Time-tag impulses with zero-crossing circuit
Edited by Bill Travis
Elio Rossi, Itesre-CNR, Bologna, Italy -- EDN, January 24, 2002
A "constant-fraction discriminator" usually performs the time-tagging of impulsive events, which have a peaking time of the signal amplitude. The implementation of this technique requires a delay in the input signal of approximately the same amount as the signal's rise time. You can attain this delay by using a coaxial cable of an appropriate length. For many applications, in which the rise time for impulsive events is 1 to 10 µsec, you must consider alternative solutions, because of the length of the cable you'd require. Figure 1 shows the typical output from a spectroscopic amplifier, where the presence of a large amount of detector noise with Gaussian distribution is a limiting factor for system performance in amplitude and timing resolution. The time-tagging of such pulses is subject to two well-known types of errors: the jitter related to the noise and the "walking time" arising from the amplitude variation of the signals. You can eliminate the walking time by differentiating the signal and detecting the zero crossings. The jitter is related to the noise around the zero-crossing line.
In Figure 2, an arming discriminator with a fixed threshold of 100 mV (5×VRMS(NOISE)) enables the IC1, a MAX941 zero-crossing discriminator, via the first half of IC2, an HC4538 resettable monostable multivibrator. The propagation delay in these ICs allows enabling the zero-crossing discriminator when the differentiated signal is well over the baseline noise for the full range of the input signal. The positive-going output pulse from IC3, a MAX941, corresponding to the zero-crossing time, reaches the output with a fixed length of 0.5 µsec, set by the second monostable multivibrator. This second multivibrator resets the first and latches the MAX941 at the high output level until a new trigger arrives from the arming discriminator. In this way, you avoid spurious triggers at the beginnings and ends of input pulses. The upper-threshold discriminator output, with a minimum output length of 5 µsec, serves to "veto" the 0.5-µsec output and works even for highly saturated input signals. Because of the input configuration of the MAX942, it is necessary to reduce the upper threshold level to less than 2.8V. Figure 3 and Figure 4 show the timing sequences for 0.2V and 10V input signals, respectively.
You can obtain the same results using many different implementations of the circuit, depending on the ICs available off the shelf. For example, you can substitute the MAX941 with a common discriminator and a CMOS analog switch to commutate the threshold from a positive voltage to ground. A single flip-flop then completes the circuit. This design uses an amplification stage with back-to-back limiting diodes in front of the zero-crossing discriminator. Table 1 shows the results, with comparisons to the shaping-time value of 3 µsec. You measure the walking time and jitter using a pulse generator, preamplifier, shaping amplifier, time-to-amplitude converter, and multichannel analyzer. Figure 5 shows just three of the many histograms used to calculate the walking time and jitter results.
Is this the best Design Idea in this issue? Select at www.ednmag.com.
