Introduced in 1939, the Emerson CF255 was claimed to be the smallest practical radio receiver. It is of the inexpensive "midget" type, popular in the U.S. Dimensions are approximately 85mm deep x 160mm wide x 120mm high.
Being of such unconventional design (i.e., not the same old boring superhet), it was a set I had to have. They appear on ebay from time to time, and now have three in my collection.

This tiny set uses only two valves. The set at lower left is a CF255 chassis in a 540 cabinet.

The Design.
For a set not requiring the use of a regeneration control, Emerson are probably right in their claim of being the smallest practical radio. The set is based on two valves; a 12B8 and 32L7, which were also introduced in 1939.
The 12B8 functions as a TRF stage with fixed regeneration and a grid leak detector. Detected signal feeds the 32L7 pentode which drives perhaps what is the most unusual aspect of this set - a cone speaker. Instead of the usual output transformer driving a conventional loudspeaker, we are thrown back to the 1920's with a high impedance cone speaker. It's an ingenious cost cutting and space saving idea.  Power for the set comes from a line cord resistor for the valve heaters, and the diode section of the 32L7 for the B+.
A conventional slide rule tuning dial is used and the set is housed in a tiny bakelite cabinet. Although a back was originally fitted, not many survive intact.
Again, we see one side of the mains connected directly to the chassis which is fully exposed, even if the back is fitted.

Back of chassis shows the two valves and the cone speaker. Chassis is live at mains potential.

Even with the back on, there is plenty of live metal to touch.

The Circuit.
Turning now to the circuit diagram (the complete diagram with parts list is available at Nostalgia Air), we can see the simplicity of this set.

RF amplifier.
Signal from the aerial is fed into a conventional air cored tuned circuit and thence to the pentode section of the 12B8. Its plate load is another tuned circuit. Of course, both tuning condensers are ganged. Gain of the RF amplifier is controlled by varying its bias with the 75K potentiometer. The potentiometer is unusual in that it stops at 200 ohms minimum resistance, which provides minimum bias for the 12B8. Doing this eliminates having a separate resistor, which in a tiny set like this takes up valuable space. Typical of cathode bias volume controls in TRF stages, the potentiometer is so connected that when bias is at maximum, the aerial is shunted to earth. This is required if the set should be fed with a very strong signal. Operating the 12B8 at cut off is not always enough to reduce the volume to minimum.
The aerial consists of several metres of wire permanently attached to the set. It is isolated by a 550pF condenser.

Detector.
From the secondary of the second tuned circuit, the signal proceeds to a grid leak detector based around the 12B8 triode. The choice of grid leak components is interesting, because normally the resistor would be less than 2M and the condenser up to around 500pF. Here, a 10M grid leak is used with a .01uF condenser. I assume this has been done to get the best sensitivity; the trade off being fidelity lost because of the very long time constant. I would have thought that using 10M would allow contact bias to be produced, thus making the triode perform more like an audio amplifier than a detector. However, we must assume that Emerson knows best and had good reasons to use this unusual choice of components.
Apart from that, the stage is conventional. Detected audio is developed across the 500K plate load.

Announcement for the release of the 12B8 and 32L7 valves in July 1939. It would appear they were specifically designed for circuits along the lines of the CF-255.

Fixed Regeneration.
To further improve selectivity and sensitivity, fixed regeneration is employed. Note that the detector plate is not bypassed for RF. Along with the audio, some RF is present. This is fed to the third winding on T2 via a 220pF condenser, out of phase to the incoming RF, so as to provide a degree of positive feedback.
However, fixed regeneration can only provide a limited increase in gain. It can never be set so that the RF stage is on the verge of oscillation, which of course is the ideal. To do so would run the risk of breaking into oscillation at high mains voltage, connection to certain kinds of aerial, and simply tuning from one end of the band to the other. Presumably, the amount has been set so the set remains stable under all likely conditions. Having said that, for the addition of one winding and one condenser, it is certainly worth implementing.

Audio Amplifier.
The detector feeds the 32L7 pentode via a .02uF condenser and 500K grid resistor. Cathode bias is used for the 32L7 pentode by means of the 140R. Some gain is lost by not bypassing this resistor. No doubt this was done because of the space required by an electrolytic capacitor, as well as the cost. However, not bypassing it does provide a degree of negative feedback, which although I'm sure was not intended, can only improve the sound quality.
The speaker is a high impedance cone type reminiscent of that used in the late 1920's. This eliminates the cost of and space taken by a speaker transformer.
The magnet is also smaller than would be required by a permanent magnet dynamic speaker of the time. An electrodynamic speaker would of course defeat the 'made as small and cheap as possible' design.
However, in one of my CF255's, the speaker has been replaced with a conventional permanent magnet type along with a miniature output transformer. By the time that repair was done, the parts were available small enough.

The speaker looks normal from the front, except there's nothing in the centre.

Power Supply.
Typical of U.S made live chassis series designs, a half wave rectifier using the diode section of the 32L7 is used to provide B+ from the power mains. A standard RC filter is used using a dual electrolytic.
The valve heaters are in series and fed from a line cord resistor. The dial lamp is also in series, but part of the line cord is shunted across it to prevent an excessive switch on surge when the valve heaters are cold. This is required because the dial lamp warms up much more quickly than the valve heaters.
Of course, one side of the mains is connected to the chassis, resulting in the possibility of electrocution, regardless of plug polarity. When the set is switched off, there is a path to the chassis via the valve heaters. When switched on, the mains is directly connected. So, depending on how the mains plug is inserted, you have the choice of chassis live with the set off, or live with it switched on.
The safety issues are of course overcome by operating such sets via a double wound transformer. This is required in Australia anyway, to run U.S sets off our 240V mains. An auto transformer should not be used as it does not provide isolation.

My Emerson CF255's
It didn't take long for one to turn up on eBay; in fact, soon as I went looking. Being such a tiny and unusual set I thought it would go for a lot more than it did; I think I paid about $60. As it turned out, the CF255 isn't that uncommon, especially in the brown bakelite cabinet. The coloured ones are rarer and are worth more. Of course, the line cord resistor was open circuit. But, I did temporarily power the set up from external power supplies just to see what it did, and despite the original components, it did function in a kind of way. Interestingly, despite the circuit diagram and photos of other CF255's this set had no dial lamp, and judging by the appearance, never had one. For a start, there was no provision for it in the line cord resistor, nor was there any markings where the socket would have been mounted on the speaker frame.
About a year later was an auction for two sets together; one was a CF255 and the other was said to be a 540. The 540 is the successor to the CF255, using the same size cabinet, although with an egg crate grille. Internally, the design is very different, being a normal superhet with 7 pin valves. Given the low starting price I thought it would be an ideal opportunity to get another CF255, and the 540 looked cute too. I got both sets for an absolute bargain. What turned up in the mail was a surprise however. The "540" contained the chassis of another CF255! Well, no complaints there; two CF255's was very nice to have. Interestingly, the chassis mountings and dial and control positions are the same, so the cabinets are interchangeable.
The set with the proper cabinet turned out to have a functional line cord resistor, except the dial light section was open circuit. Fortunately, the dial lamp was still intact despite the overload it must have endured. The set in the 540 cabinet had its open circuit, along with a blown dial lamp. The seller had put a new looking 6Y6 into the 32L7 socket. Probably a good thing the line cord was open, otherwise the 12B8 could have been burned out from excessive heater voltage. The 6Y6 went into my valve collection, and I ordered some more 32L7's.

Ad from February 1939.

Restoration.
So far, I have only restored the second CF255, and this is the set to be described here.
First thing was to deal with the line cord resistor. Electrically, I could have used it by simply bypassing the open dial light section with a 5W resistor mounted on the dial lamp socket. However, it was physically in poor condition with failing insulation and uncovered asbestos. So, time to make one of my repro line cord resistors with shoe lace and electric blanket element wire. The procedure has been described here.
The circuit diagram does not show the resistance values, but they are easily calculated. The valve heaters require 45.1V at 300mA. The required resistance is 250R for 120V mains.  The dial lamp section is a bit more tricky to calculate. Basically, with the valve heaters cold, the voltage across the dial lamp must not be more than 6.3V. I found with the 250mA bulb fitted that 39R was a suitable value. Of course with this method of feeding the dial lamp, it does mean once the valves have warmed up, the light is dimmer, running at about 4V. This is one of the limitations of series heater circuits, and I discuss it in more detail here.

Line cord resistor prior to covering with black shoe lace.

I cut the electric blanket element to give 250 and 38 ohm sections. These were joined to lengths of flexible wire and then covered in black shoe lace. I kept the same colour code as the original cord, although the direct mains supply had to use white figure eight wire. After a few hours work, the reproduction line cord resistor was finished and ready to be tested. All voltages were spot on. The speaker cone had largely become detached from the frame, so I glued it back together.

Reproduction line cord resistor.

Next was for the circuitry under the chassis.

With so few components, restoration is quick and easy.

All the resistors, except the 10M grid leak tested OK, so were left in situ. The dual section electrolytic was actually functional for a while, but then one section shorted out, so I replaced it with two separate 22uF units. All the paper capacitors were replaced except the 12B8 cathode bypass. Given the low voltage and low impedance in this part of the circuit, leakage will not cause any problems.
Due to the simplicity of the set, it didn't take very long at all, and had it finished in a couple of hours. It worked straight away, and even the alignment wasn't far out.

Performance.
The unusual speaker was my first point of curiosity. Despite its strange construction, it sounds much the same as any other 4" speaker. In my location about 70km from the commercial transmitters, the signal is weak with the supplied length of aerial wire. The volume has to be turned right up. The ABC transmitters are closer and more powerful, giving adequate volume for a quiet room.
With an outdoor aerial, results are much improved. However, it is important to realise that the design of the set, and its alignment, is dependent on the few metres of aerial attached. Connecting a longer aerial can alter the alignment by loading down the RF coil, and detuning it with additional capacitance. So, a compromise can be had by connecting it via a low value capacitance, of around 100pF, dependent on aerial length.
I would imagine that in the intended service area of the transmitters, performance would be adequate with just the indoor aerial.

Copied by Astor.
The well known Australian company, Astor, had a connection with Emerson, taking their "Mickey Mouse" model name and applying it to numerous locally made radios. Astor didn't stop there, and created a local version of the CF255.

Astor designated their version, the "AR", which also appears as the "ARK". It was Australianised by the incorporation of a power transformer, since line cord resistors were not permitted here, and with the set's small size, it was impractical to house a dropping resistor inside the cabinet. This would have been especially difficult with the 240V supply. The transformer supplies 45V for the valve heaters, with a 6.3V tap for the dial lamp. Additionally there is a winding for the B+, which would presumably be around 110V.
Unlike the Emerson model, the Astor AR and ARK uses a conventional electrodynamic speaker with output transformer. With this, the power transformer, and larger Australian parts, it would be impossible to make the set as small as the CF255, but nevertheless it would have been the smallest set on the local market.

The circuit, apart from the power supply, duplicates the CF255 with only minor differences. For example, the 200R resistor (16) is part of the volume control pot in the CF255. Also, the volume pot is 100K, rather than Emerson's 75K - no doubt a reflection of local parts.
An interesting oddity is that one side of the speaker transformer secondary is connected to the 32L7 grid. Yet, the other side is not connected to anything. I don't know what the intention actually was, but this strange arrangement would actually function as a capacitive feedback circuit, to roll off the upper frequency response. Instead of simply connecting, say, a 100pF capacitor from plate to grid, they've used the capacitance between the transformer windings to perform the same function.
Given that the 32L7 pentode cathode is unbypassed, and there's already a .005uF plate bypass, one wonders if this novel piece of circuitry was really needed.
The 12B8 and 32L7 were never made in Australia, so had to be imported. These valves were also used in a few other Astor models.

Astor obviously liked this simple TRF circuit, because it evolved into a much more Australianised design, using conventional (and locally made) valves; 6G8, 6V6, and 6X5. The best known of these was the "GR", known as the "Football". As such, it was no longer of the midget set category, but just a cheap mantel set.