This article will show that, if properly designed, a modern regen is quite capable of direct-conversion or superheterodyne-level performance, although it does require greater operator skill. Attention to a few simple details can provide an excellent receiver for ham or general-coverage shortwave use. For beginners and veterans alike, studying, building and using regenerative circuits can add new interest, excitement and fun to the radio hobby.
The regenerative circuit was used in both commercial and ham receivers in the 1920s until the early 1930s. It was the standard ham receiver during this period, and a great deal of experimental work was done to optimize its performance. As the great depression slowly ended, the use of the regen de clined, because many people could then afford to buy commercial superheterodyne receivers.
Hams continued to use regens, usually homebrewed, through the rest of the 1930s. Although better components were then available, regens of the 1940s and '50s had been reduced to introductory sets for beginners— with generally poor performance.
By the 1960s and '70s, regenerative circuits had been replaced by the now popular direct-conversion receiver.
What's Really Happening Here?
Fig 1 shows the basic regenerative circuit, discovered by Edwin Howard Armstrong1 in 1914.
Fig 2 shows a modern equivalent
circuit. Positive feedback, termed "regeneration" can be used to dramatically increase both the sensitivity and selectivity of RF circuits. Although a regenerative circuit may look very simple, how it actually functions is not simple. Its operation is both complex and quite fascinating.
If the output of an RF amplifier is fed back to its input—in phase—any signal present within this loop will be repetitively reamplified, typically providing a thousand-fold increase in gain over the same stage without regeneration. Although the power gain of a tube or transistor is fixed, the voltage gain in a regenerative circuit (ideally) approaches infinity at the point of oscillation. In actual use, infinite gain is not possible, due to phase shifts with-in the feedback loop. The practical result, however, is that a modern regenerative detector using a single transistor or JFET can achieve circuit gains of 20,000 or more.
Regeneration introduces a negative resistance into a circuit such that its net positive resistance is reduced. Since the circuit's selectivity, or Q, is equal to its reactance divided by its overall resistance, selectivity is also greatly increased when regeneration is applied.
When set below self-oscillation, regeneration provides a stable increase in both gain and selectivity. With more regeneration, the circuit reaches a very critical state, just at the threshold of oscillation. The exact "balancing" point—where the net circuit resistance is zero—is impossible to main tain, as even the smallest random noise source, given time, will build up to a self-sustained free oscillation.
With more regeneration, the circuit exhibits a net negative resistance and oscillates. As regeneration is increased further, curious secondary oscillations of a lower frequency are often created. These break up the main oscillation into a series of groups that periodically turn off or "quench" the main oscillation. Because of the quenching action, RF input signals build up to very high levels repeatedly, providing circuit gains approaching one million in a single stage. Discovered by Armstrong,2 this phenomena is termed "Super Regeneration," and its development by radio amateurs led to the first practical VHF receivers.3'4
Common Receiver Architectures: A Quick Overview
Fig 3 shows the block diagrams of regenerative, direct-conversion and superheterodyne receiver circuits. Let's take a look at their strengths and weaknesses.
As shown in Fig 3A, a regen uses an oscillating detector to heterodyne with an incoming RF signal at nearly the same frequency. The detector provides an audio output and, at the same time, functions as a very-high-gain RF amplifier and Q multiplier. So, a regenerative circuit oscillates, heterodynes, multiplies Q and amplifies simultaneously within a single stage.
By the use of positive feedback, a regenerative detector typically provides an audio output level of hundreds of millivolts. Because of its inherent high selectivity, high Q bandpass filtering ahead of the detector is not needed; this greatly simplifies the design. Another interesting characteristic of regenerative detectors is their ability to detect nearly all types of signals, including AM, CW, singlesideband (SSB) and FM signals. Because regenerative circuits generally use fewer components, they tend to consume less power, cost less and are easier to homebrew than other receivers. Although parts reduction is always important, many regens have been built without much regard to their operating performance. Often, the addition of just a few more components and attention to a few important details will greatly improve receiver performance.
A regen can provide very high quality audio. A regen allows you full control of the circuit's selectivity, which is often desirable. In fact, this user-controlled selectivity feature allows decent-quality reception of FM signals. A regen's variable selectivity allows the operator to optimize the slope of the receiver's amplitude versus frequency characteristic. The fixed, high selectivity of most superhet receivers prevents them from effectively demodulating FM signals by slope detection.
In use, a regenerative circuit performs quite differently depending on whether it is operated above or below the oscillation threshold. When receiv-
Fig 1—Armstrong's original regenerative circuit.
Fig 2—A modern regenerative detector circuit.
Nov/Dec 1998 25
ing AM signals, the detector is adjusted just to the threshold of oscillation for the best sensitivity and selectivity. Receiver performance can be quite good but it does require frequent readjustment of the regeneration control and a certain amount of operator skill.
When receiving CW or SSB, however, the detector is set to oscillate and the receiver is tuned away from the center of the carrier to produce an audio heterodyne (beat note). An oscillating detector is far more sensitive than any other. In addition, the grid-leak biasing normally used with these circuits tends to maintain a nearly constant oscillation amplitude over wide frequency ranges and, therefore, requires very little readjustment.
Regen negatives include potential detector RF leakage out the antenna, requiring some type of isolation: typically an RF stage.5 Another inherent problem is "blocking," where an oscillating detector tends to lock to the center of strong RF signals. The cure for this is some kind of variable input attenuation. When using a regen next to a transmitter, the high transmit RF levels produce severe blocking, which prevents using the regen as a keying monitor. The solution is a sidetone circuit, Piezo buzzer or similar device added to the transmitter circuitry. A final problem is hum modulation at the higher HF frequencies, those above approximately 14 MHz. When used in the oscillating mode, the RF output from the detector finds its way back to the antenna where it heterodynes with incoming RF signals. The solution is an RF stage with good isolation.
The audio quality (low distortion, wide bandwidth, low noise) of my homebrew regens in the AM mode is very noticeable and much better than any superhet I have used. I suspect that part of the reason for this is that the normal half-wave diode detector used in most superhets can generate very high levels of harmonic distortion. This, combined with distortion in the mixer stage, can produce harmonic-distortion levels of 20% or more. Diode-detector circuits are fine for FM reception as AM harmonic distortion products are removed by the limiter and discriminator stages.
As shown by Fig 3B, the heterodyne or direct-conversion (D-C) receiver has many similarities to a regen operating in the oscillating mode. Both mix a local oscillator signal with the incoming RF to produce a frequency output in the audio range. Because a D-C receiver's RF selectivity is determined in the front end, the reception range is usually limited to just a single ham band, with the RF input fixed-tuned to the center of that band.
The key difference between the two architectures is that in the regenerative circuit, both the circuit's gain and
Fig 3—Some common receiver architectures. 26 QEX
selectivity are amplified a thousand times or more during the heterodyne process. This means that the regen has much greater RF selectivity and gain. At the same time, both the gain and selectivity of the regen circuit can drop dramatically with very strong input signals. To achieve the same gain and selectivity, a D-C receiver needs RF gain and preselection before the mixer plus a very high gain, very selective audio section. However, the D-C receiver is easier to operate than the regen.
A good analogy here are two cars, one with an automatic transmission (D-C receiver) and one with a manual or "stick shift" (regen). The regen needs good design and several controls (plus operator skill) to perform, but is capable of very high sensitivity and selectivity.
As with the regen, the D-C receiver's oscillator can leak to the antenna. It requires an RF stage or mixer with very high LO isolation. The very low AF input level can make the D-C receiver prone to microphonics. Modern D-C homebrewers have developed many clever designs to overcome these difficulties, but the low-cost commercial D-C rigs I have tried suffer from very poor selectivity. This reinforces my belief that it is not receiver configuration, but rather the design, construction and operating skills of the builder that really matter.
The superhet receiver (Fig 3C) mixes the RF signal and that of an LO to produce an IF signal. The LO tracks the received frequency such that the sum or difference between the two frequencies always equals the IF. In a superhet, most of the amplification is provided at the IF using a high-gain, high-Q, single-frequency amplifier. When a single IF is used, it is usually lower than the received frequency because LC circuits of constant Q provide better selectivity at lower frequencies. For example, a 10 MHz LC circuit with a Q of 100 provides a bandwidth (selectivity) of approximately 100 kHz (10,000,000 4-100 = 100,000), but at 455 kHz its selectivity is 4.55 kHz (455,000 100 = 4550).
So, a superhet mixes the signal frequency to a fixed IF rather than use multiple high-Q RF amplifiers to cover the entire frequency range of the receiver. Oscillator leakage is minimal in superhets because the LO usually operates far from the received frequency. Thus, the oscillator's leakage is greatly attenuated by the receiver's input circuitry.
On the negative side, superhets are difficult to homebrew unless the range of received frequencies is very small. Although the IF filter(s) eliminates the need for single-signal front-end selectivity, it's still important that the front end eliminate the IF image frequency from the passband. (Images occur because the LO can heterodyne with frequencies both above and below the IF.) This is easily accomplished for single-band receivers by choosing a fortuitous IF. For a single band, several low-cost crystals (for IF selectivity) together with a fairly high IF can provide a decent receiver.
Nevertheless, to homebrew a good multiband superhet, you need to either bandswitch among selective front ends or use multiple conversions and up-convert to a first IF well above the highest received frequency. (For example, a 75 MHz IF for a 0.3 to 30 MHz receiver.) While some kits take the first option, these requirements have made the superhet impractical for all but the most skilled homebrewer.
Figs 4A to 4E, from the 1931 ARRL Handbook,6 show several types of regeneration control methods then used. The rotating or "movable tickler" method (4A) often used a variometer to adjust the amount of positive feedback. This method is prone to very severe detuning of the input signal as the regeneration level is increased.
Figs 4D and 4E show resistive control methods. A potentiometer or rheostat controls the detector's operating voltage. As the detector's plate or filament supply voltage is raised, its gain increases, causing an increase in regeneration.
The major problem with this method is that it suffers from a hysteresis effect when you adjust the control: It overshoots and requires readjustment. It's very difficult to set the regeneration level right at the oscillation threshold, a requirement for high selectivity in AM reception. When receiving CW, there is often a serious drift in the beat note with changes in signal level, temperature or power supply.
When screen-grid vacuum tubes came into common use, a resistive regeneration control that varied the screen voltage of the detector was common. Fig 4F shows a typical circuit from the 1942 Handbook.1 Stability was better in this "electron-coupled"
oscillator circuit (as the screen was electrically isolated from the plate) but since the detector voltage was not regulated, it still varied, changing the regeneration level.
Figs 4B and 4C show the throttle-capacitor method. From a performance standpoint, this is far better than any resistive control. A wellmade capacitive throttle allows regeneration to be set "right on the edge" of oscillation, resulting in great selectivity and sensitivity. With a capacitive control, the detector's supply voltage can be regulated, using a Zener diode or other means. In actual operation, this method provides a very dramatic improvement in regenerative-detector stability.
An RF Stage Preceding the Detector?
Back when many ham receivers used vacuum-tube regenerative detectors, power levels were high, often causing interference to other receivers in the area. As an example, a tube detector operating at 250 V and drawing 4 mA is a 1 W transmitter. Even when operated at a lower supply voltage and with very loose antenna coupling, it was still very easy to cause interference. The use of a screen-grid tube as an RF amplifier ahead of the detector could provide good antenna isolation, but many (or most) regens did not use them.
However, modern semiconductor devices provide us with better performance while operating the detector at far-lower power levels. For example, the JFET detector shown in Fig 6 operates at 5 V and about 0.3 mA of detector current, which is only 1.5 mW.
Despite this reduction in potential interference, it is still good engineering practice to use an RF stage to provide further isolation between the detector and the antenna. An RF stage prevents the antenna from absorbing power from the detector at frequencies where the antenna is resonant. It also prevents "aeronautical effects," where an antenna that swings in the wind changes the oscillation frequency of the detector. Finally, the RF stage provides gain that is often needed at the higher shortwave frequencies.
Many published articles recommend a tuned RF stage for their receivers. Although this is very sensible for a direct-conversion or superhet design, a tuned stage is both unnecessary and undesirable for a regen. The typical tuned RF stage shown in Fig 5A8 uses a bipolar transistor (Ql), operating as a common-emitter amplifier. The input signal from the antenna cannot connect directly to the tuned input circuit because the antenna's impedance would destroy its Q, and the antenna's capacitance would detune the circuit. This requires a second winding (or a tap) on the coil. Since Ql's input impedance is very low at RF, it also requires an impedance-matching device, usually another tap or winding on the coil.
The output from Q1 couples to the tuned LC circuit of the detector. Since the input and output circuits of the RF stage are both tuned to the same frequency, with gain from Q1 in between, a "tuned-grid, tuned-plate" oscillator is
Fig 4—Regeneration control methods. 28 QEX
"Throttle" regeneration control condenser
"Throttle" regeneration control condenser
created. This usually requires that the RF stage be neutralized to prevent it from oscillating. Therefore, building a decent, nonoscillating tuned RF stage involves a fair amount of work and skill. JFET devices are also commonly used, but even with a JFET, instability problems usually require a tap on T1 (and/ or T2) or the use of neutralization.
The need for neutralization can usually be avoided by operating the device as a grounded-base or grounded-gate stage, as shown in the JFET circuit of Fig 5B. Yet this still requires tapping the coil(s), and gain is lower in this configuration.
In a receiver that uses a mixer, these difficulties are often justified: It is necessary to supply enough signal to overcome the mixer noise and a high-Q tuned circuit before the mixer, to improve RF selectivity and reduce images. In a regen, however, little RF gain is needed. In fact, the detector is usually too sensitive in the oscillating mode and requires an input attenuator (more about this later).
The "Junk Box Special"
Fig 6 shows the circuit for a simple but very effective receiver that can be built using a variety of "scrounged" components. Although its parts count was kept to a minimum, this circuit still follows the guidelines previously mentioned. This design provides very good sensitivity and selectivity in the oscillating mode, as when receiving CW and SSB. AM reception is also good, but it does require frequent adjustments by the operator.
This two-band receiver covers a very wide frequency range (3 MHz to 13 MHz), is very compact and draws only 8 mA from a 6 V battery. The receiver can operate from a 5.3 to 12 V dc supply. Increase the values of R4 and R1 when operating from more than 6 V if supply current needs to be kept to a minimum.
The receiver uses a bipolar RF stage, a JFET detector, inductive coupling between the RF stage and the detector, a throttle-capacitor regeneration control, a regulated detector supply voltage and a low-cost audio amplifier IC.
Q1 operates as an untuned, grounded-base RF amplifier, providing gain and isolating the detector's oscillations from the antenna. This RF stage provides ample gain and its high output impedance does not load L2 excessively. This helps provide very high selectivity.
Too many regenerative circuits suffer from poor selectivity because of excessive loading at their inputs. This causes the circuit to detune as the REGEN ADJUST control is advanced, which then allows the circuit to oscillate prematurely, at a relatively low regeneration level. With light loading, regeneration can be increased until the receiver's selectivity is a few hundred hertz. This, in fact, is an excellent operational test for any regenerative circuit.
Ql's base is tied directly to the supply: This eliminates the usual base-biasing resistors, and any variations in Beta will not effect the performance of this stage. In addition, it is now easy to set Ql's operating current. Since the base of Q1 is tied to the supply, Ql's emitter will be approximately 0.7 less positive than the supply. So, for a +6 V supply, there will be 5.3 V across R1 and
2.4 mA flowing through it (5.3 V / 2.2 k£2 = 0.0024 A).
By experimentation, I found that
2.5 mA is a good operating current for Ql. Too little emitter current allows detection of strong AM broadcast stations because they overdrive the stage. (They exceed the bias voltage across R1 and cause the emitter-base junction of Ql to detect the signal.) Excessive operating current "eats-up" the batteries. If the receiver is to be powered from a supply greater than +6 V, R1 should be increased to keep Ql's emitter current at approximately the same level. So, for +9 V operation, R1 should be about 3.5 kQ and about 4.7 kQ for +12 V. R4 should also be increased for +9 or +12 V operation, for the minimum-acceptable Zener current. Note that the input impedance of the RF stage is not R1 but much lower, as determined mainly by Ql's emitter-base junction.
CI ac couples the antenna's signal from Ql's emitter, which prevents shorting R1 should the antenna become grounded (if you short Rl, Ql will fail). LI inductively couples the output signal from the Ql's collector to the detector.
A Motorola type J310 or similar JFET can be used in the RF stage instead of the bipolar transistor. Simply
Fig 5—Typical tuned RF stages.
ground the base, connect the JFET drain to LI and connect a 200 £2 resistor between the source and ground. The antenna connects to the JFET source through C1. The JFET RF stage has lower gain but is less likely to be affected by extremely strong local broadcast stations.
JFET Q2 operates as a regenerative detector in a Hartley oscillator circuit, the tap on L2 providing the positive feedback needed for oscillation. The optimum position of the tap depends on the gain of the active device used for the detector. If a (high-gain) bipolar transistor were used for Q2, the tap would need to be closer to the ground end of L2, to maintain the same (smooth) level of regeneration control,
C3 is a standard two-section AM-radio tuning capacitor with its trimmers removed. A single-pole miniature toggle switch, connected with very short leads, allows one or both sections to be used, providing very simple band switching. A FINE TUNE capacitor, C4, is connected in parallel with C3. A small mica capacitor, in series with C4, permits the builder to set the desired bandspread and allows the use of almost any small airvariable capacitor from the junk box.
R2 and C5 provide grid-leak bias for the detector, which (together with R3) sets the total bias level of Q2. The JFET detector is operated with a very high negative bias, keeping its gain low. For an oscillating detector, it's very important to have a low-gain device (such as a JFET) as the active element within the regenerative loop; this permits very smooth regeneration control. The old-timers of the 1920s knew this and operated their tube detectors from low supply voltages to achieve the same result. However, a high-gain bipolar transistor provides the highest sensitivity in nonoscil-lating operation. So, a detector using a 2N2222 or similar transistor will make a very sensitive AM shortwave receiver that needs only a 39-inch whip antenna to get hundreds of stations. This receiver will have poor CW and SSB performance, however.
The RC time constant of R2 and C5 is long enough that AF amplitude variations in the RF carrier cannot "leak off' fast enough and therefore change the dc bias level of the JFET. These bias variations cause the JFET's operating current to change along with the modulation, providing AM signal detection. The grid-leak biasing tends to maintain the detector's oscillations at constant amplitude, which improves the
Was this article helpful?