Varicap Tuned 6AK5 Super-Regenerative Receiver.


No variable capacitor! Varicap diodes allow for potentiometer tuning remote from the receiver.

The purpose of this project was to design a super-regenerative receiver for the VHF FM broadcast band, which could be tuned with varicap diodes, and did not require a user adjustable regeneration control. This was all part of the design and construction of a radio to be installed in the Jeep.
The Jeep radio will be constructed in three units. These are: 1) Power supply and receiver unit,  2) Speaker with output stage, and 3) Control unit. This is necessary because of the available space and dashboard layout. Since the control unit is located separately to the receiver, a difficulty occurs with remotely tuning the VHF circuit.
In the past, Bowden cables were used with two unit car radios, to provide a mechanical coupling between the control unit, and the variable capacitor inside the receiver unit. In the modern day, obtaining or making the mechanical parts required would be very difficult.

I have been through this exercise before, with the Model T radio. It is also a two unit set with speaker. The obvious solution there was to use varicap diodes to tune the receiver, since all that is required is a variable DC voltage. The tuning control then becomes a potentiometer, which can be located at any distance from the receiver. Unfortunately, varicap diodes are completely out of character for something that was intended to be of the valve era, but we have to be practical, if we want a radio for the Jeep (or Model T)!

Having dealt with the tuning, the next thing is the regeneration control. Experience has shown that user adjustable regeneration controls are undesirable for a car radio. It is a major distraction to the driving, and a two handed tuning operation is obviously impractical.
Recent super-regenerative receiver designs, as described on this site, have included stabilising circuits, which go a long way to minimise regeneration control adjustment.
While the Model T radio works well, and seldom requires regeneration adjustment, I wanted to see if this could be improved on. If no improvement was possible, the Jeep radio could simply use the same receiver design.

Using a Fremodyne circuit would of course solve the automatic regeneration control problem. It was considered, but with more parts required, and the limited space available, it could end up being quite cramped.

Varicap Diodes.
Evident from the Model T radio is that varicap diodes have a much lower Q, compared to an air spaced variable capacitor. This is a problem with a self quenched super-regenerative detector, because the oscillation must be very strong. It has to be strong enough to bias the valve into cut-off. The lower Q makes this difficult, and medium gain valves like 6C4 and 7193, will not super-regenerate properly, despite working perfectly well with a mechanical variable capacitor.
The varicap diodes I used in the Model T radio, and for this experiment, were Motorola types MV2109. These are 1980's era components, which would probably be difficult to obtain now. Their Q seems to be much better than the Philips BB105G.


Characteristics of the MV2109 varicap diodes.

Varicap diodes are used with reverse bias. The more reverse bias, the less the capacitance. In this circuit, two diodes are used, connected back to back with the anodes together. This is done so that the voltage developed across the oscillator coil cannot forward bias either of the diodes. If only one diode is used, and the oscillation voltage is high enough, the diode will conduct and restrict the tuning range. Of course, when two varicap diodes are connected in series, the capacitance range is halved.


Varicap diodes can be seen connected across the oscillator coil.

It should also be obvious that the voltage fed to the varicap diodes must be very well regulated, if tuning drift is to be avoided.

Automatic Regeneration Control.
My first attempt at making the regeneration control automatic used cathode stabilising. Oscillation level is dependent on cathode current, so if the cathode current can be stabilised, then so should the oscillation level. This method uses a cathode resistor, suitably bypassed, and works the same as any cathode bias circuit. That is, if cathode current increases, negative bias on the grid increases, so reducing the cathode current, and vice versa. The Model T radio uses this method. In practice, it is rarely necessary to adjust its regeneration control.

More recently, experiments have been done with grid stabilising circuits. I first experimented with this method here. Typical of receivers using it is this 7193 circuit.
As mentioned previously, the level of oscillation is dependent on cathode current, and thus the grid bias. For this method of stabilisation, the rectified grid voltage, which occurs as part of the oscillation, is itself used as the bias voltage. A negative voltage appears at the grid because of the charging and discharging of the grid capacitor (33pF), through the diode formed between the grid and cathode. The greater the oscillation, the greater the negative voltage, and the greater the current through the grid resistor.

If the grid resistor current is used to charge a capacitor, we then have a negative DC voltage across this capacitor, proportional to the oscillation level. So that this level of DC can be adjusted, more or less current is shunted to earth by a resistor across the capacitor.
For example, if the capacitor charges to -2V, this is exactly the same as biassing the grid circuit to -2V from any other source. The greater the oscillation, the higher the bias, and the subsequent reduction in oscillation.

Somewhat coincidentally, I later discovered both methods of stabilisation were also used with the Fremodyne circuit. At the time I had not realised this, since it was prior to obtaining the patents for, and fully understanding, the operation of that circuit.

First Tests.
Seeing as one of the goals was to eliminate the regeneration control, the first test was to test one of the simple grid stabilised circuits. The first immediate difficulty, however, was covering the full FM band with the varicap tuning.
Yet, the Model T radio, using the same type of varicap diode did not have this problem. A little investigation into the characteristics of the valves I was attempting to use explained this. The test panel has a 7 pin valve socket, and valves I tried were initially types like 6AU6, 6BZ6, 6GK5, and 6AM6. While all these types will oscillate at VHF, there's the question of input capacitance.
 
7193 2.2pF
6C4 1.8pF
12AT7 2.2pF
6AK5 4pF
6CB6 6.5pF
6EW6 10pF
6GK5 5pF
6AM6 7.5pF
6BZ6 7pF
Input capacitance of several types of valve.

This explains why the Model T radio had sufficient tuning range. It uses a 12AT7 with an input capacitance of only 2.2pF. At the other extreme, 6EW6 has 10pF of input capacitance. That is enough to tune half way across the FM band! The problem here is the restricted tuning range of varicap diodes. In the case of receivers using mechanical tuning capacitors, the capacitance range is much greater, so this is not so much of a limitation.
Having discovered this, the next thing to do was choose a more appropriate valve. As mentioned before, type 6C4 or 7193 hasn't enough gain to oscillate to the required degree in the low Q circuit. However, the 6AK5 looked promising. It would also fit nicely into the WW2 theme.

Indeed, this valve worked exceptionally well, oscillating possibly more readily than any other type I've so far used. It was still self quenching nicely with under 50V B+ and varicap diodes! Because the 6AK5 oscillates so easily, it means a higher than normal plate load resistor can be used. This results in a noticeably higher audio output. Typically, 1.3Vp-p was produced by weak stations, increasing to about 3.8Vp-p on strong stations. Another characteristic is the lower cut off voltage, which assists with self quenching. Previous experiments with pentodes showed no advantage in providing a separate screen grid supply, so the 6AK5 was connected as a triode.
While not tested, it's possible that the lower Q, but more easily available BB105G diodes might work.

Circuit 1.

Grid stabilised 6AK5 circuit for first test.

As can be seen, the basic circuit is the same as others which I've described elsewhere on the site. It is a basically a modified Colpitts oscillator, relying on the capacitances across the cathode choke, and between the grid and cathode, to oscillate. It self quenches in the usual way, with the voltage across the 270k grid resistor building up high enough to cut the valve off, causing a cessation of oscillation. The process repeats, depending on the time constant formed by the 33pF and 270k. This is the quench frequency, which is in the realm of 35 kHz.
The 1N914 and RC will be mentioned later, and are not part of the original test circuit.

Tuning voltage is provided by the 100k pot, which is supplied through the 82k and 6.8k resistors from a regulated 15V supply. The pot is wired as a shunt resistor rather than a voltage divider, so that only one wire and an earth is required. This is to simplify the connection to the remote control unit, by eliminating one wire. Unfortunately, this method results in non-linear operation, with stations cramped together at the low end of the band. This can be offset to some degree by using a pot with a logarithmic curve. The 6.8k sets the lowest oscillation frequency, while the 82k sets the highest.

Because of the LC ratio of the tuned circuit, the amplitude of the oscillation varies as the receiver is tuned across the band. At the 88 MHz end, the ratio of capacitance to inductance is less than at the 108 MHz end. The more capacitance, the less the circuit wants to oscillate. As such, the regeneration requires adjustment in conjunction with the tuning, if performance is to be consistent across the band. The regeneration needs to be increased as the receiving frequency is decreased.

It has been found the best method of adjusting the regeneration is to take the grid resistor to a more or less negative voltage. Other methods commonly seen elsewhere, such as adjusting plate voltage, result in inferior sensitivity.

6AK5 with Grid Stabilisation.
With the recent success of the grid stabilising circuit, I naturally applied it to the circuit under test, using the 6AK5. Here, the current flowing in the 270k grid resistor charges the 0.1uF capacitor to a negative voltage. The voltage is set by the 2M pot shunted across the capacitor. The higher the oscillation amplitude, the greater the negative voltage.
Alas, results were only good up to a point. The problem was a marked drop off at the 88 MHz end of the band. It was found the quench frequency was increasing to virtually double what it was at 108 MHz. As quench frequency is raised, the sensitivity and audio output level drops off. Ideally, the quench frequency should remain around 30-35 kHz across the band. This effect was not greatly influenced by the value of load resistor or B+ voltage.

Measurements were taken with various load resistors and B+ voltages. Setting the regeneration control at the 108 MHz end of the band provided the following results:
 
RL & B+ Vg 88 MHz Vg 108 MHz fq 88 MHz fq 108 MHz
230k @ 240V -1.34 -3.24 57 kHz 28.6 kHz
150k @ 150V -1.21 -2.26 65 kHz 35 kHz
56k @ 100V -1.31 -2.17 59 kHz 34.2 kHz

Since Vg is entirely derived from rectified grid voltage, we can clearly see how active the oscillation is across the band. The VHF oscillation amplitude increases as frequency increases, and vice versa. The effect on the quench frequency is the opposite. As the VHF oscillation frequency decreases, the quench frequency increases. With lesser oscillation at the 88 MHz end, it takes less voltage to cut the valve off. In other words, less time is required to charge the 33pF grid capacitor, before the valve cuts off and stops oscillating. A lesser time constant, of course, means a higher quench frequency.

With varicap diodes, the situation is worsened because the Q decreases with capacitance. Thus, the Q is less at the 88 MHz end of the band, where oscillation is already difficult.
To compensate, it then becomes necessary to reduce the negative grid voltage (Vg), to decrease the quench frequency.

Since the tuning control provides an increasing positive DC voltage as the receiving frequency is increased, could we not use this to compensate? This is where the 1N914 and compensating resistor, RC, come in. By offsetting the negative grid voltage (Vg) with the tuning voltage, it should provided the desired effect. So that the negative grid voltage could not affect the tuning voltage, the 1N914 diode was included. The compensating resistor value was not measured, but was a 3M pot in the test circuit. The resistance was at least 1M.
It worked to a degree, except the compensation was not as linear as desirable. It was also felt that this solution was really masking the problem, rather than actually compensating for it. The degree of automatic stabilisation would be reduced, as the tuning voltage was increased

Circuit 2.

Cathode stabilising circuit added.

Remembering the cathode stabilised circuit used in the Model T radio worked quite well, the next test was to see how well it would work with the 6AK5. The 2M pot was reduced to 0 ohms, and the cathode stabilising components introduced; these being the 10k pot and 22uF bypass capacitor. Since the 'problem end' was at 88 MHz, the cathode resistance was adjusted to optimum operating conditions at this frequency; i.e. a quench frequency of 34 kHz. When tuned to 108 MHz, the quench frequency decreased to about 30 kHz.
This in itself was a huge improvement over the grid stabilising circuit on its own.

What if the grid stabilising circuit was also used? With the receiver tuned to 108 MHz and the quench frequency still 30 kHz, a slight adjustment of the 2M pot brought the quench frequency back up to 34 kHz. Importantly, when the receiver was tuned back down to 88 MHz, it was still 34 kHz!

Mission accomplished!

The operating conditions were optimised. Less intermodulation distortion was had by using a quench frequency of 30.4 kHz.

Circuit 3.

Final circuit.

The adjustment of Rg (the 2M pot) was not critical, and it seemed acceptable to replace it with a fixed resistor. This would mean only one preset adjustment was required. The final circuit is shown above. The 10k pot can be wired as a rheostat if desired, with the 180k resistor earthed. The result is electrically the same. The optimum resistance in the test circuit was 2.46k, which means a 5k pot could be used instead.

Quench Frequency vs. Signal Strength.
It is important to note that because the stabilising circuits also act on signal strength, the adjustment must be done with no signal. It will be found that quench frequency increases with signal strength. This is actually an ideal situation, because it provides an additional AVC effect, to that already inherent in a super-regenerative detector. A higher quench frequency is desirable for sound quality, but because of the desensitising effect, is unsuitable for weak signals. However, in this instance, the quench frequency is automatically decreased for weak signals, where it's needed, and increased for strong signals where sensitivity is less important.

It is now an interesting proposition to modify the Model T radio, with the simple addition of one resistor and capacitor, to remove any need for its regeneration control adjustment.


Test set-up for the receiver. This is how many of my projects begin.


Home