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
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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 9025

Article: 9025
Subject: Re: Devices and Prices
From: s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb)
Date: Sun, 15 Feb 1998 00:46:28 GMT
Links: << >>  << T >>  << A >>
On Thu, 12 Feb 1998 10:58:13 -0800, Gary Levin
<gary_levin@phx.mcd.mot.com> wrote:

>Thanks. That is good advice.

And it was free :-)

<snip>
>The problem with PLD and FPGA design is that many devices seem to
>require vendor dependent tools.

All require their own P&R /Fitter, but all can be designed with one
design flow. Namely HDL. Sitting on my PC is a copy of Exemplar's
Galileo. Support is there for :

Actel ACT1, ACT2, ACT3, 1200XL, 3200DX,
Altera MAX5000/7000/9000, FLEX 6K, 8K, and 10K
Atmel 6K02, 6K04
Lattice pLSI
Lucent 3000, 1C, 2C, 2A
Motorola MPA1000
Quicklogic pASIC1, pASIC2
Xilinx 3K, 4K, 5K, 7200A, 7300, 9500

Should be enough for most :-)

So the issue is really back end fitters. To be honest, I don't know
anybody who charges for their back end tool if you engage with them
sensibly.

e.g. "Hello Mr. Vendor, I'd like to use 10,000 of your devices. I have
a netlist generated by my synthesis tools, and I really don't fancy
forking out money to see whether your silicon is up to the job, and
cost effective".....

>Personally, I prefer to use Verilog because we design our ASICs here
>with Verilog.  I just did a bunch of Lattice 2064's and 1016's using
>Verilog and Synopsys. That worked out fine. But using Verilog for FPGA
>Verilog doesn't always give the best results. Although Verilog is easier
>to maintain here.

I prefer VHDL, but I guess it's a case of what you learn with first.

>For critical designs, for example, one group of designers use Viewlogic 
>for Xilinx. Another group prefers Altera parts. Our factory likes
>Lattice because of ISP and manufacturing issues. Still yet, some older
>engineers like ABEL for smaller PLD's.

This comes from the history of different vendors OEM'ing different
third party packages and hence locking designers into a particular
mindset for design.

>Our boards are designed using Mentor Graphics DA-LMS. We have older
>designs still in Cadence. So as you see, the tools can get out of
>control.
>
>What I'd like to know, before I start a new design and get locked into a
>particular vendor is some competitive pricing for a particular class of
>device. Unless of course they are mostly cost competitive and it really
>doesn't matter to compare.

They are all competitive when it comes down to it. Some will appear
cheaper than others on a pure gate counting basis, but the wide range
of real system performance, features and (worst of all) differing
speed grades, makes comparisons a minefield.

But... Just for fun, I pulled the following data for SRAM based,
nominally 10K gate part, 208 pin QFP where possible. Parts are listed
in ascending cost-per-gate.

Device			100 up	Claimed	4ip 	Registers	Cost
Percentage
			cost	Density	LUTs			per K
Increase
								gates

EPF10K20RC2084		$42.50	15000	1152	1341		2.83
0.00%
EPF6016QC2083		$22.95	8000	1320	1320		2.87
1.25%
OR2T10A2S208		$35.70	12300	1024	1024		2.90
2.44%
EPF10K10QC2084		$22.00	7000	576	710		3.14
10.92%
OR2C10A2S208		$45.30	12300	1024	1024		3.68
29.99%
EPF6016QC2082		$32.50	8000	1320	1320		4.06
43.38%
OR2T10A4S208		$51.30	12300	1024	1024		4.17
47.20%
EPF10K20RC2083		$63.50	15000	1152	1341		4.23
49.41%
EPF10K10QC2083		$32.50	7000	576	710		4.64
63.87%
XC4010XL-3PQ208C	$49.10	10000	800	1120		4.91
73.29%
XC5210-6PQ208C		$49.80	10000	1296	1296		4.98
75.76%
OR2C10A4S208		$65.20	12300	1024	1024		5.30
87.09%
OR2T10A5S208		$66.80	12300	1024	1024		5.43
91.68%
XC5210-3PQ160C		$81.30	10000	1296	1296		8.13
186.94%
XC4010E-4PQ208C		$85.70	10000	800	1120		8.57
202.47%
XC4010XL-1PQ208C	$86.15	10000	800	1120		8.62
204.06%
XC4010E-1PQ208C		$211.00	10000	800	1120		21.10
644.71%

Allowing for marketing "variations" and usability/routability issues,
it would appear that Altera and Lucent have a similar position on a
cost per gate basis, but only if you can accept a 3.3V device from
Lucent. The figure for the 6K part may be a little low, but I have yet
to see some real performance/routability information for that family.
Xilinx would appear to be well out of the frame on the cost metric,
with the "low cost" 5200 series not being so low cost after all. The
real glarer is the 4010E-1. Must be some real aggressive bin selection
going on here!

Read into the numbers what you will.

As Altera and Xilinx of these companies have net margins bumping
around the 60%+ mark, there is a LOT of room for movement here.

Stuart
--
For Email remove "die.spammer." from the address
Article: 9026
Subject: Re: Devices and Prices
From: s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb)
Date: Sun, 15 Feb 1998 00:46:29 GMT
Links: << >>  << T >>  << A >>
On Thu, 12 Feb 1998 19:42:44 -0500, Isabelle Gonthier
<igonthie@total.net> wrote:

>	Look at Xilinx.  They are most agressive in that type of quantity. 
>What density are you looking at?

And their market share has been doing what exactly?

All vendors will be aggressive where appropriate, but as many have
differing features, you will typically find that a certain design
gravitates towards one architecture or another.

eg for SRAM FPGA's

Need low cost, slow, power not an issue, medium density ?

	Try Xilinx or Altera and start the price war.

Need big, slow and cheap, power not an issue?

	Try Altera.

Need very big, fast, with DPRAM at reasonable speed?

	Try Xilinx

Need faster than the rest with bags of features, but not excessive
density (at the moment)?

	Try Lucent.

At the low end of the performance and features market where there is a
war going on for who can grab the biggest chunk of the (currently holy
grail) gate array market, then you can probably start a price war with
the two big players.

Stuart
--
For Email remove "die.spammer." from the address
Article: 9027
Subject: Re: VHDL vs schematics
From: mushh@jps.net (David Decker)
Date: Sun, 15 Feb 1998 17:51:39 GMT
Links: << >>  << T >>  << A >>
First:, sorry for my redundant post  2/13. Sometimes FreeAgent has 
trouble posting an article. It then squirels it away and for ever 
more, when ever I close FreeAgent, it tells me I have stuff to post 
and asks if I want to post now. I always try to click no, but 
eventually I use it when I'm too tired, and accidentally click yes. 
Does any one know how to kill pending posts in FreeAgent?

Second: It's a good thing I check DejaNews, because neither rk's 
post nor Stuart's post ever made it to my .jps news server. I'm
therefore posting a follow-up to my own message via .jps. 

Third: I have one point each for both rk, and Stuart Club:

rk says:
rk:
it's easy enough to break out the terms and simply tell the compiler
to add them up; i.e., x + 8x.  i just finished an fir filter done
that way and the.

dd:
Did you check, rk, whether this produced a 9bit or a 12 bit adder? 
My experience was that those tools that did synthesize an adder, 
use a 12 bit adder, when only a 9 bit adder is required. Could be your
FIR could have been smaller.

stuart wrote:
RESULT <= ext((a & "000"),12) + ext(a,12) ;

Works just fine through Galileo Xtreme. Single 8 bit adder with the
carry out being used.

dd:
I believe a 9 bit adder is really required, and that the carry out 
can not be used in this way, for 2's compliment DSP signals.

decimal	8bit/carry		sign extension
    0	 00000000		000000000
+(-1)	 11111111		111111111
= -1	011111111		111111111

As you can see, if you you use the carry out, and add an 8 bit -1 
to and 8 bit zero you get an answer of +255. 
If you use a 9 bit adder with sign extension, you correctly get -1.

Cheers,
Dave Decker

Please use only one 'h' in mush. I'm trying to reduce the spam.



"Animals .  .  . are not brethren they are not 
underlings;  they are other nations, 
caught with ourselves in the net of life and time, 
fellow prisoners of the splendor and travail of 
the earth."
Henry Beston -  The Outermost House
Article: 9028
Subject: the problem about counter.
From: PENG LIU <liupeng@isee.zju.edu.cn>
Date: Mon, 16 Feb 1998 15:53:36 +0800
Links: << >>  << T >>  << A >>
Hi:
	I used altera EPF81500-240AQC2 designing a board. There is a counter in
the fpga, but why I mesure the counter's output, I found that if I
mesuare another signal, then the counter's output is right, when I take
down the probe on another signal, the counter's lowest bit lost some
pulse. 
	who meet the same problem. I need help!!

yours.

ws.
Article: 9029
Subject: Re: Devices and Prices
From: Zoltan Kocsi <root@127.0.0.1>
Date: 16 Feb 1998 19:25:34 +1100
Links: << >>  << T >>  << A >>
s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb) writes:

> So the issue is really back end fitters. To be honest, I don't know
> anybody who charges for their back end tool if you engage with them
> sensibly.
> 
> e.g. "Hello Mr. Vendor, I'd like to use 10,000 of your devices. I have
> a netlist generated by my synthesis tools, and I really don't fancy
> forking out money to see whether your silicon is up to the job, and
> cost effective".....

"Well Mr. Customer, $500 doesn's seem too much for a very good P&R tool,
does it ? It routes all of our smaller chips. After all, you've forked 
out twenty grand for your synthesis, this is just peanuts. Excuse me ? 
You use unix ?! Well, then, the  situation is a little bit different, 
you must be rich, so $1000 is the price, plus 20% for the compulsory 
one year maintenance. Which chip ? Well, not that one, that's a high 
end chip. That will be an other grand but then you've everything."

Zoltan

-- 
+------------------------------------------------------------------+
| Reply address antispammed. Use ZOLTAN-at-BENDOR-dot-COM-dot-AU   |
+--------------------------------+---------------------------------+
| Zoltan Kocsi                   |   I don't believe in miracles   |  
| Bendor Research Pty. Ltd.      |   but I rely on them.           |
+--------------------------------+---------------------------------+
Article: 9030
Subject: Re: Why altera CPLDS are slow to power-up?
From: "Andy Whitehouse" <awhitehouse@spambucket.ndsuk.com>
Date: 16 Feb 1998 08:36:57 GMT
Links: << >>  << T >>  << A >>

First we have

> No, [all] The storage cells in CPLD are directly involved in the
operation, so
> they must be "read out" in a few nanoseconds, and they operate all in
> parallel. That's what has given CPLD their ( historic ) speed advantage.
>
> I gave a probable explanation for the slow POR in a different posting to
> this newsgroup.
>
> Peter Alfke, Xilinx


Then we have

> There is a fairly well-kept secret regarding CPLDs:
> 
> Although most of their configuration is directly EPROM, EEPROM, or
> Flash-based and thus does not require any lengthy power-on delay, some
> of the configuration bits, although stored in the non-volatile area, 
> are then read out on power-up and loaded into latches. This applies to
> the miscellaneous signals around the macrocells, which cannot be
> conveniently or fast controlled from the well-structured non-volatile
> memory area.
> 
> Hope that helps.
> Peter Alfke, Xilinx Applications


Come on Peter, which is it ?

Andy



Article: 9031
Subject: Re: Why altera CPLDS are slow to power-up?
From: Achim Gratz <gratz@ite.inf.tu-dresden.de>
Date: 16 Feb 1998 10:49:21 +0100
Links: << >>  << T >>  << A >>
Peter Alfke <peter.alfke@xilinx.com> writes:

> Achim Gratz wrote:
> > Most likely because some types of EEPROM are slow to read and even
> > slower to write to.
> 
> No,The storage cells in CPLD are directly involved in the operation, so
> they must be "read out" in a few nanoseconds, and they operate all in
> parallel. That's what has given CPLD their ( historic ) speed advantage.
> 
> I gave a probable explanation for the slow POR in a different posting to
> this newsgroup.

I've read your explanation, but either I'm missing something or you
contradict yourself.  If I read your other posting right, you say that
some CPLD are (at least partly) SRAM based and backed by EEPROM.  That
only makes sense if you use cheap EEPROM for these bits and that was
actually what I thought of when making my comment.  If the same fast
cells as in the signal path (if there are any at all) were used, it
would be a truly stupid circuit design if it took as long as has been
reported.  Sorry if I haven't been clear.  Come to think of it, if
there's a charge pump for boosting on these chips, it'll also take
it's share of the total POR time to come up.

I've just looked at the MAX9000 data sheet.  The fact that they have
a "Turbo mode" would suggest that they can boost while the drawings
for the macrocell hint that the EEPROM storage is seperate from the
select muxes and integrated into the product term matrix as Peter was
explaining.  I didn't find the POR time in the data sheet, though.


Achim Gratz.

--+<[ It's the small pleasures that make life so miserable. ]>+--
WWW:    http://www.inf.tu-dresden.de/~ag7/{english/}
E-Mail: gratz@ite.inf.tu-dresden.de
Phone:  +49 351 463 - 8325
Article: 9032
Subject: Interesting one for anyone using ORCA
From: david.storrar@gecm.SPAMMENOT.com (David Storrar)
Date: Mon, 16 Feb 1998 11:14:55 GMT
Links: << >>  << T >>  << A >>
Hi all,

I'd appreciate it if anyone could give me some advice.  I have a
schematic design (entered with Mentor Design Architect) targeted at a
Lucent ORCA device (OR2C08).

I would like to simulate this design using a VHDL testbench, but I'm
having problems with the design interface.  In order to map the
design, chip IOs can't appear on the symbol interface (correct me if
I'm wrong) otherwise the pin names are interpreted as pin locations by
the Foundry mapper.  So my chip IOs are internal nets.

The question is, how do I apply stimulus to the device IOs if they are
not on the symbol?

Thanks for any help.

Dave

--
REMOVE "spammenot." in address to reply
Dave Storrar
Development Engineer    | e-mail: david.storrar@gecm.com
GEC-Marconi Avionics    | Tel: +44 (0)131 343 4729
Radar Systems Division  | Fax: +44 (0)131 343 4091
--
Article: 9033
Subject: Xilinx Simulation in Synopsys
From: kaps@WPI.EDU (Jens-Peter Kaps)
Date: 16 Feb 1998 20:39:15 GMT
Links: << >>  << T >>  << A >>
Hi,

I have a problem with simulation an Xilinx FPGA design after place and 
route using the Synopsys tools. The Xilinx tools tell me that the minimum
clock period is 44 ns whereas a simulation with a 4.4 ns clock period is 
still fine and I could increase the frequency even further. I don't
believe that the Xilinx timing analysis is so pessimistic that it is off
by more than a factor 10.

Im using: Xilinx M-1.3.7 and Synopsys Version 1997.08 -- Jul 11, 1997
I do: vhdlan -i  of the vhdl files and then
      vhdldbx -sdf_top testbench/uut -sdf time_sim.sdf CFG_TB

unfortunately I can not use compiled mode as we don't have a c-compiler
for the HP stations.

I'd apreciate any help or pointer to the solution,

     Jens
-- 
  ------------------------------------------------------------
  Jens-Peter Kaps, 79/8 Park Avenue., Worcester, MA 01605, USA
  Home: (508) 831-9789;      Office: (508) 831-5757; Fax -5491
  ------------------------------------------------------------
Article: 9034
Subject: High-Tech Jobs/Resume
From: deepka <deepka@best.com>
Date: Mon, 16 Feb 1998 13:41:04 -0800
Links: << >>  << T >>  << A >>
Check out new site for technology jobs .....computers/IT, engineering,
design, process,CAD, application, software etc. POST RESUME FOR FREE
http://www.techjobbank.com
Article: 9035
Subject: Re: Why some CPLDS are slow to power-up?
From: jim granville <Jim.Granville@xtra.co.nz>
Date: Tue, 17 Feb 1998 10:45:10 +1300
Links: << >>  << T >>  << A >>
Andy Whitehouse wrote:
> 
> First we have
> 
> > No, [all] The storage cells in CPLD are directly involved in the
> operation, so
> > they must be "read out" in a few nanoseconds, and they operate all in
> > parallel. That's what has given CPLD their ( historic ) speed advantage.
> 
> Then we have
> 
> > There is a fairly well-kept secret regarding CPLDs:
> >
> > Although most of their configuration is directly EPROM, EEPROM, or
> > Flash-based and thus does not require any lengthy power-on delay, some
> > of the configuration bits, although stored in the non-volatile area,
> > are then read out on power-up and loaded into latches. This applies to
> > the miscellaneous signals around the macrocells, which cannot be
> > conveniently or fast controlled from the well-structured non-volatile
> > memory area.
> >
> > Hope that helps.
> > Peter Alfke, Xilinx Applications
> 
> Come on Peter, which is it ?
> 
> Andy

 I have no problem with accepting both - it fits with the behaviour I
have
seen on CPLD. ( try some brownout tests on your favourite CPLD .. )

In a CPLD there are TWO distinct areas, The AND/OR arrays, and the
macrocells.
A MacroCell has maybe 16 BITS defining all the options.
Given a well structured memory area, with fast parallel access, covering
98% of the
fuse terms for AND/OR.

If you have, a say 5nS read time on this memory, how do you 'grab' the
macrocell info
on-the-fly - very hard, and you get 5nS time penaltie(s).
Much easier to load the 16 bits into Latches ( so SRAM if you prefer ),
that control 
the macrocell HW directly - now you are into sub nS GATE delays.

There are also charge pumps, on some devices, and these will have multi
uS delays
What I have not seen, from any supplier, is a table showing
- Time when outputs are completely undefined, after powerup
- Time when outputs are defined, but other specs ( eg delays) are
settling
- Final time, when 100% spec is met.
- Brownout protection - is a full re-PowerUP necessary to glitch recover
?

That would be very informative.

-jg.

Article: 9036
Subject: Re: x86 soft cores?
From: "William M. Wiese Jr." <w i e s e @ jps[antispam]net>
Date: 16 Feb 98 21:50:23 GMT
Links: << >>  << T >>  << A >>

I believe a co.  called   "V Automation"  at  www.vautomation.com  has
synthesizable cores for the x86.  I believe the x86's they're referring to
are the 8086 and 80186;  don't believe they have coverage for 286 or above.
  Their schtick is that the cores instruction sets are modifiable for
special apps.



----------------------------------------------------------------------------
----
Willam M. Wiese Jr.
Sr. Engineer
Cignal Global Communications
2041 Pioneer Ct,  Suite 17
San Mateo, CA   94403
(650) 341 - 4761
b w i e s e  @  c i g n a l  DOT c o m       {remove antispam}
----------------------------------------------------------------------------
----


Joachim Strombergson <emwchim@emw.ericsson.se> wrote in article
<34DEB95A.2BED@emw.ericsson.se>...
> Tja!
> 
> I'm looking for soft cores for x86 processors. Does anybody develop and
> sell such beasts? Any help greatly appreciated.
> 
> -- 
> Med vänlig hälsning
> 
> Joachim Strömbergson
> ---------------------------------------
> Joachim Strömbergson UN/X
> 
> ASIC Technology and System on Silicon
> Ericsson Microwave Systems AB
> SE-431 84 Mölndal
> Joachim.Strombergson@emw.ericsson.se
> ---------------------------------------
> 
Article: 9037
Subject: Re: Why altera CPLDS are slow to power-up?
From: jim granville <Jim.Granville@xtra.co.nz>
Date: Tue, 17 Feb 1998 10:51:57 +1300
Links: << >>  << T >>  << A >>
Peter Alfke wrote:
> 
< snip >
> Certain CPLD manufacturers don't like to talk about that, because it
> smudges their pristine image of non-volatility.
> In reality, as Xilinx FPGAs have proven for the past 12 years, there is
> nothing wrong with storing configuration in latches. You just have to
> load them properly, and you have to detects Vcc dips.
> Peter Alfke, Xilinx Applications

So, what do the XILINX 95xx CPLDs do ?

Do they POR load Macrocell info, like the others ?

If you remove power, to say 1-2V, then reapply, does the part
still work ?

 We have designed various oscillator topologies into CPLD, and
found that, for example, an INVERTER does not ALWAYS invert -
esp after power supply cycle tests.

- jg.

-- 
======= Manufacturers of Serious Design Tools for uC and PLD  =========
= Optimising Modula-2 Structured Text compilers for ALL 80X51 variants
= Many reusable object modules - i2c, SPI, SPL, RC5, etc 
= Safe, Readable & Fast code - Step up from Assembler and C
= Emulators / Programmers for ATMEL 89C1051, 2051, 89C51 89S8252 89C55
= *NEW* Bondout ICE for 89C51/89C52/89C55. OptoISP for 89S, AVR, AT17K.
= for more info, mailto:DesignTools@xtra.co.nz  Subject : c51Tools

Article: 9038
Subject: Re: the problem about counter.
From: Jack Lai <jwlai@mmmpcc.org>
Date: Mon, 16 Feb 1998 18:48:09 -0600
Links: << >>  << T >>  << A >>
PENG LIU wrote:
> 
> Hi:
>         I used altera EPF81500-240AQC2 designing a board. There is a counter in
> the fpga, but why I mesure the counter's output, I found that if I
> mesuare another signal, then the counter's output is right, when I take
> down the probe on another signal, the counter's lowest bit lost some
> pulse.
>         who meet the same problem. I need help!!
> 
> yours.
> 
> ws.
I think your problem is probing techniques,are you using good 10X probes
with very short ground leads?. Have you try active probes?

Regards, Jack

Article: 9039
Subject: Re: Walace tree???
From: Rick Carmichael <rcarmichael@intellistor.com>
Date: Mon, 16 Feb 1998 19:00:06 -0700
Links: << >>  << T >>  << A >>
Victor Levandovsky wrote:
> 
> Hi!
> 
> I`m looking material about Walace tree.
> Could you help me, please?
> 
> Regardless,
> 
> Victor Levandovsky
> 
> vicNS@alpha.podol.khmelnitskiy.ua
> 
> Remove NS if reply

Two good books (I am only certain of the author - not the title) is
computer logic design? by Kai Hwang. I used that book as a reference
for writting a program for Wtree adder hardware minimization several 
years ago. This book devotes an entire chapter to the subject
and is well worth seeking out. (I loaned mine out three years ago for
this very topic and haven't gotten it back yet).

Another good multiplier reference is the very old Fairchild F100 ECL
user handbook, where an application note demonstrates a high-performance
Booth multiplier with wallace trees (at the gate level). It gives
100% of the required info needed to quickly build and understand a
multiplier. This is an old book, let me know if you have problems
finding it. I inherited mine from a mainframe ASIC designer at my
first job out of school.

Warm Regards,
rick
Article: 9040
Subject: ASIC_designers_needed_Colorado!
From: Rick Carmichael <rcarmichael@intellistor.com>
Date: Mon, 16 Feb 1998 20:03:03 -0700
Links: << >>  << T >>  << A >>
Boulder County ASIC Positions

Our chips are growing in gate counts and we need designers to join our
ASIC R&D team in Longmont, Colorado.  ASIC design engineering positions
are available, junior through staff levels.  We are looking for
engineers
who can design, develop, and test portions of large ASICs used in RAID
storage controllers.  Positions require a BSEE or MSEE or equivalent.  
Desired capabilities may include Verilog, Synopsys, Computer
Architecture, 
ASIC design experience, UNIX, C/C++, Python, perl, PCI, Fibre Channel, 
SCSI-3,SPARC, ECC and Compression.

Fujitsu Ltd. is a $36 billion technology company - the world's second
largest computer company.  The ASIC engineer positions available are
with Fujitsu Computer Products of America, Inc., Intellistor R&D 
Operation in Longmont, Colorado (EOE).  The positions offer an 
opportunity to work for an R&D division whose charter is to architect, 
develop, and test large, leading-edge Fujitsu ASICs and ASIC-based 
systems using high-end, UNIX-based tools. Applicants must possess 
outstanding technical capabilities (or potential), work well in a 
team environment, and be comfortable working with a design initiated 
with a "clean sheet of paper".  

If you would like to find out more about the company and the
opportunities, please contact me via e-mail 
(rcarmichael@intellistor.com), or via phone at (303) 682-6551.

Warm Regards,
rick
Article: 9041
Subject: Re: Why altera CPLDS are slow to power-up?
From: Kayvon Irani <kirani@cinenet.net>
Date: Mon, 16 Feb 1998 22:34:53 -0800
Links: << >>  << T >>  << A >>


Achim Gratz wrote: I didn't find the POR time in the data sheet, though.

    The POR time is not in the data sheet. They mention that in their design
     guide lines for CPLDs under the section that talks about VCC.

    Regards,
    Kayvon Irani
    Los Angeles. Ca

Article: 9042
Subject: Viewlogic/Speedwave
From: "John Huang" <hungi@tpts4.seed.net.tw>
Date: 17 Feb 1998 08:08:43 GMT
Links: << >>  << T >>  << A >>

Hi all:
	I've got a problem about viewlogic/speedwave, would
someon can tell me

	I 've a design that write in VHDL, and I split it into
many entity(use hierarchy design),  and now I want to verify
its function, but I just can find the signal and results of 
top entity, all sub-entity(components) are not found by speedwave
, I've use the VHDL manager to analyze all vhdl files, and add 
they into one library, and I use IEEE and SYSNOPSYS for sys 
library.
	Can someone tell me how to use hierarchy design in speedwave
?

	Thanks
				John Huang

Article: 9043
Subject: Re: Devices and Prices
From: Sam Falaki <falaki@nospam.videotron.ca>
Date: Tue, 17 Feb 1998 08:23:24 GMT
Links: << >>  << T >>  << A >>


Zoltan Kocsi wrote:

> s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb) writes:
>
> > So the issue is really back end fitters. To be honest, I don't know
> > anybody who charges for their back end tool if you engage with them
> > sensibly.
> >
> > e.g. "Hello Mr. Vendor, I'd like to use 10,000 of your devices. I have
> > a netlist generated by my synthesis tools, and I really don't fancy
> > forking out money to see whether your silicon is up to the job, and
> > cost effective".....
>
> "Well Mr. Customer, $500 doesn's seem too much for a very good P&R tool,
> does it ? It routes all of our smaller chips. After all, you've forked
> out twenty grand for your synthesis, this is just peanuts. Excuse me ?
> You use unix ?!

This excuse for charging high prices is starting to get old.  A $20k
synthesis tool cantarget a lot of different technologies and it doesn't have
to run on a Unix box when PCs
are starting to outperform them (not to mention the ease of maintenance and
instalation
of the software).  Now if you are in this situation and you intend to use
FPGAs, you can't
directly compare different FPGAs because you need every vendor's P&R tools
running
at about $2k each.  This has a tendency to lock you in to one or two
particular vendors
which can be very annoying, and you can't even be sure that you've made the
right choice
until after you spent your hard-earned cash.  I think the vendors
underestimate the potential
of the small players.  As ASIC methodologies are beginning to be applied by
traditional
PLD designers and as the software migrates towards PCs, I believe that a
vendor offering
free tools would capture a good part of this growing market.  I hope that
the arrogance
of some of the vendors backfires on them.  There aren't only big companies
with money to
waste out there.

> Well, then, the  situation is a little bit different,
> you must be rich, so $1000 is the price, plus 20% for the compulsory
> one year maintenance. Which chip ? Well, not that one, that's a high
> end chip. That will be an other grand but then you've everything."
>
> Zoltan
>
> --
> +------------------------------------------------------------------+
> | Reply address antispammed. Use ZOLTAN-at-BENDOR-dot-COM-dot-AU   |
> +--------------------------------+---------------------------------+
> | Zoltan Kocsi                   |   I don't believe in miracles   |
> | Bendor Research Pty. Ltd.      |   but I rely on them.           |
> +--------------------------------+---------------------------------+



Article: 9044
Subject: Re: Viewlogic/Speedwave
From: William White <will@fpga.demon.co.uk>
Date: Tue, 17 Feb 1998 10:05:50 +0000
Links: << >>  << T >>  << A >>
Did you get any warnings at the elaboration stage? Elaboration occurs
when you select the top level entity to simulate from the working
library. This is where the simulator builds the hierarchy of the design
from the VHDL source. Even though the VHDL may be correct and compiles
without error in the HDL manager, this does not mean that the
appropriate VHDL configurations have been set up, which essentially
defines the hierarchy. If this is the case you might get warnings along
the line of:

"Configuration for component XXXX is open"

If so, you will need to add the appropriate configuration statements to
the VHDL. If it is simply multiple components at a single level of
hierarchy a default configuration of the form:

CONFIGURATION conf_name OF top_level_entity IS
FOR top_level_arch
END FOR;
END conf_name; 

where conf_name is the name you assign to the configuration.
      
should do the trick. If the hierarchy is more complex then the
configuration statement or statements will also be more complex.
Reference to one of the standard VHDL texts such as VHDL by Douglas L.
Perry ( ISBN 0-07-049434-7 ) should help. After you have analysed this
configuration ( this type of configuration can be in a separate file and
analysed at the end. ), you then need to simulate the configuration
rather than the top level entity.

The interpretation of the VHDL LRM is a little ambiguous when it comes
to default configurations. Some simulators assume them, others require
them. The best thing to do is remove the ambiguity and always have them
where appropriate.

Hope this helps.

Regards

Will


In article <01bd3b7c$3ac4ee50$dec5af8b@ntjohn>, John Huang
<hungi@tpts4.seed.net.tw> writes
>
>Hi all:
>       I've got a problem about viewlogic/speedwave, would
>someon can tell me
>
>       I 've a design that write in VHDL, and I split it into
>many entity(use hierarchy design),  and now I want to verify
>its function, but I just can find the signal and results of 
>top entity, all sub-entity(components) are not found by speedwave
>, I've use the VHDL manager to analyze all vhdl files, and add 
>they into one library, and I use IEEE and SYSNOPSYS for sys 
>library.
>       Can someone tell me how to use hierarchy design in speedwave
>?
>
>       Thanks
>                               John Huang
>

-- 
William White
Article: 9045
Subject: Re: Altera 5032 programming problems
From: David Pashley <david@fpga.demon.co.uk>
Date: Tue, 17 Feb 1998 12:02:50 +0000
Links: << >>  << T >>  << A >>
In article <34E6191E.4479@xtra.co.nz>, jim granville
<Jim.Granville@xtra.co.nz> writes
>RWilson@oxfordnuclear.com wrote:
>> 
>> We have experienced problems with their classic line (EP1810LC-35T,
>> EP910-30T, and EPM5128LC).  Since the switch to the new die, the devices
>> program and verify OK but they flake out in certain applications.  We
>> have actually had them generate enough heat to burn labels off of them.
>> Altera is not really helping us and claim we are the problem.  We have
>> procured safety stock of old devices to protect our product lines until
>> we can reach a resolution.  Interestingly, the market price for older
>> devices (pre die-change) has tripled in recent weeks.  We are stumped.
>
> Problems on DIE revision are more common that many realise..
>
>1) You should check carefully that the algorithm has not changed - and
>get
>the confirmation from as reliable as source as possible.
> The reflex reaction of most FAE's is 'no, nothing has changed'
>
>2) try as many different BRANDS of programmer as possible
>

I don't think that this is the way to go.

For 70% of programmable devices, there are 4 or more die changes during
the device's market life. For some devices (especially CPLDs) there can
be as many as a dozen changes.

Leading programmer vendors work very closely with the semiconductor
vendors on this issue. That's why you can buy a programmer for thousands
of dollars from Data I/O (and some other major vendors), yet there are
dozens of vendors of $500 programmers that seem to have an OK spec.

Despite much more expensive products, Data I/O has been the market
leader for 25 years, and is many times bigger than any other programmer
company. 

The answer to this apparent paradox is also the answer to the question
being posed in this thread.

It costs a lot of time and money to track the constant die changes, and
cheap programmer vendors don't bother. However, most serious users of
programmable devices (esp. in manufacturing) know that this is the
biggest single issue in device programming, and that in this respect,
you get what you pay for.

[Direct Insight Ltd is Data I/O's UK VAR]

-- 
David Pashley
Direct Insight Ltd

Article: 9046
Subject: This is a test to see if this will post!!!
From: "bgeorge" <bgeorge@northatlantic.nf.ca>
Date: 17 Feb 1998 14:21:16 GMT
Links: << >>  << T >>  << A >>
Test....
Article: 9047
Subject: Xilinx download cable ??????
From: THIEBOLT Francois <thiebolt@irit.fr>
Date: Tue, 17 Feb 1998 16:29:17 +0100
Links: << >>  << T >>  << A >>
Hi,

Can someone tell me what are the FPGA Xilinx parts that could be
programmed using a simple download cable (like lattice parts) ???

Thanks for your help.

Francois.

-- 
-------------------------------------------------------------
THIEBOLT Francois \ You think your computer run too slow ?
UPS Toulouse III  \ - Check nobody's asked for tea !
thiebolt@irit.fr  \ "The Hitchikers Guide to the Galaxy" D.Adams
-------------------------------------------------------------
Article: 9048
Subject: ACROBAT
From: "Dmitriy A. Gorkaev" <goda2@uic.nnov.ru>
Date: Tue, 17 Feb 1998 19:37:08 +0300
Links: << >>  << T >>  << A >>
PLEASE IF YOU KNOW WHERE I CAN GET ACROBAT READER FOR WINDOWS3.1 IN .ZIP
AND NOT MORE THEN ONE DISKET PUT ME 



ADRESS
Article: 9049
Subject: Re: Why altera CPLDS are slow to power-up?
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Tue, 17 Feb 1998 09:57:23 -0800
Links: << >>  << T >>  << A >>
I am getting dragged deeper into this than I wanted to, but let me
clarify:

No single delay in any modern CMOS circuit is longer than 50 ns, most
likely not even longer than 20 ns. We just don't  waste space on large
resistors and large capacitors. So when something takes microseconds or
milliseconds it is always a sequential event. ( Or it is thermal, or the
result of a floating node ).

The "funny bits" in the CPLD are stored in the regular EPROM (EEPROM,
Flash) array and then read out into their latches, where they can
directly control certain functions of the macrocell. I have always
assumed that this read-out is sequential, somewhat like the
configuration process in FPGAs.

It is sad that there is nobody at Altera willing to stick his head out.
Of course they are monitoring this newsgroup, but there's no guts on
Orchard Parkway...

It's fun to needle somebody who is too afraid to come out of hiding, and
relies on being rescued, occasionally, from New Zealand...

Peter Alfke, Xilinx Applications



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
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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