Using FPGA to feed 80386 - Page 3

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
Re: Using FPGA to feed 80386
On Monday, December 15, 2014 9:25:12 AM UTC-5, Tom Gardner wrote:
Quoted text here. Click to load it

Because the guy who is developing this laser etcher looks like he has
high skills.  You see the multiple boards, the programming involved,
understanding a great many disciplines and bringing them together.
Plus, on his first video, he shows the filter lenses and posts a splash
screen warning about the dangers of laser light.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 15/12/14 14:30, Rick C. Hodgin wrote:
Quoted text here. Click to load it

Sure.

What about the people that make or use such a device?

There are too many ignorant and/or stupid people around.

Re: Using FPGA to feed 80386
On 12/14/2014 7:03 PM, Rick C. Hodgin wrote:
Quoted text here. Click to load it

It is not hard to make your own single sided PCBs for DIP packages and  
the larger pitch leaded surface mount parts.  But if you try to do  
anything beyond that you will find it is quite an effort.

Etching is not the hard part.  How do you connect traces through vias on  
double sided boards?  Even trying to do finer pitch parts on single  
sided boards is hard because there is no solder resist and shorts are  
very hard to prevent.

Remember, this video was an ad by Jameco.

--  

Rick

Re: Using FPGA to feed 80386
Rickman wrote:
Quoted text here. Click to load it

The solution is obvious: solder material cast
into rivets, with a special tool for automated
insertion and mechanical compression, with
even an optional heated welding as by a
separate lower power laser step.

Also, there's no reason a device couldn't be
constructed to laser etch both sides at the same
time, even post-processong for burn-thrus in
liew of mechanical drilling. Use of staggered
lasers would also allow faster throughput.

I think there's a business opportunity here.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
Rickman wrote:
Quoted text here. Click to load it

A simultaneous parallel laser could burn thru
a transparency to provide the solder mask.

An assembly line could even be created to run
them N-off in assembled form. Automated
deposition of solder paste, physical IC packages,
run through an oven, cooldown and testing.

It could all be done in a year from idea to
production.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 12/14/2014 10:59 PM, Rick C. Hodgin wrote:
Quoted text here. Click to load it

Hmmmmm.... or they could make them the way they do now and save the  
year!  :)

I was asking how *you* would do it at home...

--  

Rick

Re: Using FPGA to feed 80386
rickman wrote:
Quoted text here. Click to load it

Use an online service to get finished boards and
solder masks. :-)

If I had money to do so, I would move into this
area of profession. Reminds me of welding (you
create physical things rather than intangible
things as with software).

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 12/14/2014 11:48 PM, Rick C. Hodgin wrote:
Quoted text here. Click to load it

I expect it does take a bunch of money to make boards.  The equipment is  
rather expensive.  But doing the design can be fairly inexpensive still.  
  It's the debug that requires expensive equipment.  Scopes and such can  
rack up big bucks.

--  

Rick

Re: Using FPGA to feed 80386
ricknman wrote:
Quoted text here. Click to load it

I do have an oscilloscope. Got it in 1996. :-)

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 15/12/14 05:02, rickman wrote:
Quoted text here. Click to load it

Remarkably little, nowadays. So much so that it really is questionable
whether it is worth doing them at home. IMHSHO, the main reasons would
be "to see if I can" and "if I need it today".

Low end example: 10cm*10cm double sided, plated through holes, solder
mask, silkscreen. Cost $25 for 10 boards. (4 layer $50)
http://dirtypcbs.com/



assembly, e.g. http://www.pcbtrain.co.uk/


Quoted text here. Click to load it

Scopes can be had cheaply, especially second hand. High-speed probes typically
cost as much as the scope (>$10k isn't unusual!) Even decent 150MHz passive
probes cost $150.


Re: Using FPGA to feed 80386
On 12/14/2014 6:56 PM, rickman wrote:
Quoted text here. Click to load it
Listen to Rick: don't try to make your own boards.

I don't have the kind of professional board layout experience that many  
of the regular contributors here do, but I've played around at the  
experimenter level for many years. I have etched my own boards (very  
many years ago) and I don't recommend it. There are many way to mess it  
up, and even if you do everything perfectly you'll only produce a crude  
board. I haven't used the places Rick mentioned, but I've had good  
results from Alberta Printed Circuits (http://www.apcircuits.com /).

Regardless of how you get the board made, you'll need to design it  
first. There are lots of inexpensive (even free) CAD tools available and  
I'm sure you can get good recommendations here. You'll need to create a  
schematic as the first step in designing a PC board, even if you use an  
integrated suite those will generally be separate tools.

And finally I'm not quite clear: are you honestly considering designing  
and building a custom board for your project, or are you just curious  
about the process and the step involved? What's your experience  
designing and building hardware previously?



Re: Using FPGA to feed 80386
crisabele wrote:
Quoted text here. Click to load it

Minimal. Nearly zero with regards to ICs, and
zero related to FPGAs.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
I found these chips available for $10 each. 3.3V
fully static, 16-bit data bus.

    http://www.findchips.com/search/ng80386sxl-33

I am considering them.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 12/14/2014 11:25 PM, Rick C. Hodgin wrote:
Quoted text here. Click to load it

Considering them for what?  Are you planning to build something?

--  

Rick

Re: Using FPGA to feed 80386
On 15/12/14 04:01, Rick C. Hodgin wrote:
Quoted text here. Click to load it

If you are thinking of designing circuits and building boards
with FPGAs, then here's a couple of test questions. Good answers
are /necessary/ but not /sufficient/.

What is the highest frequency in your design?

When would you choose to use 6 layers rather than 4?

What types of decoupling capacitors would you use, where would
you place them, and how would you connect them?

What shape decoupling capacitor is optimal?

Warning: there's a *lot* of work and understanding required
to get accurate answers to those questions. Poor answers will
result in non-working boards or pattern-sensitive boards.



Re: Using FPGA to feed 80386
On Monday, December 15, 2014 8:18:16 AM UTC-5, Tom Gardner wrote:
Quoted text here. Click to load it

Probably 33 MHz to 40 MHz.

Quoted text here. Click to load it

I'm guessing:  When you need to isolate analog and digital traffic, or
when your pin counts are sufficiently high that you need access to a
common ground plane that's not closely coupled to the power source.

Quoted text here. Click to load it

No idea. :-)  Wherever the people on forums like this tell me to put
them after they review my design.

Quoted text here. Click to load it

Corn kernel shape. :-)

Quoted text here. Click to load it

No doubts.  The majority of the things I would do would be fabric
connection, and in many cases, using existing designs which already
have those things figured out, but with different programming in
the FPGA for whatever task I'm pursuing.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 15/12/14 14:29, Rick C. Hodgin wrote:
Quoted text here. Click to load it

Nope. You are probably too low by a factor of *~100*.
Hint: the clock frequency is completely irrelevant.


Quoted text here. Click to load it

Nope, not even close.
Hint: your circuit will have significant energy at GHz frequencies.


Quoted text here. Click to load it

You would be advised to read all the manufacturer's app notes.
For example Xilinx, for one family, has UG922 and UG483,
150 pages which are aimed at experienced PCB designers.

I hope by now you are considering buying an FPGA board and
adding your circuit to it as a daughter board or carrier board.
Ensure the connectors are impedance controlled and have good
grounding: many don't.



Re: Using FPGA to feed 80386
On Monday, December 15, 2014 9:54:58 AM UTC-5, Tom Gardner wrote:
Quoted text here. Click to load it

My project involves building a 386-like CPU.  I do not anticipate it
will run beyond 33 MHz to 40 MHz.

Quoted text here. Click to load it

I don't intend on running at GHz frequencies.  In fact, I'd like to
ultimately manufacture my CPU on 5,000 nm process technologies as
were used in mid-1980s.

Quoted text here. Click to load it

I appreciate your input.  My design plans are to use the Altera FPGA
I've purchased, to connect an Ethernet board I've purchased and get
it running, and then develop from there.  It will be all logic within
the FPGA at first.  That will take time, and during that time I'll be
in a position to ask questions, read, learn, etc.

We'll see though... I am currently iterating through ideas, trying to
figure out how best to serve the Lord with my efforts.

Best regards,
Rick C. Hodgin

Re: Using FPGA to feed 80386
On 15/12/14 16:38, Rick C. Hodgin wrote:
Quoted text here. Click to load it

Your problem here is that you don't know the basics of designing
schematics or pcbs for FPGAs.  People spend years at university learning
this sort of thing, then years of full-time professional work before
they consider themselves competent.  It is not reasonable to expect that
you can start from scratch and put together a half-decent FPGA board
based on nothing more than hints and advice from a Usenet newsgroup.

A key issue here is that you are thinking about digital hardware, when
in fact this is all high-speed analogue.  Designing /digital/ hardware,
such as the Verilog/VHDL code in the FPGA itself, is pretty simple -
it's all ones and zeros, and if you can get your head around clocking,
flip-flops and latches (and how to avoid them!), and keep to a single
clock domain, then it's not hard to pick up.

But pcb design is analogue.  You think your clock signals switch at 40
MHz, so your maximum frequency is 40 MHz - in fact there are relevant
parts of that signal up to 400 MHz and beyond.  You think your power
lines are at 3.3V (or whatever) - in reality, they vary and contain
significant signals at different frequencies, amplitudes and phases.
You think the 3.3V line is the same across the board, because it is DC -
it will in fact vary in DC and AC components throughout the board, in a
way that can easily be relevant if you get it wrong.  You think that
when a pin drives high or low, the whole signal track goes high or low -
on a fast pcb, the transmission time is relevant, as are reflections as
the signal moves along the track.

I've no doubt that you could learn all this stuff - or at least, learn
as much as you need to know.  But it will take time, and it will take
trial and error - which can become very expensive with FPGA's.  That is
why you are strongly recommended to buy an FPGA board, and limit your
pcb design to daughter boards for it (preferably avoid that too).  Come
back to the fpga pcb board design later when you have the time to spend
on it.





Re: Using FPGA to feed 80386
On 15/12/14 15:38, Rick C. Hodgin wrote:
Quoted text here. Click to load it

You appear to have missed my hint. Here it is again:
THE CLOCK FREQUENCY IS COMPLETELY IRRELEVANT.

If you use a modern FPGA then you *will* have frequencies
in the GHz region. How many GHz depends on the FPGA, IO
standard and drive strength.

Hint: google for risetime and knee frequency.


Quoted text here. Click to load it

So what? You *will* have GHz frequencies present.

Even if the max frequency is only 500KHz, you would benefit
from googling for transmission line, ground plane, inductance,
slot, ground bounce in the context of PCBs.


Quoted text here. Click to load it

Unless you are exceptionally perceptive, you will make several
iterations of that PCB, and unless you understand the cause of
the failures, random iterations/permutations will probably be
ineffective.


Site Timeline