Opamps - everything

Well, you can reduce the effective GBW by increasing the noise gain. In an inverting amp, if your feedback resistor is 1M, putting 1k from the summing junction to ground will make it 1000 times slower. It'll also have 1000 ti mes the offset and noise.

Slew rate is unaffected, but of course it'll need a big input voltage to ge t it to maximum slew.

For a normal unity-gain-stable bipolar op amp with pole splitting and no em itter degeneration or current-booster funnies in the input stage, the slew limit is always near 0.3V*GBW.

The National tutorial app note by (iirc) Widlar and Dobkin explains why. It s title is something like "the monolithic op amp: a tutorial guide".

In a general-purpose device, a low GBW makes high slew rates useless, since feedback couldn't control the slew before the amp railed. Making an order- unity change takes roughly a radian at the unity gain cross. In your exampl e, an amp with 1 kHz bandwidth responds in a couple of hundred microseconds , by which time the output could notionally have slewed way over 100 kV.

Cheers

Phil Hobbs

So far the LT6090 is working fine. I used an apex opamp in a design for years, they upgraded the amp (twice) It got worse each upgrade, but I didn't notice it till the last one, where the noise was enough to show up in routine testing.

George H.

Such as some ancient Burr-Brown designs which switched in extra current for a large differential input.

For GBW in MHz and "modern" architectures. My venerable MC1530 (still sold by Lansdale) has ~10MHz and 10V/us, but sports a zero in the compensation and voltage gain after the pole.

Current mode feedback OpAmps are different. For instance Apex PB64 has ~3MHz GBW, but 800V/us slew-rate for large step inputs.

...Jim Thompson

The AD8065 does something like that. When the jfets run out of gas, bipolars kick in.

Current-mode amps automatically kick in more current as the input error increases. There is no diff-pair bias current to limit internal drive. Just keep stealing more of the user's signal! A lot more.

There was a boosted 741 as well--741S, maybe?

Apart from the noise and offset, a CFB behaves a lot like a VFB with that small resistor to ground from the SJ--the bandwidth is almost independent of the gain. That's how I usually think of them.

Cheers

Phil Hobbs

I don't remember that one.

Yep. Constant risetime instead of flat slew-rate.

Got in big hassle with a customer that didn't understand that :-( ...Jim Thompson

Was that the uA748 ?

I actually specified 741 in a new design as recently as 2006 for an undemanding application simply because they were so incredibly cheap and multi-sourced.

piglet

The 748 is just a 741 with an external compensation cap. This one had (iirc) about 5 V/us slew rate, vs. 0.3 or 0.5 for the 741.

Cheers

Phil Hobbs

I reckon you were right. Just *** big pages :)

NT

Yep. I wrongly supposed that the readers here didn't need help with applications >:-} ...Jim Thompson

Yeah, it's more for complete beginners.

NT

TI have quite a nice paper by Jim Karki on the differences between the idealised mathematicians perfect OPAMP and their engineering reality.

The 'e' component in LT Spice is almost an ideal opamp.

The 'E' component is in _all_ Berkeley-compliant Spice variants. 'E' alone gives fake results because it has no roll-off, or, almost more

"Op-Amp-Config.zip" on the Device Models & Subcircuits page of my website presents a fair representation of real OpAmps... and is configurable by the user by simply inputting datasheet information.

It is several generations old... I have newer versions that are quite accurate, but require a "spreadsheet" to obtain the coefficients of the model.

I do not share the spreadsheet. only the application-specific subcircuit >:-} ...Jim Thompson

It works fine for DC analysis. If you want a rolloff, add an RC at the output.

Its advantage over UniversalOpamp2 is that it doesn't need power supplies. There used to be a UniversalOpamp that didn't need supplies, I think, but it's gone.