1/f noise corner frequency

Hi Guarana,

Thanks for your interest in AD9689.

The device does not have 1/f noise corner frequency is because it is not considered anymore in testing the device's total noise due to the domination of broadband noise in high-speed ADCs like AD9689.

Regards,

Xavier

Hello Xavier,

Ok, you don't measure it but AD9689 has 1/f noise even if you don't consider useful to test it.

I don't ask you to measure it, just give me an estimate based on a simulation. 

Is this possible?

Regards,

N.

Hello Xavier,

Ok, you don't measure it but AD9689 has 1/f noise even if you don't consider useful to test it.

I don't ask you to measure it, just give me an estimate based on a simulation. 

Is this possible?

Regards,

N.

Hi Guarana,

I will ask our design team if they have simulations for this test. 

Can I ask what is the application or motivation of finding the 1/f corner frequency? What is your target sampling frequency?

Thank you and regards,

Xavier

Hi Xavier,

We do very long acquisitions (300us) with the AD9689 at  2GS/s.

Without signal, RMS noise is OK but peak-to-peak noise is much larger than expected. We would like to understand if the 1/f noise could be the reason of this.

Thank you and regards,

Nicolas

Hi Nicolas,

Can I ask for the waveform capture of your analog input and the FFT waveform? It would also be helpful to know the analog front-end schematic.

Thank you and regards,

Xavier

Hi Xavier,

We don't do any fft on the noise.

We record (524288 samples @ 2GS/s = 262us)  We observe a noise ratio (pkpk to rms of 10) instead of 6. 

The front-end we use is a SiGe fully diff amp exhibiting very low 1/f noise corner.

Thank you and regards.

Nicolas

Hi Nicolas,

Oh I see. I am now getting picture of where the 1/f noise came.

Can your share your noise measurement and also the fully diff-amp + AD9689 schematic? What is the device name of the fully diff-amp?

Thank you and regards,

Xavier

Hi Xavier,

Sorry but I cannot share the schematic on this site.

The fully diff amp is a full custom design.

My question is very simple. What is the 1/f corner frequency of the AD9689 input referred noise? Just consider the AD9689 on its own with 100 Ohm diff connected between IN+ and IN-.

Thanks and regards.

N.

Hi Nicolas,

1/f noise has normal distribution too, so what you have found is not an indicator of the presence of 1/f noise.

I think the problem can be with the interpretation of the peak-to-peak value, although I do not know how you have determined it.

6 RMS means that on average 99,73% of the samples fall in this range. Taking your case, you have 524288 samples, so you can easily have more than 1400 samples outside the 6 RMS range and therefore have larger peak-to-peak value. Using a normal distribution calculator I’ve found, that at 9.5 RMS you can still have typically 1 sample out of this range (due to statistics it can be more or less of course). This seems to fit your observation quite well.

I agree with   that if you are interested in the frequency dependence of the noise, you should use spectral analysis (FFT) and the analog front end can affect the results.

I hope I could help.

Zoltan

Hi Zoltan and Xavier,

Thank you for your relevant input.

I should have done the calculation, sorry it is my fault.

It's true that with such a large record, it's easy to fall out of the +/- 3 sugna.

Thanks a lot and regards,

N.