Usually, this is not a problem. A PPS signal works on the rising edge, so the circuit where the signal is used senses only the positive edge. In this way no matter how long the pulse is. It will work.
But some times, the PPS signal goes to a more exotic circuit or application and a too long PPS pulse can be a problem. One of these rare uses of a PPS signal is to synchronize audio streams using the VLF Receiver Software Toolkit. It records the PPS signal along with the desired audio, and then, using the recorded PPS signal along with the computer's clock, it adds precise timestamps to the audio stream.
It works nice, but not if your PSS signal is too long:
With the pulse completely distorted (because of the low pass frequency response of the sound card) the vlftoolkit does not work at all. The solution? To shorten the pulse length. Probably there is a command for the GPS to configure the pulse length but not always it is easy to get the GPS to run that command, so the quickest way is to add an edge detector to conform the pulse to whatever length you want.
I build this circuit and worked pretty well. I used a 74HC74 and a 74HC14, but probably other series, like HTC will work. You can replace the 74HC14 with a 74HC132 NAND gates with inputs joined together. The length of the pulse at the output can be adjusted with the aid of R and C. I used 22k and 20nF and I got a 0.8ms pulse length that works nicely with the vlftoolkit software.
As a bonus, working at 5 volts, it works with 3.3 and 5 volt PPS signals.
Of course, this circuit adds a delay of a few tenths of ns, but working with sounds cards at sample rates under 1 MHz, this is not a problem at all.
Miguel A. Vallejo, EA4EOZ