872 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 50, NO. 4, AUGUST 2001 The Effects of Offset Voltage on the Amplitude and Bandwidth of Kick-Out Pulses Used in the Nose-to-Nose Sampler Impulse Response Characterization Method Donald R. Larson, Senior Member, IEEE, and Nicholas G. Paulter, Jr. Abstract—The amplitude and bandwidth of kick-out pulses used in the nose-to-nose sampler impulse response characteriza- tion method were measured as a function of offset voltage. The measuring instrument was of similar bandwidth and its effects on the measurement were not removed. The amplitude is almost linear for offset voltages from 500 mV to 500 mV, except for values from about 50 mV to 50 mV. Slight changes in bandwidth were observed for offset voltages from 250 mV to 250 mV with significant bandwidth loss observed for offset voltages outside this range. Index Terms—High-speed samplers, impulse response, sampling oscilloscopes, waveform spectrum. I. INTRODUCTION C HARACTERIZATION of the impulse response or transfer function of high-speed/high-bandwidth samplers used in digital sampling oscilloscopes can be performed with either time or frequency-domain techniques. Because of the difficulty in obtaining phase information using fre- quency-domain methods, however, time-domain techniques are typically used. These time-domain methods require having pulse generators that have significantly greater bandwidth than the samplers themselves or, if the generators have comparable bandwidth, then their pulse profiles must be accurately known so that their influence can be deconvolved. Some time-domain methods that are being explored for sampler impulse response characterization use a photoconductively-generated electrical pulse for the reference pulse, and this pulse is measured using photoconductive or electro-optical sampling [1], [2]. Conse- quently, these methods may require that the electrical impulse response of the sampling photoconductor or electro-optic element also be known or measurable. This requirement, of course, brings us to the original problem of delivering a known input pulse to the samplers from which a sampler impulse response can be derived. The nose-to-nose sampler impulse response characterization method, introduced several years ago, is a method where a sampler is used as a pulse generator. The sampler generates a pulse (the kick-out pulse) when it is operated at a nonzero offset voltage, and it is argued that the Manuscript received May 4, 2000; revised February 27, 2001. The authors are with the National Institute of Standards and Tech- nology, Gaithersburg, MD 20899-8111 (e-mail: donald.larson@nist.gov; nicholas.paulter@nist.gov). Publisher Item Identifier S 0018-9456(01)06014-4. kick-out pulse represents the impulse response of the sampler [3]–[5]. This kick-out phenomena is limited to a particular sampler architecture. Since it would be difficult to determine the kick-out pulse of a sampler using only that sampler, a series of kick-out pulses are measured using a set of three samplers. Each sampler in turn acts as a source of the kick-out pulses that are measured using the other samplers. The assumption that the sampler im- pulse response is very similar to the kick-out pulse cannot be valid unless the frequency or time response of the kick-out pulse is independent of the offset voltage. Furthermore, the utility of using a kick-out-pulse-derived sampler impulse response for purposes of calibration is dependent on the sampler’s response being linear over its signal range. II. EXPERIMENT The nose-to-nose method gets its name from the arrangement of the samplers in the measurement procedure; that is, the input connectors of two different sampling heads are connected together using a male–male coaxial adapter. The sampling heads are physically and electrically situated such that the distance between their input connectors is minimized, resulting in a nose-to-nose appearance. Two 50 GHz digital sampling oscilloscopes were used and, in our implementation, both sampling heads are triggered using a short transition duration step (17 ps transition duration, 0.25 V amplitude before a wideband splitter) with a 2 kHz repetition rate. The kick-out pulse also contains a signal contribution from the strobe pulse that couples into the sampling diodes. To minimize the strobe-coupling effect, the following procedure is used [3]. Two kick-out waveforms are acquired, one using a negative offset voltage [Fig. 1(a)] and another using a positive offset voltage of equal magnitude [Fig. 1(b)]. The polarity of the kick-out pulse changes sign with the offset voltage but the strobe pulse polarity does not. Subtracting these two kick-out pulse wave- forms will produce a difference waveform in which the strobe pulse contribution has been minimized and the kick-out pulse amplitude has been doubled. Dividing this difference waveform by two yields an estimate of the kick-out pulse with the strobe pulse contribution removed [Fig. 1(c)]. The offset voltages for the sampling head used to generate the kick-out pulse are (for the positive offset voltage) and (for the negative offset voltage), and the offset voltage of the sampling head U.S. Government work not protected by U.S. copyright.