Op-amp design: Bipolar or Cmos

To be honest I never really cared. As long as noise, GBW and whatever else we need are fulfilled the only variable that matters boils down to one symbol: $

I know this doesn't help you much here but I wanted to share what many design engineers think.

Who is "WE"?

CMOS OpAmps are more "popular" because they're cheap. They're also generally low power, resulting in high noise, high VOS, and low gain-bandwidth product

No. But I can design one for you once you have a set of specifications.

Yes. I started designing when there was ONLY Bipolar

None. Better offset, better gain-bandwidth product, better slew-rate, higher output power capability, higher operating voltage capability.

Depends. I just designed a CMOS amplifier with 0.03% linearity for a ±1V (differential) output.

...Jim Thompson

I have been staring at this post for a few minutes and I still can't decide if it is a joke....

Maybe I'm out of touch, is this something that companies can just decide to do now? Perhaps I should be getting into the diode business, or something?!

What about input bias current / noise?

If you're planning on building some chips that are _just_ op-amps this seems like an odd thing not to know. And if you _are_ planning on embedding the amp into a chip, I would think that the process of the chip that you're building would form a huge flashing arrow pointing at the process that you would have to use with your amp.

CMOS op amps have vanishingly small input current, as a consequence of the fact that you're looking into an insulated gate. In general they have significantly higher offset voltages than bipolar amps, so you have to pay close attention to your requirements and decide which amp is right for you. Keep in mind that bipolar and CMOS op-amps are happily coexisting in the market today (I dunno if there are any new JFET-input designs out there, however); if one technology would do for everything the others would have faded away.

I'm not sure that you'll find that equivalent-performance parts will have significantly different power consumption numbers -- unlike logic, CMOS analog circuitry needs some standing current to bias everything into the linear range. This standing current consumes power, so CMOS amps are certainly not "no power" parts like CMOS logic that isn't being clocked.

If you're going to be enabling the op-amp on demand, then you should probably calculate overall power consumption the way that TI advocates for their microcontrollers: count the coulombs that it takes to turn the op-amp on, wait for it to stabilize, use it, then shut it off. If you have to have a microprocessor on while this is happening, factor that in, too. A circuit that consumes 10mA but can do everything in a microsecond would be better than a circuit that consumes 100uA but needs a millisecond before it can shut down. The above numbers are absurd, but I hope you get my point.

Yeah, but at the frequency range where that begins to matter I design most everything with discrete parts ;-)

Ok, I confess that I have violated that rule once this year by designing in a THS4021. Promise repentance...

This comes about because management consists of MBAs and C programmers, neither of which has a clew about engineering.

Tam

Often lower cost as well, simply because they are older and the sheer volume is so high. I mean, where can you get a rocket like ye olde uA733 for around 30 Cents in CMOS?

I think Sphero might be talking about external modulated RF signals causing artifacts in quite low frequency circuits. E.g. thermocouple amplifier next to a transmitter.

Yes, it will be embedded in a device. We can go either way (bipolar or cmos), with least design costs and masks. So, we want to start with something well tested. I am not doing the design, just looking for contractors to do it.

If bipolar is good enough, we will go with the reference design. It might not be perfect, but save time and money.

That's what I thought, probably not much difference for analog. I am just trying to get some opinions before arguing against the well respected author.

The microcontroller will be coming out of stand-by, turning on the charge pump and op-amp. I think we have seconds to do that.

I seem to recall linnix as pretty clewful though... That's why I was puzzled.

Most important factors in buying op-amps:

1. Wide color choice. Women tend to favor bright red and green, men usually go for the brown and navy units. Get Eddie Bauer as a fashion consultant, even.
2. Cupholders. More cupholders, the more buyers.
3. Salesmen. Salesmen with white leather belts and white leather shoes seem to be the most succesful.

Good luck!

Tim.

We figure the incremental costs of 10 to 20 cents, cheaper than the external packaging costs. Of course, we have to eat the $30,000 NRE.

Have you looked at the price of "ye olde" LM358? Not the greatest opamp perhaps, but then I don't suppose the ones inside a "system-on-chip" are that great either. If that is what you are doing

- it's still not clear to me.

Input linearity can quite noticably affect EMC, and thus $.

Best regards, Spehro Pefhany

Forgot bias current ;-) Bipolar is higher bias current, BUT input referred noise is MUCH better in Bipolar... show me 2nV/rt-Hz in CMOS ;-)

Except for noise, BiCMOS is best.

...Jim Thompson

Yes, that's what we are designing out.

If I can have the layouts please (GDSII or CIF files). The dice need to be fitted into the wafer in exact places, but not necessary efficient usages.

Sorry about the misunderstanding. I am talking about adding an op-amp inside a chip. It will be part of a wafer cap anyway, so only minimum additional processing costs. It will save bonding cost and PCB spaces.

And certainly not a CLUE ;-)

...Jim Thompson

I don't want to start a flame war (probably did). The best IC designers I know are C programmers, who wrote the design programs.