Fullscreen HTML5Med Res (??MB)HTML4

While designing an oscilloscope with a 20MHz bandwidth is relatively easy, it is much harder to achieve 150MHz and even harder to get to 1GHz. In this chapter, we discuss some of the techniques which achieve this and result in the beam electronics moving at more than one quarter the speed of light. By BRYAN MAHER A good high-frequency oscilloscope is the only way to show the true shape, rise and fall times or the presence of any dis­ turbing anomalies or oscillations in your signals. A very high trace writing speed is also required, otherwise fast rising pulses will be invisible on the screen. Only a wide bandwidth oscilloscope can reveal logic circuit malfunctions or unwanted defects in input signals such as fast jitter in the sub-nanosecond range or high frequency ringing and overshoot on pulse waveforms. Typical of modern analog instruments is the Philips PM3094, a scope of 200MHz bandwidth, capable of displaying four signals simultaneously and with CRT readout of measurements on screen. Its vertical sensitivity of 2 millivolts per division is accurate to within 1.3% for large deflection in mid-screen. Main and delayed time­ bases are provided with the fastest sweep speed being 2 nanoseconds per division (when x10 horizontal magnification is used). An acceleration voltage of 16.5kV ensures sharpness and brightness of the traces. The need for wide bandwidth CROs to display the true shape of even moderate frequency signals is easily This Philips PM3094 200MHz oscilloscope can display four signals simultaneously. Vertical sensitivity is 2mV per division. Its fastest timebase speed is 2ns/div (using x10 magnification). Features are main and delayed sweep, on-screen readout and au­tomatic delta-time measurements. 6  Silicon Chip demonstrated by displaying similar waveforms on two oscilloscopes having differ­ ent bandwidths. In a particular case, the author was measuring pulse currents through a very low value resistor. Though the pulses were at the relatively slow repetition rate of 4kHz, the pulse rise time was known to be extremely fast. A simple demonstration For the demonstration, rather than use two different scopes, I used a Tektronix 7904 which has two vertical amplifi­ers, one with bandwidth of only 100kHz and one with 200MHz band­ width. When the signal was plugged into the low bandwidth chan­nel, its rise time appeared to be quite modest at about 20 micro­seconds. On the second channel though, the picture was quite different, with the pulse having a very fast rise time plus over­shoot and severe undershoot. From this demonstration it can be seen that, even with low frequency signals, it takes a scope with a really wide bandwidth to reveal the true nature of many waveforms. While there are many good scopes on the market with a bandwidth of around 20MHz or so, much lab and workshop use requires models with 10 times that figure or considerably more. Design brief What must designers do to produce an analog oscilloscope with a band- Fig.1: sectional drawing of a high performance CRO tube, capable of wide bandwidth. The total acceleration potential is 24kV, with most of that applied by the spiral post deflection acceleration (PDA) anode. width up to, say, 500MHz and with a writing speed to match? Last month, we defined “Deflection Factor” as the voltage which must be applied to the deflection plates to produce one centi­ me­tre of trace on the screen. Designers aim to keep that voltage requirement as low as possible. The deflection factor can be reduced by lengthening the vertical deflection plates and by reducing their spacing. If the vertical deflection plates are made very long and spaced close together, they must also be curved to give clearance to the deflected beam. By this means, the deflection factor can be brought down to about 6.5 or 7V/ cm which is a great improvement. It means that the vertical amplifier only needs to develop about 56V of signal for 8cm of vertical deflection. Typically, the extra capacitance of long connecting leads is avoided by bringing the deflection plate connections straight out through hermetic metal-glass seals in the neck of the CRO tube. The deflection amplifier is mount­ed adjacent to keep the leads short. However, the inevitable effect of longer deflection plates and closer separation is increased capacitance, up to as much as 16pF. That becomes a real problem at very high signal frequencies. Writing speed & brightness While the deflection amplifiers may be able to deflect the beam sufficiently at high frequencies, you still need to be able to see the trace on the screen. This is a function of the “writing speed” of the CRO tube. This is defined as the fastest speed at which the trace can travel over the screen and still be clearly visible. Consider displaying the rising edge of a high-frequency pulse signal, which has a rise time of about 300 pico­ seconds. Let’s assume that the timebase is set to 200 picoseconds/division and that the trace moves up the rising edge of the pulse 4.5 divisions vertically in 400 picoseconds. This represents a writing speed about of 100 picoseconds /division or about 1/3 the speed of light! That might seem like an extreme set of conditions but one of the photos in this article portrays this event, taken from the screen of a Tektronix 7104. This has the fastest writing speed of any scope currently available. Achieving this extreme writing speed takes some very special technology. To give a less extreme example, say we wanted to display a 500MHz sine­ wave on the screen. The period for this signal (ie, time for one cycle) is just two nanoseconds. Accordingly, with a timebase setting of 1ns/div, the trace will take 10 nanoseconds to cross the 10cm wide screen. So five cycles of the signal will be displayed, repeatedly. The persistence characteristic of a P31 phosphor screen means that the light generated by each sweep lingers for about 300 microseconds after the beam has passed so it still lingers while subsequent sweeps occur. So the display you see always consists of many thousands of sweeps superimposed. The light from those thousands of superimposed sweeps may give acceptable brightness, provided the electron beam hits the phosphor with sufficient energy. Single shot display But what happens when you want to display a very fast non-repetitive pulse? In logic circuits and many electro­physical systems, signals must have fast risetimes, yet sometimes repeat only leisurely, maybe once a minute, or less. In such cases there is no superimposition of consecutive sweeps to add trace bright­ness. Each display of the signal fades away before the next occurs. The pulse actually photographed on the Tektronix 7104 analog oscilloscope referred to above occurred only once; May 1996  7 Fig.2: electron transit time is the time taken by an electron to pass through the vertical deflection plates from A to C. With low frequency signals (a), the signal voltage barely changes during the transit time, so sufficient beam deflection occurs. With very high frequencies applied to the vertical amplifier (b), the signal voltage can change back and forth while an electron is travelling from A to C. This effect places a frequency limit on CRO tubes using solid vertical deflection plates. a one-shot, never repeated. In such a case the electron beam must be so energetic that its collision with the phosphor generates suffi­ cient light immediately, in a few picoseconds. That’s what we mean by a CRO tube capable of a fast writing speed! A high energy beam means high electron velocity. That’s one reason why wide bandwidth oscilloscopes must use very high accel­eration volt­ ages. The second reason is that to show fine detail accurately on the screen, the trace must be very thin, as well as brilliant. The light spot must be small, as little as 0.1mm in diameter. That’s difficult to achieve because the negatively charged electrons in the beam repel each other, spreading the beam. The cure for that is to accelerate the electrons to as high a velocity as possible. Typically, we need an electron beam velocity of about 90,000 kilometres per second (nearly one third the speed of light). That electron velocity requires a very high accelerating potential of about 24kV. As discussed last month, there is a conflict between accel­eration voltage and deflection factor. Increasing the accelera­tion voltage by a factor of 12, (2kV up to 24kV) will drastically spoil the deflection factor. Previously, we were concerned about the acceleration voltage measured between the tube cathode and the region near the deflection plates. The solution is to use Post Deflection Acceleration (PDA). This is shown in Fig.1. In this case, the electron beam is initially accelerated to a low velocity of about 26,000km/second, using a 2kV potential between the cathode K and the acceleration grid G3. The average potential on the vertical deflection plates Y1,Y2, rests at about the same potential as G3. Those low velocity electrons passing between the vertical deflection plates are easily deflected. So the low deflection factor obtained by long curved deflection plates and close spacing is retained. Now comes the clever part: do most of the acceleration after the beam has been deflected. Fig.1 shows that cathode K is maintained at -1850V while the acceleration grid G3 is held at about +150V. That means that the acceleration field between cathode K and G3 is 2kV. After leaving the deflection plates, the electrons come under the attraction of the +22,150V PDA potential at the screen. But before acceleration, those elec- Fig.3: distributed vertical deflection plates overcome the upper frequency limitations imposed by transit time, by segmenting the plates into many small sections. Each plate section is fed signal from a tap on a delay line. The aim is to have the signal elec­trons fly past the deflection plates at the same speed as the deflection signal propagates along the delay line. 8  Silicon Chip An oldie but a goodie: the Tektronix 7904 oscilloscope consists of a mainframe and CRO tube capable of 500MHz bandwidth, with provision for two independent plug-in vertical amplifiers and two plug-in horizontal sweep timebase units. trons must be focussed into a fine stream. This happens at G2, the “focus grid”, which is a metal cylinder. After leaving the cathode, the electrons pass through a 1mm hole in the control grid (G1), which acts as a point source. To achieve focus, we critically control the shape and strength of the electric field between the cathode, focus grid G2 and accelera­tion grid G3. Electrostatic lens The G2-G3 region is an electrostatic lens, with its focal length altered by changing the ratio of the potentials on these two electrodes. Any electrons which happen to be on the centre line when passing through G2 and G3 are equally affected by all parts of the fields here, so they pass down the centre-line of the beam. But electrons which are off centre line in passing through the region between (A) and (B), encounter the G3-G2 electric field which has a component of force repelling those electrons back towards centre. Because G2 is only a few hundred volts more positive than the cathode, the electrons near (A) are moving relatively slowly, so their path is easily affected by the fields. Thus, their track bends easily as at (B). But the path of such electrons must bend again, between (B) and (C), to prevent overshooting the centreline. This is achieved by the component of the G3-G2 field between (B) and (C), where the field is facing in the opposite direction to that between (A) and (B). Because G3 is 2kV more positive than the cathode, by the time the electrons have passed (B) and (C) they have accelerated up to 26,000km/second. So their path is bent less easily at (C) than at (B). Therefore, the bending of the path back to centre beyond (C) is gradual and progressive. The G3-G2 potential difference can be adjusted by focus control VR2 to force all electrons to come together at one small point upon reaching the phosphor at the screen. Due to the non-axial attitude of some electrons entering the electrostatic lens at (A) in Fig.1, focusing suffers from astig­matic error, causing the spot on the screen to form a tiny ellipse, rather than a circle. This is minimised by the astigma­tism control, VR3. This sets up a cylindrical electrostatic lens effect between G3 and the vertical deflection plates (Y1,Y2). Proper adjustment is obtained when the spot on screen is the best approximation of a tiny circle. Moving beyond the deflection plates, the electron beam accelerates rapidly, attracted by the Post Deflection Accel­ era­ tion (PDA) voltage of +22150V. The PDA aquadag electrode inside the screen is deposited in the form of a spi- ral which helps give a uniform electric field over the entire screen. For this and a number of other reasons, high performance CRO tubes have a very thin layer of aluminium deposited over the phosphor compounds inside the screen. The electron beam pene­ trates this aluminium to reach the phosphor. Beam electrons penetrate the aluminium layer, excite the phosphor compounds, then use the aluminium as a pathway to flow away to the aquadag layer and to the high voltage PDA supply terminal. This prevents charge building up on the screen; an important feature. By contrast, CRO tubes using acceleration voltages below 10kV often do not have an aluminium screen backing, because penetration of that metal would absorb too much of the available electron energy. Without this aluminium layer, the electrons arriving on the screen phosphor must leak across the luminescent material to reach the aqua­dag. This in turn means that, because of the poor electrical conductivity of phosphor compounds, a large number of migrating electrons will be found on the phosphor and the inner side of the front glass screen. As a result, the screen acquires a negative charge. Such a charge is undesirable, as it partially repels new electrons arriving in the beam, reducing the beam current and thereby the screen brightness. Reflecting the light The aluminium layer also acts as a reflector for the phosphor. Without it, light generated within the phosphor not only radi­ates out through the front glass but also back inside the tube, where it is wasted. In fact, some 6090% of the luminance can be wasted in this manner. An aluminium layer can reflect this light back to the screen, thereby approximately doubling the trace brightness. The aluminium backing also absorbs any unwanted negative ions which may arrive at the screen. Ions are (relatively) heavy charged atoms emitted by the cathode along with the electron beam. Any heavy ions reaching the phosphor will cause its rapid deterioration by ion-burn. Aluminising the screen prevents this problem. As well, the aluminium layer helps dissipate heat gener­ated by the impact of electrons with the phosphor May 1996  9 Oscilloscope bandwidth has a big affect on the signal displayed. Here pulses of current are being measured by the low bandwidth amplifier of a Tektronix 7904 oscilloscope. Although the pulse repetition rate is only about 4kHz, notice the severe rounding of the displayed waveform. grains. Again, this helps reduce longterm deterioration of the phosphor. Not all the effects are good though and there are some disadvantages. For a start, electrons in the beam lose 3- 5keV (energy) in penetrating the al- This view shows the same waveform as at left but displayed via the 200MHz vertical amplifier in the Tektronix 7904. Notice the pulse overshoot and severe undershoot, features which are com­pletely unseen on the lower bandwidth amplifier. uminium layer. Making the aluminium thinner would not help, as then the metal would be insufficiently reflective to light. The usual remedy is to raise the beam energy, by increasing the acceleration voltage. The aluminium backing also tends to broaden the trace seen on screen, as a side effect of the reflection of light back through the phosphor. This effect is ameliorated by making the phosphor no thicker than the electron penetration depth and using phosphor compounds having micro­ grain crystals. Another effect of the aluminium backing is to reduce the apparent contrast of the screen display. This is because ambient room light passes through the glass screen and the transpar­ent phosphor and is then reflected by the aluminium layer, to back-illuminate the whole screen. The usual remedy is to make the tube front of thick dark glass. The trace illumination then loses its bright­ ness once in passing through the glass, while any ambient room light reflected by the alumin­ium layer loses its brightFig.4: a patented “microchannel plate”, ness twice, because it makes two used in the Tektronix 2467B analog trips through the dark glass. This oscilloscope, acts as an electron multiplier technique is also used in computimme­diately before the phosphor. This er monitor and TV picture tubes. can increase the intensity of a dim The aluminium backing layer waveform a thousand times, enabling a must be thick enough to act as a very high speed trace to be clearly visible. 10  Silicon Chip light reflector, yet thin enough to allow the electron beam to penetrate to excite the phosphor. For tubes using overall accel­eration potentials between 10kV and 25kV, an aluminium backing layer 100 nanometres thick is satisfactory (100 nm = 1/4 wave­length of visible violet light). As an alternative to using very high acceleration voltages, the Tektronix 2467B 400MHz analog oscilloscope uses a patented “Bright Eye” display, a micro­channel plate behind the screen phosphor. This acts as an electron multi­ plier to increase the intensity of the trace of a dim waveform up to a thousand times. With this option, a single pulse at 500 picoseconds per division sweep speed is quite visible. High frequency limits The above techniques are very effective for increasing bandwidth and maintaining a good deflection factor and high writing speed but there are still limits. Reducing the vertical deflec­ tion factor down to 6.5V/cm, by using elongated, close-spaced deflection plates, is sufficient for analog oscilloscopes for frequencies up to 150MHz. However, the resultant increase in capacitance between the plates is prohibitive for higher frequencies, because plate charging current drawn from the deflection amplifier rises pro­ portionally to the signal frequency. At 500MHz, for example, the impedance Here, a non-recurrent pulse with a rise time of 350 picoseconds and an amplitude of 50 millivolts is portrayed on the screen of a Tektronix 7104 analog oscilloscope, at a timebase speed of 200 picoseconds/ division. This is possible only if the scope has an extremely wide bandwidth and a very fast writing speed. of a 16pF capacitor is only 20Ω and when driven by 50V or so from the deflection amplifiers, the charging current is quite consid­erable – several amps, in fact. This is a very difficult requirement at 500MHz. Worse still, above 150MHz a new effect called “electron transit time” raises its ugly head. This places an absolute upper limit on the frequencies which can be displayed on an oscillo­scope tube. Fig.2 illustrates the passage of a beam electron on its way to the screen. At low frequencies, in Fig.2(a), the potential on Y1 is positive all the time that the electron is passing through points A and B and C, so it is continually deflected upwards, as should occur. But at frequencies in the 150-1000MHz range, the signal voltage applied to the vertical deflection plates can pass through perhaps half a cycle during the time that an electron travels from A to C – see Fig.2(b). In Fig.2(b), at point A the electron is attracted upwards, at B it is heading back down, and at C it has straightened out again so the net effect is very little deflection. The result is that signals above a certain frequency cannot be displayed, no matter how the vertical amplifier is designed. This is a fundamental frequency limitation of the CRO tube itself, caused by changes in deflection signal polarity during electron transit time. The long deflection plates now defeat their original purpose which was to improve the deflection factor. Distributed deflection plates So there are two problems to be overcome with long curved deflection plates: electron transit time and high capacitance. The answer lies in the use of distributed deflection plates, as shown in Fig.3. Here the vertical deflection plates are segmented into 44 sections. The leftmost upper and lower plate segments Y1 and Y2 are supplied with signal from the vertical deflection amplifier. Subsequent deflection plate sections Y3-Y43 and Y4-Y44 all tap onto junctions of a delay line. This line consists of the series inductances L1-L41 and L2-L42, together with the distributed shunt capacitance of all the deflection plate segments. In one Tektronix design, each section (L1, L2, etc) of series inductance consists of five turns of wire. Each plate segment is 3.175mm long, with 1mm spacing between each segment. The inductive coils of the delay line and the plate segments are mounted on glass rods within the neck of the CRO tube. In Fig.3, deflection signals from Q1, Q2 travel down the line from left to right. Remember that here we are dealing with frequencies up to the UHF region, so signal reflections must be avoided. Therefore, impedance matching of source, line and load is mandatory. To achieve this, resistors R1 & R2 reduce the output impedance of Q1, Q2 to match the characteristic impedance of the delay line. Then resistors R3 & R4 terminate the delay line in its characteristic impedance to prevent end reflections. The load on the deflection amplifier Q1, Q2 is now the characteristic impedance of the delay line, about 900Ω, which is easily driven. Four leads from the deflection plate assembly at H, K, M, N are brought out through metal-glass seals in the CRO tube neck, for connection to resistors R1, R2, R3 & R4 and the deflection amplifier which is mounted adjacent. All leads are as short as possible to minimise inductive impedance at such high frequencies. Signal propagation velocity In the distributed deflection system shown in Fig.3, the aim is to have the electrons whiz through the deflection plate assembly at the same velocity as the deflection signal propagates along the line from Y1, Y2 down to Y43, Y44. Such matching of velocities results in full beam deflection at all frequencies, because each electron passing between the plates is affected by the same signal deflection voltage for the entire transit time from A to C. Unfortunately, the design of the deflection assembly and delay line results in a signal propagation velocity which is not quite constant over the entire frequency range. Velocity mismatch will eventually occur at high frequencies, resulting in reduced deflection. Early distributed deflection plates had the transmission line components outside the tube. This necessitated too many metal-glass seals for the connections through the tube neck. To overcome this problem, Hewlett Pack­ ard researchers developed a technique called “Helical Distributed Deflection Plates”. This eliminates external components since the required inductance and capacitance are built in. Each helix is equivalent to a lumped-parameter transmission line feeding the distributed plates. Each helix is a continuous strip of metal, mounted rigidly to glass rods which also support the electron gun assembly. Only four feed­ throughs in the tube neck are needed, to connect to the vertical deflection amplifier and the terminating resistors. Acknowledgements Thanks to Tektronix Australia, Philips Scientific & Industrial and Hewlett Packard for data and illustrations. Also thanks to Professor David Curtis, Ian Hartshorn, Ian Marx and Dennis Co­bley. References (1) Tektronix Aust: “XYZ’s of Oscilloscopes” and Application Notes. (2) Hewlett Packard Aust: R. A. Bell “Application Note 115”. (3) Philips/Fluke USA: “ABC’s of Oscilloscopes”. (4) Van der Ziel A. “Solid State Physical SC Electronics”, Prentice Hall, NJ. May 1996  11