A simple avalanche pulser produces a fast rising edge, possibly a short flat section and then a slow decay. Many applications need a square topped pulse.
There are several techniques to generate such shapes with these pulsers but generally around 50% of the voltage (75% of the power) is lost in the process.

The simplest method is the suicide "T'' method. The pulse is fed into a conjunction of two cables in parallel, one has half the impedance of the source and the other cable, the output, is the same as the source. E.g. a 50Ω source cable is fed into a 25Ω cable in parallel with another 50Ω output cable. The 25Ω cable is terminated with a short circuit at some distance from the conjunction. With a step votlage on the input, the output pulse will be a single pulse of half the amplitude of the step and duration equal to the round trip transit time to the short circuit and back. After another round trip another pulse will be delivered. Also there may be a reflection from the generator.
This type of pulse forming is not often used because of the reflections and late time secondary pulses. However, it is useful as a calibration tool as the output is exactly half the input. We use these on 12kV (and above) pulsers to halve the voltage before putting the pulse into an attenuator. This allows us to use a standard Barth type 142 attenuator. Note that for square pulses, this is only really suitable for short pulses as no attempt is made to compensate for the droop in voltage from a raw pulser.

A second approach which is a little more efficient is to feed the output pulse from the pulser directly into two cables in series, generally two 50Ω cables.One is allowed to float and the far end is open circuit. The other is the ouput. This delivers a pulse with a fairly flat top for pulses up to about 10ns. The pulse length is determined by the length (round trip time) of the cable.The output voltage is a bit more than half of that from a raw pulser. It is a little more efficient.

With this technique it is possible to make adjustable length pulsers by changing the length of this cable. However, the connection where the change in length is made limits the shortest pulse that can be produced. So while a fixed pulse lengths can go down to around 100ps, variable lengths using this technique start at around 0.7ns.

It is possible to use this technique and obtain more voltage by using a lower impedance cable for the pulse forming but then the falling edge will not return to zero.

A third approach that works well for pulse lengths in the range 100ps to 1.5ns is to use a modified version of the suicide "T" which has some compensation for the pulser droop.
This technique is used in fast camera systems for changing the pulse length over this range and we have even made systems where the pulse length can be changed remotely with this technique.

A fourth approach that is used in the PSP1 pulse is to use two pulse generators, one for each edge and sum them in a way that offers two fast edges and the ability control the pulse length from around 200ps to 10ns. This is even more wasteful of pulser energy at around 12.5 % by power, however, it offers very good results.

A fith method that works over the range 100ps to ~1ns is to add positive and negative pulses with a resitive network to reduce reflections. The timing of the two pulsers sets the output pulse length just like method four. This technique has the added advantage that it will go smoothly from around 1ns negative pulse to 1ns positive pulse.

There are instances where the pulse does not have to be formed electrically but instead two pulses can be combined at the load in some other manner. A typical example is a pockels cell where each side of the cell can be driven from a separate pulser and adjusting the timing can affect the net voltage accross the pockels cell.

Kentech has made all of these systems. We can advise which is the best for a particular application. Please get in touch.