A 1.2-volt Vackar-style minimalist regenerative receiver: Design, analysis, and implementation (updated 2017/05)

Abstract

A minimalist shortwave regenerative receiver circuit is presented that is powered by a 1.2-volt cell and uses exclusively 2N3904 transistors. Tuning is accomplished by a 1SV149 hyperabrupt varactor powered by a separate 9-volt supply. The receiver design emphasizes simplicity and exhibits good operational performance including smooth regeneration control, frequency stability, and sensitivity sufficient to hear HF band noise. Key aspects of the receiver's behavior are analyzed using LTspice. Several references are provided, both to explain aspects of the presented design, and to show more advanced designs.

Introduction

The goals of this project were as follows:

1. Design and implement a minimalist but practically usable regenerative receiver capable of HF reception from 3-30 MHz.
2. Analyze the key aspects of the receiver's operation (e.g. behavior of the regenerative detector, gain and frequency response of the AF amplifier) using the circuit simulation tool LTspice.
3. Write this article explaining the design and explaining the LTspice analysis techniques in enough detail such that the interested reader can reproduce the simulation results. Also, provide several references to improved designs for regenerative detectors or regenerative receivers.

For the circuit design, the specific constraints are described in the following sections. The constraints are described in two sections: one dealing with constraints to enforce minimalism, and one dealing with constraints to ensure practical usability of the finished receiver. A section on non-constraints (possible improvements that were not adopted for this receiver) is also provided.

Minimalist constraints

The following constraints were chosen to keep the design a minimalist design.

1. Only 2N3904 transistors should be used.
2. The power for the transistors should come from a 1.2-volt cell.
3. The receiver should be built as physically small as possible, preferably small enough to fit in a pocket.

The choice of these constraints reflects the personal biases of the author, so I shall explain in more detail the motivation behind choosing these constraints.

Why 2N3904 transistors?

2N3904 transistors are commonly and cheaply available, aiding reproducibility of the design. Their transition frequency is high enough to allow them to serve as oscillators over the entire HF spectrum. At the same time, their transition frequency is low enough so that they don't tend to oscillate spuriously at VHF or UHF. And, I have several on hand.

Why 1.2 volts?

A 1.2-volt AAA cell is the smallest generally-available, general-purpose, rechargeable cell, and I prefer using rechargeable rather than disposable batteries to power my receivers. Given the minimalist goals of the receiver and the additional goal of small physical size, a 1.2-volt AAA cell was the natural choice for the power supply. Restricting the receiver to be powered off of a single cell rather than a pack of several cells again helps keep the physical size of the receiver small, as well as greatly simplifies the process of removing, charging, and replacing the battery; it would be tedious to individually charge and replace several cells at once. Also, using a single cell for power might allow for the possibility of in-situ battery charging, via for example a solar cell. Finally, it is an interesting exercise to design a receiver with sufficient intput-to-output gain (approximately 100 dB) that is powered by a single low-voltage cell.

Why a physically small receiver?

The reason for constraining the design to be physically small is to allow the set to be easily taken on trips or in the field. A small physical size encourages more frequent use of the receiver in various situations. Also, designing a physically small HF receiver is an interesting challenge to ensure good RF behavior (no unwanted output-to-input couplings or unwanted oscillations) and a convenient physical circuit layout allowing practical operation of the physical controls.

Practical use constraints

To enable the receiver to be usable in practice for listening to a variety of signals on various HF bands, the following constraints were defined.

1. To allow both a wide tuning range and a comfortably slow tuning rate, in a physically small package, tuning should be done with a 1SV149 hyperabrupt varactor connected to a ten-turn potentiometer.
2. To prevent detuning effects from hand capacitance or variations in antenna impedance, an isolating RF stage should be included.
3. To make best use of the receiver's dynamic range, a passive RF attenuator should be included.
4. The operation of the regeneration control must be smooth, free from hysteresis, and free from squegging, motorboating, super-regeneration, or fringe howl.
5. To allow easy band-switching, the coil should be a 2-terminal coil with no taps or secondary windings.
6. The receiver must be able to operate from 3-30 MHz without changing any components except for the tank inductor.

Non-constraints

No receiver design is perfect. The below list enumerates some "imperfect" aspects of this article's receiver design that were considered for improvement, but that were ultimately excluded from the scope of this design.

Use of a single-active-device regenerative detector

No attempt is made to separate Q-multiplication, detection, and amplitude limiting into separate stages, instead relying on the “classical” single-active-device topology that requires that single active device to be responsible for linear signal amplification, non-linear detection, and self-limiting amplitude behavior as the regeneration control is increased past the oscillation threshold. Distributing the above functions onto separate active devices in general allows a higher-performance receiver, as each active device can be optimally biased and parameterized for its particular function (linear signal amplification, non-linear detection, or non-linear gain compression). Achieving such a high level of performance was not a goal of this project.

However, for examples of higher-performance designs, see for example Ref. 1 (references are provided at the end of this article), which separates the Q-multiplier and detector stages (using a hybrid cascode detector stage). For an even more involved design, see Ref. 2 and Ref. 3, which describes a design that separates the Q-multiplier, detector, and gain compressor into three separate stages, and furthermore implements some unique optimizations in the gain compressor, using a scale-invariant gain compression law, to allow a never-before-seen capability to control (above the oscillation threshold) both the Q-multiplier bandwidth and the oscillation amplitude independently of one another; normally, in a single-active-device regenerative detector, bandwidth and oscillation amplitude cannot be independently adjusted and are tied together due to the implicit, poorly-characterizable, and scale-varying self-limiting behavior of the active device as it enters class C operation when pushed far above the oscillation threshold.

Tank loading by the transistor

Only minimal attention is given to minimize tank loading by the transistor. The chosen oscillator topology, described later in this article, depends on transistor saturation for its amplitude-limiting behavior, which will (a) reduce tank Q and (b) introduce noise.

Regarding point (b) of noise introduction due to saturation, this essentially does not matter in a regenerative detector intended for HF use because the HF band noise will be stronger than the detector noise (Ref. 4). 

Regarding point (a) of reduction of tank Q due to saturation, this effect may matter in that it broadens the tank's Q-multiplied bandwidth at critical threshold. However, the tank Q is already being reduced due to another factor -- the design choice to use a 1SV149 varactor for the tank capacitance. This varactor, intended for MW use, introduces at HF frequencies noticeable loss into the tank (noticeable in that the regeneration level required to reach threshold is higher than when a variable capacitor is used), which is to be expected as this varactor's Q will be only about 20 at 10 MHz (Ref. 5, Ref.6). Since the tank's Q is already relatively low due to the low varactor Q, the possible improvement in tank Q achievable by reducing transistor loading is probably small in comparison to the varactor losses, though no measurements were made to confirm this assumption. Furthermore, reducing transistor loading would require a more complex biasing scheme, a more complex (tapped) inductor, a higher supply voltage, and/or a separate detector stage (due to poor detection efficiency of the rebiased q-multiplier) -- all of which go against the minimalist nature of this project.

Fidelity of AF amplifier

No attempt was made to design a high-fidelity and low-distortion AF amplifier. Instead, the AF amplifier (which, like the rest of the receiver, runs off of 1.2 volts) was designed simply to have enough gain from low to moderate AF frequencies, with no concern for fidelity.

Background and Related work

V. Polyakov, in his book on homebrew receivers (Ref. 7), presented a very simple regenerative receiver design based on a Hartley oscillator, whose design served as an inspiration for the design of this article's regenerative detector circuit.

The circuit's operation is described in the References 8, 9, and 10. Below we review some of the particular features of this regenerative detector.

Biasing of Polyakov's regenerative detector and AC amplification

Perhaps the most notable aspect of this circuit is its biasing, which allows an extremely low parts count and operation at low voltages. The biasing sets both the collector and the base at the same DC voltage. However, even in this state, the transistor can still amplify AC signals (Ref. 9, Ref. 11).

The following circuit simulation, carried out in the LTspice circuit simulator, verifies the AC amplification of this circuit (note the below circuit uses an NPN transistor in contrast to Polyakov's original PNP transistor). In the below circuit, the regenerative detector has been modeled, but the feedback loop has been broken open at the base. Normally, the base would be connected to the node labeled "feedback," which allows the base to receive its bias voltage through the small feedback coil L2. Instead, in this open-loop simulation, the base is tied directly to Vcc, and a new AC voltage source V2 is inserted at the base to simulate the effect of incoming signals perturbing the base bias. R2 has been added to represent coil loss.

First, we run a frequency domain analysis, also called an AC analysis, to measure the AC response of the amplifier. As explained above, the amplifier input for the simulation is at the base. We measure the amplifier output at the open end where we broke the feedback loop, i.e., at the node named "feedback," after the amplifier has amplified the input signal and sent it through the resonator.  The frequency is swept from 1 MHz to 15 MHz. The amplifier's frequency response curve is plotted for various values of the regeneration control (emitter resistance) from 100 ohms to 50k ohms.

The result of the AC analysis is shown below.

At low emitter resistance we see significant gain of around 30 dB (the green curve). As emitter resistance increases, AC gain of the amplifier decreases. As long as the amplifier's output at the node "feedback" is sufficient to overcome the coil losses, we will have enough gain for oscillation.

For example, the blue curve in the above graph illustrates that with an emitter resistance of 10.1k, peak gain of approximately 13 dB occurs at 4.758 MHz. To double-check the AC amplification properties of the transistor, the following transient (time-domain) analysis was run, exciting the tank with a 1 microvolt signal at the tank's resonant frequency of 4.758 MHz.

The results of the transient analysis appear below. The green trace is the input signal at the base -- a steady 1 microvolt over the entire time of the simulation. The blue trace is the output signal measured at the "feedback" node. As expected, it slowly grows in amplitude over time as the continued rhythmic variations of the input signal pulse in phase with the natural oscillation of charge in the tank at 4.758 MHz. The natural amplification of the tank is assisted by the amplification of the transistor. The final steady-state amplitude of the amplified signal is about 4.5 microvolts, which is 4.5 times the input amplitude of 1 microvolt. And a voltage amplification of 4.5 times equals approximately a 13 dB amplification (voltage gain in dB = 20 × log (V2 / V1)), which is the same result we obtained from the AC analysis. 

It is also possible to conduct more detailed analyses in LTspice such as estimating the actual value of the regeneration control (emitter resistance) required to reach the oscillation threshold, estimating the smoothness of regeneration control, or evaluating the detection efficiency of the regenerative stage. Future articles can explore these topics if there is interest.

Note that the above simulations omitted any magnetic coupling between the two coils L1 and L2. In Polyakov's original design the two coils L1 and L2 are formed by a single tapped coil, implying magnetic coupling between L1 and L2. However, since this is a Hartley oscillator, no magnetic coupling between L1 and L2 is needed. This seems surprising or even unbelievable to some experimenters, for which reason I prepared the following video demonstrating the above oscillator circuit oscillating with two separate, non-magnetically-coupled coils. In any event, the presence or absence of magnetic coupling does not affect the validity of the above simulations that illustrate the AC amplification capability of the Polyakov regenerative detector.

Video illustrating non-necessity of magnetic coupling in Polyakov-style Hartley oscillator.

Biasing of Polyakov's regenerative detector and Vce stability

An additional property of the biasing of Polyakov's circuit is voltage stabilization (Ref. 12).  The voltage across the transistor is relatively stable even as the regeneration control (emitter resistance) is adjusted. This voltage stability means that the transistor's parasitic capacitances are also mostly stabilized, meaning that there is little frequency shift when adjusting the regeneration control around threshold.

The following circuit simulation illustrates the stability of the collector-emitter voltage Vce as the regeneration control is adjusted. Below is a model of the Polyakov regenerative detector at DC. At DC, the base and collector are tied to Vcc, since the DC impedance of the tank inductor is 0 at DC. Furthermore, at DC, all of the capacitors are open circuits and can be ignored.  

The above simulation parameters are configured to vary the emitter resistance from 1 ohm to 10k ohms. (In the physical circuit described later in this article, it was verified that 10k was a plausible value to reach the oscillation threshold.) Also, the battery voltage is varied from 1.3 volts to 1.0 volts, to represent lower supply voltage as the battery is depleted over time.

For the above simulation parameters, the collector-to-emitter voltage is graphed below. 

Note that even as the regeneration control and the supply voltage vary over reasonable ranges, the collector-to-emitter voltage stays relatively constant around 600 mV. And the base and collector are tied to the supply voltage, remaining constant. So all the transistor voltages are stabilized, leading to the reduced frequency shift when the regeneration control is adjusted around threshold. 

Biasing of Polyakov's regenerative detector and saturation/noise 

One superficially detrimental aspect of Polyakov's detector circuit is that the oscillator depends on driving the transistor into saturation for part of the oscillation cycle. The transistor enters saturation if the base-collector diode becomes forward biased, which then presents a low impedance to the collector (Ref. 13, Ref. 14) and damps the tank, introducing noise and harmonic energy in the process. Reference 15 describes the saturation mechanism of a very similar oscillator.

The saturation process can also be intuitively understood by considering that as oscillation commences, the opposite ends of the tank inductor alternately go positive and negative, and since the opposite ends of the tank are both biased to be at the same Vcc voltage, the oscillating voltage necessarily causes, for half of the oscillation cycle, a forward biasing of the base-collector junction, which is equivalent to placing a forward-biased, low-impedance diode across the high-impedance tank.

The below transient simulation verifies the saturating behavior of the Polyakov detector in the oscillating condition. Current source I1 has been added to the simulation to "kick start" the oscillator by providing a momentary current pulse at the start of the simulation. This allows the oscillator to reach its final steady-state amplitude faster.

The transient simulation results are as follows. The red trace is the collector-to-base voltage, which for half of the cycle dips below 0 volts, indicating the base voltage is higher than the collector -- a saturation condition that lowers the collector impedance and introduces noise into the tank.

In an isolated oscillator this noise might be worth worrying about. But, as stated above in the "Non-constraints" section and in Reference 4, in a regenerative detector, this oscillator noise is insignificant compared to the band noise (from the antenna) that we are injecting into the oscillator anyway. So as long as we can hear the band noise, it's not worth worrying about the oscillator noise introduced by transistor saturation.

A modified Polyakov-style circuit based on a Vackar-style oscillator

For the development of the minimalist regenerative receiver in this article, the above Polyakov design was considered as a starting point. The basic concepts were kept of holding the base and collector at the same DC voltage, and of using emitter resistance to control regeneration. However, in the current design, the tapped coil (or separate coils) required by the Hartley oscillator was replaced by a single, non-tapped coil, and the feedback arrangement was altered to become a Vackar-style oscillator using a pi feedback network to achieve the necessary output-to-input phase shift.

The circuit diagram is presented below, and the following subsections describe the parts of the circuit in detail.

The Vackar-style regenerative detector Q2

The regenerative detector (Q2 in the above circuit diagram) is based on a Vackar oscillator, but differs from a true Vackar oscillator in two ways.

First, the high-impedance end of the tank inductor is fully coupled into the active device (at the collector) instead of using a fractional-coupling, capacitive-divider arrangement as in the true Vackar oscillator.

Second, the capacitor C12 at the low-impedance end of the tank inductor is of higher impedance than is recommended in true Vackar oscillators. C12 has a compromise value chosen experimentally to allow enough feedback over the entire 3-30 MHz range. Ideally, it should be 10x larger than the largest value of the variable capacitance D1, which is ~500 pF for the 1SV149 varactor, meaning C12 should be about 5000 pF to make the top of C12 (base side of L1) much lower impedance than the "hot" end of the tank (top of D1/collector side of L1). The problem with making C12 too large though is that its impedance becomes so small that likely there will not be enough feedback for oscillation (or smooth transition into oscillation) on the high band 9-30 MHz (when L1 is small). On the other hand, the problem with making C12 too small is that it raises the impedance at the base-side of L1, thus introducing an impedance mismatch between the somewhat-high impedance point at the top of C12 and the low-impedance transistor base, leading to tank damping and parasitic capacitance as the base is more tightly coupled into the tank. Another problem with making C12 too small is that the raised impedance more tightly couples the RF amp Q1 into the tank, whereas ideally this coupling should be quite light. So, it's clear that by using only a single C12 value (to cover 3-30 MHz with different values for L1), its value must be a compromise.

Dual function of radio-frequency choke L2: Floating at RF, but grounded at AF

The radio-frequency choke L2, with a value of 3.3mH, is being used above its self-resonant frequency. The datasheet at Ref. 16 shows that the typical self-resonant frequency of a 3.3 mH choke is around 700 kHz, whereas Q2's oscillation frequency will be significantly above this, somewhere from 3 MHz to 30 MHz. This means that the choke's apparent reactance in the circuit is actually capacitive instead of inductive. Fortunately, in this circuit, this doesn't matter because the capacitive reactance of L2 just appears to be an additional small (but frequency-dependent) capacitance in parallel with C12.

To enable the required RF feedback signal to be fed into the base, L2's moderately high (though capacitive) reactance at RF serves to hold the base-end of tank inductor L1 above ground at RF. At the same time, L2's low reactance at AF also serves to hold the base-end of the inductor at ground at AF. This helps prevent 50 Hz hum pickup (from surrounding household wiring) by the tank inductor by shunting such low-frequency signals to ground. For example, a 3.3 mH choke will have a reactance of only about 1 ohm for 50 Hz signals, effectively preventing hum pickup by the base-end of the tank inductor. If, on the other hand, the base-end of the tank inductor L1 were floating at AF, this could make the detector susceptible to hum pickup. Such a situation (floating L1) would occur if Q2's base were biased through high-value resistors instead of an RFC. See Ref. 17 for an example of such a hum-susceptible Vackar-style regenerative detector.

Tuning range and bandswitching

L1 is currently about 25 turns on a T50-6 (yelllow) toroidal core. With this inductor, the set oscillates from about 4-16 MHz, with a smooth transition into regeneration. By replacing L1 with an inductor having only 12 turns on a T50-6 core, it was verified that the circuit oscillated up to 30 MHz and still still behaved well with a smooth transition into regeneration and a clean oscillator signal, as monitored on a nearby receiver.

Though oscillator behavior was tested on the high band of 9-30 MHz, actual reception tests have not yet been performed on the high band with the smaller inductor. The current low-band coil is permanently soldered into the circuit, making bandswitching tedious. In the future this receiver will be modified to allow easier bandswitching of the coil L1, at which point actual reception tests will be done on the high band of 9-30 MHz.

Although it is possible to switch both ends of L1 among different coils for bandswitching, it is also possible to use a single coil with multiple taps, and to to switch only one end of L1 (preferably, the low-impedance base-end) among the multiple taps. When a tap is switched into the circuit, the extra unused turns of the inductor are simply left floating and unused. Experiments with a tapped T50-6 coil confirmed that extra floating turns (when a coil tap rather than the full coil is switched into the circuit) did not have any effect on the operation of the regenerative detector. Such a tapped-coil arrangement allows a single coil to cover multiple bands, which might make for a more compact bandswitching implementation than a scheme using multiple coils.

RF amplifier Q1

Q1 serves as a common-base RF amplifier. The collector load is provided by RFC L2, which conveniently also supplies Q1's collector voltage. The base is effectively tied to Vcc, thus allowing the emitter resistance to set the current through Q1. Emitter current always flows through the full resistance of VR1, which is 100 ohms.

Note that the voltages at the Q1 collector and base are not the full Vcc voltage of 1.2 volts, but are instead intentionally reduced by the voltage drop across R11 (described in the following subsection about the AF amp). Accounting for the voltage drop across R11, the voltages at the Q1 collector and base are approximately 936 mV according to an LTspice analysis.

With 936 mV at the base/collector and an emitter resistance of 100 ohms, an LTspice analysis indicates 2.56 mA of current continuously flowing through Q1. This closely aligns with the value of 2.5 mA recommended in Ref. 18 for a common-base amplifier. However, it should be noted that the circuit in Ref. 18 used a supply voltage of 6 volts or above, whereas the circuit of this article uses only 1.2 volts, meaning that Ref. 18's recommendation of 2.5 mA might further be optimized for the circuit of this article. In particular, Ref. 18 implies that the important value may be not the current through Q1 but rather the voltage across Q1's emitter resistor, which should be large relative to the expected maximum signal strength at the antenna in order to avoid overload. This in turn implies that the Q1's current should be set as high as needed to achieve the required voltage across Q1's emitter resistor.

The wiper on VR1 serves as an RF attenuator and allows a portion of the incoming antenna signal to be shunted to ground.

AF buffer stage Q3

Q3 is an emitter-follower stage that has high AF input impedance and serves to isolate the detector Q2 from the following common-emitter stages Q4-Q7. If Q3 is omitted, and for example C10 is connected directly to C11, then C2 will effectively be in parallel with C11, leading to a quite large capacitance across VR4. This high capacitance can lead to super-regeneration, because the high capacitance allows a high voltage to build up across VR4, with VR4 being unable to discharge the capacitance completely before the next oscillation cycle. This gradual buildup of charge across VR4 eventually raises the emitter voltage so high that the transistor gain is reduced and oscillation is, momentarily, halted. After oscillation halts, VR4 discharges the capacitance, and again oscillation commences, until the charge buildup again halts it.

In practice this periodically-interrupted oscillation manifests itself as an AF oscillation when the regeneration control is adjusted near critical threshold. Interestingly, it is possible to exploit this property to achieve an automatic regeneration control by rectifying the AF oscillation and using it as a control voltage to back off the regeneration (Ref. 19). (An alternative means of achieving automatic regeneration control, operating on the RF signal directly, is described in Ref. 20.)

Preventing the above super-regenerative behavior can also be done by inserting a resistance between C10 and C11 (e.g., Ref. 21). However, sufficient isolation requires a high value of isolating resistance, and this in turn has the undesired side effect of reducing the AF signal voltage available for subsequent amplification. Using the emitter follower Q3 causes no loss of AF signal voltage while still achieving good isolation between the detector Q2 and the following AF stages.

The biasing of Q3, achieved through resistor R10, has not been carefully designed. The quiescent voltage at the Q3 base is probably not optimal, which might limit the dynamic range of Q3. However, the input AF signal levels fed into the Q3 base will be very low anyway, so overload of Q3 is probably unlikely.

AF amplifier Q4-Q7

Q4-Q7 simply form a series of common-emitter stages with collector feedback bias. 

Gain of AF amplifier

The 4 stages should provide approximately 90 dB of gain at audio frequencies from around 200 Hz to 10 kHz, as shown in the below LTspice simulation. Gain drops off sharply after 10 kHz thanks to the 10 nF capacitors tying each transistor's base to ground at RF. Reduced gain at RF helps ensure amplifier stability by preventing amplification of stray RF signals that might make their way into the AF amplifier.

In practice, sufficient AF gain was confirmed by connecting the regenerative detector's AF output to the input of the AF amplifier. Using a piezoelectric crystal earphone connected to the amplifier output, the regenerative detector's threshold noise could be clearly heard as the regeneration control was advanced towards threshold. Being able to clearly hear the regenerative detector's threshold noise indicates in practice that the AF amp has enough gain.

The AF amplifier also can drive low-impedance (32-ohm) consumer-grade headphones. In this case, the resulting audio level is quite low, but listenable. It is also possible to connect the AF output to the microphone-in or line-in of an external amplifier, for increased output power or making a recording of the receiver's audio (as was done for the video shown later in this article).

Stability of AF amplifier

An initial prototype of the AF amplifier, constructed according to the above circuit diagram, showed some tendency for self-oscillation. This could be verified by tying the input of the amplifier to Vcc, which immediately caused an audible AF oscillation at the amplifier output.

The root cause of this instability is the later stages of the AF amplifier being able to ripple the Vcc voltage by a small amount, and in-phase with the input signal. This in-phase ripple travels over the Vcc line back to the input of the amplifier, causing positive feedback and AF oscillation. To remedy this situation requires decoupling the earlier AF stages from the later AF stages to prevent such Vcc ripple from traveling back across the Vcc line. Such decoupling can be achieved with a series resistor and capacitor at the first AF amplifier transistor, essentially forming a separate Vcc supply for the first AF amplifier and the rest of the receiver (Ref. 22).

For the receiver circuit of this article, the AF amp decoupling is achieved by R11 and C14.

Good RF design practice would place an additional RF bypass capacitor across C14 to ensure a sufficiently low-impedance path for RF signals and prevent RF from traveling into the Vcc supply line where it might cause instability. High-value electrolytic capacitors like C14 may not have sufficiently low impedance at RF. Nevertheless, in this circuit, such additional RF bypassing was omitted for simplicity, and C14 by itself seems to function adequately as an RF bypass.

Physical construction

Construction began with the low-voltage AF amp. First, the AF amp was prototyped on a solderless breadboard.

Then, a compact version of the AF amp was built. The collector feedback bias resistor and RF bypass capacitor were soldered directly to each transistor.

The four transistors were then soldered together to form the four-stage amplifier. An LM386 chip is shown as a size comparison.

The AF amp could be soldered into and out of the circuit as an integrated unit. In the below photograph, the prototype receiver's detector stage is constructed on a copper ground plane, while the AF amp is soldered as an integrated unit onto the detector.

After the detector design (including a preceding RF amp and a following AF buffer) was finalized, the additional transistors were soldered directly to the AF amp to make for a maximally-compact build. In the photograph below, the left three transistors are the RF amp, regenerative detector, and AF buffer, respectively. The right-most transistors are the AF amplifier.

With the basic receiver components in place, the tank inductor L1 and the controlling potentiometers were soldered into the circuit, and the whole circuit was placed inside a small plastic box. Ten-turn potentiometers, with turns-counting dials, were used for the tuning potentiometers VR2 and VR3. The turns-counting dials allow a reasonable degree of repeatability in tuning. A ten-turn potentiometer is also used to control the regeneration, allowing a high resolution of adjustment.

The final receiver looks as shown in the following two photographs.

Performance

The following video shows the receiver's performance when receiving CW, SSB, and AM signals. Reception starts around 7 MHz and the receiver is tuned progressively higher in frequency.

Discussion

Antenna and reverse isolation of Q1

The above video was created with the receiver connected to an M0AYF-style active, broadband loop antenna (Ref. 23). No hand capacitance effects were observed when connecting/disconnecting the antenna, when moving the antenna's coaxial cable, or when adjusting the antenna attenuator VR1.

It should be noted that the M0AYF-style active antenna requires a 12-volt supply and consumes 30 mA of current, much more than the receiver itself. In keeping with the low-power theme of this receiver, additional reception experiments were conducted with passive antennas.

The first passive antenna tried was a small, tuned loop antenna having approximately ~0.8m diameter. The small loop antenna was remotely tuned with a 1SV149 varactor, which consumes almost no current (only microamps). RF signal signal pickup was taken from a smaller coupling loop located inside the larger loop. The smaller coupling loop was connected to the receiver, at the input of the common-base amp, through 50-ohm coaxial cable. Signal levels with the tuned loop antenna were excellent; better, in fact, than with the active loop antenna, as the active loop antenna tended to provide too much signal to the detector, requiring maximum attenuation all the time. Furthermore, tuning of the tuned loop antenna had no pulling effect on the frequency of the regenerative detector, even when the antenna was tuned through the resonant frequency of the detector. This indicates that Q1 provides quite good reverse isolation between the detector Q2 and the antenna.

Similarly, when attaching a random wire antenna of approximately 3m length, no pulling of the detector frequency is noticeable when connecting, positioning, or adjusting the attenuation of the antenna. Signal levels with the random wire antenna were noticeably lower than with the passive tuned loop or the active broadband loop. This is to be expected as a short wire antenna will have a high impedance at HF, whereas the RF amplifier Q1 has a low input impedance, leading to signal loss due to the impedance mismatch.

Dual voltage supply

Main receiver power is supplied by the 1.2-volt rechargeable cell V1. Varactor reverse bias, however, requires a higher voltage 9-volt supply. This is provided by a stack of 3 3-volt CR2016 button cells, with a 100k pot wired across the 3 cells as a voltage divider.

Current draw from the 9-volt supply is only the current drawn by the 100k pot and the reverse-biased varactor, which is around 90 microamps total, meaning the 9-volt supply only rarely needs to be changed.

If a traditional tuning capacitor were used instead of the varactor D1, then the separate 9-volt supply would be unnecessary.

Lack of a ground plane

Although good RF practice recommends the use of a ground plane for stability, no instability was noticed in operation thus far. However, stray capacitances may degrade receiver performance at higher frequencies as we approach upper HF. Upper HF (nearing 30 MHz) may require a ground plane and/or more careful attention to layout of the wiring.

Partial equalization of threshold regeneration level

The regenerative stage Q2 is a Vackar-style oscillator using a pi feedback network. I had discovered in an earlier receiver that this topology requires reduced regeneration adjustment as it is tuned, due to increasing Vackar feedback at low frequencies being counteracted by increasing varactor losses at low frequencies (Ref. 24, Ref. 25).

In the new circuit of this article, I was again able to verify the somewhat-equalized regeneration level. With the current L1 value, tuning from approximately 4 to 16 MHz, the required regeneration to reach the oscillation threshold starts out somewhat high (i.e., emitter resistance is low) at the lowest frequencies, where capacitance is highest and the varactor losses are highest. A low emitter resistance allows high transistor gain, enough to overcome the high varactor losses.

As the set is tuned higher in frequency, varactor losses slowly decrease, requiring progressively less regeneration, until approximately the middle of the tuning range is reached, where the minimum regeneration is required (i.e. the emitter resistance is at its highest).

Then, as the set is tuned even higher, progressively more regeneration is required (requiring the emitter resistance to be decreased), as the varactor losses start to become insignificant; the significant feedback-determining factor then becomes the reactance of C12, which decreases as frequency increases, requiring more feedback to compensate.

The above results were observed when tuning at low-to-mid frequencies from 4 to 16 MHz. Similar results were observed even when L1 was reduced to cover 9 to 30 MHz. Again, required regeneration started out high, reduced to a minimum somewhere around mid-band, and finally again required more regeneration as the set was tuned towards the upper-most end of the band.

Biasing scheme and emitter resistance

With the Polyakov-style biasing scheme, the detector Q2's base and collector are essentially tied to Vcc at DC. This means the emitter resistance determines the current flowing through the transistor. In the final receiver circuit, the wiper value of VR4 determines the emitter resistance, and this resistance can be reduced to zero if the VR4 wiper is thusly adjusted.

An emitter resistance of zero will allow theoretically unlimited current to flow through the transistor. Fortunately, in this circuit, decoupling resistor R11 limits the maximum current draw from the battery to 12 mA (1.2 volts / 100 ohms).

If the Q2 detector is used with a different voltage supplying scheme that does not limit the maximum current, then some means should be provided to prevent excessive current from flowing through Q2 (an additional resistor in series with either Q2's collector or emitter).

Frequency pulling at higher frequencies

At higher frequencies above 12 MHz, the detector seems somewhat noisier than at lower frequencies. The suspected cause is the increased tendency for detector self-modulation as the tank capacitance is decreased. As tank capacitance decreases, any small change in the parasitic capacitances of the detector transistor Q2 will have an increasingly large influence on the tank's resonant frequency, which can manifest itself as a noisier oscillation. The easiest way to address this problem is simply to increase the tank capacitance by adding a large capacitance (say, 500 to 1000 pF) in parallel with the tuning varactor D1, making the tank more resistant to frequency pulling (Ref. 29) and effectively swamping out small-scale parasitic capacitance variations with the larger tank tuning capacitance (Ref. 30). Implementing this change will likely require adjustment of the value of the Vackar feedback capacitor C12. Also, the tuning range will decrease if a large capacitance is placed in parallel with D1.

General approach: Why not remote control?

Readers may be familiar with the author's investigation into and implementation of a remote-contolled regenerative receiver (Ref. 26). However, the current article implements a more traditional receiver scheme where the antenna is located remotely and all of the receiver circuitry is located locally. Given the potential advantages of a remote-controlled design (Ref. 26), why was remote control omitted from the current design?

Two main difficulties were encountered with remote-controlled designs that led to the more traditional design of this article.

The first problem is the need for variable RF attenuation. Due to the limited dynamic range of a regenerative detector, RF attenuation is important if high signal levels are expected (such as when receiving night-time shortwave broadcasts using a moderately-sized antenna). Ideally an RF attenuator should be a linear component (such as a resistor) and not a non-linear component (like a variable-gain amplifier). However, remote control makes it difficult to implement a remotely-located variable resistance. A PIN diode network with remotely-controlled bias (e.g. Ref. 27) might be one option, but was considered too complex for this receiver. On the other hand, it is probably relatively easy to implement a remotely-controlled variable-gain RF amplifier (whose base bias voltage is varied over a long control cable), but such a non-linear approach to attenuation could lead to intermodulation distortion as the gain is reduced and current through the amplifier is reduced.

The second problem concerns the use of a remotely-located whip antenna. To avoid the need for an RF attenuator, one option might be simply to keep signal levels low by using a short whip antenna. The short whip antenna will have a high impedance and could be connected directly to the top of the high-impedance tank. The problem with this approach is that the whip antenna's impedance depends on its surroundings, as its high impedance makes it susceptible to capacitively couple with its environment. In practice, what happens with a short whip antenna and remote control is that there is significant hand capacitance when one's hands are brought near the remote-control cable, even though the cable is grounded (to the circuit's ground plane) and should be carrying only DC. In other words, the long DC control cable actually is not at RF ground potential, and the operator's hands will affect the RF impedance of the DC control cable. Such changes in RF impedance are "visible" to the remotely-located high-impedance whip, which leads to undesirable hand capacitance and frequency pulling when adjusting the remote-control potentiometers. A large, external ground plane (such as a large copper sheet) located at the remotely-sited receiver might help hand with such remote-control hand-capacitance problems, by isolating the whip antenna from the varying impedance at the DC control cable.

Update 2017/05: Experiments with a ferrite rod antenna

To decrease the size and increase the portability of the receiver, I experimented with using a ferrite rod antenna to replace the tank inductor L1. In this case, no external antenna is used. I rebuilt essentially the same receiver circuit as presented earlier, except omitting the RF amplifier Q1 and the AF buffer Q3. Instead of L1, a small ferrite rod antenna of type SL-45GT -- designed for AM broadcast-band frequencies -- was used. A few turns of wire were wound on the ferrite rod and secured in place with double-sided tape. The few turns were confirmed to resonate the tank over a range of about 6 to 18 MHz. In the early evening, when signal levels are strong, the receiver works surprisingly well even with no external antenna.

I made some low-quality videos to give a general idea of the signal levels that were received when using the ferrite rod only and no external antenna. The videos are linked below.

https://www.youtube.com/watch?v=KyAlpqw-fGM

https://www.youtube.com/watch?v=eENrSUPxg1M

This version of the receiver, with the ferrite rod, lacks sufficient feedback to oscillate above about 14 MHz. This is likely due to two reasons: (1) increased ferrite losses at higher frequencies and (2) inherently reduced feedback at higher frequencies due to the Vackar oscillator topology. I experimented with increasing the feedback at higher frequencies by adding Colpitts-style feedback via a 4 pF capacitor between the collector and emitter, creating an additional collector-to-emitter capacitive feedback path in the style of typical common-base oscillators. This new Colpitts-style feedback network will have increased feedback at higher frequencies. Nevertheless, even the addition of the Colpitts-style feedback did not give enough feedback for oscillation above 14 MHz. More adjustment of the feedback network is needed. (References on feedback behaviour to be added later.)

In light of the low signal levels available from a ferrite rod antenna, it might be worth reconsidering the oscillator topology and the noise introduced by allowing the transistor to saturate during oscillation. A more conventionally-biased oscillator (with e.g. voltage divider resistors on the base) might be less noisy and allow weaker signals from the ferrite antenna to be heard more clearly.

Conclusion and future work

A low-voltage regenerative receiver was presented using varactor tuning. The receiver can operate from 3-30 MHz by changing only the tank inductor. Regeneration control is smooth and hysteresis-free over the entire frequency range. Effective antenna reverse-isolation is provided by an RF amplifier, eliminating any hand capacitance issues. The physical circuit construction is extremely compact, and the finished receiver including the housing and control knobs is small enough to fit inside a large pocket.

Possible improvements might include the following.

Tank loading could be reduced by winding an additional, lower-turn-count winding on L1 and tying the collector to Vcc through this winding instead of directly through L1. This was tried in an earlier receiver build (Ref. 24). The disadvantages are a more complex inductor and reduced detection efficiency (lower AF output). The latter disadvantage could be addressed by a separate detector stage, but this introduces additional complexity and subtle difficulties when running off of 1.2 volts (such as relatively low BJT detector input impedance at higher HF, which again will load down the tank and might negate any reduced-loading benefit of the low-impedance collector winding).

The AF amplifier might be more carefully designed to ensure a specified dynamic range and frequency response. Currently, the amplifier's low-frequency response starts to fall off sharply below about 300 Hz, because of the 100 nF coupling capacitors used; frequency response below 300 Hz could be improved by increasing the value of the AF coupling capacitors. Also, additional AF filter stages, such as low pass or band pass filters, might be added to aid reception of different modes such as CW or SSB.

There is currently no provision for volume control of the AF signal at the output of the AF amplifier. The RF attenuator can serve as a sort of volume control.

An RF AGC approach might be attempted (e.g., Ref. 28), to reduce the need to manually adjust the RF attenuator. However, as mentioned earlier, the use of a gain-controlled amplifier to reduce the input RF signal is prone to intermodulation distortion.

With an RF AGC stage implemented, the receiver might be used as a fixed-frequency, AGC-capable IF strip for a superheterodyne receiver.

Appendix

In the future this article may be expanded to include some of the following topics.

• Failed circuits: Various circuits that were tried but discarded as part of the design process for the final receiver circuit.
• RF stage gain: LTspice simulation showing RF amplification of Q1.
• RF stage reverse isolation: LTspice simulation showing reverse isolation of Q1.
• Measured data on the threshold regeneration level vs. frequency.
• Performance of the receiver at MW frequencies (500-1700 kHz).

References

1. Ivanenko, V. "Regenerative Receiver #4." http://qrp-popcorn.blogspot.jp/2015/02/regenerative-receiver-4.html. Relevant paragraphs: "It features a Colpitt's oscillator AC coupled to the tank for the Q multiplier plus a separate, low current FET detector. [...] The hycas detector hikes reverse isolation plus lifts the output AF voltage and impedance."
2. N., Vlad. "6AS6 regen, mk2." http://theradioboard.com/rb/viewtopic.php?p=45323#p45323. Relevant paragraph: "Any regeneration receiver has an inherent oscillation amplitude stabilization mechanism. The 6AS6 detector was interesting in two ways: (i) it has more gain in the amplitude control loop (AAC) compared to classic detectors and (ii) the loop is external and explicit, while the regeneration amplifer section operates in a linear class A mode. This made it an ideal experimental testbed for learning the effects of AAC parameters on the oscillator/regen behavior. The results were interesting and not obvious (at least to me) and are listed in the theory thread here: http://theradioboard.com/rb/viewtopic.php?f=1&t=4680 . In short, the most important practical outcome of this study is that: 1. Using a scale-invariant control law in the AAC loop eliminates the bandwidth expansion with the increase in the oscillation amplitude. 2. Narrowing the AAC loop bandwidth with a variable RC filter allows for a wide range Q-factor control above the oscillation threshold and a full recovery of the Q-factor back to the level available at the threshold."
3. Gaijin, Q. "Designing class-A solid-state regen w/explicit AAC loop." http://theradioboard.com/rb/viewtopic.php?f=4&t=5171
4. Newkirk. D. "Re: BJT oscillator saturation and phase noise." http://theradioboard.com/rb/viewtopic.php?p=53328#p53328. Relevant paragraph: "every practical regenerative detector operates with external signals and noise intentionally injected into its feedback loop. That being true, antenna-system noise will always override whatever relatively weak phase and amplitude noise may be present in a regenerative detector capable of practically useful reception."
5. Toshiba. "Toshiba variable capacitance diode / Silicon epitaxial planar type / 1SV149 / AM radio band tuning applictions." http://www.qsl.net/df7tv/datasheets/1SV149.pdf
6. Skyworks. "Application note / Varactor diodes." http://www.skyworksinc.com/uploads/documents/200824A.pdf. Relevant paragraph: "However, one may extrapolate Q to a different frequency simply by using the reciprocal relationship."
7. Polyakov, V. Техника радиоприёма Простые приёмники АМ сигналов (Russian language). Retrieved from http://www.smail.lt/~ncss/regen_info/VP01.GIF (machine translatable version at http://amfan.ru/).
8. ChipDipvideo. "How to Make Regeneration Cascade Circuit with Common Base." https://www.youtube.com/watch?v=ObETk7TN_QQ
9. Polyakov, V. Техника радиоприёма Простые приёмники АМ сигналов, section titled
   "Простой регенератор" (Russian language). Retrieved from http://www.smail.lt/~ncss/regen_info/VP02.GIF (machine translatable version at http://amfan.ru/avtodiny/prostoj-regenerator/).
10. Polyakov, V. Техника радиоприёма Простые приёмники АМ сигналов, section titled "Практическая схема" (Russian language). Retrieved from http://www.smail.lt/~ncss/regen_info/VP03.GIF (machine translatable version at http://amfan.ru/avtodiny/prakticheskaya-sxema/).
11. Wehner, C. "The short-circuit amplifying technique." http://swissenschaft.ch/tesla/content/T_Library/L_Theory/EM%20Field%20Research/Short-Circuit%20Amplifying%20Technique.pdf
12. Polyakov, V. "Super Regenerator." Retrieved from http://zpostbox.ru/super_regenerator.html. Relevant paragraph: "The regeneration control does not affect much on the frequency of the resonant tank L1C1, because the voltage across transistor VT1 remains the same (about 0.5 volts), so the inter-junction capacitances are almost not changed."
13. Rodwell, M. "Amplitude prediction." http://www.ece.ucsb.edu/Faculty/rodwell/Classes/ece218b/notes/Oscillators2.pdf. Relevant paragraph: "We can look at Vcb or Vdg to verify the limiting condition. In the case of the bipolar transistor, if Vcb = 0 or negative, the base collector junction is forward biased – a low impedance at the collector will be experienced."
14. Rodwell, M. "Oscillators." http://www.ece.ucsb.edu/Faculty/rodwell/Classes/ece218b/notes/Oscillators1.pdf
15. Pinkham, C. "Amplitude limited varactor tuned L-C oscillator." http://www.google.com/patents/US4287489
16. Bourns. "RL622 Series - Radial Lead RF Choke." http://www.bourns.com/data/global/pdfs/rl622_series.pdf
17. DrM. " Re: 1.5-volt regens." http://theradioboard.com/rb/viewtopic.php?p=46345#p46345
18. Kitchin, C. "High Performance Regenerative Receiver Design." http://www.arrl.org/files/file/Technology/tis/info/pdf/9811qex026.pdf. Relevant paragraph: "By experimentation I found that 2.5 mA is a good operating current for Q1. 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 Q1 to detect the signal.)"
19. Cool386. "Automatic Regeneration Control for Regenerative Receivers." http://members.iinet.net.au/~cool386/arc/arc.html
20. Kovalenko, S. "A Shortwave receiver with automatic regeneration control." http://zpostbox.ru/a_shortwave_receiver_with_automatic_regeneration_control.html
21. Kitchin, C. "A short-wave regenerative receiver project." http://www.electronics-tutorials.com/receivers/regen-radio-receiver.htm. Relevant paragraph: "The audio output is extracted from the JFET source and travels through resistor R4 to the audio filters. R4 isolates C10 and C12 from R3 and C8 in the detector's source; otherwise, the detector may break into super regeneration. This can occur with high levels of RF feedback when a long RC time constant is used in the detector circuit. A large increase in either R2 or C4 would produce the same effect."
22. Mitchell, C. The Transistor Amplifier, section titled "Squealing, buzzing, oscillating, and motor-boating." http://www.talkingelectronics.com/projects/TheTransistorAmplifier/TheTransistorAmplifier-P2.html#Motor-boating
23. K., Des. "A Wide Bandwidth Active Loop Receiving Antenna." http://www.qsl.net/m0ayf/active-loop-receiving-antenna.html
24. Gaijin, Q. "Re: Varactor-tuned, hybrid-feedback, low-voltage BJT regen." http://theradioboard.com/rb/viewtopic.php?p=45649#p45649
25. Gaijin, Q. "Tilt-balanced hybrid-feedback varactor-tuned regens." http://theradioboard.com/rb/viewtopic.php?t=5870
26. Gaijin, Q. "Low-complexity and low-power regenerative receiver with remote control of tuning and regeneration for use in electromagnetically noisy urban enviroments." http://qrp-gaijin.blogspot.jp/2014/02/low-complexity-and-low-power.html
27. Avago Technologies. "A Low-Cost Surface Mount PIN Diode π Attenuator / Application Note 1048." http://www.avagotech.com/docs/5966-0449E
28. Andersen, Rick. "The AGC-80 / A Regenerative Receiver with AGC for 80 Meters or 3-10 MHz Shortwave." http://www.ke3ij.com/AGC-80.htm
29. Newkirk, D. "High-Capacitance (High-C) Regenerative Detectors." http://theradioboard.com/rb/viewtopic.php?t=446
30. Rainey. M. "Cross-Coupled Receiver (excerpt from a letter to CO2KK)." https://groups.yahoo.com/neo/groups/regenrx/conversations/messages/5324

Building a 7 MHz small transmitting loop antenna, part 5

Part 5 of this series describes some mechanical design changes and first disappointing tests with the loop.

First design change: From copper pipe to copper strap

As described in parts 1 and 2 of this series, the original design for the 3m x 2m loop radiator was to construct it using several 1m-length segments of 6mm-diameter thin-wall copper pipe. However, such a structure, once built, is impossible to move or alter without tedious disassembly.

Since I wasn't completely confident that the loop would work as expected, I decided to go with a more mechanically flexible solution using copper strap instead of copper pipe. The flexible strap cannot support itself and thus requires a supporting frame to be built, on which the strap is draped. I built a supporting PVC frame, 3 meters by 2 meters, as follows (apologies for the blurry photograph):

The bottom half of the support frame is obscured by lower balcony wall, which has been cropped from the bottom of the photograph.

The above photograph shows a top supporting bar spanning the three vertical support pipes.  In the end I ended up removing this top supporting bar because its weight tended to make the whole structure sway excessively in the wind. Removing the top supporting bar means that the uppermost horizontal run of the copper strap is supported only at the tips of the three vertical support pipes. The strap then runs down the two end-most vertical support pipes and is temporarily affixed to the frame with velcro cable ties.

The above setup allows repeated and easy setup and tear down of the antenna as needed for various changes, especially around the capacitor section. As it turns out, this ability to easily tear down the antenna turned out to be vitally important, as the antenna didn't work as expected and required major changes.

Cutting and splicing the copper strap

I used copper flashing with 0.2mm thickness as the loop radiator. The skin depth of copper at 7 MHz is 24.605 micrometers, so the flashing with 200 micrometers thickness is greater than 8 skin depths thick for RF currents at 7 MHz.

The flashing is available in a single sheet of dimensions 365mm x 600mm. I decided to cut this large sheet into 18 individual strips of 20mm width, then to solder all of these strips together to form a ~10m-long radiator.

According to W9CF's strip-to-round-conductor equivalency calculator at http://fermi.la.asu.edu/w9cf/equiv/, a 20mm-wide strip should have RF resistance equivalent to that of a ~6.2mm-diameter conductor, i.e., equivalently low resistance to the 6mm copper pipes I originally planned on using. Note that the calculator assumes that the strip "is much thicker than a skin depth", which is satisfied by our ~8-skin-depths-thick copper flashing.

The following photographs show the time-consuming process to cut and splice the copper flashing by hand. This took a few hours, and my fingers ached for a few days afterwards; cutting this thickness of copper flashing with small shears is tiring on the fingers. 

After joining all the strips into a ~10m roll, the next step was to wrap the copper strap in vinyl electrical tape, to avoid risk of injury on the sharp copper edges when handling the strap.

The last step was to solder the copper strap to the butterfly capacitor and to mount the capacitor in the box.

Failure #1: Not enough capacitance

I draped the loop over the PVC support frame with the remotely-controlled capacitor at the top center of the support frame. I then created a ~2m-circumference auxiliary coupling loop and placed it at the bottom of the large loop radiator.

Unfortunately, my worst fear materialised: the butterfly capacitor did not have enough capacitance to resonate the loop at 7 MHz, the intended frequency of operation. I could resonate it from about 9165 kHz to 15270 kHz, listening for the noise peak in my receiver.

I did a quick tune up test on 10 MHz and 14 MHz and was able to get the SWR down to better than 2:1, but not much better.

To allow operation at 7 MHz, I investigated soldering another smaller butterfly capacitor inside the box, in parallel with the existing butterfly capacitor, in order to reduce the resonant frequency. However, this plan was mechanically infeasible.

As a stopgap measure to allow operation at 7 MHz I tried attaching an additional 150 pF variable capacitor (with a wiping contact, which is not recommended due to ohmic losses) to the butterfly capacitor using alligator clips (again, not recommended due to ohmic losses), and was able to resonate the loop at 7 MHz with the 150 pF variable capacitor almost completely unmeshed. However, I was unable to obtain a low SWR at 7 MHz.

My inability to reduce the SWR to 1:1 on any band prevented me from measuring the SWR bandwidth of the antenna, which might have been an indication of the antenna's efficiency. I did various tests squashing and stretching the coupling loop but was never able to obtain a good SWR. This is one disadvantage of using a coupling loop: it can be difficult to adjust the coupling loop for a good SWR, especially if the loop losses are excessive and unknown due to environmental or other ohmic losses.

Second design change: Capacitive instead of inductive matching 

One alternative to the difficult-to-adjust inductive coupling loop is to use a capacitive network to match the loop impedance to 50 ohms. There are a number of schemes for doing this, and I chose the scheme shown at http://homepage.ntlworld.com/g4kki.william/army_loop_tuner.htm, which resonates the loop with a split-stator capacitor, and feeds the loop through a matching capacitor into one side of the loop conductor (with the coax shield going to the neutral rotor of the split-stator capacitor). Note the above link refers to this as an "Army loop tuner", though it is not the same capacitive network used by the original "Army loop". G8ODE describes the commonly-seen capacitive matching schemes here: https://rsars.files.wordpress.com/2013/01/qrp-loop-tuner-80-20m-g8ode-iss-1-32.pdf.

I built a prototype capacitive matching network on a polypropylene cutting board, and soldered it to my loop conductor. For the loop resonating capacitor, because the 50pF butterfly capacitor had insufficient capacity, I used a dual-gang 2x440pF variable capacitor in a split-stator configuration.

The loop tuning and matching circuit construction looks as follows:

Note that the capacitive matching scheme is not balanced, and so there is a danger of common-mode currents on the outside of the shield of the coaxial cable. To suppress these, I wound a ferrite-cored common-mode choke on a FT250-43 core, using 12 turns of RG58A/U. According to G3TXQ's page at http://www.karinya.net/g3txq/chokes/, this choke should exhibit >4k ohms impedance at 7 MHz.

Although it is preferable to mount the capacitor at the top of the loop (to keep it and the intense electric field away from lossy surroundings), the heaviness of the new capacitor assembly and the choke precluded top-mounting with my somewhat fragile PVC support frame. Therefore, I mounted the capacitor, feedline choke, and feedline at the bottom of the loop, which was still about 20cm above the concrete balcony floor.

Partial success: Perfect 1:1 SWR achieved

After hooking up the loop to my rig, I tuned the split-stator capacitor and quickly found the capacitor setting (approximately 25% meshed) that would peak the receiver noise at 7 MHz. Then, I experimented with different settings of the 150pF matching capacitor. It turned out that setting the matching capacitor to fully-meshed (150pF) gave a perfect 1:1 SWR.

This was my first experience with a capacitively-matched loop, and it was quite easy to tune the loop for minimum SWR. Adjusting the matching capacitor is much easier in practice than adjusting the size of an inductive coupling loop.

There is some interaction between the tuning and matching capacitors, due to the matching scheme requiring that the loop be tuned slightly off-frequency to provide the inductive reactance needed to effect the impedance transformation; see here for details: http://www.eham.net/ehamforum/smf/index.php/topic,100081.msg802902.html#msg802902. In practice, this interaction really isn't a problem. You just tune the main capacitor for maximum noise, adjust the matching capacitor to some approximate appropriate value, and retune the main capacitor to again peak the noise.

Failure #2: Ridiculously high bandwidth

My joy at attaining a perfect 1:1 SWR was short-lived, as I quickly discovered that the SWR was still 1:1 even after retuning several tens of kHz. In fact, the SWR was less than 1.5:1 over the entire 200 kHz of the Japanese 40m band!

This is quite a bad sign indeed. A previous vertical dipole I built for 7 MHz (2m long) had a 1.5:1 SWR bandwidth of about 40 kHz, and that was broader than the 4nec2 prediction of 20 kHz. A small transmitting loop should have even narrower bandwidth. Therefore, a 1.5:1 SWR bandwidth of 200 kHz is ridiculously high and indicates significant losses within the system.

I can think of a number of possibilities to explain the loss:

1. My split-stator capacitor is too lossy. 
2. The matching capacitor is too lossy.
3. The copper strap is inherently too lossy and/or the electrical tape is increasing the losses.
4. The environment around the loop is too lossy, with the loop running close to the concrete walls.
5. The environment around the capacitor is too lossy, with the loop capacitor near the concrete floor.
6. One or more solder joints is faulty. The solder joints include those that connect the short copper strap segments together, those that connect the strap to the capacitor, and those around the matching capacitor.

Possibility 1 I think is unlikely, as I used the split stator capacitor in a previous small transmitting loop that exhibited a narrow bandwidth. Possibility 2 also seems unlikely because the matching capacitor is not part of the resonant circuit and isn't carrying that much current; its losses should not adversely affect the loop efficiency. And, possibility 3 also seems unlikely because I am inclined to believe W9CF's calculations on the RF resistance of strip conductors.

I can think of the following ways to proceed:

• To attack possibilities 4 and 5, a smaller-diameter loop (reducing radiation resistance) with a wider conductor (reducing loss resistance) could be used. That could allow comparable efficiency to the larger loop while allowing the loop to be positioned farther away from the concrete walls.
• Alternatively, and still for possibilities 4 and 5, an experiment could be done erecting the loop in an open area away from concrete walls, and observing if the SWR bandwidth decreases. If so, then the environment is a major source of loss. An experiment with top-mounting of the capacitor might also yield insight.
• To address possibility 6, I could re-do all the solder joints. This will be quite tedious on the shorter copper strap segments, as these have all now been wrapped in and sealed with electrical tape.
• Alternatively, to address possibilities 3 and 6, an experiment could be done with a continuous conductor instead of my pieced-together copper strap. The continuous conductor could be a wire, a coaxial cable shield, or a long strip of unbroken copper tape. None of these is an ideal conductor for a low-loss small transmitting loop, but using one or more of these continuous conductors with the existing capacitive loop tuner might reveal if the pieced-together and tape-insulated copper strap is excessively lossy in comparison.
• To address possibility 2, the matching capacitor - currently an air variable capacitor with a wiping contact - might be replaced with the high-Q, no-wiper, butterfly capacitor that proved to be unsuitable as the resonating capacitor.
• Finally to address possibility 1, a vacuum variable capacitor could be used as the loop resonating capacitance. This would require two matching capacitors rather than just one. And, again, the weight of the vacuum variable capacitor would preclude top-mounting on my current support frame, requiring a bottom-mounted capacitor.

Most likely I will proceed next by re-doing the easily re-doable solder joints. If that does not improve (reduce) the bandwidth, then I will investigate using a continuous conductor.

My current hope is to achieve a narrow-bandwidth with a capacitively-matched loop and bottom-mounting of the capacitor. If I can achieve narrow bandwidth (and reasonable efficiency), then I will motorise the variable capacitors for remote control of tuning and matching.

Building a 7 MHz small transmitting loop antenna, part 4

Part 4 of this article series focuses on the PWM circuit for precise and slow control of the motor that drives the capacitor.

Attempt 1: Analog-to-analog motor control circuit 

Before discussing the final PWM circuit I used to control my motor, allow me to discuss another highly interesting circuit: the 7N3WVM "Remote Motor Controller" described here: http://www.qsl.net/7n3wvm/DC_mot_control.html. Please have a look at that page before proceeding.

The 7N3WVM circuit is attractive because it seems to offer an intuitive interface to control the remote motor: a locally-operated DC motor that is turned to generate an analog DC control voltage, which is then amplified by a power amplifier and sent to the geared motor. If clockwise rotation of the control motor generates a positive voltage, then counter-clockwise rotation generates a negative control voltage. Either polarity is properly amplified by the power amplifier and is sent along to the remote motor, allowing bidirectional control by turning the shaft of the local control motor left or right. What's more, faster rotation of the local control motor generates a higher voltage, leading to a higher voltage at the remote motor and faster turning of the remote motor.

The system sounds ideal, and initial tests were quite encouraging. I actually wired up a previous loop antenna using this motor control method. However, there is unfortunately a subtle but crippling flaw in this system: the motion of the remote motor is unreliable (more on that in a moment), so when turning the local control motor shaft indoors, you never know if your local knob rotation had no effect, some small effect, or a too-large effect at the remote motor. So you're operating essentially "blind", randomly turning the local control knob left and right, fast and slow, but never really knowing if the remote capacitor is moving at all, moving slowly, or moving too fast.

So what causes this unreliability in the motion? The problem is that rotating the local control motor's shaft generates a varying analog DC voltage depending on how quickly and how smoothly you rotate the shaft. For very slow rotations of the control knob, the generated voltage, after amplification, is too small to turn the remote motor. So nothing happens, and worse yet, you have no way of knowing that nothing happened (unless you implement some means of sensing the remote motor's current angular setting, via e.g. a potentiometer as mentioned in Part 3).

Then, if your steady hand rotates the control knob at just the right speed, then the remote motor will rotate slowly and smoothly. However, if you rotate the control knob too quickly, the local voltage jumps very high, leading to a very high amplified voltage at the remote motor, leading to a quick and jerky rotation of the remote motor. And again, with no feedback at the local operating position, you have no way of knowing when this happens.

I experimented with the supply voltage of the power amplifier; 7N3WVM recommends 3 volts. But I found that 15 volts worked best (note that the NJM2073D chip is rated for a maximum of 15 volts supply voltage), using a 3V motor for the local control motor, and a 3V geared motor (with 661.2:1 reduction) for the remotely-controlled motor. With 15 volts supply voltage for the NJM2073D power amplifier chip, even relatively small voltages, caused by slow rotation of the local control knob, were amplified enough to cause rotation of the remote motor. But even in this best-case scenario, I still could not completely solve the issue of unreliable, jerky rotation. Sometimes, the remote motor would not rotate at all; sometimes, it would rotate too fast.

I think it is very important that any remote control scheme work reliably: if you take some action locally (push a button, flip a switch, turn a knob), then it is imperative to know that something well-defined definitely did happen at the remotely-controlled device. As elegant as 7N3WVM's scheme is, it unfortunately fails this test, at least in my setup with my choices of motors.

I think the 7N3WVM scheme might be made to work well if some of the following ideas are applied:

• Use a higher-voltage motor for the local control motor, to ensure that even slow rotations of the local control knob generate relatively high DC voltages. This should help ensure that slow rotations more reliably generate enough voltage to drive the motor at the remote end.
• Use a higher gear reduction ratio on the geared motor at the remote end. My motor used a 661.2:1 reduction ratio; a further reduction ratio would physically limit the maximum rotational speed of the remote motor's output shaft, which would make jerky motion more tolerable.
• If possible, add a potentiometer to the remote motor shaft to sense the remote motor shaft's current rotation. Then, display the potentiometer value with an ohmmeter at the local control position. This will allow visualisation at the local control position of the effect (if any) that the control knob rotation had at the remote motor.
• When operating the local control knob, make a series of short rotations, separated by a brief and full stop of the control knob, instead of one smooth slow rotation. The intermittent nature of the series of short rotations is probably generating a pulsed output voltage which seems more reliable in ensuring motion of the remote motor.

Attempt 2: Transistor PWM control circuit

The previous paragraph mentioned that in the 7N3WVM control scheme, a pulsing voltage (generated by a series of short control knob rotations) was more reliable in achieving remote motor motion, compared to a steady low voltage (generated by a continuous, slow rotation of the control knob).

Indeed, a well-known technique for achieving reliable slow rotation of a motor, without stalling, is to use a pulse-width modulated (PWM) voltage. The technique and its motivation are described in many places, so I won't repeat the explanation here; see for example this page for a good introduction: http://www.electronics-tutorials.ws/blog/pulse-width-modulation.html.

The above page presents a PWM circuit using a 555 timer IC. I wanted a discrete transistor solution, and found a simple PWM circuit here: http://www.techpowerup.com/forums/threads/make-your-own-fan-speed-control-under-1-usd.124633/#post-1932968.

After building the above transistor PWM circuit and hooking it up to my geared DC motor, I was very pleased to see that it did allow reliable and very low-speed rotation of the motor. Unfortunately, after a few minutes, the smell of burning electronics drifted through the air, and I realised that the last driver transistor was smoking hot and certainly on the verge of failure. I used 2N3904 transistors throughout, and the motor was obviously pulling too much current through the driver transistor.

An obvious solution would be to use a power transistor capable of handling the motor current, but I had none on hand. And anyway, it seemed rather pointless to think about power transistors, when I already knew that I had a component on-hand capable of handling the motor current: the NJM2073D power amplifier chip from my previous experiments.

Final circuit: BJT PWM + power amplifier to drive motor

So, the logical next step was to use the transistor PWM circuit to generate a pulsed voltage, that would serve as the input to the NJM2073D power amplifier, which would then drive the motor. The final circuit as as follows.

Q1 and Q2 form an astable multivibrator. The values of the base bias resistors, R3 (for Q2) and R4/R10 (for Q1) were determined experimentally to yield a low-frequency pulse train. (An easy way to verify the pulse train frequency is to connect a piezoelectric earphone between ground and the Q2 collector and listen for the clicking or buzzing sound.) Switch S1 allows shorting out one of the Q1 base bias resistors, R4, which reduces the total Q1 base bias resistance from 570k to 100k. This has the effect of increasing the frequency of the pulse train, which leads to faster rotation of the output motor. With S1 open, very slow motion of the output motor is possible; with S1 closed, faster motion is possible.

The Q1/Q2 circuit was designed for operation off of 1.2 volts and was initially powered off of a separate 1.2 volt battery. Later, when reworking the circuit to use my main 12 volt power supply, I originally used a few silicon diodes to form a simple voltage regulator to supply 1.2 volts to the Q1/Q2 circuit. But during testing I empirically found a simpler solution: simply inserting R12 (10k) between the 12 volt positive rail and the Q1/Q2 circuit yielded about 1 volt available for Q1/Q2, and the Q1/Q2 multivibrator functioned properly in this simpler configuration. This is a rather ad-hoc solution, but it works.

Q3 is the driver transistor. Originally, the motor was connected directly between the Q3 collector and the positive rail. Now, instead, the Q3 collector does not drive the motor, but instead only provides input voltage (from the multivibrator pulses) to supply to the NJM2073D power amplifier. In 7N3WVM's original NJM2073D circuit, driven by the voltage generated by a DC motor, the input pins 5 and 8 are floating and connected to the DC motor terminals. 7N3WVM's "floating input" configuration allows both positive and negative polarities to appear at the output pins 1 and 3. However, in my circuit, instead of a floating input at pins 5 and 8, I have permanently tied pin 5 to the positive rail (through R13) and pin 8 to the negative rail (through Q3). This means that the output voltage at pins 1 and 3 will have only one fixed polarity and that the output motor can only be driven in one direction. Therefore, I added a DPDT switch at the output (S2-S5) to allow reversing the motor direction by flipping the DPDT switch to reverse the motor connections to the output pins 5 and 8. Physically, the DPDT switch also has a "neutral" position in the center where it is not connected to any output, which allows stopping the motor once the appropriate capacitor setting has been reached. For operating convenience, I chose a spring-loaded DPDT switch that automatically returns to the central neutral position when pressure is removed (an "auto-return" DPDT switch). This allows momentary pressure on the switch in either direction to rotate the motor in either direction, and automatic stopping of the motor when one's finger is removed from the switch.

Resistor R13, between input pin 5 and the positive rail, was determined empirically to limit the input voltage available at pins 5 and 8 such that the geared motor was just barely able to rotate. If R13 is too small, the input pulses at pins 5 and 8 are so large that they result in a strong, jerky motion of the output motor for each pulse. If R13 is sized appropriately, the input pulses (and output pulses) are reduced in strength such that the motor smoothly and slowly rotates. If R13 is too large, then no rotation of the motor is possible because the input pulses and output pulses are too small.

Due to the number of empirical adjustments made to the circuit, it is likely that the circuit will need to be adjusted if used with a different motor and/or a different capacitor (whose shaft stiffness will determine how much power the motor needs in order to rotate it).

Video of PWM-controlled motor in action

The following video shows the operation of the PWM-controlled motor. Notice the slow rotation of the motor gears and the extremely slow movement of the capacitor vanes, which are important to allow fine adjustments to the capacitor setting.

The video starts out with the slowest PWM pulse train. Later in the video, I close switch S1 which leads to faster motion. I also switch the DPDT switch to show reversal of the motor motion.

There is some backlash when reversing direction; it takes several seconds for the capacitor shaft to start rotating in the opposite direction. This was expected and is tolerable.

Most importantly, the circuit (after the empirical adjustments described above) is very reliable. When voltage is applied, the motor always rotates, at a slow and reliable speed. This makes remote operation easy; in particular, it is easier than the 7N3WVM analog-motor-to-analog-motor circuit, which was unreliable and prone to both under-responding (motor stalling) and over-responding (jerky rotation). But with the PWM circuit, it is possible to know with certainty that the remote motor will rotate in response to the local switch positions: with the DPDT switch in the upper position, the motor will definitely rotate clockwise with a well-defined maximum speed and no jerkiness (due to the speed limit imposed by the multivibrator pulse train); with the DPDT switch in the lower position, the motor will definitely rotate counter-clockwise, again with no jerkiness; with the DPDT switch in neutral, the motor will stop.

The system of course is not perfect: as with any mechanical system, there will be some unpredictabilities in the operation like slightly varying rotational speed, temperature-dependent expansion of parts or heating of grease, slow startup and overshoot on stopping due to friction and inertia, and gear backlash when reversing directions. Nevertheless, I expect that the system will be sufficiently convenient, accurate, and reliable for tuning the loop antenna.

https://www.youtube.com/watch?v=-gtEbqFzXX4

Building a 7 MHz small transmitting loop antenna, part 3

Part 3 of this article series focuses on the mechanics of the motorised variable capacitor.

Mounting the capacitor and motor

I had originally planned to use hot melt glue to mount the butterfly capacitor and the motor directly inside a small plastic box. However, working within the cramped confines of the box interior would make precise placement and alignment difficult. I therefore decided it would be easier to mount the capacitor and motor on a small flat board, which would allow me easy access to all the parts from all angles. Once the parts placement is finalised, I can simply mount the finished board inside the plastic box.

The board I used for mounting the capacitor and motor is a 1mm-thick polyethylene cutting board. I specifically chose polyethylene because of its low RF loss. The electric field in a small transmitting loop antenna is concentrated around the capacitor area, so it is important that any materials around the capacitor have low dielectric loss. As a counter-example, it would probably be a bad idea to mount the capacitor on a wooden board, because wood is a poor RF dielectric. Even though the capacitor losses, caused by poor dielectrics around the capacitor, will be small and likely in the milliohm range, in a small transmitting loop the radiation resistance is also small, making it important to minimise even seemingly trivial losses like dielectric loss. For more reading on the effects of dielectrics on the capacitor losses, here are a few links. This link describes an improvement in the Q of a variable capacitor by replacing the phenolic insulators with HDPE (high-density polyethylene): http://theradioboard.com/rb/viewtopic.php?t=1525#p12346. And, this link describes in detail various loss mechanisms of air variable capacitors, including dielectric losses: http://g3rbj.co.uk/wp-content/uploads/2013/10/Measurements_of_Loss_in_Variable_Capacitors_issue_2.pdf.

The capacitor is mounted on the polyethylene board as follows.

When mounted vertically, the height of the capacitor shaft was lower than the height of the motor shaft, so I mounted the capacitor on top of a stack of three scrap pieces of the cutting board, raising the height by about 3 mm. The scraps were affixed to the main board with hot melt glue, and the capacitor was also held in place by hot melt glue. A shaft coupler was attached to the variable capacitor shaft to enable interfacing with the motor shaft. The variable capacitor's shaft was slightly too thin for a snug fit into the coupler, so I wrapped some paper a few times around the capacitor shaft to ensure a snug fit. A snug fit is important to ensure that the center of the capacitor shaft's rotation is the same as the center of the shaft coupler's rotation (simply tightening the coupler's screws against a too-thin capacitor shaft would lead to eccentric rotation of the shaft coupler).

The motor looks as follows.

The motor shaft is very thin, much thinner than the capacitor shaft, and does not fit snugly into the shaft coupler. Unlike the solution for the thin capacitor shaft, where paper was wrapped a few times around the capacitor shaft to fatten it, with the motor shaft the thinness is so extreme that wrapping paper is not a viable solution for fattening the motor shaft; the paper would simply slip as the shaft turned. Therefore, to expand the small radius of the very thin motor shaft, I applied hot melt glue liberally to the shaft and allowed it to harden into a bulb. Then I carefully snipped away the hardened glue until the diameter of the glue bulb would just fit into the shaft coupler. The hardened glue, being a form fit for the motor shaft and being adhesive by design, grips the motor shaft well enough that it should not slip and can deliver the rotational power from the motor shaft to the load.

The glue-encased motor shaft was then inserted into the shaft coupler, the coupler's screws were tightened around the hardened glue, and the motor was glued in place to the polyethylene board.

Use of hot melt glue

I found hot melt glue especially suitable for this kind of hardware layout because it allows quick and easy initial positioning of elements, without requiring any drilling of screw holes or the like. Repositioning of elements, if needed, is also easy.

When applying the glue, I carefully apply a neat blob or seam to the key connection points required for strength, taking care not to apply too much glue that would be difficult to remove later if needed.

If repositioning is needed, I can pull away the glue with either my fingers or a pair of pliers.

Possibility of remote sensing of capacitor rotation

The current motor setup allows remote control of the capacitor by applying voltage to the motor, but it is not currently possible to remotely determine the current setting of the variable capacitor.

Notice, however, that in the above image, the motor shaft extends not only to the left side towards the capacitor, but also extends to the right side, which is currently unused. In the future, it might be possible to couple the right-hand shaft to a potentiometer, which would then allow remote reading of the potentiometer value to sense the current angular setting of the variable capacitor.

This sensing capability might be exploited to implement an automatic tuner, where for example the transmitter's frequency is automatically read via some interface and a computer activates a control voltage to rotate the motor appropriately, monitoring the potentiometer value to gauge the capacitor's rotation and stopping when the rotation reaches some previously-saved value appropriate for the current frequency. Of course, implementing such a system requires consideration of real-world robotics issues like backlash and overshoot, and would likely require some sort of a PID controller implementation in software.

RF choking, or lack thereof

By choosing to mount the motor directly next to the capacitor, I am placing the motor and its control wires in the vicinity of the capacitor's strong electric field. A more conservative approach would place the motor at or near the loop's zero-voltage point (diametrically opposite the capacitor), using a very long shaft to couple the bottom-mounted motor to the top-mounted capacitor. In a previous loop design I did exactly this, using a one-meter plastic coupling shaft. The problem is that a long coupling shaft will usually introduce extra backlash into the system, making the capacitor slower to respond to the motor rotation, due to time it takes for the long plastic shaft to twist until it has enough tension to drive the capacitor. And in the current loop design, a 2-meter shaft would be necessary, which would introduce even more backlash and would be mechanically challenging as well due to the long shaft length.

For the above reasons I chose to mount the motor directly next to the capacitor, to simplify the mechanics. Also, it's worth noting that the MFJ 1786 loop antenna (generally regarded as a well-performing, properly-constructed small transmitting loop) also places the motor directly at the capacitor.

However, the mounting of the motor and its control wires near the capacitor could conceivably cause RF currents to flow on the motor control cables, which could lead to problems both on transmit and receive: on transmit, the transmitted energy might make its way back into the motor control circuitry causing equipment damage or posing a shock hazard; on receive, the motor control wires might form part of the antenna system and allow induced noise on the control wires to couple into the capacitor and the loop antenna, where it then would be passed back into the receiver as unwanted noise.

The way to solve this problem is to use an RF choke or chokes on the motor control wires to prevent RF currents from flowing on the control wires. The MFJ 1786 loop antenna uses RF chokes for this purpose. Also, N4SPP's page on small transmitting loop construction (http://www.nonstopsystems.com/radio/frank_radio_antenna_magloop.htm) shows how he uses two ferrite chokes on his motor control cable.

I considered using ferrite chokes, but was concerned about how to decide on the specifics of the choke design: the proper core material, core size, number of turns, choke location, etc. Before diving too deeply into the choke design, I decided to run a 4nec2 antenna simulation to see if choking was even necessary or not.

I made a 4nec2 model of the loop and tried running a disconnected wire down from near the capacitor to ground, to represent the motor control wires (which are near, but not connected to, the capacitor). The separation between the capacitor segment and the motor wire segment was 1 cm. Also the motor wire deliberately ran away in an asymmetrical fashion (which makes it easier for unwanted currents to flow) and was running very close to the main loop conductor (1cm spacing). Even in this asymmetrical condition, the current induced on the unchoked wire, with 5 watts transmitting power, was only about 20 milliamps maximum, which I consider to be low enough not to be concerned about.

Although the simulation results lead me to believe that the unchoked motor cables will not carry significant RF current, it is possible however that for certain specific lengths, the control cable might be more prone to coupling into the capacitor and becoming an unwanted part of the antenna system. This will be determined during actual usage of the antenna. I plan to adjust the antenna for minimum SWR, then alter the routing of the motor control cables to see if the SWR changes. If it does, then choking of the motor wires may be needed.

Abrasion of the hardened glue

After testing the ability of the motor to drive the capacitor, some slippage started to occur after a few minutes. This was caused by the coupler's screws not being tightened enough around the hardened glue; the motor shaft and glue bulb were rotating, but the shaft coupler and the capacitor shaft were not rotating.

To fix this, I first tightened the screws even harder against the hardened glue. I cannot see inside the shaft coupler, but I believe the screws are now so tight that they are biting into the surface of the hardened glue. One concern is that the hardened glue is still quite soft compared to the usual metal shafts for which the shaft coupler is intended. Over time, as the motor is repeatedly rotated both clockwise and counterclockwise, I can imagine the shaft coupler's screws slowly grinding away at the hardened glue until the glue is completely stripped away from the motor shaft, leading to slippage and inability of the glue bulb to deliver power to the load. The expected failure mode will not be immediate; as the glue is slowly worn away around the contact points with the screws, I expect a widening hole in the glue to form, which will first lead to excessive backlash when the motor reverses direction, and may eventually lead to the wearing of a groove all around the glue bulb's circumference, at which point the motor-driven rotation of the glue bulb will no longer be able to push against the coupler's screws and will no longer be able to rotate the capacitor shaft.

In an attempt to delay this mechanical abrasion of the glue, I added more hot melt glue at the shaft coupler's opening where the motor shaft enters.

The hope is that the additional hot melt glue will allow the motor shaft to grip the outer surface of the shaft coupler, and that the required torque to turn the capacitor shaft will be partially provided by the grip of the glue on the coupler's surface. This might reduce the pressure, while the motor is running, of the coupler's screws on the inserted glue bulb, which in turn might lengthen the lifetime of the inserted glue bulb. Time will tell.

Addendum 2015-01-18: The glue seems to become harder after a 24 hours; it seems hard enough that moderate pressure from the screws will not significantly wear away the glue. The slippage mentioned above was caused by insufficient insertion depth of the hardened glue bulb into the shaft coupler, meaning that the shaft coupler's screws were only barely grazing the tip of the hardened glue bulb instead of firmly gripping the main body of the hardened glue bulb. With a sufficient insertion depth, 24-hour-hardened glue, and moderately-tightened screws on the shaft coupler, no further slippage has been observed.

Motorised capacitor in operation

The following video shows the operation of the motor when run from 4 AAA batteries.

https://www.youtube.com/watch?v=_CNhot_yUv8

The motor's rotational speed is still several RPM, which is too fast for precise tuning of a narrow-bandwidth small transmitting loop antenna.

The problem is compounded by the fact that a butterfly capacitor covers its complete capacitance swing in only 90 degrees, as opposed to the 180 degrees of a normal air variable capacitor.

The next post in this series will present a pulse-width-modulated circuit that can achieve a much slower rotational speed of 1 RPM or less.

Addendum: A note on motor lifetime

At the following page, some interesting data is presented on the lifetime of the type of motor I am using: https://www.pololu.com/docs/0J11/all#2. It can be seen that the lifetime for continuous operation is, in the best case, on the order of tens of hours, and in the worst case (when running at higher voltages) is less than ten hours.

This has some serious implications for the overall system design:

1. The voltage used to run the motor should be kept within limits to lengthen the motor lifetime.
2. In case of over-voltage, the motor will fail fairly soon. Assume for example that the motor is run at 6 volts. The expected lifetime will be only about 5 hours or 300 minutes. Then, assume the antenna is used such that the total motor on-time per day is 10 minutes (which is realistic if the antenna is frequently re-tuned, for example for short-wave listening purposes on various frequencies). Then, the motor will fail after only 30 days of operation.
3. In the current loop design, it will be quite tedious to replace the motor because:
• The loop dimensions (3m x 2m) are too large to allow transportation of the loop or laying the loop down flat (in the limited space on the balcony) for access to the capacitor box
• The entire loop must be de-soldered in place, piece by piece (see part 1 of this article series), to access and dismantle the top-mounted capacitor box
• After motor replacement, the entire loop must be re-soldered together
4. To better accommodate the eventual necessity of replacing the motor, it may be better to use mechanical connections to connect the capacitor to the main loop. The mechanical connections could then be easily undone to completely disconnect (mechanically and electrically) the capacitor box from the loop, where it could then be taken inside and the motor replaced. Such mechanical connections could be wing nuts, hose clamps, or similar. These connections will introduce additional ohmic loss compared to a solder connection, but with a large contact area the resistance should be able to be kept within manageable bounds. Additionally, the radiation resistance of the loop is relatively high due to the quarter-wavelength circumference, making tolerable the additional ohmic loss through mechanical connections. 

Building a 7 MHz small transmitting loop antenna, part 2

I did some more exploratory work on the mechanics of constructing my 7 MHz small transmitting loop.

Fitting copper pipes together

The copper pipes I am using are 6mm in diameter with a wall thickness of 0.5 mm. This means that the inner diameter should be 5mm.

My plan on fitting the copper pipes together was to insert a short length of 5mm copper pipe into the 6mm pipe, as an inner segment to provide some structural stability. Then the adjoining 6mm copper pipe would also be slid snugly over the 5mm copper pipe, and butted against the other 6mm pipe. The butt joint, reinforced by the interior 5mm pipe, would be heated and soldered.

The problem in reality is that the 5mm pipe cannot quite be inserted into 6mm pipe due to manufacturing tolerances. The outer diameter of the 5mm pipe is slightly greater than 5mm and/or the inner diameter of the 6mm pipe is slightly less than 5mm.

Therefore, I needed to swage open the ends of the 6mm copper pipe, creating a flared opening at the pipe end, so that the 5mm pipe could fit inside.

My hand tools are limited, but I found that I had a screwdriver of just the right diameter to swage (somewhat crudely) the 6mm pipe opening wide enough to accommodate the 5mm pipe.

My swaging process is quite crude and requires me to wiggle the screwdriver in the 6mm pipe opening to flare it open. This process is not precisely repeatable, so each time I swage a pipe, the exact depth of the flare (i.e. the maximum depth into which the 5mm pipe can be inserted before it is stopped by friction against the unflared 6mm pipe wall) is different. This means that the required length of the inner 5mm connecting pipe segment will be different for each set of pipes to be connected due to the inexact and differing depths of each flared-open pipe end. This is illustrated in the above diagram by the different lengths of flared end A and flared end B.

To proceed, each end of each pipe to be connected must be individually flared open such that the 5mm pipe can be inserted to a depth sufficient to support the joint. Then, for each pair of pipe ends to be connected, a custom length of 5mm pipe must be cut specifically for that pair of pipe ends so that when inserted, the custom length of 5mm pipe allows the 6mm flared pipe ends to exactly butt against each other. 

Then, the pipes will be soldered together.

Copper soldering with a torch

I have no experience soldering copper with a torch. As an experiment, I grabbed some spare copper flashing I had lying around and tried soldering it with my small torch. The results, as you can see, were highly unsatisfactory, though I at least did get the solder to wet and bond in some places.

Clearly, I will need much more practice with the torch before I will have confidence soldering the pipes together.

The above experiments were done without flux. My next try at torch soldering will use flux applied to the copper surface.

Soldering flux to be used in future copper soldering experiments with the torch.

As an alternative, it may turn out to be easier to use a large-wattage soldering iron instead of a torch to solder these small copper pipes.

Attaching copper pipes to the capacitor box

The top-mounted and motorised capacitor will be mounted in a small plastic box for protection from the weather. Another mechanical problem to solve is how to attach the top-mounted capacitor box to the pipes at the top of the loop.

Due to the top-mounting of the capacitor box and the lack of a central spine in my loop design, the points at the top of the loop where the pipes connect to the mounting box may be subject to moderate stress if the loop flexes, as may happen in wind or during manual repositioning of the loop. And, indeed, we want the pipe-to-box connecting point to bear all of the stress caused by loop flexing, because otherwise that stress would be borne by the stator shafts of the delicate butterfly capacitor, which in my case is quite small and might easily be torn apart, cracked, or warped if the stator shafts are subject to too much stress.

Hot melt glue seems appropriate in this case to bond the copper pipes to the capacitor box. It seems strong enough to bear the expected stresses, yet it can be pulled apart for disassembly by a firm tug with a pair of pliers.

In the following image, I am applying hot melt glue to attach one pipe (that will eventually be at the top of the loop on one side) to the left half of the capacitor box. Later, another pipe will need to be connected to the right half of the capacitor box.

Applying hot melt glue to affix the one copper pipe to the plastic capacitor box. The glue forms a large blob completely covering the copper pipe and the surrounding plastic.

After the glue solidified, it formed a moderately strong bond that served to hold the copper pipe in place against the plastic box. In the following image, I am holding the capacitor box only. The full one-meter length of the pipe is unsupported and hanging freely in air, exerting maximum leverage on the hot glue joint. The joint held and showed no signs of breaking; I estimate the pipe would bend before the glue joint would break. 

The capacitor box is supported, and the one-meter length of copper pipe is hanging freely in air. The glue joint does not break.

Also, holding free the end of the pipe and allowing the capacitor box to hang freely, again the glue joint showed no sign of breaking. The joint still held even when the capacitor box was filled with the extra weight of the capacitor and motor.

The end of the pipe is supported, and the capacitor box hangs freely in air. The glue joint does not break.

These tests indicate to me that the hot glue joints at the capacitor box will be able to bear the stress of the loop flexing.

After these tests, I could break the glue joint by pulling it apart with a pair of pliers.

The above describes the solution for the mechanical attachment of the pipes to the capacitor box. For the electrical connection of the pipes to the capacitor, located inside the capacitor box, I will solder a flat strap to each of the two top-most copper pipes. Each flat strap will then be routed from outside the box to inside the box underneath the box lid, where each strap will be soldered to one end of the butterfly capacitor.

As with all aspects of this loop design, the mechanical and electrical connections should be capable of being reasonably easily disassembled for loop maintenance, transportation, or storage. To achieve this objective, the general approach for assembling the capacitor area will be:

1. Solder copper straps to left and right copper pipes.
2. Electrical connection: Solder each strap to one side of the butterfly capacitor.
3. Insert capacitor in box, routing straps and the connected pipes outside the box.
4. Mechanical connection: Glue (with hot melt glue) the pipes to the top of the capacitor box.
5. Place lid on capacitor box.

In particular, note that this assembly sequence is easily reversible and does not require any tricky operations like torch soldering inside the plastic box.