FPGA > ASIC

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.

My question is how much can this venture cost in general .

A general estimate will be very good .

Is it realistic to search for an investor ?

Thanks in advance
EC

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.

So what quantities are we talking about?
Are you sure that an ASIC will be able to give You roi?

> 
> My question is how much can this venture cost in general .
> 
> A general estimate will be very good .

Depends on which way you are going. Are you using vendor specific IP-blocks?

> 
> Is it realistic to search for an investor ?

Think about Your story and then you might guess if that idea is worth 
some millions for an investor.

> 
> Thanks in advance
> EC
> 
> 

Maybe You'll want to have a look at Altera's HardCopy design-flow. ASICs 
created with that flow are not the cheapest ASICs, but are cheaper then 
FPGAs (and You won't need extra logic for configuring the FPGA). For 
quantities I suspect you are talking about HardCopy might be quite 
interesting.

Regards,

Lorenz

On Sun, 28 Dec 2008 11:12:05 +0200, "RealInfo"
<therightinfo@yahoo.com> 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.

An ASIC is only needed for very high volume and/or low margin
products. Do you really think that your product qualifies and it's
worth the effort to make an ASIC? You can probably use a low cost FPGA
to implement your audio implementation and in your market margins are
hopefully quite high. What's the cost of the FPGA you're using now?
What is your projected annual sales for the next couple of years? 
>
>My question is how much can this venture cost in general .
>
Assuming that you don't need a high-end process and 180nm would be
acceptable for die size/performance, your mask costs would probably be
around $250K and another $250K for converting your current RTL to ASIC
(including some hard IP like PLL & IO etc). Then you should be able to
buy 8 inch wafers for $1500 each which might give you 1000 5mm by 5mm
dice. Add another buck for packaging & test so you're looking at $2.5
per chip.
Now your real challenge starts: how are you going to advertise,
market, sell, distribute and support your chips? You can probably all
of the steps in the previous paragraph without getting off your chair
but the actual task of getting the chips to the customers and getting
them to use it is the main challenge. And this step is the same for an
ASIC or an FPGA. I'd suggest you start with an FPGA, make a product,
sell it and then depending on your volume see if an ASIC is justified.

>A general estimate will be very good .
>
>Is it realistic to search for an investor ?
>
Probably not for making an ASIC but maybe for getting your design into
the hands of your customers in any form. You have to find a market for
your product, make reasonable sales figures and have a good business
plan first and then search for an investor for your product whether it
uses an ASIC or an FPGA.

>Thanks in advance
>EC
>

Muzaffer Kal

DSPIA INC.
ASIC/FPGA Design Services
http://www.dspia.com

The last time I looked into this myself, about five years ago, there
were two main price regions. I found three manufacturers who had so
called 'low-NRE' product lines, two Japanese and one Korean. The lowest
price came in at about EUR50.000 for a minimum order of 25.000 pieces
and a smallish die size.
Just in case your not familiar, NRE 'non-returnable engineering' costs
is what you pay up front for mask manufacturing etc. At the time the
low-NRE product lines were restricted in what you could do (number of
metalisation layers, clocking schemes, max die size etc.) For audio
frequencies though I would expect it to be OK for you, assuming these
lines are still offered.

For the more mainstream product lines I would say you are talking about
EUR250.000 upwards. There are a lot of variables, principally die size,
quantities, whether you use third IP etc. Some manufacturers won't even
have a serious conversation with you if you are not planning a minimum
of 250.000 pieces per year. Although with a crisis looming they might be
temporarily more friendly.

As a rule of thumb, my personal feeling is that if you are not planning
more than 25.000 pieces per year or do not need some special features
like fancy power management, special IP, clocking schemes etc. migrating
to an ASIC is not cost effective.

You still have a number of options to market your chip. As another
poster suggested, Altera offer hardCopy, others possibly as well (I
haven't looked). Lattice for instance offer encrypted FPGAs in the low
cost segment (ECP2M/S. Here, you could sell your customers the FPGA with
an encrypted bit-stream for somewhere below EUR20,00 (for more accurate
figures, talk to a Lattice guy) per piece. I'm not sure if they even
charge you for the encryption key, probably depends again on the
quantities. I know of at least one video-IP provider who currently uses
this model.

RealInfo schrieb:
> 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.
> 
> My question is how much can this venture cost in general .
> 
> A general estimate will be very good .
> 
> Is it realistic to search for an investor ?
> 
> Thanks in advance
> EC
> 
> 

Muzaffer Kal <kal@dspia.com> wrote:

>On Sun, 28 Dec 2008 11:12:05 +0200, "RealInfo"
><therightinfo@yahoo.com> 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.
>
>An ASIC is only needed for very high volume and/or low margin
>products. Do you really think that your product qualifies and it's
>worth the effort to make an ASIC? You can probably use a low cost FPGA
>to implement your audio implementation and in your market margins are
>hopefully quite high. What's the cost of the FPGA you're using now?
>What is your projected annual sales for the next couple of years? 

I think FPGA isn't the proper route either. I've made quite a
complicated audio processor around an ARM7 microcontroller and a
codec. The microcontroller costs a few dollars in larger quantities.
That's hard to beat!

-- 
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
                     "If it doesn't fit, use a bigger hammer!"
--------------------------------------------------------------

Many thanks to all who took time and effort to answer.

Be well .

EC


"Charles Gardiner" <invalid@invalid.invalid> ??? 
??????:gj7odu$lku$03$1@news.t-online.com...
> The last time I looked into this myself, about five years ago, there
> were two main price regions. I found three manufacturers who had so
> called 'low-NRE' product lines, two Japanese and one Korean. The lowest
> price came in at about EUR50.000 for a minimum order of 25.000 pieces
> and a smallish die size.
> Just in case your not familiar, NRE 'non-returnable engineering' costs
> is what you pay up front for mask manufacturing etc. At the time the
> low-NRE product lines were restricted in what you could do (number of
> metalisation layers, clocking schemes, max die size etc.) For audio
> frequencies though I would expect it to be OK for you, assuming these
> lines are still offered.
>
> For the more mainstream product lines I would say you are talking about
> EUR250.000 upwards. There are a lot of variables, principally die size,
> quantities, whether you use third IP etc. Some manufacturers won't even
> have a serious conversation with you if you are not planning a minimum
> of 250.000 pieces per year. Although with a crisis looming they might be
> temporarily more friendly.
>
> As a rule of thumb, my personal feeling is that if you are not planning
> more than 25.000 pieces per year or do not need some special features
> like fancy power management, special IP, clocking schemes etc. migrating
> to an ASIC is not cost effective.
>
> You still have a number of options to market your chip. As another
> poster suggested, Altera offer hardCopy, others possibly as well (I
> haven't looked). Lattice for instance offer encrypted FPGAs in the low
> cost segment (ECP2M/S. Here, you could sell your customers the FPGA with
> an encrypted bit-stream for somewhere below EUR20,00 (for more accurate
> figures, talk to a Lattice guy) per piece. I'm not sure if they even
> charge you for the encryption key, probably depends again on the
> quantities. I know of at least one video-IP provider who currently uses
> this model.
>
> RealInfo schrieb:
>> 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.
>>
>> My question is how much can this venture cost in general .
>>
>> A general estimate will be very good .
>>
>> Is it realistic to search for an investor ?
>>
>> Thanks in advance
>> EC
>>
>> 

On Sun, 28 Dec 2008 11:12:05 +0200, 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.
> 
> My question is how much can this venture cost in general .
> 
> A general estimate will be very good .
> 
> Is it realistic to search for an investor ?
> 
> Thanks in advance
> EC

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 ASIC 
(40nm, 65nm) costs millions of dollars, using a non-state of the art 
process is cheaper but your gains vis-a-vis an FPGA are less. The FPGA 
companies are the process drivers for their foundries so their current 
offerings are the first devices (with the exception of Intel which has a 
12 month lead over the rest of the industry) that come out in any 
process. The Altera Stratix 4 is a 40nm part for example. While random 
logic is inefficient in an FPGA other important features, like RAM and 
multipliers are as good or better than you are going to get in any ASIC. 
If you have a significant amount of RAM in your design you might find 
that the die size of a state of the art FPGA is smaller than the die size 
for an ASIC that's built in an antique process.

Your best bet would be to figure out how to fit your design into the 
cheapest FPGA that you can. 

RealInfo <therightinfo@yahoo.com> wrote:
 
> 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.
 
> My question is how much can this venture cost in general .
 
> A general estimate will be very good .

Probably better to keep it in an FPGA.  That is being done
now for many that would previously go ASIC as mask costs
are rising.  Unless your market is in the millions, FPGA
is likely a better choice.

-- glen

> RealInfo <therighti...@yahoo.com> wrote:
> > I have a litle dream to turn a FPGA based design into a real ASIC

viasic.com is another low cost metal mask programmable ASIC vendor.
Still, take the others advice if you need an FPGA based design cost
reduced
and shop for other programmable ways to get there first.  Then
concentrate on
your sales methods before worrying about selling 800,000 pieces to the
music makers
market...hmmm... could anyone sell >800K to musicians?  How many
musicians are there?

Less than kids that buy a guitar-hero toy, I bet.

How about programmable
analog and digital  PSOCs from cypressmicro for guitar effects?  Then
sell $45 units with a $15 BOM
to musicians.

John

john_griessen wrote:
>   Then
> concentrate on
> your sales methods before worrying about selling 800,000 pieces to the
> music makers
> market...hmmm... could anyone sell >800K to musicians?  How many
> musicians are there?


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.

     -- Mike Treseler