power opamp?

Your suggested approach is probably the least fraught with possible stability problems.

But when I need lots of near-rail-to-rail swing I put resistors in the VCC and VEE lines and add the transistors there (e.g., base to VCC pin, emitter to rail, collector to output point... if you don't need rail-to-rail, tie OpAmp output to collectors).

...Jim Thompson

Single:

Slower dual:

The 8272 looks good, especially so since I'll be essentially driving a bypassed power rail, and it doesn't mind capacitive loads. It's $1.30 at 1K, but it would save me a lot of parts, 16 times. Boosting a small opamp into a bypassed load would add, say, 7 more parts.

John

Like this....

v+ | v+ R | | e \VCC----------------------b pnp \ c \ | \--------------------+-------OUTPUT / | / c /VEE---------------------b npn | e R | | v- v-

...Jim Thompson

Yeah, that's neat but harder to analyze for stability. One would

I can imagine way too many oscillation modes!

We recently did something similar, using opamp supply currents to drive transistors sort of like that, but current-mirror mode, to turn the opamp into a fairly precise current source. The sim we did didn't work at all, apparently because the opamp model doesn't actually pull current from the rails!

John

If you tie "Output" back to the OpAmp out it resolves to just a compound follower.

Spice models can be treacherous, which is why I roll so many of my own.

...Jim Thompson

You can't use a regular amp with biased npn/pnp into the rail and then sum the 16 others into its input?

Why not? I once had to cram over 50 parts onto the size of a standard

If I had trusted SPICE every time many of my design would not exist today. Can't remember the last time I fired up SPICE. Some time last year I guess.

But you aren't building a very complex structure from individual devices. My chip designs are almost always way over 1000 devices.

...Jim Thompson

True. However, some designs are, let's say, highly unorthodox. For example SPICE often isn't very useful when you have to use the saturation behavior of a custom-mix ferrite as a design parameter. Or in detecting and handling mode jumps of a laser diode. Or mapping sound wave reflections from a biopsy needle in a diseased area of human tissue.

It sounds like, pretty soon, you and Jim are going to have to collaborate -- your scales of construction are constantly converging!

Tim

-- Deep Fryer: A very philosophical monk. Website @

Anybody got decent models for ferrite beads?

John

I seem to recall seeing one... maybe Google ?:-)

...Jim Thompson

=A0|

=A0c

=A0c

=A0|

The supply voltage is lower than you want - absolute maximum +/-6.5V - and the guaranteed output current is only 50mA, but it has a surprisingly potent output stage.

-- Bill Sloman, Nijmegen

For pure frequency domain stuff, yes:

However, saturation is another matter. I've not seen much there that I could really use. In the end you have to design a circuit that can find the onset and slope and auto-calibrate. Ferrites are like a pasta dish. When Giuseppe cooks it tastes different than when Antonio is the chef.

ad826 is close I think, and dual

-Lasse

Have you looked at the audio headphone or speaker drivers? National makes dozens. If they are stereo, you only need to do it 8X.

Tam

Check out if an audio amp might fit. For example, one version of the LM386 goes to 18V single supply if you really need 15V total:

Those are really cheap, well under 50c in qties.