production spread of opamp GBW

What level of GBW production spread should I assume for worst-case testing? the opamp is a TLV274, GBW 3MHz or so.

Cheers Terry

I read somewhere recently (LTC appnote on thermocouple acquisition, I think) that opamp gbw and open-loop gain can have bad tc's, so it's best to way overkill on gain when precision matters.

John

I don't know what process that part is on, but on-chip resistors and capacitors typically have +/- 20% or so tolerance for each type of component. In practice the variation is likely to be much less unless they switch to a different fab or shrink the die etc. Whether you would believe that it is worth planning for a worst case variation in both R and C simultaneously is up to you.

Chris Jones

I dont care about Aol, only GBW. Its easy to do the simulations, but I still kind of need a guesstimate for GBW range. 2-6MHz is perhaps enough?

Cheers Terry

Easily I would say. +/- 1MHz would be my guess.

Nice to have that data though. Just been using an Infineon CoolMOS part and the data sheet shows both typical Ron and another 'worst case' curve for 98% of all devices. They don't say how bad the other 2% are though !

Graham

The compensation corner is sufficiently below the transistor effects that they can be ignored.

Resistors are typically +/- 20%

Capacitors are typically +/- 15%

GBW is proportional to 1/RC

So you're looking at 0.725x < GBW < 1.47x

BUT, The data sheet will usually specify a minimum, so max GBW would be ~2x the specification minimum.

(TC's ignored, but typically inconsequential.)

...Jim Thompson

FWIW I was once privy to "real" Hitachi electrolytic cap data. The lifetime figure they quoted on the datasheet was 3 sigma *below* the production mean, which was IIRC twice rated.

Cheers Terry

Jim's simplification, GBW proportional to 1/RC, still holds.

Note that GBW is usually determined by:

Av=gm.Xc

where, gm is 40Ic, Xc is the capaciter reactance. Gm is not a resister.

However, this current is ultimately set by a resister, somewhere.

Kevin Aylward snipped-for-privacy@anasoft.co.uk

"Somewhere" = the bias setup. The better designs have the tail current set by a PTAT, keeping gm relatively flat.

...Jim Thompson

Hi Jim,

thanks for that - exactly what I/m looking for. Archived with thanks :)

Cheers Terry

Yes, but thats what I meant by "this current is ultimately set by a resister".

Kevin Aylward snipped-for-privacy@anasoft.co.uk