OPA818: frequency dependence of input current noise

Hello Davorin,

I am going to research the best way to answer your questions and add clarity.  While I am doing so, please take a look at these two TI resources available off our TI site:

Noise Analysis in Op-Amp Circuits

Noise Analysis for High Speed Op-Amps

Best,

Alec

Hi Alec,

I have read the application notes and find what I consider inconsistency in 'Noise Analysis for high speed op amps".  The figure 3 show frequency dependence of input noise voltage and input noise current densities.  However, the Section 6 applies only 1/f form of the input noise voltage density to calculate the total noise although it does the calculation up to 10 MHz where the input noise current density is 100 times larger than its low frequency value.  Why such very different treatment of the 2 input noise densities?

If one assumes that the the application note's plots and calculations are correct, that would mean that the input noise voltage density curve represents an actual frequency dependence of E(sub NI) noise voltage source, but that the input noise current density curve does NOT represent an actual frequency dependence of the I(sub BN) and I(sub BI) noise current sources.  However, the two noise curves are treated as equal in the Figure 3 and no comment to differentiate them in any way is made anywhere in the application note.

Could you please clarify the discrepancy and explain the situation?

My personal guess is that the noise calculation applies to BJT input op amps that have only 1/f dependence for both voltage and current noises as shown in the Fig. 2. I need  your guidance and explanation whether the same procedure applies to a JFET input op amp.  The frequency dependence of the input noise current density and its treatment in the total noise over a bandwidth calculation are the cause of the issue that should be addressed.

Best,

Davorin

Hello Davorin,

Thank you for providing your detailed response and study.  I am going to look into your discovery.  Please allow some time to look into the complexity of the noise analysis.

Best,

Alec

Hello Davorin,

I can confirm the current noise density increases with frequency on FET transistor designs.  Inherent characteristics of the FET transistors, including the resultant input capacitance of the amplifier, contribute to the behavior you discussed above.

Unlike BJT amplifiers, with a high 1/f noise followed by low consistent broadband noise, FET amplifiers have a relatively low input current noise at low frequencies.  As higher frequencies are approached, the FET input current noise increases until it rolls off at -20dB/decade once the amplifier no longer has gain available.  

The analysis of current noise in FET amplifiers is difficult to find in official collateral.  I commend your analysis and your results suggest you are able to accept the difference as I restated above.

I encourage you to continue to find additional resources and investigate if you have further questions.  I have shared what I can on this subject.

Best,

Alec

Hi Alec,

Thank you for your kind words but this does not resolve my issue.  I am not sure about "official collateral" but I dare to suggest that TI builds a TIA with OPA818 and measure its noise up to, say, 40 MHz and then check whether the measured data fit the BJT based noise calculation or not.  I believe many potential users of the part will be very interested in the application note that would results from this effort.

Best,

Davorin