Op-amp design: Bipolar or Cmos

I just completed another laser circuit design. Lotsa CD4000 in there and it worketh :-)))

I bet Supertex' foundry could build a CMOS opamp design with a few dozen volts VCC. Of course it'll be big, but ain't everything bigger in America anyway?

I don't care, I got silicon to burn (3000x3000 sq.um). I saw a lm358 compatible design with 1600x800. I might just scale up the 800x800 design and see what happen.

Problem is I don't have much up-front bucks, but back-end license bucks are possible. The customer would pay whatever we ask, if the chip works. This will either be my first or last chip, depends on how it turns out.

To use Supertex as a foundry will cost you dearly.

...Jim Thompson

If this is a serious project we should talk, so that you make what you _actually_ need rather than what you _think_ you need.

Wait until the last week of June or later... I'm up to my eyeballs in work right now.

...Jim Thompson

The customer is serious enough to start prototyping (3000 at a time).

OK, will check with you later. I am also busy with the bottom wafer anyway.

I know. Designing ultrasound machines for over 20 years now ;-)

Do you know what became of Telmos back in the good old days? They had some really nice HV capabilities.

AT&T had a HV, like 400 volts, cmos line. They used to fab some of the Apex hv opamps. DC specs were bad, and noise was horrendous. Interestingly, some of the Apex amps are now just parts on small boards.

John

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

Doesn't AMI still run High Voltage CMOS Processes?

Yes, it seems they do.

They claim up to 100V on 0.7um CMOS.

And they claim a "cost sensitive" C5X Process at 0.5um with 20V Supplies.

Robert

What? Didn't the profession redesign the 1970's Fairchild (CMOCFairchild) to come out with the weekly new integrated wonder of which the 700 series are still with us. It was fun to be in electronics then.

Stanislaw

Yep, that's worth a look-see. But Linnix was saying "3000", which isn't exactly high volume. Maybe a MOSIS run?

I only seem to get to Pocatello in the winter :-(

...Jim Thompson

MOSIS run is 10,000.

Except for Telmos back in the old days I somehow kept ending up with automotive but that worked. Not opamps, this was custom muxes with our own amps plus pulsers and some logic on there. The kicker was a 64 to 16 mux where the design was a bit of a white-knuckle ride. Then they came and bingo, worked right out of the box. That night we all went out to a pub ;-)

It still is fun! Or actually more fun than ever because now we have places like Digikey.

I always liked AMI, never been disappointed. They did some tough chips for us. So yes, Idaho does produced something other than potatoes. If it just wasn't so freaking cold up there.

And cheap PCB proto houses.

robert

Friend of mine built this:

May give you details on how to design your op-amp.

JB

Thanks Jim, for the slow learners : The voltages defines the insulation thickness, and a thicker insulation is to be compensated with area to get the same capacitance. Since the insulation thickness is linear to the voltage, as the area is to the capacitance, IMO, the area should scale linear with the voltage. Meaning double the voltage, and the area is also doubled.

Sounds reasonable. No ?

Rene

I'm not quite sure about the ratio, but...

In CMOS, thicker "insulation" as in gate oxide equates to more device area to get the same transconductance. Also the drain-source spacing must increase to prevent punch-through.

In bipolar, the epitaxial thickness goes up linearly with voltage, but isolation diffusion horizontal spread during the vertical diffusion increases area as well. Base to isolation dimension must increase linearly with voltage, etc.

I'm guessing it's at least a "square" function for both process types :-(

...Jim Thompson

Without saying too much to break my NDA with the fab, the target voltage is 15V. So, no point in going more than 18V to 20V.

By the way, what's the best way to make 5 to 10pF capacitors?