On/Off Volume is at left, with radio/external input switch at right.

This set was built back in 1997 to determine the performance of a TRF set with only one RF amplifier.
I had seen from reading Wireless Weekly issues from the late 1920's to early 1930's that this was not an unusual arrangement in good signal areas. Most TRF sets used a three or four gang tuning condenser, along with the same number of extra valves and coils. By omitting all but one stage, a much cheaper receiver could be built. At this point I departed from the 1920's designs by using TV valves and an infinite impedance detector. As with many of my projects I used TV valves here because they're common, cheap, and I have a lifetime supply of them.
The power supply was also novel, using a modern power transformer I had on hand.

RF Amplifier
A conventional tuned circuit formed by the aerial coil and one gang of the tuning condenser feeds the input of the RF amplifier valve in the usual way. The valve is a 6BX6, or as Europeans would know it, EF80. This is an RF pentode developed in the late 1940's for use in TV IF amplifiers, but like others ended up being used for many different applications.
As such, it is of fairly high gain. For a 7Mc/s channel bandwidth, three 6BX6's are required in a typical television IF stage. It can be imagined that down on 1Mc/s with say 10Kc/s bandwidth, the gain will be very high indeed. To put things into perspective, the 6BX6/EF80 has a transconductance of 7.4ma/V, whereas a typical RF pentode normally used in a MW set, such as a 6K7 has a transconductance of 1.5mA/V. Later valves such as 6BA6 have around 4.4mA/V.
Your eyes might light up at the thought of replacing the IF and RF valves in your MW sets with TV IF pentodes in view of this tremendous increase in gain. Imagine all the distant stations you'd be able to receive! Unfortunately, it doesn't work out this way. The gain is actually too high and due to stray capacitances and other couplings you end up with an unstable whistling mess. So, when used for narrow band RF amplification on medium waves, it is necessary to reduce the gain.
One obvious way to reduce gain is to leave the cathode resistor unbypassed. However, I still found the gain too high. By reducing the B+ supplying this stage, stability and good performance was achieved.  It may surprise those who can only understand valves working at 250V to find the RF stage in this set working at 21V. The gain is still very high at this voltage, but not so much to cause instability. Not surprisingly, the 6BX6 is another valve useful for low voltage designs.
The source of 21V was simply taken from the output valve cathode.

Detector
Back in the 20's and 30's, the detector would usually be an anode bend type, or if regeneration was added, a grid leak type. However, I decided to use an infinite impedance detector instead. This type of detector appeared much later in the scene and was usually associated with hi-fi receivers due to its very low distortion.
Anode bend detectors are essentially a sharp cut off pentode biassed almost to cut off. It is connected as a normal resistance coupled audio amplifier, but with a higher than normal cathode resistor. Incoming RF will cause the valve to cut off on negative excursions, and conduct much more heavily on positive excursions. Hence, it detects AM. This type of detector was standard in TRF and superhet receivers up until diode detectors took their place in the mid 1930's, when valves such as 6H6,75, and 6B7 appeared. An advantage of this detector is that it does not load the circuit feeding it as grid current never flows. This means good selectivity as the Q of the preceding tuned circuit is not reduced. It also can handle large signals without overloading. However, it suffers from distortion as the detector valve is also functioning as an audio amplifier which has too much negative bias.
Another common detector of the time, which is even older, is the grid leak circuit. It is still the standard today with regenerative receivers, and can be used with just about any triode or pentode valve. Here, the grid and cathode form a diode. The grid leak condenser and resistor simply form the diode load. Because the plate of the detector valve is bypassed at RF, only the audio component developed across the grid-cathode diode is passed on. Such a detector must not be operated with any negative bias. To do so would desensitise the detector as the incoming RF voltage would have to overcome the bias voltage before the diode could conduct. In fact, in many battery circuits, the grid resistor is actually returned to the positive filament supply to do the opposite and make the detector more sensitive.
As a grid leak detector operates with no, or slightly positive bias, it is essential that excessive plate current cannot flow as the valve will be damaged. Hence, such detectors operate at only 22.5 or 45V if feeding headphones or audio transformers, or through a resistor of 100K or more when operating from higher voltages.
While this type of detector has good sensitivity, it does load the tuned circuit, thus reducing the Q and selectivity. When used in a regenerative circuit these losses are made up and become irrelevant. Likewise, in a TRF set with no regeneration, a couple of preceding tuned stages will improve this. The other disadvantage is that, as the valve is again also an audio amplifier, it is this time operated with no bias. So, the signal handling is limited before it starts distorting. Grid leak detectors are thus best used with weak signals. A few grid leak detector circuits, such as this are available elsewhere on this site.
Another kind of detector eliminates the loading and distortion problem. It is called the "Infinite Impedance Detector". The grid is never driven positive and thus does not load the tuned circuit, allowing full selectivity and gain to be obtained. In terms of the audio signal, it is operating with 100% negative feedback due to the cathode follower configuration. As can be imagined, audio gain is not as high as the other types and needs to be made up in successive stages.The type of valve used is not critical and most triodes can be used.
This kind of detector was often used in various TRF and wideband superhet circuits throughout the 1940's and 50's and was used in many Radio & Hobbies circuits. In view of its advantages, this is the detector I chose for this receiver.
As can be seen, it uses another 6BX6/EF80. This time the valve is triode connected as it is operating as a cathode follower. The 100K and 270uuF form the diode load, with additional filtering by the 47K and 100uuF. A line level signal is available at this point.

Terminals from left to right are: Aerial, Earth, Speaker. RCA socket is for external audio input. Note the shielding between the two stages. Trimmer capacitors adjacent to the coils are for alignment at the top end of the band. The top cap of the 21A6 is live at about 180V.

Audio Stages
Yet another 6BX6 is used as the first audio amplifier and is used in the conventional way. I provided an audio input socket for use with a CD player or FM converter as an added convenience.
For the output stage, things became unconventional as I'm using a 21A6 (PL81) line output pentode. This is not a hugely common valve in Australia as series heater TV sets were in the vast minority. However, it was used in some Ekco models and a few others. The 6CJ6 (EL81) which is the 6.3V heater version was used in a number of Astor models. However, I do have some in my collection, and as I was tied to using a 300mA series heater circuit, it was a perfect choice. As I've discussed at length here, TV line output pentodes generally are a poor choice for single ended audio, at least when used in conventional circuits. Not all are as bad as the 6CM5/EL36 in this application, and I found the 21A6 to provide a couple of watts of good quality audio. In view of the heater current, it would be wasteful to use this circuit on a 6.3V heater supply with the 6CJ6 when much more efficient audio output valves are available.
The output transformer is an Altronics M1120 100V line transformer with the 4K primary tapping used. See here for more details on this.
Negative feedback with a slight amount of bass boost is achieved by the 820K and .001uF network. The RC network across the primary of the output transformer is for this particular type and helps improve frequency response.

Power Supply
I have a number of Ferguson PL80/60VA transformers available. These are rated at 60VA with two 40V secondaries. I had often thought about how they would go for powering a valve radio by connecting the secondaries in series and then following with a voltage doubling rectifier. For the heaters, they would have to be in a series circuit powered from 40V. This receiver was the test for this, and it worked perfectly.
As it happened, I didn't need a dropping resistor as all the heater voltages added up to 40. It is because of the 6BX6's that the heater current had to be 300mA. This meant the output valve also had to be 300mA and was one reason I chose the 21A6. I could have used a 16A8/PCL82 with a dropping resistor instead. Note the order in which the heaters are connected. It is important to have the valves more susceptible to hum at the earthy end of the heater string. While I could have used a full wave voltage doubler, it would have meant the whole heater chain floating at about 90V above earth which is not a good thing.
Once again, we see just how easy it is to use valves with modern transformers.

As with some of my other MW receivers, I've made the coil formers to plug into octal sockets. Power supply circuit is to the right, RF to the left, and audio at the top.

Performance
Sensitivity is surprising given there's only one RF stage. I can easily get 2LT from Lithgow and 2BS from Bathurst in the day time, even though their signals are directed away from where I live. All Sydney stations come in with good strength and there are no problems with selectivity. At night, the usual interstate stations are receivable.
The most notable thing is the excellent sound quality.  As I have said elsewhere, if people heard low distortion, wideband AM, they would not tolerate the $5 squawk boxes with their nasty 2Kc/s bandwidth, class B output stages, and 2" speakers. Neither would they hold the view that FM is essential for music listening.

Triode RF Amplifier. (July 2014)
Many "radio people" have an aversion to TV valves, and throw them away.  When building a TRF set, they'd use some low gain pentode, like a 6J7 or maybe a 58. As a TV enthusiast, I know the virtues of these unwanted valves, and as can be seen throughout the site they are good for all purposes.
In view of the huge amount of twin triode VHF amplifier valves I have, it seemed worthwhile to see how they would perform as a medium wave RF amplifier. Now, triodes are not new for this purpose. Triodes were all that were available until the late 1920's. To use them as an RF amplifier is complicated by the grid to plate capacitance which can cause oscillation.
At the time, one way out of the problem was to use the Neutrodyne circuit, or if you didn't want to pay Hazeltine patent royalties, resistors in series with the valve grids would stabilise the circuit. Despite this, triodes are less noisy than screen grid or pentode valves. By the mid 1950's, TV tuners were using triodes because of their improved noise characteristics. The first method was to connect two triodes in 'cascode' style in order to overcome the stability problem. Typical valves for this service are 6CW7, 6BQ7, 6ES8. Then, in the early 1960's, the neutralised triode method became popular. The most common of type used in this circuit is the 6GK5.

The four valve TRF set described here seemed to be the ideal test for the concept. I chose a 6ES8/ECC189; this being a very common frame grid valve. The circuit was made up as follows:

Essentially, the 6BX6 was substituted directly with a 6ES8 cascode circuit, using the same bias resistor and supply voltage. In fact, I simply made an adaptor that plugged into the 6BX6 socket with the 6ES8 and circutry built onto it.
As can be seen, it is a standard cascode circuit. The lower triode is biassed in the usual way with a 330R resistor. Normally, this would be of lower value, and would be bypassed, but in this receiver, excessive gain causes instability. The upper triode is biassed in the usual way to about half the supply voltage. The upper triode functions as a grounded grid amplifier and thus provides the isolation between the two tuned circuits. With a tetrode or pentode, the screen grid provides the isolation.
More astute readers will wonder about the heater current and the series heater circuit this receiver uses. The 6ES8 requires 365mA at 6.3V. This means in a 300mA series heater circuit, the 6ES8 will be slightly under run, and the other valves very slightly over run. To correct this, the heaters of the other valves need to be bypassed with a resistor to increase the heater current by 65mA.  Alternatively, the 300mA series heater version,  7ES8/PCC189 can be used.

Improved Performance.
The circuit worked extremely well with the 6ES8. In fact, 2ZB from Wellington, NZ, was received at room filling volume the first night I tried the new RF amplifier. The gain is higher than the 6BX6. I also noted that gain across the band seemed to be more even. As with the 6BX6, full B+ cannot be used because of instability. No doubt, if the coils were shielded or perhaps mounted at right angles to each other one could raise the voltage. However, it does illustrate that, again, many normal valves work well at low voltage.
Having proven the concept, it opens possibilities for new designs. For example, a high gain, low noise all triode TRF receiver without having to use the Neutrodyne approach. This will be on the list of future projects.

Other valve types.
There is a wide choice of valve types that could be used apart from the 6ES8/ECC189. Types such as 6DJ8/ECC88, 6AQ8/ECC85, 6FC7/ECC89, 6CW7/ECC84, 6BQ7, 6BK7, etc., and their series heater variants. The older 12AT7/ECC81 should work but will probably require higher B+. Note that none of these other types has actually been tested and some experimenting may be required. The 6ES8 is the highest gain with its frame grid construction. As mentioned before, modifications may need to be done to the heater supply if this receiver is built with a series heater circuit. European constructors will find the PCC189 easier to get than the ECC189, and this requires no changes to the heater string.