I've found some conflicting information about how the I/Q waveforms look compared to the 4-PAM signals before the modulator.
Here it is shown that the 4-PAM digital signals are filtered, so the analog I/Q components look like the 4-PAM levels:
A confirmation for this method can be found in the paper "FPGA implementation of software defined radio model based 16QAM. Naghmash, Majid & Ain, M.F. & Hui, C.Y. (2009)". From this paper, the image that best exemplifies the filtering is this:
But then I found this: Source: https://www.slideshare.net/aknigin/digital-communicationunit-3
So the output of the balance modulator in the above picture does not look like the filtered 4-PAM. The schematic for the transmitter can be seen at page 76 of the presentation.
Some observations I've made:
-to get the filtered 4-PAM look approximately the same as the digital signal in terms of amplitude, there must be no ISI (intersymbol interference), thus a higher bandwidth must be occupied by the filtered signal than the absolute minimum required. Specifically, the raised cosine filter must have a higher rolloff, which in the case of beta parameter equal to 1 instead of 0 means double occupied bandwidth. (https://en.wikipedia.org/wiki/Raised-cosine_filter)
-in the second case, where the signal does not look like the 4-PAM levels, the bandwidth also appears to have doubled because of the phase shifting. The main lobe bandwidth is doubled compared to the PAM signal.
- Are these different methods for the same 16-QAM?
- How are these compared in terms of bandwidth efficiency?