A novel method for generating multichannel quasi

REVIEW OF SCIENTIFIC INSTRUMENTS 86, 086110 (2015)

Note: A novel method for generating multichannel quasi-square-wave pulses C. Mao, X. Zou, and X. Wang Department of Electrical Engineering, Tsinghua University, Beijing 100084, China

(Received 25 March 2015; accepted 19 August 2015; published online 31 August 2015) A 21-channel quasi-square-wave nanosecond pulse generator was constructed. The generator consists of a high-voltage square-wave pulser and a channel divider. Using an electromagnetic relay as a switch and a 50-Ω polyethylene cable as a pulse forming line, the high-voltage pulser produces a 10-ns square-wave pulse of 1070 V. With a specially designed resistor-cable network, the channel divider divides the high-voltage square-wave pulse into 21 identical 10-ns quasi-square-wave pulses of 51 V, exactly equal to 1070 V/21. The generator can operate not only in a simultaneous mode but also in a delay mode if the cables in the channel divider are different in length. C 2015 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4929870] Square-wave pulse generators are commonly used as trigger-timing units and response-time calibrating sources in the pulsed power experiments. In these cases, one or two square-wave pulses are normally enough. However, in some other experiments,1,2 many square-wave pulses or even hundreds of identical pulses are needed to inject simultaneously into a device. Although the commercial available pulse generator such as DG535 can deliver as many as 8 squarewave pulses simultaneously into its loads, the amplitude of these pulses is only 4 V. In this paper, we described a novel method by which a pulse generator was constructed to simultaneously produce 21 identical quasi-square-wave pulses of 50 V. The quasi-square-wave pulse generator consists of two stages. The first stage is a high-voltage pulser that produces a square-wave of 10 ns in pulse width and about 1.1 kV in amplitude. The second stage is a channel divider that divides the high-voltage square-wave into n identical quasi-squarewaves but with lower amplitude. The circuit of the high-voltage pulser is shown in Fig. 1 and it is the typical circuit for forming nanosecond squarewave.3 T0 is a 50-Ω polyethylene cable used as the pulse forming line (PFL) that is 1-m long which corresponds to a one-way transit time of 5 ns. First, the PFL T0 is charged to about 2.4 kV by a high voltage DC source V1 via a currentlimiting resistor R0 of 2 MΩ. Then, the switch U1 is closed. Since an impedance-matching load, a 50-Ω polyethylene cable T1 used as the pulse transmission line (PTL), is connected at the output of PFL, a square-wave of about 1.2 kV in amplitude and 10 ns in full width at half maximum (FWHM) will be produced and delivered along the PTL to the load RL. In order to get a square wave, the rise time of the output voltage from the PFL should be as short as possible. This rise time is mainly determined by the closing time of switch U1. We used an electromagnetic relay as switch U1. The electromagnetic relay is shown in Fig. 2. A DC voltage of 2.4 kV is applied to an air gap between a moving electrode and a fixed electrode. When a driving current flows through the relay coil, the relay is triggered and the moving electrode moves fast onto the fixed electrode. During the moving process, the

shortened air gap breaks down. Due to the time delay of the breakdown, an overvoltage across the shortened air gap is achieved, which leads to a rise time shorter than 2 ns. This was detailed elsewhere.4 As the amplitude of the output voltage from this generator can be as high as 1000 V, a resistive divider with a voltage ratio of about 50:1 is needed for measuring it with an oscilloscope. Fig. 3 shows this resistive divider as well as its circuit. The resistive divider was designed with a compact and coaxial structure to reduce the stray inductance. The high-voltage input port is at the top, while the low-voltage output port is at the bottom. Both ports are BNC connectors. To make the impedance matched at the input port, it is required that R1 + R2 = 50 Ω. In consideration of the commercial available resistors, we chose R1 = 50 Ω and R2 = 1 Ω, so R1 + R2 = 51 Ω, which is nearly matched to the 50-Ω polyethylene cable. R1 is composed of 25 50-Ω resistors. First, all the resistors were divided into 5 groups each with 5 resistors in parallel, then 5 groups were connected in series. R2 is composed of 10 10-Ω resistors in parallel. The resistive divider was calibrated with a square wave from DG535 (Stanford DG535 Digital Delay and Pulse Generator). The results showed that the resistive divider is capable of measuring a square wave with a rise time as short as 1 ns and has a voltage ratio of 51:1 exactly as that was designed. Being recorded with a digital storage oscilloscope (DSO9254A from Agilent), the typical waveform of the output voltage from the high-voltage pulser is shown in Fig. 4. Since the waveform of the output voltage from the highvoltage pulser changed slightly from shot to shot, the waveforms of 10 shots were recorded. The parameters averaged over these 10 waveforms were given in the following: 1.55 ± 0.1 ns in the rise time, 10.10 ± 0.1 ns in FWHM, and 1070 ± 20 V in the peak value. The vibration on the flat top of the waveform is about 65 ± 10 V, which is about 6.1% of the peak value. Therefore, it is a quasi-square wave. The circuit of the channel divider is shown in Fig. 5. A network consisting of (n + 1) resistors was inserted in between the cables. The condition of the impedance matching for the

0034-6748/2015/86(8)/086110/3/$30.00 86, 086110-1 © 2015 AIP Publishing LLC This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 101.5.223.117 On: Mon, 31 Aug 2015 14:08:57

086110-2

Mao, Zou, and Wang

Rev. Sci. Instrum. 86, 086110 (2015)

FIG. 1. Circuit of high-voltage square-wave generator.

circuit can be expressed as R+Z = Z, (1) n where R is the resistance value, n is the total number of output ports, and Z is the characteristic impedance of the cable. Under this condition, there will be no reflection no matter which port of this channel divider a voltage wave comes from. Take the input voltage wave, for example. The characteristic impedance on the left side of point A is Z, while that on the right side of point A is R + (R + Z)/n, the characteristic impedance is matched at point A according to Eq. (1). Therefore, when the input voltage wave coming from the input port transmits to point A, no reflection will appear. The same thing happens when a voltage wave comes from any cable other than the input cable. In this way, we can avoid the reflection at any input point of the resistor network and thus avoid the distortion of the waveform. From Eq. (1), we can obtain that R+

n−1 · Z. (2) n+1 From the circuit shown in Fig. 5, we can see that the input voltage Uinput is divided twice before it arrives at the output port. The first time is between the n-way output branches in parallel and the resister next to point A, while the second time is between the cable and resister in each output branch. Therefore, taking Eq. (1) into consideration at the same time, it can be easily derived that the ratio between the amplitude of R=

FIG. 3. Resistive divider: (a) photo, (b) circuit.

the output voltage Uoutput and Uinput is that Uoutput = Uinput

Z +R n Z +R n +

R

·

Z = Z+R

Z +R n

Z

·

Z 1 = . Z+R n

(3)

While n output pulses were obtained from one input pulse, the amplitude of the output pulse was reduced to 1/n of that of the input pulse. The channel divider is shown in Fig. 6. In order to avoid the distortion in the waveform of the output voltages, the most important issue is to reduce the stray inductance and capacitance. All ports are BNC connectors that were mounted on a circular plate made from aluminum alloy. The input port is placed at the center, while all output ports are uniformly placed at the outer circumference. This arrangement makes the divider axisymmetric; thus, all output voltages are identical. Obviously, more output ports require a larger circular plate and longer lines for connecting the resistors from the center to the output BNC connectors, which will increase the stray inductance that we want to reduce. Making a balance between the more output ports and the less circuit inductance, we chose n = 21, 21 output BNC connectors that were closely

FIG. 2. Electromagnetic relay used as switch U1. Reprinted with permission from Rev. Sci. Instrum. 86, 034705 (2015). Copyright 2015 AIP Publishing FIG. 4. Typical output voltage of the high-voltage pulser. LLC. This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:

101.5.223.117 On: Mon, 31 Aug 2015 14:08:57

086110-3

Mao, Zou, and Wang

Rev. Sci. Instrum. 86, 086110 (2015)

FIG. 5. Circuit of the channel divider.

FIG. 7. Typical waveform of the output voltage from 1-channel to 21channel divider.

FIG. 6. Picture of the channel divider.

and uniformly mounted at the outer circumference of a circular plate with a radius of 5.7 cm. A 1-channel to 21-channel divider was constructed. All the cables are 50-Ω polyethylene cables. From Eq. (2), it was calculated that n−1 21 − 1 ·Z= × 50 = 45.45 Ω. (4) n+1 21 + 1 Instead of 45.45 Ω, we chose R = 45.3 Ω from the commercial available resistors. The typical waveform of the output voltage from the channel divider is shown in Fig. 7. Comparing to the waveform shown in Fig. 4, the waveform shown in Fig. 7 was distorted a little bit at the leading edge, especially on the top, which may be caused by the stray inductance and capacitance in the channel divider. However, it is still a quasi-square wave. Ten waveforms of the output voltage from the channel divider were recorded. The parameters averaged over these R=

10 waveforms were given in the following: 2.25 ± 0.2 ns in rise time, 10.15 ± 0.1 ns in FWHM, and 51 ± 1 V in peak value. It was experimentally confirmed that the peak value of output voltage, 51 V, is exactly 1/21 of that of the input voltage, 1070 V. The vibration on the top of the waveform is about 5.3 ± 0.5 V, about 10.4% of the peak value. It should be indicated that our generator can output multichannel quasi-square-wave pulses not only in a simultaneous mode but also in a delay mode if the cables in the channel divider are different in length. In conclusion, we proposed a novel method by which multichannel square-wave nanosecond pulses could be obtained. Comparing to the commercial available square-wave generators, the generator constructed with this novel method can output much more channel square-wave pulses with a significantly higher amplitude. The authors would like to thank the National Natural Science Foundation of China under Contract No. 51277109 and the Key Laboratory of Pulsed Power of China Academy of Engineering Physics (CAEP) under Contract No. PPLF2014PZ02 for supporting the research. 1R.

A. Petr, W. C. Nunnally, C. V. Smith, Jr., and M. H. Clark, Rev. Sci. Instrum. 59(1), 132–136 (1988). 2C. Mao, X. Zou, and X. Wang, Laser Part. Beams 32(4), 599–603 (2014). 3I. A. D. Lewis and F. H. Wells, Millimicrosecond Pulse Techniques (Pergamon Press, New York, 1959). 4X. B. Zou, H. T. Shi, H. Xie, X. X. Wang, and G. X. Zhang, Rev. Sci. Instrum. 86, 034705 (2015).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 101.5.223.117 On: Mon, 31 Aug 2015 14:08:57