a low noise level opamp

Sorry, I use it as a output not a input.

It seems ok but not appropriate for my application. and I wanna have a output with noise level less than 20 micro V < 5kHz, the load will be Gohms.

Thanks

10uV rms with a 5kHz bandwidth implies an output noise density of vn = 10uV/sqrt(5kHz) = 141nV/sqrt-Hz, assuming white noise. This is an extremely high noise level you are allowing. Most opamps have voltage-noise density, e_n, well under 35nV, and many are under 5nV, some are even down to under 1nV.

Two other noise sources you may encounter are Johnson noise and opamp current noise, both involving your feedback resistors. If you use resistors below 150k-ohms the Johnson noise, sqrt(4kTR), will be under 50nV/sqrt-Hz. And with 150k resistors, if you keep your opamp's current noise, i_n, below 50nV/150k = 0.33pA/sqrt-Hz, then that source will also be below 50nV. Opamp current noise is just shot noise from the input bias current. If the bias current is below Ib = i_n^2 / 2q = 0.34uA, that source will be well under control. As you know most opamp bias currents are far below that. If you use a JFET opamp, that term will disappear entirely.

If you have four 50nV noise sources, you'll have 100nV of noise density, which is 7uV over a 5kHz bandwidth (the fourth source is a placeholder for your signal source or voltage reference). You may find that aspect harder to achieve, but if the first three terms each contribute under 50nV, you'll be allowed 111nV/sqrt-Hz for your signal source, while still meeting your 141nV budget.

--
 Thanks,
    - Win

There is nothing lower noise than a base-input, small collector current, npn transistor.

If an op-amp has such an input transistor then all well and good.

For an even better performance - freeze it.

I guess you are calling the following "nothing":

A medium power NPN transistor running at a modest collector current, A low noise JFET such as the IF3601, A varactor diode, A SQUID, A MASER.

I suspect that we could also add a power MOSFET to that list at low frequencies.

--
--
kensmith@rahul.net   forging knowledge

** Another Google Groupie.....

** If that 10uV is an EIN figure - even a 741 should do.

What is the source impedance ?

.......... Phil

"Reg Edwards"

** Huh ???

Has Reg never heard of source impedance issues ?

Never heard of low noise FETs either ?

Never heard of paralleling BJTs for low impedance, low noise usage ?

........ Phil

:) very thanks

the analysis is very very useful, and can you show me where I can get some sources on noise analysis? I checked some of electronic books and got little information. thanks again

The Art of Electronics has 38 good pages of noise discussion.

--
 Thanks,
    - Win

a general but usful one.

got it:)

thanks!

thanks

thanks

oh, your book:) my pleasure to enjoy it.

There are 2 equations that you are likely to need regularly for noise.

Where a device has a specified input noise voltage density ( typically measured in nV / sqrt Hz ) the wideband input noise *in a badwidth B* is simply given by

Noise ( rms ) = noise_density * sqrt ( B )

When making noise calculations you are also likely to need to consider the effect of thermal ( Johnson ) noise from resistors.

Thermal noise for a ressitor of value R at an absolute temperature T ( degrees Kelvin ) over the Bandwidth B ( in Hertz ) is given by.

Noise ( rms ) = sqrt ( 4 * k * T * R * B )

Where k is Boltmann's constant = 1.37 * 10^-23

Also when *adding* multiple noise sources be aware that due to its 'randomness' noise doesn't add arithmetically. Instead it adds as the root of the sum of the squares.

So for several noise sources N1, N2 N3 etc.....

Noise total = sqrt ( N1^2 + N2^2 + N3^2...... ) etc

Graham