and PCI bus seemed like a way to go.
I guess I shouldn't be a nay-sayer, but I get the impression that if you
can use someone else's PCI IP core, you won't have to learn how PCI's
protocol works.
Just because someone claims that their PCI IP works doesn't mean that
you should totally trust that, and ultimately it will be your
responsibility to make sure the Opencores.org PCI IP core meets setup
timings (Which seems like it does according to their website.), and
doesn't violate PCI bus protocol.



> Can someone please help me with finding a right fpga (type and/or
> manufacturer) and how to program this device? Like the cable I have to use,
> the software for compiling the core and the software to get the code into
> the fpga.


        Like the Opencores.org PCI project, I used an Insight
Electronics Spartan-II PCI Development Kit to make sure that my PCI IP
core actually works in a real computer.

http://208.129.228.206/solutions/kits/xilinx/spartan-iipci.html


The board alone without a license for Xilinx LogiCORE PCI PCI32 was only
$145, and even with two other related option boards and a parallel port
JTAG download cable, the whole thing cost me only about $370, but I was
told recently that Insight Electronics discontinued the board.
You can call the number on the website (Not sure if you can call a toll
free number from overseas, but it won't hurt to try.) to make sure if
that is true or not.
That person also said Insight Electronics supposedly is going to release
a new PCI board some time in the future. 
A company called Nallatech has another Spartan-II based board that has
two Spartan-II XC2S150-5 on it, but I don't know how much it costs.
Probably not as cheap as Insight Electronics Spartan-II PCI Development
Kit.
I will guess that it will cost something like $500, but it can cost more
than that.

http://www.nallatech.com/products/dime_select/strathnuey/index.asp


        Speaking of what FPGA you should use for PCI, I will say that it
should be whatever the fastest FPGA available at the time, but one thing
you should know is that most desktop computers still use 5V PCI, and
some newer FPGA no longer support 5V PCI.
For Xilinx FPGAs, Virtex/Spartan-II are the last FPGAs that support 5V
PCI without external voltage level conversion chips.
Newer FPGAs like Virtex-E/Virtex-II/Spartan-IIE support only 3.3V PCI,
and not 5V PCI by itself (Unless external voltage level conversion chips
are used.).
For Altera FPGAs, FLEX10KE/ACEX1K/APEX20K are the last FPGAs that
support 5V PCI.
Yes, there are other FPGA vendors out there, but very few people use
them, and often tools are not free or useless, so I will not discuss
them here.
Because I primarily use Xilinx software/device for my PCI IP core
development, my opinion is biased, but still I find Xilinx devices
easier to work with than Altera's devices when it comes to PCI.
So, for 5V PCI applications, I recommend Spartan-II because boards based
on it tends to be fairly low cost.
        Assuming that you are poor, and cannot afford to pay for tools
like me, you should download a copy of free Xilinx ISE WebPACK 4.1.
I find Xilinx's ISE WebPACK 4.1 more generous and stable than Altera's
free Quartus II 2.0 Web Edition + LeonardoSpectrum-Altera when a Windows
98 PC is used.
When I say generous, Xilinx lets you use ModelSim XE-Starter for free
with some restrictions (Slows down quite a bit after 500 lines of code,
but will continue to run.), but Altera only offers a simple waveform
simulator.
        A JTAG parallel cable should be fine to program a Configuration
PROM on the PCI card.
The necessary software (I believe it is now called iMPACT) comes with
ISE WebPACK 4.1.



> Maybe there's someone out there who's actually used thos core in one of hos
> projects who can help me.
> 
> I am really new at this, but I'm always in to learn :).
> Hope someone can help me woth this. Any help is appreciated.
> 
> Tim

        Opencores.org PCI project forum should be able to help you.

http://www.opencores.org/forums/pci/


Going back to my opinion as a nay-sayer, the observations I made from
looking at their PCI IP core source code is that, do they (the authors)
expect other people to fix problems/update code with the way they wrote
it?
The level of abstraction of the code seems to me like something very
close to gate level, therefore it is pretty hard to follow what they are
doing in the IP core.
My guess is that they wrote it that way to get the PCI IP core to meet
7ns setup time, which the designer is not being careful, won't be met
even with Spartan-II (I had tons of problems with setup time several
months ago.), but after several months of struggle, I am now able to
meet 7ns setup time with easier to understand RTL level code after some
manual floorplanning.
If the Opencores.org PCI IP core is not buggy, that is fine, but if it
is, I will guess that it will be pretty hard for someone other than the
original authors to fix the problems.
        Another problem you face in my opinion is that like almost any
other Europeans, you use VHDL, but the Opencores.org PCI IP core is
written in Verilog.
I personally don't understand why Europeans almost always use VHDL
rather than Verilog which is a simpler and easier to learn language, but
regardless, you will either have to learn Verilog, which shouldn't be a
big problem if you already know some VHDL, or you will have to do a
mixed language design.
To do a mixed language design, you will have to create a "blackbox" in
the Opencores.org PCI IP core for your backend application, and
synthesize it without the backend application.
Separately, you will synthesize the VHDL backend application likely
without I/O pads, and merge the two netlists (in ISE WebPACK, ".ngc"
files) after synthesis (In ISE WebPACK, NGDBUILD handles the merging of
netlist files.).
A few weeks ago, I did a Verilog/VHDL mixed language design with my PCI
IP core, where I merged a VHDL backend application netlist to my PCI IP
core netlist written in Verilog.
So, it is not impossible to do a mixed language design, but for a
beginner, it is probably easier to do the backend application design in
Verilog.




Kevin Brace (Don't respond to me directly, respond within the
newsgroup.)

Article: 40724
Subject: Re: digital video PLL
From: John_H <johnhandwork@mail.com>
Date: Thu, 14 Mar 2002 00:57:03 GMT
Links: << >>  << T >>  << A >>
The Ultralock is all digital but the application is different.  By oversampling
with a jitter-free clock, the analog data can be reformed to the precise timing
based on a "virtual" recovered clock.  What the person who started the thread
appeared to be doing was generating a pixel clock directly - I may be wrong in my
interpretation.  If the design intent was not to produce a DPLL recovered clock
output to somewhere else in the system then there are polyphase methods that can
be used in sampling (decimation) or output (interpolation) of analog signals with
low jitter clocks on the outside of the system.  The "virtual" internal frequency
that controls the polyphase filter isn't subject to the same kind of jitter issues
that a DPLL clock output would generate.  DDS techniques could be used to push
jitter into the noise floor and get a true, clean DPLL clock output.  The DDS
system also costs a bit of extra cash and involves precision analog components
(converters and filters) beyond the digital realm.

Give ultralock a jittered clock and you'll compromise your results.



Frank Vorstenbosch wrote:

> Oops!  The Bt8x8's "UltraLock" is alleged to be all digital, though.




Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search