The working principle of the first (I assume no load and ideal diodes):
The first circuit is a clamper followed by a peak detector.
When V1 is negative, C1 is charged through D2. At the end of the first half period, the voltage on C1 is positive from right to left, and the value is |Vpeak(V1)|. When V1 is positive (second half period), the voltage at the node C1-D1-D2 is brought to V1 + |Vpeak(V1)|. D1 turns on, and there will be some charge-sharing between C1 and C2, therefore the voltage of C2 will increase each cycle. After "many" cycles, the voltage of C2 will reach the peak voltage of C1, i.e. 2*|Vpeak(V1)|
The second circuit is made of two-half wave peak detectors (one positive and one negative), in series (C4-D3 and C3-D4).
When V2 is positive, it charges C4 to |Vpeak(V2)|, through D3. When V2 is negative, it charges C3 to -|Vpeak(V2)|, through D4.
- The first circuit shares the ground with the input. The second doesn't.
- In the first circuit, the maximum voltage might be reached after many cycles, because C1 is charged and then subsequently discharged on C2.
In the second circuit, the full double voltage is reached in one full cycle. In other words, the second might provide a better voltage stability and lower output impedance, given the same capacitor values.
simulate this circuit – Schematic created using CircuitLab