Huge savnings for all, collaborative ASIC

[bitrex]

On 2/26/20 10:21 AM, Klaus Kragelund wrote:

On Wednesday, February 26, 2020 at 2:06:05 AM UTC+1, bitrex wrote:
On 2/25/20 8:00 PM, bitrex wrote:
On 2/25/20 7:47 PM, bulegoge@columbus.rr.com wrote:
Who is going to do the design? That NRE has to go to a person who is
going to get it right the first time.....which is unlikely unless you
have the infrastructure of a big company that you are trying to beat.
I would not be looking for here to make your dreams come true.


I think mixed-signal OTP chips like these Dialog "GreenPAK" arrays are
going to make a number of the proposed applications hard to justify
except in truly enormous quantity. The configuration software is easy
and fun to use, drag-n-drop more or less.

http://www.silego.com/uploads/Products/product_606/details/AN-1187%20High%20Efficiency%20Adjustable%20Buck%20Converter.pdf

I think a more reasonable goal for a "collaborative project" for the
group if that's what someone want to do would be to come up with
mixed-signal "programs" ready-to-run that could be used for some
purposes which are needed but aren't available at reasonable cost or
have problematic implementations.

I can't think of a reason off the top of my head why there'd need to be
a bespoke SED buck converter controller design that could be loaded into
chips in quantities of thousands instead of 100ks or if that would be
any cheaper than some other option but who knows.

Maybe I did not come out clear, it was not meant for the SED users. It's for the big players

Right, I don't know who "we" is. Fab prices have come down somewhat
tho, group-buys seem plausible at this point even for not-so-big players
if the group is 30 or 40 people perhaps...maybe if not a full chip at
least get someone to make a fast PNP again!

The volumes required to successfully vertically-integrate and beat the
chip companies on things like regular op-amps and buck converter
controllers would seem staggering.

 

[Klaus Kragelund]

On Wednesday, February 26, 2020 at 4:45:17 PM UTC+1, bitrex wrote:

On 2/26/20 10:21 AM, Klaus Kragelund wrote:
On Wednesday, February 26, 2020 at 2:06:05 AM UTC+1, bitrex wrote:
On 2/25/20 8:00 PM, bitrex wrote:
On 2/25/20 7:47 PM, bulegoge@columbus.rr.com wrote:
Who is going to do the design? That NRE has to go to a person who is
going to get it right the first time.....which is unlikely unless you
have the infrastructure of a big company that you are trying to beat..
I would not be looking for here to make your dreams come true.


I think mixed-signal OTP chips like these Dialog "GreenPAK" arrays are
going to make a number of the proposed applications hard to justify
except in truly enormous quantity. The configuration software is easy
and fun to use, drag-n-drop more or less.

http://www.silego.com/uploads/Products/product_606/details/AN-1187%20High%20Efficiency%20Adjustable%20Buck%20Converter.pdf

I think a more reasonable goal for a "collaborative project" for the
group if that's what someone want to do would be to come up with
mixed-signal "programs" ready-to-run that could be used for some
purposes which are needed but aren't available at reasonable cost or
have problematic implementations.

I can't think of a reason off the top of my head why there'd need to be
a bespoke SED buck converter controller design that could be loaded into
chips in quantities of thousands instead of 100ks or if that would be
any cheaper than some other option but who knows.

Maybe I did not come out clear, it was not meant for the SED users. It's for the big players


Right, I don't know who "we" is. Fab prices have come down somewhat
tho, group-buys seem plausible at this point even for not-so-big players
if the group is 30 or 40 people perhaps...maybe if not a full chip at
least get someone to make a fast PNP again!

Full wafer cost is about 900 USD, and if you make a chip that is 1x1mm (which has a LOT of functions), then the cost at Xfab would be something like 0..2 USD, Smaller designs, like a small opamp, can be made in 0.2x02mm, so a lot less die cost, but still package cost

The volumes required to successfully vertically-integrate and beat the
chip companies on things like regular op-amps and buck converter
controllers would seem staggering.

I am maybe optimistic. If you can make it for a lot lower cost, you have a way into the marked

 

[bitrex]

On 2/26/20 11:03 AM, Klaus Kragelund wrote:

On Wednesday, February 26, 2020 at 4:45:17 PM UTC+1, bitrex wrote:
On 2/26/20 10:21 AM, Klaus Kragelund wrote:
On Wednesday, February 26, 2020 at 2:06:05 AM UTC+1, bitrex wrote:
On 2/25/20 8:00 PM, bitrex wrote:
On 2/25/20 7:47 PM, bulegoge@columbus.rr.com wrote:
Who is going to do the design? That NRE has to go to a person who is
going to get it right the first time.....which is unlikely unless you
have the infrastructure of a big company that you are trying to beat.
I would not be looking for here to make your dreams come true.


I think mixed-signal OTP chips like these Dialog "GreenPAK" arrays are
going to make a number of the proposed applications hard to justify
except in truly enormous quantity. The configuration software is easy
and fun to use, drag-n-drop more or less.

http://www.silego.com/uploads/Products/product_606/details/AN-1187%20High%20Efficiency%20Adjustable%20Buck%20Converter.pdf

I think a more reasonable goal for a "collaborative project" for the
group if that's what someone want to do would be to come up with
mixed-signal "programs" ready-to-run that could be used for some
purposes which are needed but aren't available at reasonable cost or
have problematic implementations.

I can't think of a reason off the top of my head why there'd need to be
a bespoke SED buck converter controller design that could be loaded into
chips in quantities of thousands instead of 100ks or if that would be
any cheaper than some other option but who knows.

Maybe I did not come out clear, it was not meant for the SED users. It's for the big players


Right, I don't know who "we" is. Fab prices have come down somewhat
tho, group-buys seem plausible at this point even for not-so-big players
if the group is 30 or 40 people perhaps...maybe if not a full chip at
least get someone to make a fast PNP again!


Full wafer cost is about 900 USD, and if you make a chip that is 1x1mm (which has a LOT of functions), then the cost at Xfab would be something like 0.2 USD, Smaller designs, like a small opamp, can be made in 0.2x02mm, so a lot less die cost, but still package cost

If one needed large qs of something very specific perhaps that wasn't
available on the market at reasonable cost? I'm thinking like a
high-voltage op-amp where DC performance wasn't critical; that could be
like 8-10 transistors or FETs to make a perfectly usable device. even
with allowance for the larger-size trannies for the higher breakdown
voltage sounds like that would surely fit in 0.2x02mm

The volumes required to successfully vertically-integrate and beat the
chip companies on things like regular op-amps and buck converter
controllers would seem staggering.

I am maybe optimistic. If you can make it for a lot lower cost, you have a way into the marked

 

On Wednesday, February 26, 2020 at 7:23:10 AM UTC-8, Klaus Kragelund wrote:

On Wednesday, February 26, 2020 at 3:45:36 AM UTC+1, bitrex wrote:
On 2/25/20 9:01 PM, edward.ming.lee@gmail.com wrote:
On Tuesday, February 25, 2020 at 5:09:54 PM UTC-8, bitrex wrote:
On 2/25/20 8:02 PM, Winfield Hill wrote:
Klaus Kragelund wrote...

Just a good idea, my humble opinion: ASIC is expensive,
and the chip companies earn good profits on selling
commodity chips: opamps, regulators, diodes, ...

Some IC houses make very good designs, and are quite
affordable. For example: 3Peak, purchase from LCSC.
TP09 dual 6MHz opamp, 400uV Vos, 6 cents, quad 10 cents.



I just got some of these by courier, 3 days from Shenzen to the likely
much more Third World-looking DHL service point in Pawtucket, RI.

https://lcsc.com/product-detail/General-Purpose-Amplifiers_3PEAK-TP10-2-SR_C130325.html

A small warehouse by the train tracks with three tractor-trailer bays,
and two dented yellow trucks compose the DHL infrastructure for all of
Rhode Island I think.

You are lucky. I am still waiting for some stuffs i ordered a month ago.


Unfortunately having worked sales at some Internet retailers many years
back I came to understand that "in stock" is often bullshit, like what
didn't we have "in stock."

They can't ship since no planes are entering china and local post service is suspended also

I ordered my stuffs before Shenzhen lock down, but after WuHan lock down. Hoping to catch the window, some made it but some did not. Right now, the only way to get stuffs is to send a courier from Hong Kong to get them.

 

[Klaus Kragelund]

On Wednesday, February 26, 2020 at 5:10:43 PM UTC+1, bitrex wrote:

On 2/26/20 11:03 AM, Klaus Kragelund wrote:
On Wednesday, February 26, 2020 at 4:45:17 PM UTC+1, bitrex wrote:
On 2/26/20 10:21 AM, Klaus Kragelund wrote:
On Wednesday, February 26, 2020 at 2:06:05 AM UTC+1, bitrex wrote:
On 2/25/20 8:00 PM, bitrex wrote:
On 2/25/20 7:47 PM, bulegoge@columbus.rr.com wrote:
Who is going to do the design? That NRE has to go to a person who is
going to get it right the first time.....which is unlikely unless you
have the infrastructure of a big company that you are trying to beat.
I would not be looking for here to make your dreams come true.


I think mixed-signal OTP chips like these Dialog "GreenPAK" arrays are
going to make a number of the proposed applications hard to justify
except in truly enormous quantity. The configuration software is easy
and fun to use, drag-n-drop more or less.

http://www.silego.com/uploads/Products/product_606/details/AN-1187%20High%20Efficiency%20Adjustable%20Buck%20Converter.pdf

I think a more reasonable goal for a "collaborative project" for the
group if that's what someone want to do would be to come up with
mixed-signal "programs" ready-to-run that could be used for some
purposes which are needed but aren't available at reasonable cost or
have problematic implementations.

I can't think of a reason off the top of my head why there'd need to be
a bespoke SED buck converter controller design that could be loaded into
chips in quantities of thousands instead of 100ks or if that would be
any cheaper than some other option but who knows.

Maybe I did not come out clear, it was not meant for the SED users. It's for the big players


Right, I don't know who "we" is. Fab prices have come down somewhat
tho, group-buys seem plausible at this point even for not-so-big players
if the group is 30 or 40 people perhaps...maybe if not a full chip at
least get someone to make a fast PNP again!


Full wafer cost is about 900 USD, and if you make a chip that is 1x1mm (which has a LOT of functions), then the cost at Xfab would be something like 0.2 USD, Smaller designs, like a small opamp, can be made in 0.2x02mm, so a lot less die cost, but still package cost

If one needed large qs of something very specific perhaps that wasn't
available on the market at reasonable cost? I'm thinking like a
high-voltage op-amp where DC performance wasn't critical; that could be
like 8-10 transistors or FETs to make a perfectly usable device. even
with allowance for the larger-size trannies for the higher breakdown
voltage sounds like that would surely fit in 0.2x02mm

Yes, there is a lot of possibilities with ASICs. Just not many companies willing to take the step into that domain. If this is split across several companies, the business case will have an extremely short payback time, and it's a no brainer

There is examples of hobbyists doing designs, ordering with MOSIS, just 5-10 samples, for a total NRE of less than 1000 USD for the first samples

But, then you need to be proficient in ASIC design, and not many people are.. So one needs to use the consultants and that benefits from drawing on the long experience they have

 

On Wednesday, 26 February 2020 14:39:21 UTC, Winfield Hill wrote:

tabbypurr wrote...

max output swing 5/15mV. Kinda low for an opamp.

That means within 5mV of the rails, 15mV worst case.
And no more than 1V at 50mA loads, not bad at all.