Integrator Opamp Choices.

[greylourie]

When looking at various schematics, the integrator choice is a fet-based opamp. As a novice,this is something I completely overlooked !
When using the parametrics on ele14 website (UK), there is currently a choice of 4488 products. Looking at the selection, only 31 products are listed as FET/JFET input based.
Is FET the only way ? (LT offers lots of alternatives at the bottom of each of their product pages, and some of those are suggested for sample/hold - but dont mention the "FET".)

What are considered the main points required for integrator opamps for P.I ? How great does the slew rate need to be ? (What if we choose to use CMOS/TTL for our sample switching ? Are the two (slew/switching) somehow related ?)
What about GBW/bandwith product ? Too low is not good, but how low or high can we go ? CMMR ?

Be nice if to have a table perhaps listing the choices and their characteristics. I made a crude attempt a while ago, but it was based around the popular TL074/2, and I scrapped it in frustration as the number of opamps listed grew-and the characteristics diverted. It listed 5 parameters : BW, Slew, Offset, Supply/and consumption per stage,CMMR/PSSR. What other parameter/s should be included ? (maybe noise ?)

Maybe I should just calm down, and take one minipulse plus -as per Doctors instructions.

[Davor]

There are integrators and integrators. My favourite would be "2C integrator" as it does not suffer from fast phenomena at its input. A more common inverting integrator (with 1C) may suffer from glitches, as the 0V point is created actively, and the active element (the op amp) may not be fast enough to compensate.

Common ground for all of them is that you wish no bias current at the op amp inputs, or otherwise your integrator will integrate the bias current as well. Therefore the FET input op amps are usually used. MOSFET based op amps could do fine with rail to rail operation and all, but they tend to be awfully noisy. Trouble with noisy op amps is that the most of the noise falls in a 1/f region, and it means that the noise rises towards the lower frequencies below a certain "corner frequency". At 1Hz you'd have a lot of noise, but heck, 1Hz noise is just like waving your hand in random fashion, and in effect it performs as if the offset between input pins is unsteady. Hence it would present itself as a chatter in audio.

Some bipolar op amps have bias current compensation circuitry, but they seem to be half way there in all directions.

[Kev]

Hi Grey,
I'm no expert, but I believe on an integrator a FET amp is also used because it has a very high input impedance, so they don't load the sampling circuit and integration capacitor and skew the result. Additionally they have a low output impedance so will deliver a strong result to successive circuits.
Cheers

[Silver Dollar]

If you poke around here a bit you can find some threads on opamp selection. I was just reading one
the other day where the famous Ferris Toes was trying a few different ones in a super fast PI. Some
of the better specs ones didn't do too good as they had trouble coming out of saturation. There are
a few things that are not on the data sheet that help sample early on a PI unit...

Have a look here; http://www.geotech1.com/forums/showt...nt-End-for-P-I

[greylourie]

Thanks for the input, guys. Going to read the thread linked by Silver Dollar.

[Carl-NC]

As Davor said, FET opamps are used mostly to get zero input bias current to minimize droop error.