FPGA > ASIC

On 28 Dez., 19:28, General Schvantzkoph <schvantzk...@yahoo.com>
wrote:
>
> There are two reasons to build an ASIC, either your volumes are huge (in
> the millions) or an FPGA can't come close to handling the problem because
> of speed, size or power consumption. Building a state of the art

There are other good reasons to use an ASIC. Fpgas tend to force you
on several constraints. You can't expect to get an Fpga in very small
package (e.gt) less than 100 pins) and lots of resources, because you
could get either small package or high performance. You might have an
design with small logic and lots of ram which forces you to the high-
end fpga, but the design itself could be done on an small and cheap
asic, etc.
As long as your design is mainstream, you are right with the reasons
to change to ASIC. But some designs have constraints that are badly
supported from fpga vendors.

bye Thomas

Mike Treseler wrote:

> US Labor statistics
> 
> http://www.bls.gov/oco/ocos095.htm
> 
> says there were
> 196,000 professional musicians and singers in the US in 2006
> 
> Let's guess that 30,000 are guitarists
>                  10,000 use an effects box
> 
> May not need that asic.

For every "pro" musician there are gazillions of weekend warriors that 
buy effect boxes. There is sizeable market for this sort of thing.

But I agree an asic NRE does not make economic sense when you can buy 
something like an analog devices SHARC DSP that can give you 300 million 
MACs/sec for under $10.

-Jeff

RealInfo wrote:
> Hi all
> 
> I have a litle dream to turn a FPGA based design into a real ASIC so I can 
> sell this ASIC in the market .
> This is an audio processing design aimed at the guitar effects market.
> 
We have a couple nice little analog signal processors (with a fair 
amount of digital control logic, too) that we have made in the 0.5 um 
AMI CMOS process through MOSIS.  A minimum run of 40 of these chips 
costs us about $11,000.  A larger run of a couple hundred parts runs to 
about $45,000.  Remember that not all of these components are good, 
although the yield has been remarkable, over 90%!

Perhaps your DSP, all-digital design could fit in a smaller die, and 
therefore cost a bit less.  One thing to watch out for is that with the 
smaller feature size processes, the mask costs go up exponentially.  If 
you make 100K parts, that is not a concern, if you make 1000, it will 
KILL you!

I doubt you can find anyone that can put a single waffle tray of parts 
on your desk cheaper than MOSIS.  But, they won't layout the chip for 
you.  There are various packages that can do the layout, such as Mentor, 
Cadence, etc.  Note that there are restrictions on what can be done with 
the educational versions of their software, specifically, you can't use 
the educ. version to make commercial chips!

Also, note that there are REAL risks in ASIC design.  Digital stuff is 
"fairly" straightforward, but you still need rigorous and conservative 
design rules, and take all representations from the foundry with a large 
grain of salt, or you could end up with a pile of sparkling trinkets 
instead of functional chips.

Why do you think you can't sell an FPGA design?  I've been selling 
boards with FPGAs in them for years to a niche market, and certainly 
have NO plans to go ASIC on them!

Jon

john_griessen wrote:
>> RealInfo <therighti...@yahoo.com> wrote:
>>> I have a litle dream to turn a FPGA based design into a real ASIC
>=20
> viasic.com is another low cost metal mask programmable ASIC vendor.

I did not know this company so I went there and downloaded a PDF,
that writes :

"
Available processes include:
    =B7 TSMC 90nm
    =B7 ST 0.13
    =B7 TSMC 0.13
    =B7 DongBuAnam 0.13
    =B7 IBM 0.13 Rad Hard
    =B7 AMS 0.35 4ML
    =B7 Sandia 0.35 Rad Hard
"

I also read on another page :
"
Two mask high density configurable logic fabric built on
silicon proven standard cells that can be used as an
embedded block on an SOC or as a full chip fabric for
building structured ASICs.
"

So this seems (to me) to be just a provider of large standard
multi-purpose logic cells...


By coincidence, I discovered another company yesterday
that provides "via-layer electron-beam programmed FPGA-like chips"
(my own interpretation, not theirs) see http://www.easic.com/
They seem to provide the chips, not layouts/designs.
It seems interesting for low-volume ASIC-alike stuff.
I asked for pricing but all the reps are away...
I don't even know if anyone will answer
but their stuff looks interesting.


Happy new year everyone,

> John
yg

--=20
http://ygdes.com / http://yasep.org