By adding a single-channel pulse-height analyzer (SCA) to the output of a TAC, the time-to-amplitude converter can be used to identify coincident events between two detectors. To appreciate the power of this method, one must compare it to the alternative technique, the simple overlap coincidence circuit. Figure 2 illustrates the principle behind the overlap coincidence function offered in the ORTEC Models CO4020, 414A, and 418A. The overlap coincidence circuit is simply a two-input AND gate. As depicted by the waveforms in Fig. 2, the AND gate generates a "logic 1" output only when "logic 1" pulses are present on both the A and B inputs. In fact, the output is generated only for the time during which the A and B pulses overlap. This is the reason the circuit is known as an overlap coincidence.

Detecting truly coincident pulses places special restrictions on pulses A and B. First, the delays through the electronics producing the pulses must be the same for both detectors, so that both pulses arrive at the AND gate at the same time. Second, the width of each pulse must be equal to the maximum timing uncertainty for its respective detector. If the pulse width is too narrow or the delays are not quite matched, some of the truly correlated pulses will not overlap, and the C output will be missing. This represents a loss of coincidence detection efficiency. If the A or B pulses are too wide, uncorrelated events will have a higher probability of generating an output due to accidental overlap, and that is contrary to the purpose of the scheme. Choosing the proper pulse widths and delays to achieve 100% efficiency for identifying correlated events, while minimizing the sensitivity to uncorrelated events, requires a laborious series of trial-and-error measurements. Experimenters often avoid this task by making the pulse widths much larger than the "best guess" for the detector's timing uncertainty. Of course, the quality of the experiment will suffer if these pulses are either too wide or too narrow.

Figure 3 shows how a TAC with an SCA (i.e., Model 567 TAC/SCA) can be used to simplify the selection of the optimum coincidence resolving time. The prompt timing pulse from the germanium detector operates the start input of the TAC, while the delayed pulse from the scintillation detector triggers the stop input. When the analog output of the TAC is analyzed by the multichannel analyzer, the spectrum in Fig. 4 is observed. There is a peak formed by the correlated gamma- ray events from the two detectors. This peak sits on an essentially flat background caused by the uncorrelated events from the two detectors. (See the comments following Equation 2.)

By connecting the logic output of the SCA to the gate input of the MCA, only those TAC pulses which fall within the SCA window will be analyzed by the MCA. With minimal effort, the SCA thresholds can be adjusted to ensure that only the events in the peak are accepted. Subsequently, the SCA output is used as the coincidence gate when analyzing the energy spectrum from the germanium detector on the MCA. By replacing the overlap coincidence with a TAC and SCA, the optimum coincidence resolving time can be selected quickly and with full knowledge of the intrinsic time resolution of the system.

Note that the SCA window for "correlated events" in Fig. 4 includes a background contribution from "uncorrelated events". The contribution of these uncorrelated events to the energy spectrum can be assessed by setting another SCA window of equal width in the uncorrelated background region of the time spectrum. This second SCA is used to gate a second MCA, which will record the energy spectrum corresponding to uncorrelated events. Subtraction of the two energy spectra will yield a spectrum free of the uncorrelated events. (NOTE: a minor correction to the second SCA window width based on Equation 2 may be required at high counting rates.)

At extremely high counting rates the processing time of the TAC and SCA may contribute noticeably to the dead time losses of the coincidence spectrometer. In this rare case, an overlap coincidence with updating inputs and outputs is the better choice because of its inherently lower dead time for identifying coincident events.