how can i analyze the fft of different length signals.?

Hi guys i am a bit confused about taking fft of a signal consist of different length signals. Let me explain it.

let say i am measuring a signal from a source [Signal_A]. The oscillations in Signal_A changes in different conditions(which is the spectral analysis part).

Let say I measure Signal_A for 100 seconds with 1000 sampling per second(which is my sensor frequency).

Let say,

Signal_A(t(0:10))=sin(2*pi*50*t),
Signal_A(t(10:30))=2*sin(2*pi*20*t),
Signal_A(t(30:80))=3*sin(2*pi*90*t) and
Signal_A(t(80:100))=4*sin(2*pi*100*t)

so my signal is :

Signal_A=[sin(2*pi*50*(t(0:10)')); 2*sin(2*pi*20*(t(10:30)')); 3*sin(2*pi*90*(t(30:80)')); 4*sin(2*pi*100*(t(80:100)'))]

(i gave the time series for each signal like (t(20:30)) but the signals are not continuous. For instance the third signal can occur multiple times like t(2:5),t(10:20),t(33:40) for a total of 20 seconds.)

My problem is that i can find each signals occurrence time, but i am not sure about the frequencies and amplitudes. I am taking fft same as this http://www.mathworks.com/help/matlab/ref/fft.html . I find each signals length and then draw them on a 500Hz Frequency scale graph separately. I want to compare the signals amplitudes and frequencies. I think i am right with the frequency part but i an not sure about the amplitude part.

Can anybody help me? (experienced in dsp and fft, people will definitely know this )