Noise floor changes with timebase

Post discussions on projects you are working on
Post Reply
justinis
Newbie
Posts: 0
Joined: Wed Oct 27, 2021 11:04 pm

Noise floor changes with timebase

Post by justinis »

I am collecting data using a 5444D picoscope. The input on the scope is shorted. I used python to make two different data collections (600K samples each) that were identical except the timebase for one was 6 and the other was 34. I created a PSD of the data using scipy.signal as follows:

Code: Select all

fs = 1/np.abs((t[1]-t[0])) #calc sample rate
freq, psd = signal.welch(volts, fs=fs, nperseg=20000, window='boxcar')
plt.semilogx(freq, np.sqrt(psd))
In the resulting plots, the noise floor of the measurements are an order of magnitude different from one another. Why is this?

I tried collecting the data with the picoscope software instead of python, changing only the ms/div between measurements. I got a similar result.

MicheleFBK
Newbie
Posts: 0
Joined: Mon May 04, 2020 5:09 am

Re: Noise floor changes with timebase

Post by MicheleFBK »

Hi justinis,
I have observed this behavior in other devices and I think it is due to the noise of the wideband input amplifier being compressed in the 0-Fsample/2 band (i.e. aliasing of the input noise).
If this is the case, you should observe increasing noise as you reduce the sampling frequency.
To reduce the problem you should sample at the highest possible frequency, then apply a digital filter and decimate the data.
Michele

justinis
Newbie
Posts: 0
Joined: Wed Oct 27, 2021 11:04 pm

Re: Noise floor changes with timebase

Post by justinis »

Thank you! I have come to similar conclusions. I think the scope drops samples to reduce the sample rate, which makes noise measurements inaccurate.

Post Reply