How to calculate brainwave amplitudes for a moment in time?

Question

.NET

Good time, the
initial data is as follows: there is an EEG signal (sampling frequency 512Hz), there are ranges for different types of brain waves (alpha, beta, delta, etc.). That is, each wave is not a specific frequency, but a certain range (for example, 0.5 - 4 Hz are delta waves). In general, they cover the range from 0.5 to 50Hz.
It is necessary to calculate the amplitudes of these rhythms. You can normalize, i.e. the sum will be 1. The problem is that this needs to be done for a point in time. If we were talking about a minute of the signal, we take all these readings, build the FFT, and break it into pieces and calculate the area under the curve (then we divide by the total area under the curve). But in a minute the waves can change many times. It is necessary to produce the result 5 times per second.
Obviously, it is impossible to construct a spectrum for a point in time. I understand that the maximum frequency in the spectrum is half the sample rate of the signal (44100Hz gives a bandwidth of 22050Hz). But what is the minimum window width for a correct FFT calculation? How many samples do I need to take in order to be sure to say: yes, at this moment in time, the spectral pattern from 0.5 to 50 Hz is such and such?
Thank you.