Op Amp Dirty Little Noise Secrets

I immediately presume a chopper stabilized and/or auto-zero type op-amp, but the datasheet does not say. Unless I fell asleep during class so many years ago, after the 1/f knee the noise is considered to be white noise a.k.a broadband noise. The noise density is equal in a given bandwidth; i.e., the noise density is the same 10kHz - 20kHz or 22kHz - 32kHz so on and so forth (each bin being 10kHz).
I am totally unaware of a phenomena of gain/phase margin which would be a product of the internal compensation, would cause such a hump in noise; especially if we are talking about your typical 2-pole model of an op-amp. The gain/phase plot on pg. 8 suggest some funny business, but that occurs around 2MHz and does not correlate to the 400kHz hump in noise. 
I guess the long and the short of it is that Linear is taking measurements out that far because this is 1) ultra low noise, 2) wide bandwidth 10MHz!! amplifier. So any type of chopping or autozeroing may very well happen within the bandwidth of the amplifier.

Kudos to Linear for their datasheet. Lots of graphs and data. None of this "cherry picking" some vendors do.

If I am wrong, I would love to know what else would cause that large jump in noise.

-Ken

Ken:

Here is the only history I could find on TI's recognition of the issue. This was pasted from a thread over two years ago:


"..........The phase compensation in the OPA211 causes a rise in the noise density in the 1MHz to 100MHz range. This will be a contributor in your 20MHz measurement. It peaks to around 3.8nV/rt-Hz at 20MHz. This may be what you are seeing. The folks who characterized the device and wrote the data sheet had a noise measurement system that went to 100kHz and did not see this effect. Furthermore, most applications for the device operate in substantial gain, limiting the bandwidth so this effect is not seen.

Regards, Bruce" (Trump)

This seems to be a common problem in low noise op amps from all manufacturers. I am hoping TI can provide data for their low noise op amps so I can evaluate their potential for a low noise amplifier to 10 MHz.

Thanks, Cris

Cris,

It is common in most op amps that at lower frequencies the input voltage noise is dominated by the noise of the input stage since any noise contribution from the subsequent gain stages get divided by the gain of the input stage.  However, at the higher frequencies, especially close to unity gain bandwidth, the input stage runs out of its gain and the noise of other stages directly contribute to the input referred noise.  That is the reason why at higher frequencies there is a hump in the input voltage noise spectral density plots before it gets rolled off by the bandwidth limitation of the op amp.

You may simulate OPA211 input voltage noise and current spectral densities to any frequencies of interest using our macro-model.  I have done this here - please see the graphs below.

BTW, OPA211 and OPA1611 are NOT chopper stabilized op amps. 

Well, that certainly makes sense and obvious; running out of loop gain.

In my experience the most popular pieces of equipment used to measure noise is a spectrum analyzer or one of those QuanTech noise analyzers. I believe the QuanTech only measures certain frequencies; it does not sweep. The QuanTech is limit is 100kHz, and depending on the spectrum analyzer 100kHz is also the limit

If you ask me, there is something else going on with the Linear part so they would want to take the data. Ask them to reveal their secrets!

-Ken

Marek:

I understand that this noise behavior is a common occurrence in op amps, but it certainly is not well publicized.  Linear is the first company that I have seen disclose it on a data sheet.  Most op amp experts/designers have said they simply weren't looking out that far when the op amps were designed....and when they did, they were surprised at the levels they were seeing.

Do you have any data on the LME4990 or the OPA1611 for the noise further out.  Does TI have an op amp that does not exhibit this kind of noise peaking?

Thanks, Cris

Cris,

This is exactly correct. In all op amps, the input voltage noise spectral density increases at the higher frequency as the op amp runs out of loop gain. This is just part of the physics of this phenomenon – the input referred noise in multi-stage op amps increases as the gain of the first stage decreases. Needless to say that the higher the op amp bandwidth, the higher frequency where the noise hump occurs. The limitation of the spectrum analyzer used in noise characterization is the reason why in most cases the noise spectral density shown in the datasheets is limited to 100kHz.

For more details about the noise, please watch the TI Precision Labs series under link below:

training.ti.com/ti-precision-labs-op-amps-noise-1

I do not have the data on LME49990 since it comes from another group (I believe TI High Speed Product Line) but judging by its Gain Bandwidth Product of 110MHz, I would expect the noise hump to begin above 10MHz. As far as OPA1611 goes, it is based on the same chip as OPA211 and thus its input voltage noise spectral density is the same as the one I showed in my previous post.