The concept of using a Circular Buffer is pretty straight forward:
Acquire continuously till a signal is found, then send the digitized signal to the computer.
This approach allows to have the incoming signal shape also before the trigger event.
I prepared some diagrams to make myself clear. The following points are referring to the images.
On the first image we can see what I mean with continuous acquisition. We define a buffer that will store the data, in my case an array with 1280 slots, then we start to continuously read the ADC output register (ADCH) ad filling the buffer with the data. When we get to the end of the buffer we restart from the beggining without clearing it. If we immagine the array arranged in a circular way it is easy to see what I mean.
When the signal surpasses the threshold, the Analog Comparator Interrupt is activated. Then we start a waiting phase in which we continue to acquire the signal but keep a count of the ADC cycles that passed from the Analog Comparator Interrupt.
When we waited for N cycles (with N < 1280), we freeze the situation and stop the ADC cycles. So we end up with a buffer filled with the digitization of the signal temporal shape. The great part of this, is that we have also the shape prior to the trigger event, because we were already acquiring before that.
Now we can send the whole buffer to the serial port in a block of binary data, instead of sending the single ADC reads. This reduced the overhead required to send the data and the bottleneck of the sketches that I found on the internet.