Hi. I have a question regarding peak amplitude consistency on different resolution averaged FFT curves.

We have to measure according with any of the following protocols:

PROTOCOL 1:

Frequency band: 0.1 to 50 Hz

Frequency resolution: 0.125 Hz

Frequency average a minimum of 20 linear averages (about 2 minutes samplig, with the given frequency band and resolution).

Hanning window must be applied to the entire spectra.

PROTOCOL 2:

Frequency band: 0.2 to 70 Hz

Frequency resolution: 0.2 Hz

Frequency average 20 curves (2 minutes sampling)

Picoscope software doesn't allow us to have such a narrow frequency band, so we have to adapt to another protocol:

USED PROTOCOL:

Frequency band: 0.1 to 152 Hz

Frequency bins: 1024 (which means a frequency resolution of 0,148 Hz)

To keep 2 minutes samplig period, we have to average about 12 curves, not 20.

Hanning window is applied to the entire spectra.

Average FFT mode is used during acquisition.

Our main concern is to know if such adapted protocol is roughly equivalent to those indicated previously.

Our second concern has to do with the following:

If we change bins number, increasing frequency resolution, in such a way that sampling period remains the same (ie, taking lesser curves for average), we observe that main peaks amplitude is LOWER, and ambient noise is greater. We are not concerned about ambient noise but peaks losing amplitude is a great concern for us (see attached curves). We have compared:

- A 12-curves average with 1024 bins resolution (2 minutes total acquisition time)

- A 6-curves average with 2048 bins resolution (2 minutes total acquisition time)

- A 4-curves average with 4096 bins resolution (2 minutes total acquisition time)

- A 2-curves average with 8192 bins resolution (2 minutes total acquisition time)

- A single shot curve with 16384 bins resolution (2 minutes total acquisition time)

We can't assert which curve's peaks amplitudes are the ones representing the observed phenomena, nor have we any criteria to decide which acquisition protocol would be better. Could you give us any advice? Is there any problem with averaging algorithm in PicoScope?

Thanks in advance,

Lucas