Controls from left to right are On/Off, Regeneration, Quench Frequency, and Volume.

With the exception of the 7193 receiver, all the valve super-regenerative receivers described so far have used miniature valves. For some time, I had contemplated the construction of a WW2 style 'all bakelite base' valve VHF receiver. The key to its practicality was the super-regenerative detector. Beyond that, the audio and rectifier circuits are easy with ancient valves. With the success of the 7193 receiver, I was finally motivated to begin construction, especially, as I was in the process of designing a radio for the Jeep. For that, I wanted to keep in with the WW2 theme where possible. Building up a complete receiver would enable me to test the design as a whole, before getting into the intricacies of the Jeep receiver. The main electrical difference was of course, the test receiver was to be mains operated.

As such, this receiver is designed around valves which were standard in WW2. The 7193 was actually developed for WW2 VHF use. It was a development of the 6J5. The audio amplifier I used is a 6H8C, which is the Russian equivalent of the 6SN7. The output valve is an 807, which appeared at the beginning of the war, and is a development of the 6L6. Finally, the rectifier is the ubiquitous 5Y3.

Despite the WW2 design, there would be little point in building the receiver to cover the 42 to 50 Mc/s FM band, which existed at the time. Instead, it receives the 'new' 88 to 108 Mc/s FM band, introduced in 1946.

Rear view shows speaker terminals and aerial socket.

To save repetition, readers are referred to the two main circuits which make up this receiver:

• 7193 receiver.
• 807 output stage.

Under view. Even two of the potentiometers are the correct vintage. Note the sub-chassis for the earthy end of the 7193 circuit.

I decided to use all "Junk box" parts, since it was really just a test receiver. However, it worked so well that I decided to put a bit more effort into it. I'm very pleased with the results; not just in appearance, but the performance and sound quality. The chassis came from the free pile at the HRSA some years ago, and the top was quite rusty. It once a had a 2" CRO built on it; I suspect a kit, given the quality of the labelling and hole punching. Wire brushing it and then painting in grey epoxy made it quite presentable. All the other parts were randomly taken from my stock. The dial is period correct, being a Jackson 6-36. This incorporates a dual ratio drive; 6:1 and 36:1. Tuning is exceptionally smooth..

The Circuit.
 
 

Circuit of the receiver. Note the older style valves.

Super-Regenerative Detector.
Heart of the receiver is of course the super-regenerative detector, which is built using the successful 7193 circuit. This is the recently improved version, which includes chokes for the heater supply. These provide a much stronger oscillation, which allows a tighter aerial coupling and more effective operation of the stabilising circuit. The three chokes are the usual quarter wave design. That is, 75 cm of 26 (approx.) gauge wire wound on a 6.5 mm or thereabouts, plastic former.

My now standard grid leak regeneration control is used here. The automatic stabilising characteristic means that the adjustment is non-critical, and the detector performs consistently over a wide supply voltage without requiring adjustment. Also, for most of the time, no adjustment is required, as the receiver is tuned from one end of the band to the other. In its present form, it does not allow for complete cut-off of the oscillation, but I have not found any reception conditions which actually require that. Astute readers will note the regeneration pot is 1M instead of the 2M usually used previously. The 1M pot was to hand when I built this receiver, and so far its control range has been found to be adequate. One difference I have incorporated was an adjustable grid leak. This allows quench frequency to be adjusted over an extra range, which the normal regeneration control does not allow. The previously used 270k resistor is now comprised of a 150k isolating resistor and 250k pot. Adjustment is thus 400k to 150k. With the control in mid position (125k), the grid leak is 275k. This mid position setting provides the same performance as the original fixed resistor. In actual fact, the resistor in the test receiver measured 300k. As it happens, the use of a 1M regeneration pot means that it too is roughly centred, to obtain the original test receiver performance. In that set, the total grid resistance was 811k for best reception. We can see that with both controls in mid position (500k + 125k) and the 150k isolating resistor, the total grid resistance is 775k, which is close enough to the operating conditions of the test receiver.

The reason for including the 150k resistor, and not just using a pot alone, is because of the layout of the receiver. To extend a wire from the 7193 grid down the chassis, to the pot, could cause all sorts of problems, and would be poor VHF practice. The grid is live at VHF, and the inductance and stray capacitance could cause strange results. By including a portion of the required resistance right at the grid connection, only the quench frequency travels down to the pot. This being virtually in the audio spectrum is not critical. Beyond the frequency control pot, the quench frequency is bypassed by the 0.1uF, so the wiring to the regeneration control is also non critical.

Close up of the VHF part.

It took quite a lot of thought as to how to mount the 7193. There wasn't enough chassis room to mount it horizontally, as it had been in the test unit. The grid cap of the 7193 needs to be near the tuned circuit. It all came down to the position of the tuning capacitor being determined by the dial design. From here, a sub chassis was made to support the tuning capacitor and coil, mounted to the side of the 7193. Under the chassis, another sub chassis supports the valve socket, and the three chokes. A short length of braid earths this to the chassis. A connection also exists through the brass spacers which support the sub chassis.

Voltage Amplifier.
Since the 7193 only outputs a few hundred millivolts of audio, a reasonably high gain audio stage is needed to drive the output stage to full power, with some reserve. For the correct 'feel' of the volume control, the output should start to overload with the volume control a bit over half way up. This gives sufficient reserve for very weak signals.

Keeping with the WW2 theme, my first idea here was to use a 6J8 triode hexode frequency converter as a two stage audio amplifier. Radio & Hobbies had been using this valve in a similar manner with some of their small receiver designs. The hexode was operated as a pentode regenerative detector, with the triode as an audio amplifier. Of course, the 6J8 was never designed to operate like this, but the more appropriate 6F7 was not made in Australia, and with the onset of war became very difficult to get. Hence, the use of valves in applications they weren't intended for. In essence, the scheme apparently works well enough, because the triode grid coupling into the hexode electron stream (at grid 4) is not high enough to be problematic. At first glance the 6K8 could be used similarly, until one takes a closer look. Here, the triode grid injects into grid 1 of the hexode. In this instance. the interaction between the triode and hexode would likely be too great to make the scheme practical.

Anyway, a circuit was made up with the 6J8. The hexode section worked quite well, but troubles were found with the triode. The triode grid was producing a positive voltage with nothing connected to it! To cut a long story short, there was an internal short between the triode grid and cathode. Since I had very few 6J8s in my collection, I felt it unwise to pursue this, since I might need them for more important replacements. A single 6J7 pentode worked quite well, but there wasn't the reserve of gain required for a car set. What other two stage valve could we use? The obvious choice was a 6SN7. The only disadvantage is twice the heater current of the 6J8. It's true that a 6SL7 could be used with its 300mA heater current, but again, that's an obscure type in my collection. I've got a lot more 6SN7s.
The actual valve used is the Russian version of the 6SN7, designated 6H8C. I used this simply because I had a couple within easy reach.

As to the design of the circuit, I referred to the RCA Receiving Tube Handbook, which contains charts for the plate, cathode, and grid resistors of various valve types used as resistance coupled amplifiers. A circuit was built up using the 6SN7 data. Gain per stage is about 15, so the entire stage gain is about 225.

Normally, I would put the volume control at the input to the voltage amplifier, but with the high gain, I felt it better to put it between the two triodes. Importantly, with the first stage always operating at full gain, the output of the 7193 is not high enough to overload it.
The advantage of placing the volume control in the middle is less noise and hum. If the volume control is at the input of the first stage, you are getting the full noise and hum of that stage, even with the volume control at minimum.
The choice of cathode bypass and coupling capacitors are not critical. I used what was to hand. Too high of a value can lead to low frequency instability, and too low can result in reduced low frequency response. The values I used are very typical.

This stage worked well immediately, without having to change anything or dealing with any instability problems. The gain measured exactly as per the RCA data.

Output Stage.
This used my recently designed 807 circuit using a 100V PA line transformer, with cathode feedback. Two changes were made, however. Since I had no more M-1120 line transformers, I had to use something else. Going through my speaker transformer collection, I found a couple of unbranded 100V line transformers rated at 8W. I can't remember exactly what they came from, since I've had them for about 30 years. An interesting feature is that power output is determined by tappings on the secondary, rather than the primary. As long as the ratio is correct, this should work just as well. The 8W rating would suit the 807, provided a suitable impedance ratio was available.

Top view shows all the above chassis parts. Note the 100V line transformer.

A slight change to the 807 operating conditions was necessary because of the power supply limitations. The original circuit required 265V at 80mA. With the power transformer I decided to use, the B+ voltage was likely to be around 230V, and limited to 60mA.
With the particular line transformer, the highest non distorted output was with the 2W tapping feeding an 8 ohm load. Here, we had 11.7V p-p into the load, which equates to 2.13W.
This is quite a drop from the 5W of the original circuit, but not surprising. The difference between 235V and 265V might not seem like a lot on paper, but the power output is very voltage dependent. I first learned this with the Model T radio, using 6AQ5s. As well as the B+ voltage reduction, the cathode current had to be also reduced from 80mA to at least 60mA. Furthermore, the 100V line transformer I used is not an M-1120. The available impedance ratios are not identical, and it is not known to what extent the DC saturation problem might exist.

A recommendation for the 807 is screen and control grid stopper resistors to prevent any oscillation. Here, I have used 100 ohms for the screen grid, and 5.6k for the control grid. With the lower power operating conditions, it probably wouldn't be a problem, but is good practice to include these components. Remember, the 807 has a transconductance of 6 ma/V, which is fairly high for this era of output valve. Incidentally, anyone following my notes on the 807 will know that the 6L6 can be used in this circuit without any electrical changes.

Ultimately, best performance with the 235V B+ supply, was with a 5k load impedance, and -14.4V bias. As a result, the cathode resistor now needs to be 360 ohms. Since this is not a preferred value, I used 330 ohms. Two 180 ohm resistors in series would provide the exact value. Normally, it would be bad practice to err on the lower value, but there is absolutely no danger of exceeding any ratings of the 807. Total cathode current is 43mA.
So, we can see that with an actual plate to cathode voltage of 221V and cathode current of 43mA, the 2.13W into the load is pretty good, especially with the 'wrong' transformer. It is well to remember, this is the power into the voice coil, instead of the speaker transformer primary, as valve data usually states. This is why, for example, you'll never actually get 4.5W from a 6V6 into a voice coil, despite what the data says.

Overall sensitivity of the amplifier from the grid of the first triode is 42mV rms. It might seem overkill to use such a large output valve, when a 6V6, 6F6, 42, etc. would do just as well with these operating conditions. But - I have a lot more 807's than any of those other types!
I also wanted to prove the 807 circuit in a built up form. One other thing, which might see me using more 807's/6L6's in the future is the sound quality. I first noticed this when I built up a six valve TRF receiver about 30 years ago, using a 6L6. I certainly don't believe in most of the pomp that the audiophool crowd puts forth, but the 807/6L6 does sound very nice. In fact, I would say this is the best sounding medium power amplifier circuit, which works with PA line transformers, that I've yet heard.

Power Supply.
The power transformer is a N.O.S. type of unknown brand. Sometimes you wonder about manufacturers not being proud to identify their products. Like the line transformer, there's no name on it. It has a 240V primary winding with no tappings. This suggests it was made in Sydney, or possibly Brisbane.
The B+ secondary is 225V a side, at 60mA. The rectifier heater winding is 5V at 2A, and the 6.3V heater winding is rated at 3A.
To keep the 807 plate voltage as high as possible, the speaker transformer is fed from the 5Y3 cathode directly. While the ripple voltage is fairly high at this point, it does not mean the output is full of hum. That's because of the high plate resistance of a beam tetrode (or pentode). Instead, the filtering of the screen grid supply is what counts. Here, an ordinary RC filter is used with a 2.2k resistor and 32uF capacitor.
While some hum is evident with headphones, it is virtually inaudible with a speaker. An extra stage of filtering would be required to remove it completely, but the trade-off then is a further reduction in B+ voltage.

The 6.3V heater supply is run around the chassis as a twisted pair. One side is earthed at the 7193 socket. Using the chassis as an earth return for the heaters is not good practice, since a 50 cycle current then flows through the chassis, and can be induced into the earth returns of other circuitry in the receiver. The chassis is not of an infinitely low resistance as one might think.

Adjustable Quench Frequency.
This is the first receiver where I've made the grid leak a user adjustable control. Previously it has been necessary to select this resistor on test, with the value being selected to provide minimal intermodulation distortion, with the regeneration control adjusted normally. Certain program material might require the quench frequency to be shifted a few kc/s, to remove an annoying beat. On its own, the regeneration control will provide this adjustment, with the quench frequency increasing with an increase in regeneration. The question was, is it better to have a high value grid resistor, with a low quench frequency as default, and then increase the regeneration to increase the quench frequency as necessary? Alternatively, should we have a low value grid leak, with a high quench frequency as default, and use the regeneration control to reduce the quench frequency? Providing a user adjustable control would allow this to be determined over a range of signal strengths, and receiving frequencies.

Sound quality improves with an increase in quench frequency. The trade-off is that as quench frequency is increased, the sensitivity and audio output drops off. Generally, for FM stereo, a quench frequency of around 36 kc/s is about the lowest, to give reception reasonably free of the intermodulation products associated with the stereo subcarriers. Around 50 kc/s gives better sound quality with less distortion.

A further observation is that the stronger the signal is, the more problematic the intermodulation distortion and stereo subcarrier beats become. In other words, a higher quench frequency gives better sound quality for stronger signals.

Regeneration.
Another factor is that adjusting the regeneration for highest sensitivity can result in overload on strong signals. This shows up as more broad tuning, unable to find the best point for good slope detection. This means the regeneration has to be increased. And, as it happens, the quench frequency increases, improving the sound quality.

The method of regeneration control I've used with all my valve super-regenerative receiver actually works out very well. That is, regeneration is controlled by adjusting the grid bias of the detector. Since quench frequency increases with the amount of regeneration, it so happens that with weak signals the regeneration is minimum, which provides a low quench frequency, and highest output. Being a weak signal, the intermodulation artefacts resulting from a low quench frequency are not problematic. Conversely, with a strong signal, when the regeneration is increased to prevent overload, the quench frequency is raised, so that the intermodulation artefacts which would be problematic are automatically reduced. The otherwise reduced audio output is counteracted with the increased signal strength.

With all that being said, my conclusion is the adjustable grid leak is not really necessary. If the grid leak is selected for good reception on a weak signal, it is only necessary to adjust the regeneration for best reception. With this receiver, the quench varies from 25 to 70 kc/s with adjustment of the regeneration control. This is with the quench frequency in the minimum position (grid leak = 406k). With the quench frequency set to maximum (grid leak = 150k), the regeneration control varies the quench frequency from 23 to 155 kc/s.

Performance.
Frequency coverage with the 4.7 to 32pF tuning condenser, and the 4 turn 10mm diameter coil, is 85 to 147 Mc/s. Sensitivity is such that 1uV with a deviation of 40 kc/s can just be heard. The signal needs to be up around 10uV to become listenable.
There is a slight reduction in sensitivity at the low frequency end, because of the capacitive aerial coupling.
Putting some effort into a good quality audio amplifier, and using a good speaker, has paid off with this being a very good sounding receiver.