I am trying to understand a previous design that uses Amplitude Modulation. Instead of just using the real part of the signal, the designers put the voice signal on both I and Q. Both I and Q have the same data.
How would this affect the spectrum of the resulting AM signal and how would this affect the performance of the demodulator (assuming we are demodulating using a simple envelope detector)?
It doesn't really affect anything at all. If you modulate a signal on the I channel only, then it would look something like:
$$ x_1(t) = m(t) \cos(2\pi f_c t) $$
If it were on the Q channel only, then it would look like:
$$ x_2(t) = m(t) \sin(2\pi f_c t) $$
$x_1(t)$ and $x_2(t)$ only differ by a phase shift of 90 degrees on the carrier. Likewise, if you were to split the signal evenly in power across the I and Q channels, it would look like:
$$ x_3(t) = \frac{1}{\sqrt{2}} m(t) \cos(2\pi f_c t) + \frac{1}{\sqrt{2}} m(t) \sin(2\pi f_c t) $$
Again, in the complex baseband form that is often used for analyzing signals, this corresponds to just a phase shift of 45 degrees from the original signal $x_1(t)$. In most systems, you will have a random phase shift between the transmitter and receiver anyway; this should not have any effect on the performance of your receiver.
