6GK5 Super-Regenerative Receiver for 88-108Mc/s.


Test chassis for the 6GK5.

The inspiration for this project came about from a reader who had constructed my 12AT7 VHF receiver, but had used 3GK5's instead. This got me thinking: The 6GK5 is a common TV valve in Australia. I have already described a simple single triode VHF receiver using a 6C4. However, 6GK5's are much more common than 6C4's in the present day, so why not see if they're suitable for super-regenerative use?
Additional interest in using this valve is because it's available with different heater voltages. There is the 2GK5, 3GK5, and 4GK5 for series heater circuits, as well as the ordinary 6.3V 6GK5. These lower voltage heaters would be convenient where the heater voltage needs to be regulated with a simple resistor and zener diode circuit, and the supply is already 6V DC. In particular, I had been thinking of improvements to the Model T Ford car radio where battery voltage fluctuation can sometimes affect performance of the 12AT7 super-regenerative detector.
As it turned out, the 6GK5 is an excellent performer in a super-regenerative circuit.

Introducing the 6GK5.
The 6GK5 belongs to a family of frame grid triodes especially designed for RF amplification in VHF TV tuners. It was released in 1961. The intention was to simplify the RF amplification stage, which until now had either used a pentode, or a twin triode in a cascode circuit.
By the time TV came to Australia, pentodes were obsolete for this application, and the twin triode cascode circuit was standard. In 1956, most Australian TV sets were using 6CW7 or 6BQ7 cascode RF amplifiers. Triodes have superior noise performance to pentodes, hence the popularity of the cascode circuit. Ordinarily, using a triode as an RF amplifier is difficult, because of the grid to plate capacitance causing it to oscillate. By using a cascode circuit, oscillation is prevented because the 'lower' triode is feeding a very low impedance plate load, which is the 'upper' triode. Because the 'upper' triode is connected in grounded grid formation, it is also prevented from oscillating.

The 6GK5 and its variants is used singularly, but because of its special design, and the circuitry used with it, it is stable for VHF amplification. This is achieved by a shield between the grid and plate, reducing the capacitance here to 0.5pF. Because the shield is reminiscent of the beam forming plates in a beam tetrode, this kind of valve is sometimes called a 'beam triode'. It has also been called, particularly in U.S. literature, a 'guided grid triode'.


The grid-plate shield is visible, adjacent to the plate.

Additionally, the RF amplifier stage is used in a neutralised circuit - a throwback to the Neutrodyne circuit of the 1920's. The frame grid construction results in very high gain, which overcomes the loss of one triode in the normal cascode circuit.
The result is a simpler and cheaper TV tuner design, with good noise performance.


Cross section of the PC97 which illustrates the internal shielding.

Other valves in this family include the previously mentioned 2GK5, 3GK5 and 4GK5. In Europe the PC97 (4FY5) was standard, with the PC900 appearing later. The PC97 and its 6.3V heater version EC97 (6FY5) were used in certain Australian made Thorn sets, but otherwise, Australian sets used the 6GK5 and variants. National and General Electric with their series heater circuits used 3GK5's  or 4GK5's in their tuners. 6ER5 and 6ES5 are similar types, but uncommon in Australia.
The 6GK5 is also equivalent to 6FQ5, and is sometimes sold under both type numbers. For a good article on these valves, see Practical Television, October 1961. This is available on the American Radio History (now World Radio History) site.


From the RCA Receiving Tube Manual.

2GK5 has a 2.3V 600mA heater. 3GK5 has a 2.8V 450mA heater, and 4GK5 requires 4V at 300mA. The 4GK5 is not shown in the RCA data, presumably because it is a 300mA series heater variant, and U.S. made series heater TV sets used only 450mA or 600mA heater strings. Pin 6, designated "IS" (Internal Shield), is the shield between grid and plate; not a shield surrounding the plate as is usually the case.
Note that the 6GK5 is a variable mu (remote cut-off) triode since it is designed to work with AGC circuits. This also opens the possibility to use it for audio applications where the gain needs to be controlled by a DC voltage.


Using the 6GK5 as a Super-Regenerative Detector.
As one who uses valves for circuits which they were never designed for, the 6GK5 seemed a bit of a paradox. A super-regenerative receiver is supposed to oscillate, yet the 6GK5 is designed not to!
Of course, there's no actual reason it can't be used in a conventional oscillator circuit, since we won't be neutralising it, and the low grid to plate capacitance is unimportant here anyway. An interesting thought is if the variable mu characteristic would affect operation as an oscillator.
To start with, testing was based on my faithful 12AT7 circuit, which has now proved itself for something like 18 years. It is stable, predictable, has high sensitivity, and provides good audio quality. Many published single valve super-regenerative circuits are comparatively poor performers, such as this. Furthermore, many such designs require that both sides of the tuning capacitor be isolated from earth, which causes constructional difficulties. Those who extoll the virtues of these designs never seem to provide an actual measured sensitivity test, and assume that since it can pull in various stations on a short aerial, that it must be sensitive, when it is in fact not. For local station reception, sensitivity does not actually need to be very high, as evident by the various FM crystal set designs, and the Fremodyne which requires 200uV for noise free reception.
It was because of this frustration that I did a lot of work developing the 12AT7 circuit, which has been used in all my valve super-regenerative receiver designs since.

The first test was simply to try a 6GK5 in place of the 6C4 in the Improved One Tube FM Receiver. The valve socket connections were suitably changed, and a 6GK5 inserted. Unfortunately, it was a very poor performer with severe modulation hum, and oscillation seemed very poor. Evidently, the 6GK5 was not a simple drop in substitute. The poor oscillation was curious since the 6GK5 has a much higher gain than the 6C4. It must be remembered that the Improved One Tube FM Receiver is not really following good VHF construction methods, with no proper ground plane.

A proper test chassis was made up out of aluminium with everything rigidly mounted and solidly earthed. This time things worked a lot better, and it seemed that if the modulation hum could be removed, the 6GK5 could be used.
Disconnecting the 6.3V AC heater supply immediately removed the hum (a 50 cycle sine wave with the audio superimposed on it), and the audio was clear until the cathode cooled down. Trying a DC heater supply provided hum free reception. Increasing the heater bypass capacitance and introducing an RF choke into the heater supply did not fix the problem. One forum post I found mentioned the 6GK5 might be prone to heater-cathode leakage. This could certainly cause the problem, and so I tried another 6GK5. Hum gone!
This does not imply the first 6GK5 is faulty. It must be remembered that this valve is designed to have the cathode solidly earthed. In fact, two pins are provided for the purpose. In this situation, any heater to cathode leakage would have no effect on performance.
However, in the oscillator circuit I'm using, where the cathode is live at RF, and is part of the feedback circuit, then heater to cathode leakage could be problematic. This means that to use the 6GK5 successfully, it may be necessary to try another if hum is evident, or to use a DC heater supply.

The internal shield is not required for this circuit to operate, but it is not good practice to simply leave it floating, since it could acquire an indeterminate charge. There seemed to be no difference in operation when the shield was earthed, or if it was connected to the plate. Given that the plate pin is adjacent, this connection was chosen for convenience.

This was also a good opportunity to test some other aspects of the circuit. One of these was to see how feedback, using a tap on the aerial coil, compared to using the RF choke, would work. If successful, it would allow simplification of the circuit by eliminating the choke. Construction would also be made easier, since it's one less coil that has to be wound up. A further object of the tests was to see if the hitherto used component values; e.g. grid and plate resistors, were optimum or needed to be changed.

To date, valves successfully used in this super-regenerative circuit are 12AT7, 6C4, and the triode of a 6DX8. (The 6C4 is equivalent to one triode of a 12AU7). So, how do they compare to 6GK5 in the valve data?
 
Type Transconductance (umhos) Amplification Factor
6C4 2200 17
12AT7 5500 60
6DX8 (triode) 4000 65
6GK5 15000 78
Comparison of valve types used in super-regenerative receivers described on this site.


Circuit Configurations.
Initial tests varied the plate voltage to adjust regeneration. While this works, performance is inferior to grid control, with regards to sensitivity and audio output level. This is exactly the same conclusion that was come to during the development of the original 12AT7 receiver. No further work was done using this method of regeneration control, and grid voltage control was returned to once again. Once the basic 6GK5 circuit was optimised, various configurations were tried; each with advantages and disadvantages. However, sensitivity and audio output remained the same for each variation. Output level is suited to a triode pentode audio amplifier, with Rin of 500k to 2M.
Not shown on the following circuits is the heater, connected to pins 3 and 4. One side of the heater was earthed, and the active side bypassed with a 2200pF capacitor.
Briefly, the component functions are as follows: The circuit is a simple self quenching type, with the quench frequency set by the 270k and 33pF grid leak circuit. Plate load where the audio is developed is 150k. The plate is bypassed at RF by the 1000pF. The 36KHz quench frequency is filtered by the 100k and 560pF.
Oscillation is caused by a quarter wavelength RF choke in the cathode circuit, along with the grid to cathode capacitance. Alternatively, positive feedback can be provided by the cathode connected to a tapping on the aerial (oscillator) coil. Oscillation frequency is set by a four turn coil which is resonated by the adjustable parallel capacitance. Regeneration is controlled by taking the grid resistor to a variable negative supply. As the grid is taken more negative, oscillation is weakened to the point of cut off. A variation of this is to instead take the cathode positive, since the grid will again be negative, with respect to the cathode.

Circuit 1.

Regeneration control using cathode rheostat.

Since many of the previously built receivers use a cathode rheostat to adjust regeneration, this was tried first. Performance was good, and as expected. There is, as with the other receivers using this method, some degree of backlash with the regeneration control. If the receiver is adjusted for maximum sensitivity, and it drops out of oscillation, the control has to be advanced further until oscillation re-commences, and then backed off to its original position. The degree of backlash seems greater than the 12AT7 circuit using the same method of regeneration control. This is probably due to the higher gain of the 6GK5.
It is, however, the simplest of all the configurations, and does not require a negative voltage supply. Until now, the rheostat has been bypassed by separate 1uF and 1000pF capacitors. The reasoning behind this was that where the 1uF is electrolytic, the 1000pF is also required for full RF bypassing, since electrolytics are poor performers at RF. It has been found that a .22uF ceramic type provides sufficient bypassing for the RF, quench, and audio components.


Construction of test set up with cathode rheostat control. Notice the 1.5pF aerial coupling capacitor made from insulated wire twisted together.

Circuit 2.

Based on the original 12AT7 circuit, grid voltage controls regeneration directly. Providing the -20V supply is clean, the .22uF could be omitted.

Given the backlash characteristic of the previous circuit, the method used with the original 12AT7 receiver was tried, since this has a very smooth regeneration control with no backlash. Indeed, the same was found with the 6GK5. However, a negative 20V supply is required. Bleed current through the 10k pot is 2mA. If less bleed current is desired, the pot value can be increased. In the original 12AT7 receiver, a 500k pot is used.


Construction showing coil tapping instead of the RFC. The 3.3M resistor was at this stage connected to a variable power supply. The output filter resistor is 150k, but was reduced to 100k to improve treble response.

Circuit 3.


Simplified version of circuit 2. The .22uF should be retained since the resistance of the 10k pot would otherwise alter the quench frequency, when it is adjusted.

Instead of feeding the 0 to -20V into the grid via a 3.3M resistor; the control voltage can be fed directly into the earthy end of the 270k grid resistor. Since there is no longer the voltage divider effect of having 3.3M in series with 270k, the negative control voltage is considerably reduced. Now, only -2.5V is required for the negative supply. The .22uF ensures that the grid resistor does not 'see' the resistance of the negative supply and the 10k pot, which would affect the quench frequency. As far as could be ascertained, performance is identical to circuit 2, and there is also no regeneration control backlash.

Circuit 4.

This circuit eliminates the RFC.

Tests were done to see if the RF choke could be eliminated by using a tapping on the aerial coil instead. The idea was successful, and performance appeared identical to the circuits using the RFC, except some degree of backlash did exist, although not as much as circuit 1, using the cathode rheostat.

Circuit 5.

Simplification of circuit 4.

Again, it was found that the regeneration control could be simplified by eliminating the 3.3M resistor, and the negative supply reduced to -2.5V. The negative supply current is 250uA, which is the bleed current through the 10k pot.
Performance appeared identical to circuit 4.

Circuit 6.


Self generating control voltage.

Given that a negative voltage is already developed in the grid circuit, thoughts were that it might be possible to use this voltage by itself to control the regeneration. Doing so would eliminate the need for a separate negative supply. It was successful, and this method of control appears to be quite legitimate. As typical with this type of oscillator, the grid leak components, along with the diode formed between the grid and cathode, cause a negative voltage to be developed while ever the circuit oscillates. By increasing the DC resistance of the grid circuit, negative grid voltage increases. Because increasing the grid resistance also decreases the quench frequency, it's necessary to provide a bypass so the quench time constant does not see the extra resistance. This is the purpose of the .22uF.
The circuit was found to be completely free of regeneration control backlash, and there appeared to be some element of automatic stabilisation. It was found that with the regeneration adjusted optimally at 108Mc/s, that sensitivity was much the same at the 88Mc/s end of the band.
The circuit can be used with a cathode tapping on the aerial coil instead of the RFC. One thing noted was the quench frequency was lower when the RFC was used, requiring the existing grid leak resistor to be reduced to 180k.
It was noted that the heater voltage was more critical with the cathode tapping instead of the RFC. While the circuit continued to oscillate with a low heater voltage (e.g. 5.7V), sensitivity had dropped off. Quench frequency is also dependent on heater voltage. With the RFC, the sensitivity was not affected by the heater voltage, and normal reception was still obtained with 5.8V on the heater. However, the regeneration and tuning controls had to be readjusted when the heater voltage was decreased.

One might wonder, how cut off can be achieved since there will be no negative voltage when the circuit stops oscillating. Indeed, complete cut off can't be obtained, but this isn't actually necessary. What happens if the 1M pot is taken to full resistance, is that as the valve approaches cut off, a fast ticking sound is heard. This is the time constant of the 1M and .22uF causing the circuit to squeg at a frequency much lower than the normal quench frequency. The 'feel' of the regeneration control is different to the previous circuits, with a less defined control characteristic.
An interesting observation is that the voltage across the .22uF shows a dependence on signal strength. Because of this, there is a limit to the apparent automatic stabilisation. Setting the regeneration for maxium sensitivity on a weak signal, and then tuning to a strong signal will increase the negative control voltage, which may drive the receiver into cut off. Similarly, because there is an audio component in the grid circuit, strong audio peak voltage can also cause cut off. This is not a limitation, since with such a strong signal, the regeneration does not need to be set close to cut off.
Fed straight into a 6M5 output valve, driving an 8" speaker, circuit 6 provided reasonable volume for a quiet room. In fact, the audio quality and sensitivity was quite good with the weak test station (2NUR-FM, 135km distant). Given a choice, the use of the RFC is preferred over the cathode tapping, especially if the supply is unregulated.

Circuit 7.


Cathode pot with bleed current.

For the sake of completeness, regeneration control using a cathode pot with bleed current was tried. This method has been used in some of the 12AT7 receivers. The idea is that by having a lower value of pot; e.g. 1k instead of 10k, the backlash might be less. Since 1k on its own does not allow full control voltage to be developed, bleed current provided by a 47k resistor is provided. The voltage across the pot is now about 3V, and in the test receiver, optimum regeneration was with the wiper at about 1.8V.
Results were that this circuit is not worth using. Backlash is about the same as circuit 1, and the extra current consumption provides no benefit. This circuit requires an extra 3.1mA from the 150V supply. For circuits 1 to 6, B+ current is less than 1mA.

Which Circuit to Use?
This is going to depend on whether or not a negative supply is available, and if it's desired to eliminate the RFC for reasons of space, or that the constructor has an aversion to winding coils.


Coil Data.
Aerial coil:  4 turns of 18 gauge tinned copper or silver plated wire. 10mm diameter, air cored. Tap at 2 turns for circuit 4 and 5.
RFC:  75cm of 26 gauge enamelled copper wire wound on a 6.5mm plastic former.

Tuning Capacitor.
A 15pf tuning capacitor is preferred. In the test set up, the 60pF local oscillator section of a MW air spaced tuning capacitor was used, with a 33pF series capacitor. While this provides full band coverage, the operation is non linear, with stations at the 108Mc/s end of the band closer together than at the 88Mc/s end. The test receiver tuned 74-112Mc/s. Air spaced tuning capacitors are preferred over plastic dielectric types or varicap diodes, since these have a lower Q.

RF Amplifier.
With the success of the 6DX8 One Valve FM Receiver, tests with the 6GK5 were done without an RF amplifier. The aerial is coupled into the aerial coil via a 1.5pF capacitor. This can be a ceramic type, or as in the tests, two pieces of insulated wire twisted together to obtain 1.5pF. So that the aerial sees something like the correct impedance, a 100R resistor is connected at the input. Ideally, this should be carbon to avoid inductive effects.
As will be noted with all the 12AT7 circuits, one triode functions as an untuned grounded grid amplifier. This has little gain, but provides isolation between the aerial and tuned circuit. In terms of sensitivity, there is little to choose between having an RF amplifier and not having one. The main effect is that the regeneration is more consistent across the band, and not dependent on the aerial. Additionally, with no RF amplifier, the aerial characteristics can affect the performance by the amount of capacitance presented to the receiver. For example, it may be observed that connecting an aerial causes the stations to shift slightly, and the receiver has to be re-tuned. Also, particularly in the case of portable receivers, touching the aerial or bringing it near other objects can cause the receiver to drop out of oscillation.
Thus, this method of aerial connection works best with fixed aerials, such as that which might be mounted on a car or house.
Some may argue that not having an RF amplifier allows the receiver to radiate. This may well be the case, but such radiation does not interfere with the reception of the receiver causing it. The chances of someone listening to the same station on a nearby receiver is minimal anyway.
If it's desired to include an RF amplifier, see the 12AT7 circuits. Another 6GK5 can, of course, be used to provide this function by using it in the same grounded grid mode. It probably matters little as to what the internal shield is connected to, but earthing it would be the 'correct' thing to do, when the valve is used as an RF amplifier.

Regulated Power Supplies.
For best performance, the B+, heater, and negative supply should be regulated. This is particularly so when the receiver is adjusted for maximum sensitivity, and only just oscillating. A slight drop in B+ and/or heater voltage can cause the oscillation to stop, as can an increase in the negative supply (where used). Another characteristic worth noting is that the grid voltage does affect oscillation (i.e. receiving) frequency slightly. Since the required B+ is 150V, an obvious way to provide this in a regulated form is with a type 0A2 valve.
A regulated B+ supply (BWD 215) was used for the tests, but when the receiver was set for maximum sensitivity, ordinary fluctuations in the mains voltage were enough to upset the heater voltage, to the point where the receiver would drop out of oscillation (except for Circuit 6). This was eliminated when the heater was connected to a regulated 6.3V supply.
Circuit 6 seems to be less affected by B+ supply voltage variation than the previous circuits, probably due to the degree of automatic stabilisation provided. The complete absence of regeneration control backlash also means that if oscillation stops, it will recommence at the same level once the supply voltage is restored, without having to re adjust the regeneration.


Waveforms.

1. Plate at Quench Frequency.

This shows the quench waveform at the plate. Frequency is 36kHz at 5.4Vp-p.

2. Plate at Audio Frequency.

This shows the waveform at the 6GK5 plate, and as can be seen the quench is of much greater amplitude than the audio. For this reason, filtering is required to prevent overload of the following audio stages.

3. Filtered Audio.

Here the filtered audio is visible. Connection is across the 560pF capacitor. Output is roughly 600mVp-p. If more effective filtering is required, an active filter should be used, as with the Mains Operated 12AT7 Super-Regenerative Receiver.

4. RF at Quench Rate.

This is the RF oscillation. The CRO was connected to the aerial terminal. As can be seen, the oscillation is occurring at the quench frequency (36kHz). This is a good illustration of how a super-regenerative receiver has good noise immunity. Noise that occurs between the bursts of oscillation are ignored, since the receiver is not sensitive during this time. This waveform also shows why a basic super-regenerative receiver radiates.

5. RF burst.

Here, the previous waveform has been expanded out to show the shape ot the RF burst. Note that it builds up and fades out gradually.

6. Expanded RF Burst.

The RF burst has been expanded here to show the RF oscillation, and thus, receiving frequency. This is at the start of the burst, and the waveform can be seen to be rising in amplitude.


A Trap for the Unwary!
It was subsequently discovered that the two 6GK5's used to develop this circuit were both weak. This was noted when the receiver was being built up in a permanent form, and another 6GK5 used. The problem was that the quench waveform amplitude was excessive, and the audio level was not as high as the test receiver. Several more 6GK5's were tried, which confirmed that the 'test' 6GK5's were different. The valves were put through the AVO valve tester which measures transconductance.
Only one of the test valves was available, and its plate current was 3mA, with the transconductance 3mA/V. The specifications are for a plate current of 11mA and a transconductance of 15mA/V.
Effectively, I had optimised the circuit for weak 6GK5's! The other valves which I had since gathered from my collection, with the exception of one extremely weak example that couldn't even super-regenerate, had much higher plate current and transconductance. For comparison, two new 6GK5's were also tested and were within specifications. There was quite a spread of characteristics with the valves tested - which except for two, were all used.
It goes to show that during circuit development, more than one valve should be used to test operation. It's just unfortunate that both valves I used were weak, which means one should try a lot more than two.

What to do about it? Obviously, the higher gain of the good valves was too high. A simple way to reduce the gain is with an unbypassed cathode resistor. Fortunately, this did solve the problem. By comparing the plate waveform of the weak, but correctly operating 6GK5, the resistance was adjusted so that with a strong 6GK5, the waveform was the same. It turned out that 470R was a suitable resistance.
Therefore, all of the circuits shown above should have a 470R resistor in series with the cathode choke, or between the cathode and coil tapping, as the case may be.

With the 470R in circuit, any heater to cathode leakage is going to be more apparent. One can either select valves to avoid this problem, or simply use a DC heater supply so that any 6GK5 can be used.
It is worth noting that of the valves tested, most were devoid of type labels, and were identified by appearance, internal construction, and heater characteristics. All the types which had an enclosed plate did not have any heater to cathode leakage problems.
Quite possibly, these are actually 6FY5/EC97 types. The pin connections are the same, and the test characteristics are virtually the same, so it is not possible to differentiate between a 6GK5 and 6FY5 using a valve tester.


Performance.
 
Input Signal 103.5MHz, AM 50% mod. Quality.
1.5uV Detectable but not intelligible
5uV Noisy but intelligible
15uV Noisy but entertainment value
20uV Almost noise free
50uV Noise free

Sensitivity was tested initially at 103.5MHz. Since the receiver is AM, this mode of modulation was used to determine the best case scenario, and the results are tabulated above. The question of FM sensitivity is harder to define, since the receiver is not tuned to the peak of the response curve, for optimum reception. However, it was found that the minimal detectable signal with FM of 40KHz deviation was 1.9uV. Above 5uV or so, the difference between AM and FM in terms of input signal to quiet the receiver was not huge. Nevertheless, FM sensitivity will always be less than AM because of the slope detection.
At 88.5MHz, sensitivity was about 3uV down from 103.5MHz. This is to be expected with the simple form of aerial coupling, since the reactance of the 1.5pF capacitor is a lot higher at the low end of the band. This can be improved upon by increasing the coupling capacitor, but an RF amplifier is then required to isolate the aerial loading. Again, see the 12AT7 circuits to see how this is done.


How FM is converted to AM by slope detection. This illustration assumes a superhet with diode detector and a 10.7MHz IF, but operation is the same for any frequency and any AM detector.

In conclusion, the 6GK5 has turned out to be very suitable for super-regenerative receivers. For utmost simplicity, it is possible to feed headphones from the audio output via a transformer. A multi-tapped transformer was tried with modern stereo headphones. As expected, the audio level increased with higher impedance tappings, and was greatest at the highest tap (30k). Since quench filtering is not required in this situation, a slight increase in volume was had by removing the 100k and 560pF. No real improvement was had by substituting the transformer for the 150k plate load. Even so, the volume is low, and no better than a crystal set fed with an average signal. It's probably best described as barely entertainment value, and more of a novelty to prove that VHF FM can be received and heard using only one valve.

Circuit 6 is preferred, and so far is proving to be an excellent performer. A tune across the band received all the Newcastle stations with good quality. These are 135km distant. Similarly, many south coast stations were well received. Needless to say, all Sydney mainstream stations came in strongly. The regeneration control required little adjustment, and was not critical. It appears that in the ordinary scheme of things, an RF amplifier is not necessary with the 6GK5 circuits. The aerial used for testing was a five element Yagi as shown in this article.
 


Home