Fullscreen HTML5Med Res (??MB)HTML4

Vintage OSCILLOSCOPE Valve-based Calibrated Oscilloscope from Radio, TV & Hobbies magazine I was pretty surprised when a fellow Historical Radio Society of Australia (HRSA) member turned up at one of our meetings with not one but two examples of Jamieson (“Jim”) Rowe’s outstanding oscilloscope design. It’s a fully-calibrated oscilloscope based on a three-inch (~75mm) diameter round CRT screen. With no exotic components or tricky construction, it was a well-designed and highly practical instrument that any enthusiast could build. The oscilloscope is effectively an X/Y plotter, plotting an input signal (Y-axis) against a time base (X-axis). That might sound simple, but the Y-axis amplifier must be able to reproduce the input waveform accurately, demanding a broad frequency response. Another challenge is that the timebase generator must be linear and 96 Silicon Chip adjustable over a wide range of speeds to suit signals of different frequencies. As for the Y-axis amplifier, let’s consider low-frequency inputs. While audio frequencies rarely extend below 20Hz, what about electrocardiograph signals, or signals from seismic monitors? What if we need to determine the DC component of a complex signal, such as a television waveform? Ideally, the low-frequency response should go all the way to DC. Early designs did not do this, either for cost-saving reasons, lack of perceived demand, or lack of design experience. Once such designs escaped the laboratory, designers implemented direct coupling and other improvements. What about the high-frequency end? There must be a practical limit to the highest frequency that a wideband amplifier can reproduce without loss. Australia's electronics magazine Common radio valves can easily work above 30MHz in tuned amplifiers, as their internal capacitances can mostly be incorporated into tuned circuits. A wideband amplifier usually has a resistive load, meaning that valve capacitances become a limiting factor. You will find a detailed description of how the circuit works in Jim’s original Radio, Television and Hobbies articles from June to August 1963. The circuit is shown in Fig.1, with some added voltage readings (green, peak-to-peak) and valve designators (yellow) to aid in troubleshooting and restoration. The overall sensitivity is governed by the required bandwidth and the high output voltage demanded by the cathode ray tube (CRT) screen. For conventional tetrode types with the deflection plates as the next-to-final siliconchip.com.au Radio and Hobbies (R&H), later Radio, Television and Hobbies (RTV&H), was Australia’s premier hobby and radio/electronics magazine from April 1939 until it was renamed Electronics Australia in March 1965. This clever oscilloscope, designed by Jim Rowe, was published in RTV&H’s June to August 1963 issues. It’s a brilliant circuit with one small flaw that I decided to address. By Ian Batty electrodes in the electron stream, sensitivities of some 20V/cm demand voltages approaching 150V peak-topeak for full deflection. As Jim noted, advanced post-­ deflection acceleration designs can bring full-screen deflection voltages down to tens of volts. The necessary expense and extra high-voltage power supplies were not judged appropriate for this design. This design settled for a -3dB bandwidth of 3.75MHz and an input sensitivity of 100mV/cm for fullscreen deflection. The vertical amplifier Vertical amplifiers have evolved logically. The first single-stage, AC-­ coupled amplifiers were developed into multi-stage versions. These commonly had limited bandwidths and provided up to 200V peak-to-peak siliconchip.com.au output to drive the CRT to full deflection. Adding a push-pull output stage halved the output voltage needed for full-scale deflection. By about this point, design theories that would extend amplifier bandwidths were being considered and implemented. Research in radar and pulse techniques during WWII had established techniques for wideband amplification, and RTV&H’s design team readily adopted them. The New Wide Band Oscilloscope in RTV&H, February 1957, p70, is the canonical design, with a bandwidth exceeding 3MHz. With a push-pull output and high-frequency peaking, the final step would be direct coupling throughout. Jim’s design is nicely tailored to give all the desirable features in an Australia's electronics magazine economical design. The cleverest part is the connection of the preamplifier and output stages in DC series, allowing a main HT supply of just 270V compared to the 400V found in Hewlett-­Packard’s model 150 from around the same time. With a -3dB bandwidth of 3.75MHz, it’s certainly good enough for most work, including black-and-white television. While the 3.75MHz limit is less than the full 5MHz bandwidth of PAL colour, the ‘scope usefully resolves the colour bar waveforms and displays the colour burst. This instrument is certainly good enough for most repair and alignment work. Previous RTV&H designs, using ex-disposals CRTs such as the venerable 5BP1, needed some 100-plus volts peak-to-peak for full-screen May 2024  97 98 Silicon Chip Australia's electronics magazine siliconchip.com.au Fig.1: the complete circuit of the Calibrated Oscilloscope from Radio, Television & Hobbies, June to August 1963. It uses just eight valves, six triode/pentodes, one twin triode (V5) and one pentode (V7), with the latter acting as a timebase oscillator. The red circle (at upper left) indicates the area where the changes noted in Fig.8 were applied. deflection. Jim chose the DG7-32/01; with its high deflection sensitivity, it only needs some 30V peak-to-peak, supplied in antiphase to its vertical deflection plates. This permits the clever design of the preamplifier and output amplifier in DC series from the HT supply noted above. The timebase, extending from 1s/ cm to 1μs/cm, is certainly suited to domestic electronics, including analog colour television. I could easily display the colour bar output from my Arlunya PG100 and observe the duration and positioning of the colour subcarrier burst with its 4.7μs duration. This showed that the Arlunya’s output, while adequate for testing, does not fully conform to the CCIR/PAL timing standard. The timebase While all vertical amplifiers look vaguely similar, timebase design is a bit of a zoo. Apart from special applications, the timebase waveform is a sawtooth wave with a linear ramp during the active display time and a rapid ‘snap’ back to zero during the blanked-out retrace period. The repetition rate must be adjustable, and it needs to offer synchronisation to either the signal being displayed or an external reference. Otherwise, the displayed waveform will not be steady on the screen. A neon lamp will go into conduction once the applied voltage reaches a particular value, typically 70V. It’s simple to take a power supply of perhaps 100V, string a series resistor to the lamp and pop the neon in parallel with a capacitor. On applying power, the capacitor will charge up until the neon strikes. It will then discharge the capacitor until the capacitor voltage drops below the neon’s extinction threshold. Once the neon extinguishes, the capacitor will begin to charge again, repeating the cycle. While this does give a continuous waveform (with frequency adjustable by changing either the capacitor or resistor), the waveform is exponential rather than a true linear ramp. This gives a less-than-­ linear time base, ‘crowded’ towards the right-hand end. The neon has finite ionisation and deionisation times, so the maximum operating frequency is limited to around 50kHz. This simple circuit siliconchip.com.au is difficult to synchronise, so it was reworked to use a gas-filled triode thyratron. The thyratron has strike and extinction characteristics similar to a neon and responds to synchronising signals on its grid. This makes it a practical circuit, but still with the limitations of non-linearity and only a moderate maximum operating frequency. R&H used such designs up to March 1950 (p52). These ‘soft’ timebases could be improved by replacing the timing resistor with an adjustable constant current source, giving a linear output waveform (R&H, April 1950, p64). The added complexity pushed designers to new circuits that were inherently linear. Various forms of multivibrator, bootstrap, and other switching circuits were used in high-performance instruments, but the circuit of choice became the Miller Integrator/Transitron, also known as the Phantastron. The Miller effect describes how a voltage amplifier effectively amplifies its own anode-grid (or collector-base or drain-gate) capacitance. The Miller effect can be used to create a repetitive linear waveform. There’s a complete description of how it works in R&H, September 1956, p32. Jim’s description (with the added detail of the synchronising circuitry) is in a separate RTV&H article in September 1962, starting on p44. The Phantastron exploits what is otherwise a serious problem inherent to tetrode valves. If the screen voltage is held constant and the anode voltage is reduced, there is a critical point below which the screen current skyrockets and the anode current falls. Fig.2 shows the effect, with the transition beginning around 100V on the anode. We need to add one more characteristic that is not commonly considered. The suppressor grid, invented to counteract the tetrode’s undesirable characteristics, can be used to control anode current. Its authority is much less than the control grid, needing some -50V for cutoff in the EF50. Now, let’s consider the basic circuit: a high-gain valve with the timing capacitor connected from the anode to the control grid and the timing resistor from the grid to a positive bias supply, shown in Fig.3. When power is applied, the valve will draw anode current through RL, and the anode voltage will begin to fall. But that will drive the grid negative via timing capacitor CT, which will tend to reduce the anode current. The circuit settles into a balance, where the tendency for the anode voltage to fall almost instantaneously to zero is balanced by the fact that such a fall would cut the valve off. The circuit will produce a ramp with a slope determined by timing capacitor CT and timing resistor RT. Varying the DC bias via the Time Cal potentiometer varies the waveform period. A simplified Phantastron If we left the circuit there, we would have a linear ramp but not the repetitive waveform we need for a timebase. Repetition is provided by the screen-suppressor circuit. As the anode voltage gets close to zero, the screen suddenly takes up the valve’s cathode current, the voltage drop across screen resistor RSG increases, and the screen voltage drops to zero. Fig.2: the sudden change in plate and screen currents at lower anode voltages is usually a problem, but it is taken advantage of in the ‘Phantastron’ oscillator. Fig.3: the basic configuration of the Phantastron oscillator. It generates a linear voltage ramp at its anode that’s periodically reset to a lower voltage over a short duration, thanks to the property shown in Fig.2. Australia's electronics magazine May 2024  99 The underside of the busbar version (one of two I received). It was the hardest to work on. This rapid drop is conveyed to the suppressor by CG3, forcing the suppressor sufficiently negative to cut off all current to the anode. When cut off, the anode circuit will rapidly rise to the full supply voltage. Once the screen comes out of its ‘bottomed’ state, the circuit resets, anode current rises, and a new downward ramp commences. The free-running circuit can be synchronised easily by applying synchronising pulses to cut off the control grid before the end of the active period. So, we have everything we need for an adjustable, synchronisable horizontal timebase waveform for the CRT from a single valve and a handful of other components. Restoration As mentioned earlier, I got my hands on two oscilloscopes built from the Scope 1: after calibrating the vertical amplifier it still had a poor high-frequency response. Scopes 1 & 2 are from my Parameters 5506 bench oscilloscope. I took them during testing to get a better idea of the exact waveform shapes than I could get from the smaller RTV&H ‘scope screen. 100 Silicon Chip Australia's electronics magazine articles. One used impressive busbar construction with solid wire insulated with sleeving, while the other had ‘just put it down and solder it in’ construction. I started with the busbar version as it had the full set of valves, but ran into a few problems. First, the main filter capacitors were drawing excessive current and would not reform. I popped in a pair of substitutes and started to test the rest of the circuitry. There was an extra voltage doubler stacked on top of the existing -300V supply for the CRT (it’s visible on a tag strip at the extreme right of the chassis underside). I have no idea why, and it was messing up the CRT voltages, so I removed it. Next, the main HT was low everywhere. I seemed to have some current drains that I couldn’t locate. I was struggling with the whole instrument – while the busbar construction looked neat, it was pretty near impossible to trace the circuit or get test probes past the wiring and onto actual valve socket connections. So I moved on to the other version, which was much easier to work on. siliconchip.com.au The other oscilloscope was messier, but easier to work on. However, it didn’t have a full set of working valves. Better yet, its electrolytic capacitors all tested OK. I ‘liberated’ the valves from the busbar instrument, tested them all, plugged them into the other unit and got into testing proper. Apart from the usual noisy switches and pots, the restoration was going well until I hit the timebase. The coarse time selector (1 sec, 100ms, 10ms etc) checked out OK, as did the fine time selector (×1, ×2, ×5). However, the variable time selector did nothing. The variable control works by pulling down the voltage divider reference, but it was having no effect. Checking both ends of the variable pot showed identical voltages, around 42V. The wiring is obscured behind a metal shield plate, but I was able to make out a green wire going from the bottom end of the variable pot. Instead of going to a grounded tag on a tag strip, it went to one with no other connection. Connecting the green lead to ground fixed what had been an original wiring fault. gain calibration, then adjusting the five frequency-­compensation trimmers. With a 1kHz square wave input, I found a conflict of settings, so I substituted a stair-step. The stair-step display showed sharp transitions without significant overshoot on all ranges except 100mV/cm. It showed much slower rise times on this range, as seen in Scope 1, so this setting (and just this one) was suffering from a poor high-frequency response. Given that the 100mV/cm range connects the input signal directly to the vertical amp’s input grid, what was causing this loss of bandwidth? Now for the vertical amp. It was working OK, so I went ahead with calibration. This required setting the Fig.4: without compensation, parasitic capacitances will cause a resistive divider to slow down rapid voltage transitions (Cin is the unavoidable grid/input/wiring capacitance). siliconchip.com.au Australia's electronics magazine Vertical amplifier A simple resistive attenuator works fine for DC measurements. Still, circuit capacitances will cause AC voltage measurement errors even at the higher end of audio frequencies and slow down the rise and fall times of square waves and other pulse waveforms. The 6BL8 pentode has an input capacitance of 5.5pF. Circuit wiring will add to that, but let’s stick with a known value. While this capacitance would have a negligible effect at audio frequencies, its capacitive reactance at 1MHz is only 30W. That will give slow rise/fall times, as shown in Fig.4. Fig.5: adding a compensation capacitor across the input resistor forms a capacitive divider with the parasitic capacitance, Cin, flattening the frequency response and speeding up transitions. May 2024  101 Fig.6: in the original Oscilloscope circuit, the compensation capacitor was over-compensating to account for the pure resistance of the calibration potentiometer. Fig.7: however, on its most sensitive setting, the compensation capacitor was shorted out, so we were back to an uncompensated divider and the resulting signal rounding. Fig.8: by adding another compensation capacitor across the calibration resistance, we no longer need the first capacitor to overcompensate, and it compensates on all sensitivity settings. The solution is to modify the attenuator to make it a capacitive divider, as well as a resistive one, as shown in Fig.5. The added capacitance in the ‘top half’ of the divider compensates for the inherent capacitance in the bottom, giving a division ratio that is (theoretically) flat with frequency. Valve input impedance falls significantly at frequencies above about 20MHz, which can add loading to the attenuator. More complex attenuator/input stage designs will be accurate over wider bandwidths, but the RTV&H circuit gives accurate attenuation for audio and video frequencies of its time. Given that the input attenuator in the ‘scope has such compensation, what was wrong, and why on only one range? The calibration potentiometer is not compensated, so it will degrade waveform rise and fall times. The 3~30pF master compensation trimmer was used to compensate for this and therefore null out the under-compensation in the calibration pot, as shown in Fig.6. On the 100mV/cm range, though, the 3~30pF compensation capacitor was shorted out by the range selection switch, and could no longer apply the overcompensation that was masking the calibration pot’s under-­ compensation, as shown in Fig.7. I dislike ‘fixing’ other peoples’ designs, but I decided to add a compensating trimmer across the pot, from its top connection to the wiper, as shown in Fig.8. After adjusting that, Scope 3: the stair-step on its own CRT. Scope 4: a colour bar waveform. Scope 2: after adding a calibration resistor and compensation capacitor, the oscilloscope was finally producing a proper stair-step display on all ranges. Scopes 1 & 2 also confirm, being from a much better-performing instrument, that (i) the asbuilt RTV&H scope did not fully resolve the issue of the input circuit’s design regarding frequency response, and (ii) when corrected, the input circuit - and the entire instrument - did work correctly. The final screenshots from the RTV&H screen (Scopes 3 & 4) confirm the RTV&H’s correct operation as a complete instrument. 102 Silicon Chip Australia's electronics magazine siliconchip.com.au Table 1 – Test point readings Test point Peak-to-peak I got a proper stair-step display on all ranges, shown in Scope 2. In hindsight, it would have been possible to accept the input signal directly to V1A’s grid and perform gain calibration by adjusting the cathode-­ to-cathode coupling of the long-tailed pair input stage. That is how the companion horizontal amplifier is calibrated. The CRT on the working set showed a strangely shaped ‘black hole’ around the middle of the screen. Being irregular, I wasn’t sure if it was screen burn-in, so I’ll leave it with the screen filled by an unsynchronised display to see if it self-heals somehow. The restored ‘scope also lacked a proper engraved graticule and dial illumination lamps (despite having the pot installed), so I pinched them from the busbar version. I’ve previously covered the hazards of unsecured mains cords, and both of these units were offenders. Putting a cord anchor into the chassis may demand enlarging the cord hole in the chassis. Using an ordinary drill or a file can risk damaging under-chassis components. In this case, using a stepped drill bit with a cordless driver gives you complete control over your work – mains-powered drills can take too long to spin down if anything goes wrong. A few other bits and bobs How good is it? We have a saying in the restoration world: “Buy two, get one working”. After my ‘tweak’, I was now able to display a PAL stair-step (greyscale) V1A G1 100mV V1A anode 1.2V V2A anode 30V V2B cathode 30V V5A G1 4.5V V5A anode 25V V7 G1 150mV V7 G2 15V V7 G3 5V V7 anode 13V waveform easily (Scope 2), the colour bar waveform (Scope 4), and the horizontal sync period. These three are complex, high-­ frequency waveforms with a lot of high-frequency content, multiple voltage steps from 0V to 1V and narrow pulse widths. As such, they are good tests of vertical amplifier bandwidth, linearity, and timebase synchronisation and stability. The blurriness of Scope 3 & Scope 4 is more due to my photography than the instrument itself; in use, the display is much more crisp. Voltage readings If you are lucky enough to acquire one of these instruments, I have added my DC analysis to the circuit diagram, Fig.1. The test point readings in Table 1 should also help with checking and calibration. Purchasing advice I already have a complete test bench, but if you see one of these, why not grab it? You’ll have an example of classic Aussie design that’s still highly usable. And it’ll fit just about any service bench! A top view of the oscilloscope chassis. Different units will vary somewhat depending on how the individual constructor has gone about doing things. siliconchip.com.au Australia's electronics magazine May 2024  103 More details on valve-based oscilloscopes by Ian Batty A basic thyratron-based timebase circuit is shown in Fig.9. HT is applied to the circuit via two resistors, VR2 & R3. Together with the selected timing capacitor (C3C5), these form the timing circuit. Note the small circle inside the valve’s symbol, denoting a gas-filled valve. The bias voltage (applied to the grid via R1) sets the thyratron’s strike voltage, restricting the maximum charging voltage of C3-C5. This uses the most linear part of the exponential charging curve, giving an acceptably linear sweep on the oscilloscope screen. More on that later. With no synchronising input, the circuit oscillates at a frequency determined by the selected ‘coarse’ timing capacitor (C3-C5) and the ‘fine’ variable resistor (VR2) in the anode supply circuit. The displayed waveform will drift across the oscilloscope screen in the absence of synchronising pulses. The thyratron is cut off during the positive-­going sweep period as the timing capacitor is charging, and only conducts during the negative-going “flyback” period. Applying synchronising pulses will force the thyratron to go into conduction early. As a result, the sweep frequency will match the incoming synchronising pulses, as long as it is set to run a bit too slow in the ‘free running’ mode. The displayed waveform will appear stationary on the screen. Thyratron behaviour The thyratron (‘door valve’) is a thermionic triode filled with low-pressure gas; hydrogen is commonly used in low-power tubes. When power is applied to the heater, we get the usual space charge cloud of electrons surrounding the cathode. If the grid is made negative to the point of cutoff, the space charge will be confined between the grid and the cathode. No current flows if voltage is applied to the anode as the valve is held in cutoff. So far, the thyratron is no different from any other vacuum triode. If the grid becomes less negative and voltage is applied to the anode, some electrons will pass through the grid and travel to the anode. This is also what we expect, but in doing so, they collide with hydrogen atoms. If the collisions are sufficiently energetic, some hydrogen atoms will become ionised, splitting into negative ions (electrons) and positive ions (the nuclei of the atoms). We now have a stream of electrons heading for the anode: the original electrons emitted from the cathode, augmented by the negative ions liberated from the hydrogen atoms. There is also a corresponding stream of positive ions heading for the cathode. As the positive ions enter the cathode’s space charge, they absorb space charge electrons and become neutral atoms. This ion-electron absorption destroys the space charge. Remember that it’s the space charge that limits the maximum current in any vacuum triode; it creates a high internal resistance between the cathode and the anode. Removing that space charge means that the valve’s internal resistance falls dramatically. The conducting thyratron can pass very high currents with a voltage drop as low as 15V. Large versions, used in high-power radar sets, could switch up to 40MW! Once conducting, the thyratron cannot be switched off by grid voltage. This can only be achieved by reversing the anode voltage polarity or taking it below the ‘keepalive’ (sustaining) voltage. Readers may recognise a similar action in the Thyristor/ SCR (silicon-controlled rectifier). Linearisation The charging curve for a series RC circuit (Fig.10) is clearly exponential over five time constants. The grid bias voltage controls a thyratron’s striking voltage as the anode goes positive. Setting the grid bias to, say, -30V allows a small amount of the space charge Fig.9: how a thyratron can be made to generate an almost linear ramp waveform with an adjustable frequency. 104 Silicon Chip Australia's electronics magazine to penetrate the grid wires and stream towards the anode. This electron stream must be highly energetic to cause ionisation, so such a grid voltage would prevent a type 884 (as used in R&H designs) from striking until its anode voltage reached some 300V. Dropping the grid bias to around -11V allows the type 884 to strike at just 100V. Now we can use a 400V supply and set the grid bias to -11V. This sets the anode strike voltage to 100V, and the valve will extinguish when the anode voltage falls to +15V, using just 85V of the potential 400V of charge. The resulting RC curve looks like Fig.11; it appears to show a linear ramp. Close examination reveals some non-linearity, but such a timebase waveform would be adequate for servicing audio and other common equipment. The thyratron has a particular deionisation period. It must expire before the valve can be made active again; typical times are in the low to high tens of microseconds. The type 884, used in R&H’s designs, could oscillate up to around 100kHz. While its lowest frequency could be set to a period of seconds, oscilloscope timebases worked fine with a minimum frequency of 20Hz. The R&H timebases were modelled on the RCA data sheet for the type 884. This design offered a continuously variable frequency ratio of 3:1. This demanded seven switched ranges (with some overlap) to cover 20Hz to 114kHz – see https://frank. pocnet.net/sheets/049/8/884.pdf Wideband amplifiers A wideband amplifier’s high-frequency response is mainly limited by circuit capacitances. The capacitances we can be certain of are the stage’s own output capacitance and the input capacitance of the following stage. For the 6BL8 pentode driving its triode, we get 3.8pF + 2.5pF = 6.3pF. That doesn’t sound like much, but that is a reactance of only about 7kΩ at the oscilloscope’s top end of 3.5MHz. With the 6BL8 pentode 10kΩ load resistor, the gain will be reduced by about 60% by 3.5MHz (about -8dB). Such a circuit would have a -3dB point of only about 1MHz. The simplest fix is to increase the stage’s load resistance with frequency. Since XL=2π × f × L, a suitable ‘peaking’ inductor (560μH in series with the 10kΩ anode load) will work just fine, as shown in siliconchip.com.au Fig.10: a standard capacitor charging curve with a resistor limiting the current. Fig.12. This is the most common method used. The simplified circuit omits all biasing. V1’s anode load comprises the usual load resistor (R2) and the peaking/compensating inductor, L1. V1’s output capacitance and V2’s input capacitance are lumped together. It’s also possible to use a cathode resistor bypassed by a low-value capacitor. Let’s say the cathode resistor is 470Ω and we shunt it with a 330pF capacitor. At low frequencies, the cathode circuit will appear purely resistive, applying degenerative feedback to reduce the stage’s potential gain. At around 1MHz, the capacitive reactance will be about equal to the cathode resistor’s resistance, and the stage gain will be increased to counteract the effect of valve capacitances. Fig.11: the thyratron charges a capacitor over a smaller portion of the curve, with the result being almost linear. Fig.13 shows a nominal amplifier’s high-frequency response from zero compensation (Lp = 0, no inductance) to an inductor with a reactance equal to the circuit capacitance (Lp = C1 × Rp2), where Rp is the total plate (anode) resistance. The circuit can become resonant, as the pronounced peak for the Lp = C1 × Rp2 curve shows. However, the stage’s load resistor strongly damps the circuit. Such a level of compensation is rarely used, as the excessive high-frequency response causes ringing on rising and falling transitions and creates undesirable phase errors. Notice that an inductor value of Lp = 0.5 × C1 × Rp2 gives an acceptably flat response and triples the upper -3dB point frequency (a gain of 0.7071; from f ÷ f1 = 1.0 to f ÷ f1 > 3). Conclusion & further reading Wideband amplifier design is complicated, but many texts on Radar and Television treat the matter thoroughly. The most authoritative source is the MIT RadLabs series, compiled at the end of WWII, to ensure their groundbreaking wartime work would be preserved. I was going to state, “they wrote the book”, but they actually wrote 27 books, available as PDFs from www.febo.com/ pages/docs/RadLab/ An extensive mathematical treatment of wideband amplifiers appears in Volume 18 of Vacuum Tube Amplifiers. For a basic description, consider reading Zworykin, V. K. & Morton, G. (1954) Television (2nd edition), John Wiley & Sons/ SC Chapman & Hall. Fig.12 (above): the roll-off in response due to unwanted capacitance in a wideband amplifier can be compensated for by a choke in series with the anode resistor. Fig.13 (right): a nominal wideband amplifier’s frequency response with no choke (green) and three chokes of different values. The red curve is as close to flat as can reasonably be achieved. siliconchip.com.au Australia's electronics magazine May 2024  105