Pissed off with Xilinx - Spartan 3

Ben Twijnstra wrote:
> What's your problem with Quartus on Linux? Since 4.0 it's been running
> flawlessly (well, as flawless as its Windows counterpart) and quickly on
> RedHat, Gentoo, Suse, and even Knoppix (i.e. Debian) in this little
> computer room. Four completely different distributions.
> Also, the tools are TCL scriptable, generate Makefile-friendly .done files -
> so what's your problem?

I'm doing amateur hobbiest stuff, so there is no Linux version for me. 
I've heard around c.a.f that Webpack will be coming out for Linux 
"soon", but afaict it doesn't exist yet.  Once that's available, I will 
definitely be setting it up and playing around with integrating its 
toolchain into my development processes.

That would then make q.1 parts availability just about the *only* issue 
for my particular case in deciding between Xilinx and Altera... 
Eventually for larger projects, when I know a lot more about FPGAs, I 
might be able to make a distinction between the parts on their technical 
merits, but like I said, I'm very new to all this stuff and have no such 
clue yet.

Erik,

OK, thanks for the explanation.

Spartan 2 today is based on a very mature process, and the sequence is 
'don't care', and the power on surge is within the specifications 
clearly outlined in the datasheet.

The Spartan 2E is a much better choice, and is based on the "EA" shrink 
(hybrid .18/.15u process), and also has no 'sequence issues', and an 
even smaller power on requirement than Spartan 2.

'Sequence issues' for any Xilinx part all relate to their use:  what 
happens when you plug them into a live bus, anbd what do you expect to 
happen while you power them ON, OFF, or do nothing.  The support through 
the web technical answers detail all that stuff.

Some parts are more hot plug friendly than others, and none sustain any 
damage to themselves as long as the absolute maximum voltages and 
currents are adhered to.

Very few people in your situation are designing boards that are going to 
be hot-plugged, and even fewer care what happens when you do.  For those 
who do care, we have a full range of documented proven solutions.

Xilinx admittedly does not cater to the home hobbiest, as we don't see 
much revenue from that source.  I am happy to see Xilinx FPGAs being 
used for software defined radios by amateur radio hobbiests (being that 
I have the call sign AB6VU, and I am also very interested and involved 
in SDR for our commercial customers).

http://www.gnu.org/software/gnuradio/doc/exploring-gnuradio.html#intro

I am sure that must be tough going, but most hobbiests realize that 
buying exiting development boards (like the Spartan 3 student Digilent 
pcb for $99) is a better choice than trying to etch your own pcb's in 
the sink.

http://www.digilentinc.com/info/S3Board.cfm

We do cater to Universities and Schools, as training generations of 
engineers (since 1985) to use our parts is a very viable business 
decision.  Perhaps stepping on board with those suppliers (like the one 
mentioned above) is much easier.

http://www.xilinx.com/univ/

Austin

Austin Lesea wrote:
> Very few people in your situation are designing boards that are going to 
> be hot-plugged, and even fewer care what happens when you do.  For those 
> who do care, we have a full range of documented proven solutions.
The design I have in mind right now is hot-plugged, but only on the USB 
side.  The FPGA will be glue between the USB chip, DRAM and flash.  I'm 
hoping that the power-up requirements of an XC2S50E or -100E won't 
require any significant control circuitry.  Eventually I hope to reduce 
the logic requirements to a lot fewer pins and gates, into CPLD realm, 
but for experimentation I want to have every pin of every device routed 
through the FPGA, even if half the logic is just crosspointing.

> I am sure that must be tough going, but most hobbiests realize that 
> buying exiting development boards (like the Spartan 3 student Digilent 
> pcb for $99) is a better choice than trying to etch your own pcb's in 
> the sink.
I have this board, and am going to try to do my initial development by 
building boards to plug into the 40pin connectors, but among other 
things that limits me afaict to an x8 DRAM instead of x16, due to lack 
of pins.  A x16 JEDEC pinout for 256 or 512Mb has 16 data lines, 13 addr 
lines, BA0, BA1, RAS#, CAS#, WE#, CLK, CKE, DQML, and DQMH.  Total of 
39.  I'm still figuring out how SDRAM works, but I'm only sure I can tie 
CKE, DQML, and DQML.  That gets me to 36, but there are only 34 I/O's on 
connectors A2 and B1 of the Spartan3 board...

Flash is easier, only 15 pins.  USB I've already done hand-wired using 
the ELRAsoft EZ-USB FX2 board, and in a manner IMO far superior to the 
board Digilent sells to do the same thing (I actually connect all the 
pins... x16 data and control).  Already done a basic test design all the 
way through to user-space software in Linux on that setup 
(host-controlled PWM of the 8 LEDs).

Erik,

How is USB hot plug to the FPGA?  USB may be hot plug to itself, but the 
FPGA isn't involved.

And, if you pull the USB plug, you disconenct the link, and it stops. 
If you plug it in, it starts to work.  This is not considered hot plug 
in the industry.  Hot plug is when you plug into something already 
working, and try not to disturb it (generally).

What you are talking about is better called "live insertion" where you 
are concerned about something breaking if you attach something while the 
power is on, or detach it when it was powered.

I can't see any issues here with any of our parts, past or present.

If you are attempting to power the FPGA from the USB 5V @ 500 mA, that 
will be an issue with any part that has a startup current requirement.

We made a USB pcb with a switching regulator (from 5Vdc to 1.5V) for a 
2VP4.  That is 2.5 W total power available from the USB, and at 1.5V, 
that is 2.5/1.5 ~ 1.5 amps (derating for efficiency).  At 1.8V it is 
less, and at 2.5V, less than an ampere.  This might be an issue for the 
old parts (see the data sheet).

The USB pcb is the "Logic Vault" board for security experimentation 
(with 3DES key, 20+ year life lithium coin cell, and a true random 
number generator IP that runs when you plug it in).

We have passed these out to people who promised to hack and tell us the 
key, or the bitstream.  No one has been able to crack them (yet). 
Evidently 3DES in V2 and V2P is still quite secure....

Austin

Erik Walthinsen wrote:

> Austin Lesea wrote:
> 
>> Very few people in your situation are designing boards that are going 
>> to be hot-plugged, and even fewer care what happens when you do.  For 
>> those who do care, we have a full range of documented proven solutions.
> 
> The design I have in mind right now is hot-plugged, but only on the USB 
> side.  The FPGA will be glue between the USB chip, DRAM and flash.  I'm 
> hoping that the power-up requirements of an XC2S50E or -100E won't 
> require any significant control circuitry.  Eventually I hope to reduce 
> the logic requirements to a lot fewer pins and gates, into CPLD realm, 
> but for experimentation I want to have every pin of every device routed 
> through the FPGA, even if half the logic is just crosspointing.
> 
>> I am sure that must be tough going, but most hobbiests realize that 
>> buying exiting development boards (like the Spartan 3 student Digilent 
>> pcb for $99) is a better choice than trying to etch your own pcb's in 
>> the sink.
> 
> I have this board, and am going to try to do my initial development by 
> building boards to plug into the 40pin connectors, but among other 
> things that limits me afaict to an x8 DRAM instead of x16, due to lack 
> of pins.  A x16 JEDEC pinout for 256 or 512Mb has 16 data lines, 13 addr 
> lines, BA0, BA1, RAS#, CAS#, WE#, CLK, CKE, DQML, and DQMH.  Total of 
> 39.  I'm still figuring out how SDRAM works, but I'm only sure I can tie 
> CKE, DQML, and DQML.  That gets me to 36, but there are only 34 I/O's on 
> connectors A2 and B1 of the Spartan3 board...
> 
> Flash is easier, only 15 pins.  USB I've already done hand-wired using 
> the ELRAsoft EZ-USB FX2 board, and in a manner IMO far superior to the 
> board Digilent sells to do the same thing (I actually connect all the 
> pins... x16 data and control).  Already done a basic test design all the 
> way through to user-space software in Linux on that setup 
> (host-controlled PWM of the 8 LEDs).

Austin Lesea wrote:

>
> All parts are in stock here at the factory (I checked).  That a 
> distributor doesn't order them is of course troubling, and another 
> issue.  That Digikey doesn't carry the latest and greatest products 
> may be because they offer little to no support.  We prefer our 
> distributors to earn their percentages (that they make on the parts).

What the hell do I need (or WANT) anyone looking over my shoulder?  Let 
me have the
parts without the overhead.  After the Xilinx web site, I don't NEED 
anything else but
my simulator.

>
> Other distributors do not want to be undersold by someone who offers 
> no services (and we don't see any benefit in that either).
>
> If all you need are some samples, you should develop a better 
> relationship with your local X distributor or X representative.

What does a (high priced) distributor actually deliver to me?  I'm not 
working for Boeing
or other huge corporation.  Digi-Key is the right-sized distributor for 
me.  I look up the
part, and order it.  Very simple.

Jon

Hi Steve,

To clarify it for me, what is the likely lead times AFTER August for
the lower density Spartan3's, eg., xc3s200?

Thanks,

A long time Xilinx customer

Steve Knapp explained: Append 0974 to the order code, and every
distributor should be able to quote  short leadtimes. If not, contact a
better distributor. They are not Xilinx employees, but rather
independent businessmen. Xilinx has lots and lots of parts, especially
of the 3S200. This is just a logistical snafu, not a real availability
problem.
Sorry for the confusion.
Peter Alfke

Steve Knapp explained: Append 0974 to the order code, and every
distributor should be able to quote  short leadtimes. If not, contact a
better distributor. They are not Xilinx employees, but rather
independent businessmen. Xilinx has lots and lots of parts, especially
of the 3S200. This is just a logistical snafu, not a real availability
problem.
Sorry for the confusion.
Peter Alfke