Using ifft() to deconvolve signal and frequency response

I have a measured time varying signal which is known to have a variable frequency response (I have freq vs. sensitivity data for the measurement device). I want to deconvolve using:

pt = ifft(fft(vt)./Mf)

where vt is the time varying signal and Mf is the frequency response.

I take the magnitude

abs(fft(vt))

and divide by the freq. vs. response curve Mf (interpolate to the same frequency range).

However if I set Mf=1 (for all values f=0 to X MHz) I don't get the original signal vt when I do the ifft.

My understanding is I need to perform the division (vt./Mf) on the fft magnitude but then use real/complex fft data for ifft to get the original signal but I'm unsure how to get this to work correctly.

[Baltam]

Are you sure you didn't do anything wrong when setting Mf to 1?

x = 0:0.001:0.1-0.001;
y = sin(2*pi*50*x);
Mf = ones(size(y));
plot(x,y,'o'), hold all, plot(x,ifft(fft(y)./Mf),'-')

This seems to work fine with me.

[2one]

I'm not sure this is exactly the output I require. I need to divide fft(y)./Mf at each point which must have the same frequency value that is if I plotted

plot(f1,fft(y)); hold all; plot(f2,Mf); hold all; plot(f2,fft(y)./Mf)

I first would need to interpolate f1 so it is the same as f2. But I think I need to use abs(fft(y)) rather than fft(y) in the division?