http://lists.intercom.it/pipermail/btcgroup/2012-August/000015.html[BTCGROUP] Update
Ken Shiring ken at shiring.com
Mon Aug 13 20:15:07 CEST 2012
Previous message: [BTCGROUP] Update
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Thanks for asking Paolo, this is an important issue we should discuss. Here
is what I know so far:
We have 2 options for funding schedule. Each option has pros and cons.
Let me explain a little first what our remaining design flow is, which will
give you some perspective on the process:
1. We receive the technology library from our services provider. I think
we will be using Globalfoundries' 65nm process.
2. I synthesize the design to the given tech library. I will have to spend
some time on this step to get best quality of results (QoR).
3. I handoff the post synthesis netlist to our services provider.
4. They perform several back-end steps, like scan chain insertion, place &
route, and DRC checks.
5. They produce a GDSII file (this is the file which represents physical
geometry at the mask level).
6. We handoff a GDSII file to the fab (Globalfoundries).
7. The fab produces a physical mask.
8. The mask is used to make wafers.
9. Wafers are cut, tested, and packaged into our specified packaging.
10. We get chips back from the foundry.
This is a quite a bit of detail, but I thought you should understand the
process so we can talk about how the different options affect our schedule.
1. Fast option: Highest risk. We give the design to our services provider,
and we do steps 1-10 as quickly as possible. In this case, we will produce
a mask that is usable directly for production. The mask that is made can
be used for high volume (10k+ parts/month). The fab does a small run of
wafers, and sends them back to us for testing. We test the parts, and we
find out if they are good or bad. If they are good, we use the existing
mask and go to volume production right away. There will be some fab queue
delays involved, but I don't yet know how long (we are still waiting on the
details from our provider). We can also pre-order any number of production
parts before we get prototypes (known as a hot run). This preorder is
another direct risk; we would get these parts before knowing for sure
whether the parts come back good. The size of the hot run is totally up to
us (we decide how much risk to take).
2. Slower option: Less risk. We give the design to our services provider
after doing steps 1-6. When the mask is made (7), we use a process called
a multi-project wafer (MPW). MPWs put many different designs on a single
wafer from many different companies. This means the cost of a mask (and
wafer) is split among many different companies, and the price goes down
considerably. However, only a small run of chips can be made. This
process is used for prototyping, and most of the large companies use this
model at 65nm or lower to mitigate risk. This option is considerably less
risky financially. However, if the design comes back fully functional, a
2nd mask set must be made for volume production. Step 7 must be repeated
to make a production mask set. There is a turnaround delay to make the 2nd
mask, which adds to the total project time.
With option #1, we would need nearly all the funds up front. With option
#2, we would need much less up front. However, option #2 still requires
the same total funding (slightly more in total due to the 2 mask sets) in
the end. I don't have specific cost numbers yet, but option #2 should be
~250k to get to prototypes. Option #1 is about ~1m to get to prototypes.
I think option #2 is the best idea from an engineering view. It puts the
smallest amount of capital at risk, and it also allows us to potentially
get started sooner if you are willing to fund the smaller sum up front and
raise the rest later. I need to call my services provider back and remind
them we are waiting on some numbers, so hopefully in a day or so I will
have concrete details for you.
Ken
On Mon, Aug 13, 2012 at 7:18 AM, Paolo Agazzone <aga at intercom.it> wrote:
> **
>
> Hello all,
>
> I hope that we finally found a working configuration for this list! :-)
>
>
> on August most people/companies are on summer vacations, so things are a
> bit slowed down.
>
>
> However I'll have a meeting with Luca on thursday where we will also
> discuss about this project.
>
>
> Meanwhile I have a question:
>
>
> in case we will successfully find all the required funds, in which steps
> will be they employed/required? All upfront, at the beginning, or there are
> several phases? If so, which would it be (ballpark figure) the amount
> needed for each step?
>
>
> Regards,
>
> Paolo.
>
>
> --
>
> ------------------------------------------------------------------------
>
> Paolo Agazzone aga at intercom.it
http://blog.agazzone.it/>
> Linux Reg. User: #205873 Vaprio d'Agogna (NO) - Italy
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Btcgroup mailing list
> Btcgroup at lists.intercom.it
>
http://lists.intercom.it/mailman/listinfo/btcgroup>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.intercom.it/pipermail/btcgroup/attachments/20120813/839f1518/attachment-0001.html>
