Optimum OpAmp Noise Performance

[jackinnj]

The current issue of EDN carries a Design Note from Linear Tech with graphs relating input resistance values AND noise performance -- giving suitable LLTC opamps -- it isn't always the correct decision to use an LT1028 or LT1115.

Of course, it is Linear so the ADI and BB parts aren't included.

Here's the link:
http://www.linear.com/pc/downloadDoc...21,P2440,D6539

[scott wurcer]

Try using a MM phono cartridge as your source resistance, pretty easy to actually do (I have). Removing the stylus gets rid of vibrational problems if any. The results may be surprising.

Properly load the cartridge too of course.

[mfc]

Heres a question on noise.

I've seen JFET input stages where the gate bias resistor can
be huge (sometimes like 100Gig).

Since the article says resistor noise is .13*sqrt(R) then
for a 100Gig resistor that would be 41uV/sqrt(Hz).

Does that mean that this is a real bad idea to have a bias
resistor that large...or is this resistor in parallel with the source
resistance, so it is essentially not a factor?

Thanks

[scott wurcer]

Originally posted by mfc
Heres a question on noise.

I've seen JFET input stages where the gate bias resistor can
be huge (sometimes like 100Gig).

Since the article says resistor noise is .13*sqrt(R) then
for a 100Gig resistor that would be 41uV/sqrt(Hz).

Does that mean that this is a real bad idea to have a bias
resistor that large...or is this resistor in parallel with the source
resistance, so it is essentially not a factor?

Thanks

This would usually be in an application with a capacitive transducer such as a condenser mic. The capsule capacitance rolls off the noise at a low frequency. For instance a Nakamichi CM300 capsule measures 55pF and uses 1GOhm bias on a 2SK118 JFET. That's about a 3Hz roll off, but your working against the 4000nV/rtHz of the 1GOhm R so the mic will show a rising low frequency noise spectrum. Now if you think about it since the noise of the resistor goes up as the square root of R and the pole goes down as R, you win with a bigger R (i.e. a 100GOhm R will have 1/10th the noise at some low frequency).

[peranders]

Originally posted by jackinnj
-- it isn't always the correct decision to use an LT1028 or LT1115

... exactly, you most know the info behind the figures also. Mosfets are very low noise in HF circuits but not for mic amps... it depends.

[mfc]

Originally posted by scott wurcer

...(i.e. a 100GOhm R will have 1/10th the noise at some low frequency).

I think you got this result because the capacitor forms a low
pass filter in parallel with the bias resistor, so a larger resistor value provides more attenuation...

At the same time the 100Gig contributes sqrt(10) times more
noise of ite own. Since this goes up by a half power, the net effect
is that the filter effect dominates and provides lower noise?

If there is no capacitor, would a really large bias resistor
be a bad idea? Also would LTSpice noise analysis model
the effect of large bias resistors?

Thanks

[scott wurcer]

Originally posted by mfc



If there is no capacitor, would a really large bias resistor
be a bad idea? Also would LTSpice noise analysis model
the effect of large bias resistors?

Thanks

You can create a non-physical abstraction with any simulator. A 100GOhm resistor has a 3Hz pole with .5pF so any real board or circuit would show noise roll off. The simulator will work fine with large resistances, but any real amplifier will have a pF or so of stray capacitance.

[mfc]

Ahhh...now it makes sense. Thanks Scott!

[mfc]

Playing around with LTSpice, I got this (somewhat)
non-intuitive result for noise in a resistive voltage
divider.