This was the natural follow-up to my one tube direct conversion set. I had the vague idea that a superheterdyne version of the direct conversion set might be possible, but hadn’t any immediate plans to build one until the question was put to me by Vladimir Novichkov on the Antique Radios Forum (where I had posted info about the direct conversion set). We discussed various circuit arrangements. This is the result. (Vlad also built one with some differences in his circuit.)
Debugging and alignment of this receiver was considerably more involved than with most receivers, including other superhet. Ironically, some of the characteristics which make its performance so good, also make the alignment more difficult. However, once everything was properly tuned up, its performance has been nothing short of incredible. I have been able to receive DX from 2280 km away, using a 30” dia. indoor loop antenna, driving a speaker with enough volume to be heard clearly across a small room.
Here is a preliminary schematic showing colour-coded signal paths:
Assume the radio is tuned to a station at 1000 kHz, and for simplicity we will assume that it is modulated with a 1 kHz audio tone. The RF (signal #1 - magenta) enters the antenna and passes through the primary of antenna coil T1. The tank circuit consisting of the variable capacitor and 51 turn secondary is tuned to resonate at 1000 kHz, and acts as a high impedance, essentially an open circuit. The signal passes to the 20 turn tertiary winding. For signals other than 1000 kHz, the tank circuit looks like a very low resistance load and shorts out those signals. Think of it as a shorted winding on a multi-winding transformer. Hence, the 1000 kHz signal is the only RF that comes out on the tertiary winding. It passes to the grid to be amplified. The other end of the winding, of course needs a ground return to be a complete circuit. It finds its way to ground through the 750pF capacitor which looks like a very low resistance at radio frequencies.
Now back to the grid. The RF signal is amplified in the tube.
Meanwhile the local oscillator (signal #2 - red) is busy oscillating at 1455 kHz. And its signal appears across the deflector plates. (More about the oscillator later.)
The amplified RF passes the deflectors and is mixed with the local oscillator signal to produce difference and sum frequencies of 455 kHz (the one we want) and 2455 (which we don't want). We'll ignore the 2455 kHz signal for now.
The 455 kHz signal appears at the tube plates as a differential (push-pull) signal (#3 - purple). It arrives at the push-pull IF transformer IFT-1a where it passes through to the ceramic filter where it is no longer a differential signal. No longer being a differential signal, we change its colour to light blue. Back on the primary side, the two phases of the differential IF have met at the center-tap and cancelled each other out. So, the purple IF signal disappears at that point.
Now, back at the ceramic filter, the IF passes through to the secondary of IFT-1b where it finds itself connected in series with the RF signal which we just discussed. Since voltages in series add up, this IF signal passes back (along with the RF) to the grid where it is amplified by the tube. It is also mixed with LO signal at the deflectors to produce sum and difference signals which we will ignore for now. But, the original 455 kHz IF arrives at the plates, amplified as well. However, this time through the tube it's not a differential signal, the signals at the plates are in phase. (I will explain why later.)
Since the parallel IF is in phase, it doesn't generate any signal in the secondary of IFT-1a, and it doesn't cancel itself out when it reaches the center-tap. In fact, these two in-phase IF signals add together and continue on to the 2nd IF transformer. It passes through to the 2nd IF secondary, where the signal is detected by the diode. The resulting detected signal consists of audio (signal #5 - green) and a DC component (signal #6 - brown) which will be used as AGC (automatic gain control). These two signals separate briefly. The audio passes through the volume control, and the AGC passes through a low pass filter. They recombine, and pass through the low pass filter consisting of the 33k resistors, and the 750pF capacitors which remove any remnants of IF. (The IF has been through the tube twice already, we don't want it going through again.)
At the output of the low pass filter, the audio and AGC finds itself added in series with the IF and RF which we've already seen. The audio and AGC goes along with them to the grid. The AGC provides variable grid bias for gain control, while the audio is amplified. The amplified audio appears at the plates as a parallel signal. And since it is parallel, it passes through the primary of IFT-1a and does not cancel itself out, but passes on through the primary of IFt-2 which is a low impedance to audio signals. It ends up at the primary of the audio output transformer where it is transformed and impedance matched to the headphones.
That covers what happens to the signals we are interested in. However, I mentioned that there are some other mixer products. In fact, every signal that enters grid 1 (including the DC grid bias) is mixed with the local oscillator. As well, all of the original signal components get amplified and end up at the plates of the tube at their original frequencies. The components which are at their original frequency are in-phase on the plates, while the sum and difference components will be 180 degrees out of phase the plates. Working through the standard trigonometric formulae for multiplying sines and cosines will demonstrate that this is true. However, I won’t go through those details here. Here are all the main components that are generated:
All of the original frequencies appear in-phase on both plates. The 1000 kHz RF is of no interest, and is filtered out by the IF circuits. They appear as a low impedance to the RF, and the 470pF capacitor across the audio output transformer also appears as a low impedance.
The 455 kHz IF and the audio have already been discussed. The DC bias accounts for the net plate current.
Now for the sum and difference components. The first thing to remember is that since they are out of phase, they will all be cancelled out by the time they get to the center-tap of IFT-1a. So, we don't need to consider them beyond that point.
Dealing with the sum components, their frequencies are all too high to affect the 1st IF, and disappear. The 1455 kHz signal is the same frequency as the local oscillator, and this is the positive feedback signal to the oscillator that keeps it oscillating.
Dealing with the difference components:
The 455 kHz difference signal is the desired mixing product that was discussed earlier. All of the remaining difference frequencies are too high to affect the 1st IF, and disappear. The 1455 kHz difference signal, as with the 1455 kHz sum signal previously discussed, is used as feedback to operate the local oscillator.
It's worth discussing the 1454 and 1456 kHz components that are generated by mixing the audio with the LO. These could contribute to some distortion in the final audio signal. They are close enough to the local oscillator frequency that when they pass through the oscillator coil primary windings, they could lead to a bit of frequency modulation of the local oscillator, resulting in FM distortion. I haven't noticed any significant distortion of the audio signal so far. Of course, this was never intended to be a hifi receiver.
Now, what happens in the cathode circuit?
The cathode current is equal to the sum of the two plate currents, plus some small DC components due to accelerator grid current. Because cathode current is the sum of plate currents, there will be negligible current components due to differential signals. Hence we don’t expect to see any local oscillator signal, or differential IF signal. However, there will be audio, common mode IF and RF.
The cathode is the ideal place from which to take an RF regeneration signal. Given that this circuit is already capable of IF regeneration using the plate balance adjustment, one might question the sanity of adding RF regeneration as well. Even so, I went ahead and tried it. The simple shunt potentiometer shown in the above schematic was not satisfactory, as it shorts out the tickler winding and kills the RF coil Q, when set to minimum resistance. However, A slightly different arrangement was tried, and was able to provide additional gain on weak stations in addition to what the IF regeneration was able to provide. There is some inevitable interaction with audio gain because there is no practical way to bypass audio from cathode to ground and still be able to direct RF through the tickler winding. However, the arrangement shown here in the partial schematic, allows maximum audio gain at both maximum and minimum settings of the regeneration control, with a slight amount of negative audio feedback in the intermediate positions. With proper component selection, this works very well.
The receiver was built on a chassis constructed of two side pieces of 5/8” thick MDF (medium density fiberboard), a piece of sheet metal for the top and rear, and a front panel of 1/8” hardboard.
The first IF filter is in the upper left corner, with IFT-1a at the top, and IFT-1b below it. the ceramic filter and the two trimmer capacitors are mounted on the underside of the two pieces of PC board. There are small holes drilled in the boards to allow screwdriver adjustment of the trimmers. The oscillator coil is at the top middle, with the extension shaft of the RF regen control passing through its centre. There is no particular electrical reason for the shaft extension to be positioned this way. Due to lack of space, it was the only convenient place to locate it. The audio/AGC circuitry is mounted on the terminal strip on the right side. The RF transformer T1 is the smaller toroid partly hidden under the regen control shaft coupler. The audio output transformer is at the bottom left. The power supply is external to the main receiver chassis.
These are parts of the homemade tuning dial assembly. They give about 16:1 reduction which aids in fine tuning of the stations.
Here, the tuning dial parts are partially pre-assembled to the back of the front panel.
Here is a rear view of the receiver with the front panel attached, speaker mounted, and tuning dial assembly installed. The speaker is a common 8 ohm unit, 2.5”x4.5” salvaged from an old television.
A closeup of the tuning dial assembly after installation.
In the course of getting everything to work, and then aligning the receiver, there were some minor changes made which are not reflected in the original colour coded diagram shown above. This will be updated sometime in the future. However, the actual schematic, as built is here: Final Schematic. The changes were minor. However, one change that deserves comment is the deletion of the 10 k variable resistor that was shunted across the output of the first IF stage. This had been added to reduce the IF gain in order to solve a severe instability problem where the receiver wanted to break into oscillation. By keeping the shunt resistance low enough, it stabilized the circuit and prevented the oscillation (which was slightly below the intermediate frequency). Unfortunately, this also reduced the IF gain drastically. With some further analysis, I concluded that the problem was related to the interaction between the tuned circuits of the first IF output, and the audio output filter. I isolated them by adding a 10 k resistor between them, and the problem was solved. With that change, the receiver became very stable, and IF gain increased significantly with no significant attenuation of the audio signal.
Alignment of the receiver was generally as follows:
Local Oscillator Adjustment - With the tuning capacitor plates fully meshed, the deflector bias potentiometer was adjusted for maximum amplitude of the local oscillator as measured with a scope probe at deflector 2 (pin 2). (Connecting the scope to this point of the circuit caused negligible disturbance, whereas connecting to either of the plates disturbed the circuit significantly. An unbalance between the tube plates of just a few of picofarads, can send the receiver into serious oscillation.)
First IF Alignment - The first IF input and output transformers had been given an initial adjustment to peak them to 455 kHz before they were wired into the circuit. So, they were expected to be reasonably close to optimum. The audio section of the receiver was disabled by grounding the output of the audio low pass filter (the point between the 10 k resistor and the secondary of IFT-1b), and the output of the first IF was isolated by breaking the connection from IFT-1b and the T1 winding. The scope was connected to the IFT-1b output at the point where it was disconnected. A low level 455 kHz signal was injected into grid 1. Unfortunately, this is where the amazingly good IF rejection of the beam deflection tube made the alignment a bit more onerous than it would have been in another receiver. In any other receiver, it would simply be a matter of injecting a 455 kHz signal almost anywhere and then peaking the trimmers on the IF filter. However, in this case, with the plates properly balanced, there was virtually no IF signal getting through. The solution was to make sure the plate balance differential capacitor was adjusted until the plates were slightly out of balance and then proceeding with the IF alignment. This was an iterative process, because as the IF was peaked, the IF gain would increase, and the circuit would want to oscillate. The plate balance capacitor would then have to be readjusted back towards balance to stop the oscillation. the IF trimmers would then be re-peaked, increasing the IF gain again, which would lead to oscillation again. This process was repeated until there was no further increase in IF gain, indicating that the input and output transformers were exactly on frequency. At this point, the plate balance capacitor could then be readjusted back to the exact balance position to ensure maximum stability of the IF stage
Second IF Alignment - The second IF alignment was much more straightforward. With the 455 kHz signal source connected to grid 1 as before, the IF appeared as a common mode signal at the plates and did not cancel out at the centre tap of IFT-1a. Therefore, it was simply a matter of adjusting the slugs of IFT-2 for peak response, (with the scope connected to the secondary of IFT-2).
Local Oscillator Tracking - Superhet tracking can be rather involved, but in this case it was relatively simple because I had used a tracking optimizer program to determine the optimum trimmer, padder and inductance values for the oscillator circuit. Prior to wiring it into the circuit, I had adjusted the turns on the oscillator coil until the inductance was within one or two microhenries of the calculated value, and I hand picked a padder capacitor that measured within a couple of picofarads of its optimum value. The result was that I only had to adjust the trimmers on the tuning capacitor to complete the tracking.
With the circuit in its final state, and everything aligned, it is sensitive enough to pick up stations as far as 2800 km away using an indoor loop antenna. At the same time, it had virtually no problems due to overloading from local stations.
In February 2009, it was entered in the Birmingham Crystal Radio Group 2009 One Active Device Contest. Over the ten days of the contest I was able to log 157 stations using the loop antenna, and listening only to the built-in speaker. No headphones were used.
In operation, the volume control is normally set to maximum, except when tuned to local stations. For distant stations there is always adequate signal to drive the speaker, but it is never loud. However, when using headphones, the volume must be turned down. The RF regeneration control can be set to approximately the 80% position, and generally doesn’t require any readjustment, except when dealing with extremely weak signals. The plate balance differential capacitor is used as the IF regeneration control, and is normally adjusted just slightly off balance. In the correct position, a distinctive regenerative type hiss can be heard, and this produces the optimum selectivity and sensitivity. The balance capacitor requires a small amount of adjustment when tuning across the band for best reception. Overall, the receiver is quite easy to tune despite all of the knobs.