🚀 go-pugleaf

iqbal@pawnee.eel.ufl.edu (M.Iqbal) writes:


>I made the following two diagrams that I have not seen in any 
>communication book. This is what I think about the PAM natural
>sampling and flat-topped sampling. Please, if someone disagrees with
>me send me an email, though, I will appreciate any comments.


>1) The following is natural sampling PAM diagram that I made with a LPF
>and an impulse train. What I have seen in almost every communication book
>is a pulse train at the mixer. However, I modeled pulse train with an 
>impulse train and a LPF. I think that in doing so one can understand 
>flat-topped PAM better which is shown in the figure of #2.


>                            Mixer 
>                            
>   Analog(message) ----------(X)------> PAM with natural sampling
>    Signal input              |
>                              |
>                              |
>                           -------
>                           | LPF |
>                           -------
>                              |
>                              |
>         Impluse train--------|



>2) All I did here is moved the LPF in the output branch from the input
>branch. Since LPF transfer function is similar to an integrator, 
>any product of the impulse and the analog-message siganal will be held. 
>Thus, we will have sample and hold circuitry in the output branch.




>                                   -------
>Analog(message) ----------(X)------| LPF |----> Flat-topped PAM
> Signal input              |       -------
>                           |
>                           |
>                           |
>                           |
>      Impluse train--------|


>I have eliminated math as it is not possible to the summation 
>signs and other symbols in ascii.
Boy have you eliminated math!
The first circuit lowpass filters the impulse train. What remains
is an assortment of frequencies being multiples of the impulse
train rate. With that you multiply your signal, getting a highly
modulated mishmash. If you use a short-time integrator as your LPF,
the impulse train gets changed to a constant, and your circuit does
NOTHING whatsoever.

The second one is the possibly more correct version (depending on
what you want to do). IFF your signal is already bandlimited (the
possible range of frequency less than half the sampling rate), the
sampling with the impulse train singles out values representing
the original signal completely. By using a short time integrator,
you get flat topped PAM. For analog transmission, however, a better
LPF would lead better results.

Your fundamental mistake in your diagram might be the assumption
that multiplication with a signal is the same as using a filter.

It is not.