yay....simulink+matlab :/
ideal sampling
as long as fs>=2B and mag(G) = 0 for mag(f)>B
ideal lowpass filter has impulse response of h(t) = 2H0Wsinc(2pi*Wt)
zero crossings of h occur at t=+/-n/2W for n!=0
a sampled signal passed through an ideal lowpass filter yields the original signal
sinc approximation ftw
practical reconstruction: need a different function in place of sinc, since sinc is hard to deal with

consider sig recon by use of zero-order hold sigs
p(t) = rect((t-0.5Tp)/Tp)
recon filter will gen g&(t) = sum<inf>(g(nTs)rect((t-nTs-0.5Tp)/Tp)
since p is a time-shifted pulse, it has freq response P = Tp*sinc(fTp*pi)e^(-jpi*Tp)
basically reconstructing out of stair step response
will pass freqs above fs-B, so shifted replicas of G will corrupt the reconstructed sig
freq response of P is not flat for freqs f<B which will distort sig
to elim distortion, use equalizer filter with freq response: Ts/P on mag(f) <= B, 0 on mag(f)>=(1/Ts - B), flexible on other regions, because it doesn't matter
equalizer passband response is E=Ts * fpi/sin(fTp*pi) * e^(-jf2pi*t0) on mag(f)<=B

for finite filter gain, require Tp<1/B
if Tp chosen very small, E = Ts*fpi/sin(fpi*Tp) ~~ Ts/Tp

practical sampling
aliasing is fun..., causes frequency content distortion
going into PCM...getting into the stuff I know :D