This project was a practical development
of the tests using a 6GK5
as a super-regenerative receiver. I was so excited by the results,
that a proper receiver had to be made up. The 6GK5 circuit is simpler than
the 12AT7 design, which has been the basis of all the valve super-regenerative
receivers so far described. Yet, it seems that there is very little, if
any, reduction in performance by not including an RF amplifier stage. The
new method of regeneration control also seems less critical and easier
to use, with less adjustment required.
The 12AT7 design has been proven over the last 18 years, and time will tell how the new simplified circuit compares.
The plan was to build a small receiver in plastic case that can easily be carried around and plugged into any available power point. A telescopic aerial would be fitted to eliminate the need for connection to an external aerial.
For the valve line up, a 6GK5 would be the detector, and initial thoughts were for a 6HG8/ECF86 for the audio stage. The power supply would use a 140V 40mA transformer, which I have a few of, and which have been used in other projects. The use of a plastic case would actually allow for a live chassis design, using just a filament transformer or even a capacitive dropper. This idea was dropped however, since it would take more work to turn up the necessary insulated extension shafts for the controls, than it would be to mount a power transformer which I already had, and which there was ample room for. A capacitive dropper for the heaters is a novel idea, but the receiver would then be unsuitable for operating off anything but the power mains. The low power factor and frequency dependent characteristics would make it unsuitable for use off an inverter.
Construction would be on an aluminium chassis which would engage in the slotted sides of the plastic case. For neatness of appearance, it was decided that the speaker would be mounted on the chassis. This would eliminate the mounting screws being visible on the front panel, as well as making servicing more convenient, by not having the speaker tethered by wires to the panel. In fact, the only screws visible would be those securing the front panel to the case. Although there would be a telescopic aerial, allowance would be made for an external aerial where required.
With the limited space available as well as limited heater current, thoughts for the audio stage were to use a 6HG8. This is a common Australian TV tuner valve, otherwise known as ECF86. Europeans would be more familiar with the series heater equivalent, PCF86.
It is a triode and frame grid pentode, intended for use as a frequency converter. As I have prolific amount of these valves, one of them would appear to be ideal for a low power audio stage. One limiting factor with using the 6HG8 for other applications is the common cathode, but this need not present a problem for audio use, since separate bias supplies could easily be arranged.
Since the use of a 140V power transformer would provide about 150V B+ after filtering, tests were done at this voltage. Unfortunately, the results were disappointing, with only 87mW produced by the pentode, and 103mW for the triode. Admittedly, the 7k load impedance was not ideal, but I felt that more output should be obtainable.
Given the high output from the 6GK5 detector, the next thought would be to use a very high gain frame grid pentode on its own. A Special Quality (SQ) type, E180F was tried next. This actually worked very well, with 240mW output. It needed an input of 1.95Vp-p to fully drive it, which meant that to fully drive it from the 6GK5, an extra stage would be required. This would be getting away from the "2 Valve FM Receiver" concept, but nevertheless, the E180F will be kept in mind for future projects.
Valves like 12AU7 and 12AT7 were discounted, since previous experience shows they really need 250V to produce sufficient output.
Back to the miniature triode pentodes - the old faithful 6BL8/ECF80 was tried next, which I have used as an audio output before. Alas, with only 150V B+, the best that could be obtained was 150mW from the triode, and 52mW from the pentode.
The Homemade Fremodyne uses a 6AW8 on 150V with good performance, so this was also an option. Its 600mA heater current would still be within the power transformer ratings. The pentode could provide 226mW, or with the pentode connected as a triode (as per the Fremodyne), power was reduced to 195mW. (It should be noted the Fremodyne set has an output transformer which is more ideally matched, and thus the output is higher).
At this stage, things were looking good for the 6AW8, but before finalising the design, I considered the two Decal triode pentodes, 6U9 and 6X9. While both these look the same, and have the same pin connections, the triodes and pentodes in each of them have different characteristics. It was found that the 6U9 gave quite a low output, but the 6X9 pentode was more what I was looking for, with 225mW output.
Introducing the 6X9.
The 6X9/ECF200 was one of a series of new 10 pin (Decal) valves introduced by Philips in 1964. Europeans would be more familiar with the series heater version, PCF200. These valves were very popular in Australia, and most commonly used by HMV, Kriesler, and Philips in their TV designs. The pentode is a frame grid type with a transconductance of 14mA/V, with the triode having 4.8mA/V. The pentode was intended as an IF amplifier, but as it turns out, it makes a good low power audio output stage. The triode is typical with a mu of 55. Together, the two sections make an ideal miniature audio stage with high gain.
The use of a 7k plate load may seem at odds since it is not properly matched. The reason for using it was to avoid buying a new transformer, when I already have plenty of 7k types in my collection. Ideally, for the operating conditions I used, the load should be 11k.
Bare chassis with few parts.
Major parts mounted prior to wiring.
Most of the work in building this set was the prolific amount of brackets required to get everything into just the right position. Everything did fit without being unduly crowded, but it took some thought to position everything. The aerial switch was also time consuming to mount. I used a slide switch here to avoid any protruding toggles, which would likely get damaged. Mounting slide switches are always difficult to mount, and the cut out is never perfectly rectangle. Nevertheless, I has pleased with the end result.
The power transformer was a type originally used in a valve millivoltmeter. It has a 140V 40mA secondary winding, and a 5.8V centre tapped 1A heater winding. If the heater voltage seems a bit low, that's quite normal for this kind of application. A slight reduction in heater voltage improves hum characteristics - important when measuring millivolts of AC. Since the valves in the meter were not being run at full power, there was no danger of cathode poisoning. While 5.8V is within 10% of 6.3V, it was noted in the initial 6GK5 tests that 6.3V was really required for proper operation. Simply winding eight turns of copper wire around the outside of the transformer, but within the core, provided the extra voltage required. This was connected in series with the 5.8V so that the voltages add.
The 6X9 circuit and power supply were installed
first. I had thought that the high mu of the 6X9 triode would allow the
use of contact bias, using just a 10M grid resistor. However, with only
15V on the plate, this was obviously insufficient, and cathode bias was
With the 6X9 working as it should, the 6GK5 circuit was wired up next. I decided on the grid rheostat method of regeneration control since it's simple and has no backlash. Also, it's a new part of the design, and this receiver would be an ideal test in various conditions.
Wiring for the 6GK5. Green wire connects to the telescopic aerial.
Trying the 6GK5 from my test prototype
in the new receiver answered the question, because things now were as they
should be. So what was different? Putting this 6GK5 and a selection of
others into the valve tester revealed that I had fallen into a trap!
The 6GK5 that worked so well had a plate current of 3mA and a transconductance of 3mA/V. Apart from the very weak 6GK5 which wouldn't super regenerate, all the others had higher plate current and transconductance.
It had transpired that I had developed this circuit with a worn out 6GK5! It's an irony, but it's what the circuit worked best with.
The next challenge is what to do about it. It would be poor design to only be able to use 'worn' 6GK5's, so more development was needed to get it working with 'normal' 6GK5's.
As it turned out this was simple. An obvious way to reduce gain of a valve is to insert an unbypassed resistor in the cathode circuit. Monitoring the plate waveform showed that by careful selection of this resistor, performance could be made identical to that of the 'worn' 6GK5.
The resistor value turned out to be 470R.
Of course, when the 'worn' 6GK5 was put into this modified circuit it didn't work properly, but all the 'good' ones did. Thankfully, the problem was easily overcome, and the 6GK5 was still a good super-regen detector. Immediately, I wondered if the same method would allow 6ES8's to work. That's for future testing...
With the success of that, I wondered about
eliminating the cathode choke. The test circuits using a cathode tapping
worked very well, but with only a hint of regeneration control backlash.
Seeing as we now had quite a bit of negative feedback (and extra stabilisation)
with the cathode resistor, I tried the cathode tapping instead. It was
completely successful and the choke was removed.
In fact, with the results at this time, I see no advantage in the cathode choke and will no longer use it. It's one less coil that has to be made up, and a resistor takes up a lot less space. I know from email correspondence that these cathode chokes are also a source of confusion for some constructors. Furthermore, the magnetic coupling effects between the choke and aerial coil are eliminated. The circuit is made much more predictable without the choke.
The only catch with introducing the cathode
resistor is now the cathode no longer has a low resistance DC path to earth.
As was found in the previous tests, some 6GK5's have heater to cathode
leakage of varying degrees, and this becomes apparent in this circuit.
Two valves tested were free from the effect, and I suspect they were actually EC97's. Most of 6GK5's had hum to a lesser or greater degree, which was completely absent when DC heating was used. Two options: Use only EC97's, or use a DC heater supply so the more common 6GK5 can be used as well.
DC Heater Supply.
The first attempt was to use a bridge rectifier and capacitor to rectify the 6.3V AC. The best that could be produced was 6.12V DC. Although in theory the DC should be 1.4 times the AC, and thus nearly 9V, there is the voltage drop of the rectifier to consider. With at least 1.4V dropped across two diode junctions, as well as the loading of the 220mA heater, the voltage was barely adequate. (Yes, a 4GK5 or 3GK5 could be used with a resistor, but the point was to use more common 6GK5's).
The next option would be a voltage doubling rectifier. Two possibilities were available here; a half wave type or a full wave type. Both use two capacitors and two diodes, but the half wave type allows for a common earth. The regulation of the half wave type is not as good, but the only way to find out if it would work was to try it. It worked better than expected, with between 10 and 11V DC under load. Since a dropping resistor was now required to run the 6GK5 heater, it was an ideal opportunity to install a zener diode regulator, and eliminate the heater voltage fluctuation effects. This worked perfectly, and the heater voltage remains stable down to 215V mains input. It was noticed that with the heater voltage fixed, that variations in B+ voltage did not cause any noticeable problems.
Two valves provides speaker reception with good sensitivity.
The design of this has been covered here. The 6GK5 is operating as a Hartley oscillator with cathode feedback to a tapping on the coil. The 470R reduces the gain of the 6GK5 for optimum performance in this circuit. Using a cathode tapping eliminates the RF choke which has been, until now, used to provide feedback.
The oscillator quenches by virtue of the time constant formed by the 33pF and 180k grid components. Although the triode is oscillating at VHF by virtue of the tuned circuit, it also goes in and out of oscillation at the lower quench frequency. Ideally, this is 28 to 36Kc/s. Lower quench frequencies provide higher output, but the beat between the 19Kc/s stereo pilot tone and the quench becomes apparent. Higher quench frequencies provide better audio quality but with reduced output. See the other super-regenerative receiver articles on this site for further details on this.
The amount of regeneration needs to be adjustable to obtain the best operating conditions. Previously, this has been done with the grid resistor taken to a variable negative supply, or by taking the cathode positive. In the new circuit, the negative voltage developed at the grid is used for the control voltage itself. By adjusting the DC resistance of the grid circuit, so too is the DC voltage at the grid. The 1M rheostat provides this adjustment. The .22uF bypass ensures that the 1M is isolated from the quench time constant components, and does not directly affect the quench frequency.
A point of note is that the voltage across the .22uF gives an indication of relative signal strength. While I have not tried it, an interesting experiment would be to connect a magic eye to this point.
As with other super-regenerative receivers, the grid voltage is made negative enough so that the stage is only just oscillating. This is the point where sensitivity and output is greatest. The quench frequency is controlled to some degree also, which is useful for reducing beat problems with the stereo subcarrier or certain program material.
The quench amplitude is about 5Vp-p at the 6GK5 plate. If not filtered out, it will overdrive the following audio stages, preventing full output. Simple low pass filtering is used which consists of the 100k and 560pF.
It may be noted that the volume control is 220k instead of the usual 500k or 1M used in previous circuits. This was simply because I had more 220k's than 500k's. It's true that the audio developed across the 560pF will be lower with the 220k pot, and also the quench filtering slightly less effective, but in practice the difference is hardly noticeable. However, anyone wishing to duplicate this circuit should use a 500k or 1M pot if using new components.
The 6X9/ECF200 has been described previously, and suffice to say the circuitry is conventional. The triode is a voltage amplifier driving the pentode as the output. Cathode bias is used for both sections. The 1000pF across the pentode grid resistor provides further quench filtering - necessary if full output is to be obtained.
I used a 7k to 3.5R Rola speaker transformer to drive the 4R 4" speaker. If one is buying new parts, a transformer with a primary closer to 11k will provide a better match.
The 140V AC from the transformer is full wave rectified using a 1A bridge. Previously, with this transformer in other projects, I used half wave rectification (as the transformers were used in their original application) but it produces a slight but annoying buzz in the transformer laminations. Full wave rectification also provides a higher B+ and is more easily filtered. The filter capacitors were obtained from dismantled equipment. Simple RC filtering is used. B+ current is about 18mA. Voltage at the first filter is 173V, with the mains at 240V. At 215V mains supply, the B+ falls to 154V.
Extra turns visible on the power transformer add to the 5.8V to obtain 6.3V.
The 6X9 heater is fed straight from the 6.3V AC. A half wave voltage doubler provides -11V which is regulated to -6.5V with a 6.8V 10W zener, 1N1602A. To drop the 200mV down to 6.3V for the 6GK5, a 1R resistor is used. The reason for the negative voltage is simply because the zener diode case is the cathode, and it avoids having to insulate the body from its chassis mounting. This negative voltage could also be used to bias the valves, but there would be no saving in components with the necessary voltage dividers, as well as decoupling capacitors being required. There is about 300mA flowing through the 15R, and regulation is maintained down to 215V mains supply. In case someone questions why there is only 6.5V across the 6.8V zener, it's because of the spread in characteristics, and also the zener is not operating at anything like its full current.
My preference is to run separate earth wires to each valve heater, so that the chassis does not carry any heater current, and possibly cause hum problems.
Under chassis wiring. DC heater supply components are to the right.
This view shows how the chassis slides into the cabinet.
Rear panel view.