Radar TV Replacements was a manufacturer of transformers, power supplies, and inverters. Some of their inverters have been described elsewhere on this site.
This inverter came to me from a very kind reader who had used it to power a telescope of U.S. manufacture. As it happens, the 12ST40 has a variable frequency output, from 40 to 60c/s, which takes care of the motor speed (dependent on the supply being 60c/s), and a Ferguson step down transformer was used to provide 115V. The complaint was that, apart from being a cumbersome set up, with battery, inverter, and transformer, the inverter ran very hot. Also, 240V was usually available from the mains where the telescope was used which made the battery unnecessary. What was really needed was a something to convert 240V 50c/s to 120V 60c/s. Such things do exist in the modern day, but as a lower cost option, I suggested that a 12V power supply be used to power a modern sine wave 12-120V inverter. This would eliminate the battery and transformer, and the motor would appreciate the sine wave. Square wave inverters are not ideal for operating inductive loads, such as synchronous motors, since the output is full of harmonics. At the time when these Radar inverters were being produced, most designs produced a square wave, since this was the most efficient way to use the available technology. In the modern day, with switchmode techniques well established, pure sine wave inverters are readily available and inexpensive. And with the convenience of internet purchasing, it is a simple matter to obtain one with a 120V 60c/s output, for operating U.S. appliances in Australia.

The 12ST40 had been well used. Most of the decal on the cabinet top had been worn away, but I was able to make out the model number. Additional labels had been applied, stating that it had been sold by the "Amateur Astronomer Supply Company" in Sydney. I noticed that the 12V supply leads had been extended. They were terminated in battery clips. There is an inline fuse holder which contained a blown fuse. This had been bridged out with a piece of fuse wire; probably 8 or 10A judging by the gauge. Evidently, the inverter had been unhappy about something at some time.

Introduction to the 12ST40.
The first inverters produced by Radar were 40W types in the late 1960's. An interesting feature was their variable frequency control, which was intended to allow fine speed adjustment for turntables or tape recorders that might be used with the inverter.
A phase shift oscillator produces a sine wave of 40-60c/s, which is then squared up and fed to the output transistors via a driver transformer. The output transistors then switch the step up transformer in the usual way, producing the 240V output.

This uses the same PCB as the first generation of 100W inverters.

My experience with Radar inverters is that they're conservatively designed and very reliable, so it was no surprise that it worked straight away. With a 40W light bulb as a load, some flicker was evident. The original supply leads were running warm. Looking at the output waveform, there was noticeable rounding off of the square wave on one half of the cycle, and one output transistor ran slightly hotter than the other.

The burned PCB can be seen at the top left.

As soon as I looked inside, it was obvious that the 100W design had been developed from this. The PCB with oscillator and driver circuit is identical, and as it turned out, the circuit is virtually identical.
Similarly, the output transistors are driven directly from a driver transformer. Perhaps the most curious thing was a very overheated 10R resistor, and a burnt PCB around it.
The output transistors are types AT301. As far as I'm aware, they are Anodeon's version of the 2N301. This is a PNP germanium transistor in a TO-3 package.

AT301 output transistors. Case functions as the heatsink.

Driver transformer at left with driver transistor mounted on it. Output transformer in middle. The two middle connections at the top of the transformer are a mystery, since they have windings connected, but are not continuous to anything.

The Circuit.

Being of 1960's design, all the transistors are PNP germanium types. As such, note that the 'earthy' side of the supply is in fact positive. This is not a problem in terms of using the inverter in a car, because the case is not connected to anything.
The inverter's timebase is a phase shift oscillator based around an AC125 and another indeterminate AC type (the transistor cases have oxidised sufficiently to obliterate some of the printing).
A phase shift oscillator works by virtue of a capacitor-resistor network. Each time a signal is passed through one such network, the phase shifts by 90 degrees. Therefore, if we have an amplifier whose input is fed from its output via three resistor capacitor stages, then the phase shift will be more than 180 degrees, and the amplifier oscillates.
As can be seen, each stage is identical, with the exception of the intermediate stage, where the 6.8K resistor is replaced with a preset 2.2K and the panel mounted 5K pot. The 2.2K preset sets the limit of the 5K pot so that the user adjustable frequency range is 40 to 60c/s. In the case of the 100W inverter, the 5K pot is replaced with a fixed resistor, since the frequency is not user adjustable.

Phase shift oscillator output at collector of second transistor.

The output is a sine wave, albeit with some distortion, and is available at the collector of the second transistor via a 2.2K preset. The supply voltage for the phase shift oscillator is stabilised at 9.1V with a zener diode.
The sine wave then passes through a 100uF capacitor into the next two stages of amplification, using another unidentified transistor and an AC128.
Form the emitter of the AC128, the signal passes into two more transistors. Both these have heatsinks, so the type number was obscured, but are likely to be AC128's. The transistor which drives the driver transformer primary is mounted on the transformer case for heatsinking.
Note that this transistor, and the one preceding it, have a common emitter resistor - this being the overheated 10R referred to previously. This connection causes a degree of positive feedback between the two stages. It is in fact a Schmitt trigger circuit. Current flowing in the driver transistor will cause a voltage drop across the 10R. Since the emitter of the preceding transistor also sees this increase in voltage, its collector current drops, increasing the base current to the driver transistor, via the 100R and 330R resistors. This in turn increases the driver transistor current even more until it saturates.
In effect, it operates as a very fast acting switch, operating in only one of two states. Thus, the sine wave is converted to a square wave, and for this reason the distortion present in the sine wave is not important.

Because the input of the Schmitt trigger is voltage sensitive, the level of the sine wave will affect the duty cycle of the square wave output. In this regard, the 2.2K preset at the phase shift oscillator output operates as a duty cycle control. For maximum efficiency, this is set to 50% so that both output transistors conduct for an equal period.

Waveform across full primary of the output transformer.

The AT301 bases are fed directly from the secondary of the driver transformer, which drives each transistor alternately. Being a class D output stage, there is no initial bias. The output transistors are either fully on or fully off.
From here, the output transformer steps up the collector voltage to 240V. A neon indicator shows that there is output.
Like all the other Radar inverters, the secondary of the transformer feeds the live and neutral pins of the output socket with no earth connection. The earth pin does not connect to the case or anything else.

A rather puzzling thing was to do with the output transformer. There are two unused tags for the primary winding, which have wires terminated to them, but there is no connection between them or anything else.
Looking closer showed that the one end of each was unconnected and taped down over the bobbin insulation. So, we had two extra windings, but what for? Seeing as they appear to have only a few turns, and were of much lighter gauge, it occurred to me that this transformer may have been used in a self oscillating inverter circuit, whereby these other windings were used to drive the output transistor bases. I am not aware of any Radar inverters (yet!) which are self oscillating, so it remains a mystery for now.

Getting it Going.
First thing to deal with was the rounded off waveform for one of the AT301's. This was easily dealt with by readjusting the duty cycle control. As expected, the input current dropped slightly, and the transformer buzz was less, since there was now no core magnetisation. Also, the flicker in the lamp was no longer visible.
Output frequency measured 42 to 62c/s.
The overheated 10R was replaced. It was found that its resistance had dropped to about 7R. While this resistor runs warm, it certainly shouldn't be enough to char the PCB. It is not clear why this happened; and can only guess the inverter had been used under unsuitable operating conditions in the past.

The warm power leads were of concern. It turned out that there was 1V drop across them when the output was fully loaded! That was clearly unacceptable, and looking at how thin the conductors were, it was perhaps not surprising. I would not run more than about 2A through wire of that gauge, but they were clearly original to the inverter. Conveniently, the wires which had been used to extend them were about twice the thickness.
I simply removed the original leads and used those which had been added (they had the same colours). Despite the longer length, the voltage drop was now down to 400mV at full load. The fuse holder was transferred across, and a 5A fuse installed.
While it seems to be commonly done, it is not recommended to extend the leads of the 12V supply because of the resulting voltage drop. It needs to be remembered that in the case of a 12 to 240V inverter, the voltage step up ratio is 20 times. If there is a voltage drop of, say 1V, between the battery and inverter, then the output will be 220V instead of 240V. If the appliance needs to be located at a greater distance from the battery, the best way to do this is to locate the inverter at the battery, and extend the 240V wiring. The current in the 240V wiring will be 1/20 of that in the 12V wiring, and the voltage drop will be insignificant.
 

Front panel shows switch, neon, socket, and frequency control.

Performance.
Input was maintained at 12.6V at the lead ends, and the following results obtained. Loads were incandescent lamps.
 

Load	Output Voltage	Input Current
No load	302	700mA
15W	272	2.1A
25W	255	3A
40W	233	4A

Regulation is very typical for this kind of inverter. Efficiency at full load is 83%, which is as expected. As can be seen, with low loading the voltage could be damaging for some applications. In fact, I wonder how the telescope motors survived, since they may have been fed with up to 150V - assuming the stepdown transformer was not saturating with an extra 50 odd primary volts.
It is unfortunate that no output level control is provided, as with the larger inverters, but it really is necessary if loads of less than about 25W are to be used.

Inverter operating 40W light bulb.

Provided its limitations are kept in mind, it is nevertheless a well designed and useful inverter. The subject of square vs. sine wave output has been discussed in various other inverter articles on this site. Briefly, a square wave inverter will produce a peak voltage of 240V. Here, the peak voltage is the same as rms voltage. For a sine wave, the peak is 340V for 240Vrms. Where the peak and the rms values are the same, incandescent lamps and heating elements work normally, since these loads are dependent only on the rms (which is the heating power). The problem can be with electronic loads because the filter capacitors charge up to the peak voltage, and not the rms.
In the case of loads where the AC is rectified for a DC supply, the voltage will be lower when the supply is a square wave. Take for example, a transistor radio which works off 240V AC, but internally runs off 9V DC via a transformer and rectifier. When run off a square wave inverter, the peak input voltage is 240 instead of 340, and the radio will now be operating at about 6.3V DC.
An ironic situation exists with a poorly regulated inverter, however. If we take this Radar for example - operating this transistor radio (which is almost no load), the AC supply is around 300V, and so the radio will actually be operating at closer to 8V. That is, provided its power transformer does not object to this higher voltage being a square wave.
While square wave inverters are very efficient, they must not be thought of a direct substitute for the mains supply, and that anything will run off them within the specified power rating.

Some might wonder why, since the phase shift oscillator used in the Radar produces a sine wave, that the output isn't also a sine wave. Well, it could be, and inverters have been made that way, but it comes down to efficiency.
To make the output stage produce a sine wave, the transistors would have to work in class B, and over the linear part of their curve. The power dissipation would rise markedly, since the transistors would effectively work as variable resistors. This power dissipation requires more heatsinking, especially with germanium transistors, and of course this heating power is reflected by an increase in supply current.
Modern sine wave inverters use switchmode techniques, where the output devices are still used as switches for efficiency, but their drive is such that the duty cycle varies instead. Provided the duty cycle varies at the correct rate over each cycle, the output will be, once averaged by a low pass filter, a sine wave.
As it is, the output transistors on the Radar barely run warm even at full load.