A high-speed oscilloscope for real-time use

256 PHILlPS TECHNICAL REVIEW VOLUME 30 A high-speed oscilloscope for real-time use G. Andrieux and C. Loty Introduetion The oscilloscope is probably the one instrument that the electronic engineer regards as an absolute essential for his various tests and measurements. Since the oscilloscope is so vital a measuring instrument, its development should not just match the general progress of electronics; it should always be one step ahead. Maintaining such a lead requires a great deal of effort, since the oscilloscope must at the same time remain a general-purpose instrument and yet give a performance equal to that ofthe most sophisticated electronic equipment. Progress in electronics has usually been associated with faster action or response. Thus, thermionic valves gave pulses which had a duration of tens of nanoseconds, present-day solid-state devices give pulses measured in fractions of a nanosecond, and the laser offers the promise of pulses in the picosecond region. Such high-speed signals can only be displayed on an oscilloscope if it has a very large bandwidth. Now, the best amplifiers that can be produced today have a rise time of the order of a nanosecond. Beyond this, two techniques are available: real-time operation without amplification, and sampling oscillography. The sampling technique, in which normal amplification can be used, has permitted remarkable progress to be made, but it is not applicable to all signals, particularly those that only occur once, for which realtime observation is the only possible method. High-speed oscilloscopes are thus divided into two types. With the first, the sampling oscilloscope, recurring signals can be displayed in the usual way in which oscilloscopes have been used since their inception. The other type, the real-time high-speed oscilloscope, was developed for a different procedure: it enables fast and unique signals to be studied by means of a photographic record, whereas recurrent signals can be displayed as a subsidiary function. This procedure, together with the absence of amplification, imposes fresh limitations on the real-time oscilloscope. In the classical oscilloscope, the cathode-ray tube is no more than a display element whose own performance does not have much effect on the performance of G. Andrieux, Ingenieur E.R.B. and E.S.E., L. èssc., and C. Loty, Ingenieur C.N.A.M., are with Laboratoires d'electronique et de Physique Appliquée, Limeil-Brévannes (Val-de-Marne), France. the equipment. Similarly, the ancillary circuits for the time-base, for synchronization, beam switching, etc., are only of secondary importance. The actual measuring element is really the signal-channel amplifier. (This may seem a strange statement, but its truth becomes apparent to anyone who has to write out an order for one of the modern oscilloscopes with plug-in units.) On the other hand, in the real-time high-speed oscilloscope the cathode-ray tube is the vital element whose intrinsic qualities determine the overall performance of the instrument. Moreover, the ancillary circuits, which have to satisfy very difficult conditions because of the high speed ofthe signal to be examined, acquire a much greater importance. This is particularly so for the delay line - an element which is not used in an ordinary oscilloscope, but whose characteristics can have a marked effect on the design of the time-base circuits and the cathode-ray tube itself. Development of the cathode-ray tube [1] We have already stated that the essential characteristics of the oscilloscope will be those of the cathoderay tube. It is therefore useful to list these characteristics before going on to consider how to improve them together rather than separately. In fact, we shall show that they are not independent, and that as in other situations a figure of merit can be applied. There are four essential characteristics: the bandwidth, the sensitivity, the useful scan on the screen and the maximum writing speed. The first three are obviously related to the signal channel of the tube. The fourth, which refers to the ability to record a non-recurrent fast signal, is directly dependent on the energy density in the electron beam. The bandwidth is of prime importance since we are interested first and foremost in high-speed oscilloscopes. In fact, the rate of response to a signal is limited fundamentally by the transit time of the electrons between the deflection plates. This transit time can be reduced by making the plates shorter but if this is done the sensitivity is also reduced. The same thing happens if the velocity of the electrons is increased by increasing the accelerating voltage. Finally, the spacing of the plates and their length can be reduced together; the sensitivity would then remain the same but there is then - less room for the beam to pass between the plates. This

1969, No. 8/9/10 REAL-TIME HIGH-SPEED OSCILLOSCOPE 257 means that the current has to be reduced, and hence also the maximum writing speed. We have thus the first indication that there is a figure of merit relating bandwidth, sensitivity and writing speed. One obvious solution to the problem of the bandwidth is to arrange a series of pairs of small deflection plates, for which the transit time is very short, and to incorporate these in a transmission line in such a way that the velocity of propagation is the same as that of the electrons. A series of additive deflections is then obtained and the total sensitivity becomes normal again. This is a well-known principle which is applied in every travelling-wave tube, and it has already been applied in the most advanced oscilloscopes of the classical type. This has been done by making use of a cathode-ray tube with divided plates which are linked by external inductances to form a lumped-element transmission line. To carry this development one step further, it is necessary to put the inductances inside the tube and to reduce them to simple loops (fig. Ja). The delay line formed in this way is in fact a constant-k filter, and has a cut-off frequency. It also has a phase dispersion, which limits the rise time well before the value corresponding to the cut-off frequency has been reached. In fact, it can be shown that the rise time as limited by the phase dispersion always exceeds the transit time per element. In order to obtain an appreciable increase in bandwidth, it is therefore necessary to divide the deflection line into a large number of elements each of which is very small. The capacitance and inductance becomes more or less distributed and in the limit the line becomes a zigzag line (fig. lb). The practical limit of such a line is in the region of I GHz. The zigzag line is not a line whose constauts are perfectly distributed: because of the alternating change in the direction of propagation there is an appreciable coupling between adjacent turns, and this coupling may be taken as being responsible for the phase dispersion just mentioned. The dispersion is much smaller for a deflection line wound in the form of a helix (fig. I c). It can in fact be made negligibly small and the bandwidth is then only limited by the transit time for a single turn. However, although the bandwidth that can be obtained in this way is very much greater than before, a new limitation appears: in order to make the phase dispersion sufficiently small the transverse dimensions ofthe line have to be reduced and this reduces the space available for the passage of the beam, and hence the maximum writing speed. Moreover, to increase the sensitivity, the deflection line has to be made longer by giving it more turns, but its losses at the higher frequencies cannot then be neglected. We thus find another, more sophisticated, relationship between the sensitivity, the bandwidth, and the maximum writing speed. We should also add that although making the deflection line longer increases the sensitivity, it reduces the useful area of the screen by intercepting the beam. As a result, the four principal characteristics are linked by a figure of merit whose best value is usually obtained when a helical deflection line is used. A mathematical analysis of the problem shows that there are only two independent means of increasing this figure of merit: reducing the diameter of the spot or increasing the supply voltages on the tube electrodes. For the first case, it would for example be possible to reduce the distance between the screen and the deflection line; other things being equal, this would reduce aij the dimensions of the image, including the spot diameter. If the sensitivity and the useful scan are expressed in terms of the spot diameter, these characteristics remain unchanged, but since the same luminous flux is distributed over a smaller area, the brightness is increased and hence the maximum writing speed. In the second case, the length of the tube could for example be increased, with the main objective of increasing the useful area of the screen. Even though the tube will be longer, the spot diameter and the beam current can remain unaltered since the higher voltage red uces the effect of the space charge; nevertheless, the Fig. 1. a) Deflection line for a large bandwidth. E earthed deflection electrode. The inductances linking the small deflection plates have been reduced to simple loops. B electron beam. b) Zigzag line. c) Helical line. [lj See the articles in Acta Electronica 10, No. 4, 1966, which all relate to the real-time oscilloscope. - This study was supported by the Commissarlat à l'energie Atornique, Direction des Applications Militaires.

258 PHILIPS TECHNICAL REVIEW VOLUME 30 Fig. 2. 5 GHz cathode-ray tube 50 D 13 BE, developed for the Direction des Applications Militaires du Cornrnissariat à l'energie Atomique of the French government. This tube is used in the prototype of an oscilloscope referred to in this article (fig. 6) and also in a series of special-purpose oscilloscopes made by Société F.E.R.I.S.0.L. for the Cornrnissar iat à l'energie Atomique. Total length of tube 48 cm. second approach never permits an increase in sensitivity. A helical deflection line like the one discussed is used in the cathode-ray tubes 42 D J 3 BE and 50 D 13 BE. The latter tube is shown in.fig. 2, and the deflection line of the 42 D 13 BE is pictured in jig. 3. We have mentioned earlier the influence of the ancillary circuits on the concept of the oscilloscope as a whole. One ancillary device whose i nfluence is quite marked is the delay line which has to be inserted in the signal channel between the trigger circuit and the tube input to compensate for the delay introduced by the time-base circuits. This line should have a delay time of several tens of nanoseconds, i.e. a length of several metres. Even if the very best coaxial cable is used, its losses cannot be neglected when the bandwidth is of the order of 5 GHz. Taking these losses into account a well-defined transit time per turn is necessary for the deflection line of the cathode-ray tube in order to optimize the pulse response of the equipment. This shows how impossible it is to separate the study of the cathode-ray tube from the study of the circuits associated with it, if the best performance is to be achieved. Problems connected with the associated circuits The principal circuits that have to be associated with a cathode-ray tube to form a high-speed oscilloscope for real-ti me use are: I) In the signal channel: A triggering coupler which serves to supply the synchronizing circuits with timing information related to the incident signal. A delay line whose function has been explained earlier. Fig. 3. Helical deflection line for the 42 D 13 BE tube.

1969, No. 8/9/10 REAL-TIME HIGH-SPEED OSCILLOSCOPE 259 2) Outside the signal channel: A time-base circuit producing a signal for highspeed horizontal sweep. An unblanking circuit which provides a pulse that switches the beam on for a time equal to that of the useful part of the sweep. Synchronizing and trigger circuits for obtaining exact synchronization between signal, unblanking pulse and horizontal sweep. The triggering coupler has to tap off a fraction of the energy of the incident signal, but it is important that it should not distort the signal, however fast it is. This condition can only be satisfied by a resistive type of coupling whose reponse is practically independent of frequency. Moreover, the amount of energy tapped off must be small enough not to introduce an appreciable reflection of the signal. A wide-band coupler designed at LEP is shown in fig. 4. We mentioned earlier that the losses of the delay line have a marked effect on the response of the oscilloscope. These losses cannot be made negligible by increasing the cross-sectional dimensions of the cable, because it would then become too bulky. For cable of the highest quality the attenuation is mainly due to skin-effect losses and therefore varies as the sq uare root of the frequency. Since the attenuation is of course proportional to the length of the cable, it follows that for a given attenuation the intrinsic bandwidth of the cable is inversely proportional to the square of its length. In the same way, the effect of the cable on the rise time of a pulse signal will be proportional to the square of its length. It can thus be seen that it is vitally important to keep the length ofthe delay line as small as possible, and consequently to reduce the triggering delay of all the circuits linked with the tube. The time-base speed should be directly related to the bandwidth ofthe oscilloscope: for example, the accurate determination of a 100-picosecond leading edge requires a time-base speed of the order of 2 cm per nanosecond. This means that the time-base signal should have a slope ofup to 150 V per nanosecond and a linear range of about 200 V. Classical time-base circuits are Fig. 4. Schematic diagram of a wide-band coupler designed at LEP. The metal film M deposited inside the ceramic bushing K forms a 2.5 ohm resistance Rs in the outer conductor of the coaxial line carrying the signal S1 to be displayed. A fraction of SI appears across this resistance and is used as a synchronizing signal S2. The ferrite toroids F prevent the resistance Rs from being short-circuited by the walls of the coupling cavity. not capable of giving this kind of performance and new techniques have to be brought into play. A method used at LEP consists in forming a high-voltage step with as steep a leading edge as possible, and then shaping this edge by means of a series of filters which permit different time-base speeds to be obtained. The voltage step can be generated by ultra-high-speed switching circuits using avalanche transistors. This method allows extremely high slopes to be obtained at sufficient amplitude. A time-base generator based on this principle is shown in fig. 5. Fig. 5. Time-base generator with choice of four speeds.

260 PHILIPS TECHNICAL REVIEW VOLUME 30 I" The introduetion of the filter increases the delay of the time-base signal, and this delay becomes more pronounced as the desired time-base speed is reduced. Now we saw earlier that this delay should be kept within reasonable limits to avoid limiting the bandwidth of the oscilloscope. This explains why only high time-base speeds can be obtained in this way: unless oscilloscope should have, at the highest speed, an amplitude of about 150 volts, and a :fiat top of about 1.5 nanoseconds duration with leading and trailing edges of the order of 0.5 nanosecond. Special precautions have to be taken in transmitting such a pulse to the control grid of the cathode-ray tu be: in fact, in the region of the maximum writing speed, any accidental q\ J,," ~ ~.. " I fig. 6. Experirnental oscilloscope with a bandwidth of 5 GHz. extra circuits of classical design are included to give low time-base speeds, a high-speed real-time oscilloscope only has a limited range of speeds. Associated with fast time-base speeds is the necessity of producing very short beam-switching pulses with a very steep leading edge. By way of example, the beamswitching pulse for the cathode-ray tube of a 5 GHz overshoot of the unblanking pulse results in an unacceptable mod ulation of the intensity of the trace. It is also important to avoid any coupling between the beam-switching circuit and the deflection circuits of the tu be. lt is obvious that the synchronizing circuits should be capable of exceptionally accurate triggering. We

1969, No. 8/9/10 REAL-TIME HIGH-SPEED OSCILLOSCOPE 261 have spoken earlier of time-base speeds reaching 2 centimetres per nanosecond. On the other hand, the thickness of the trace on the screen is always very small, a typical value being 50 [Lm. If the maximum accuracy corresponding to this trace thickness is to be obtained, it can be seen that the maximum error in the triggering should be of the order of 2 picoseconds. Since there is a limit to the repetition rate of the time-base circuits, it is also necessary to introduce scaling circuits if it is desired to examine signals with very high repetition freq uencies (several G Hz). These various req uirements, including the short delay time specified, result in synchronizing or triggering circuits that are much more complex than in classical oscilloscopes. Fig. 7. Photograph of two closely spaced pulses (time difference 25 ps) as displayed on a 5 G Hz oscilloscope at a time-base speed of 2 cm/ris. The rise time of each pulse was 100 ps, the amplitude 12 V. The photograph is enlarged about 2 x. Conclusions We have briefly discussed the varied problems en- is reproduced in fig. 7. countered in the design of a high-speed oscilloscope for With the progress made in the last few years the real-time operation, and described the close interplay real-time oscilloscope has become a measuring inbetween them. The solution of these problems requires strument which is indispensible in the study of the the existence and close cooperation of a team whose many transitory phenomena which are found in the expertise ranged over fields as widely separated as various branches of modern science. vacuum physics, electron optics, and thermionic-valve techniques on one hand, and ultra-high-speed pulse techniques and microwave circuits on the other. The fact that LEP was able to build up a close-knit team with these capabilities has led to the production today of a prototype 5 GHz oscilloscope that has no equivalent anywhere else on the world market, and has already been found extremely useful at these Laboratories. This oscilloscope is shown infig. 6. An oscillogram of two closely-spaced pulses recorded by the equipment Summary. Since real-time amplification is not yet practical for signals of less than I ns duration, these signals can only be displayed by applying them directly to the cathode-ray tube of the oscilloscope. The cathode-ray tube must at the sarne time have large bandwidth, high sensitivity, high definition and a maximum writing speed high enough to permit the display of a very fast non-recurrent signal. The high-speed time-base circuit and the synchronization and unblanking circuits raise problems which are not encountered in conventional high-speed oscilloscopes and require new solutions. Method for growing single crystals of cuprous chloride J.-J. Brissot and A. Lemogne Cuprous chloride Cu Cl in its cubic modification shows a strong Pockels effect, and for this reason is an interesting material for making electro-optical modulators in a spectral region extending from 0.4 to 20.5 microns. J.-J. Brissot, lngénieur E.N.S.C.P., L. ès Se., is with Laboratoires d'electronique et de Physique Appliquée, Limeil-Brévannes (Va/- de-marne), France; A. Lemogne, D. ès Se., was formerly with these Laboratories. it solidifies at 422 C, but at this temperature a hexagonal wurtzite structure is formed, which changes into the cubic blende structure at 407 C [11. Since the phase transition takes place in the solid state, it is extremely difficult to obtain a cubic single crystal of useful dimensions from a pure melt [21. It is therefore necessary to [1] M. R. Lorenz and J. S. Prener, Acta cryst. 9, 538, 1956. [2] F. Sterzer, D. Blattner and S. Miniter, J. Opt. Soc. Amer. 54, 62, 1964.