In a coincidence experiment, a laser shot only counts when it gives exactly one electron and exactly one ion. Shots with two of either are set aside, because there is no way to tell which electron belongs to which ion.
That strict rule keeps the data clean. Under it, an extra pulse can turn a good shot into a rejected one.
The problem
On our PEPICO spectrometer, the electron detector channel gave a false second pulse 31 ns after real ones. About one real electron in six carried such an echo. A sharp repeat at one delay points to the detector or the electronics, not the sample.
An echo like this does not have to distort a spectrum to do damage. Under the strict rule it mostly removes good events instead of adding bad ones. That makes it easy to miss. A lower count rate looks like a weak sample or a tired detector, not like a fault.
Finding the cause
A fixed delay is the most useful clue an echo gives you. Real arrival times spread out with the energy and mass of the particles. An artefact comes back at the same delay every time. Detector afterpulsing does this too, so rule it out first; then the question becomes which circuit puts it there.
There are a few usual suspects for a false pulse at a fixed delay. A signal can reflect from a badly terminated cable end and come back. An amplifier or discriminator can ring after a large pulse. A nearby circuit can also couple into the signal line.
I found ours in a histogram of the time between two electron hits in the same laser shot: a sharp spike at 31 ns. A scope on the detector output, triggered with our high-voltage switch, showed the pickup.
What changed
Once the pickup was removed in the hardware, the echo was gone and our coincidence count rose by 25%. Put the other way round, the echo had been costing us one coincidence in five.
Coincidence experiments run slowly on purpose. We keep the ionisation rate low, so most laser shots ionise nothing. Every good event is expensive as a result. With 25% more of them per hour, the same statistics need a fifth less measurement time.
What I would check on any coincidence setup
Histogram the gaps. Plot the time between consecutive hits on one detector channel. Real hits give a smooth spread. An echo shows up as a sharp spike at one delay.
Compare the delay with your cables. A signal in coaxial cable travels at roughly two thirds of the speed of light, about 5 ns per metre. If the delay matches the round trip of a cable, suspect a reflection first.
Look beyond the signal chain. High-voltage supplies and switches often share grounds and cable runs with detector lines.
Count before and after. A real fix should raise the rate of clean events at the same laser settings.
The detectors and the coincidence idea are drawn step by step on my Lab page.