Volume and tone at left, tuning at right.

This radio was manufactured by Motorola for various 1956 model Fords and is also known as the Ford model number FDR-18805-B1. The particular set to be described came via a Model T acquaintance who had obtained it while on a trip
to the U.S. Fortunately the knobs were on the set - usually car radios are missing knobs and dial parts. In fact, the set was complete except for one of the bottom covers, and also a 9 pin valve was missing.

Bottom cover is missing but appears not to affect performance. Note the FoMoCo stamp - Motorola is not stamped anywhere on the radio.

All the valves were miniatures, and the vibrator was a Cornell Dubilier Electronics 12V type. This narrowed the manufacture date to sometime between about 1955 and 1959, as Ford and other U.S. car manufacturers had by now changed to 12V electrical systems.
From around 1959 most radios made for U.S. cars were hybrid types. The valves were the standard line up; 12BA6, 12BE6, 12AU6, 12AV6, 12X4, but the missing valve had me curious. Obviously it was the audio output, but a 9 pin miniature is unusual for this role with a 12.6V heater.

The 12AB5.
Firstly a circuit would have to be obtained to find out what the missing valve was. About the only place I could get this from was Radiomuseum.org. Here I obtained the Howard Sams circuit, and also a poor copy of what appears to be the John Rider circuit. While some ebay sellers have the complete manual, the postage costs reduce my enthusiasm for purchasing one. There are also various Eastern European sites that purport to have circuit diagrams and manuals, but my experience is you can never actually download anything from them. Anyway, the valve turned out to be a 12AB5. I had never heard of this valve and it had certainly never been used in any Australian car radio. In the 1950's and 60's there ware a lot of new American type numbers introduced, which were improvements on existing types, and no doubt the 12AB5 was one of these. I was curious why they didn't just use a 12AQ5 or even a 12V6. After all, an audio output stage for a radio is nothing complicated.
As it happens, the 12AB5 appears to be based on the 6V6 and its equivalents. The operating conditions are the same. However, what makes it different is the data specifically mentions operation in car radios with a supply of 10 to 15.9V.
It would appear that the 'improvement' over the 6V6/6AQ5, etc. would be to reduce the possibility of grid emission at high heater voltages.
There is actually a 9 pin equivalent of the 6V6, the 6BW6, but this was a British valve not used in U.S. designs, and it never had a 12.6V heater version.
What to do? While the 12AB5 was undoubtedly popular in the U.S., it certainly wasn't here. Of course, they're a dime a dozen on the U.S. ebay, but the thought of paying about $30 postage just to get one here put me off that idea. Various schemes went through my mind - replacing the socket with a 7 pin type and using a 12AQ5, rewiring the socket and using a 6BW6 with a resistor in series with the heater, or even using a 12BY7 and accepting reduced output.
Just maybe I had one in my valve collection, since I do have a few less than common U.S. types. Much to my surprise, I had not just one, but five 12AB5's in my car radio section of the collection. Where they came from I can't remember, but they were all new in boxes.
That solved that problem nicely and I could begin restoration.

Note! There are mistakes on this circuit diagram regarding the AVC feed and 12AB5 cathode bypass.
 

RF Amp & Converter.
As one would expect, the aerial input socket is the ubiquitous and now standard Motorola type. This is nice to see after dealing with various bayonet and screw type plug connections used for many other car radios of the era. While there is nothing electrically wrong with such connectors, the plugs are now obscure.
There is an RF amplifier using a 12BA6 in the conventional way. As the aerial is connected directly to the first tuned circuit to provide maximum coupling, its capacitance affects the tuning range. Therefore, as is typical, a trimmer capacitor is provided to compensate for different aerial capacitances. The 12BA6 plate load is actually a 10K resistor and this feeds the 12BE6 converter. Resistance coupling of RF amplifiers in car radios is not that unusual. In Australia it was used by Astor and AWA at least.
The second RF tuned circuit is connected to the signal grid of the 12BE6. It is unclear what the series choke with shunt resistor is for, but it can be assumed the choke requires the resistor to lower its Q. The oscillator circuit for the 12BE6 contains nothing unusual being a typical cathode feedback type circuit. Perhaps the only non standard thing here is the 47K grid resistor is not returned to earth directly, but instead through the oscillator coil secondary. Electrically, this should make no difference to operation.

IF Amplifier.
This is the first unusual aspect of design. A 12AU6 is used in a fairly conventional circuit, but there's a number of interesting aspects. Firstly, the 12AU6 is a sharp cut off pentode. Thus, it is not suitable for an AVC controlled amplifier, except in a limited way. High AVC voltage applied to a sharp cut off pentode can result in distortion, as the valve is taken out of linear mode and into class B. For this reason, remote cut off types are normally used. As it happens, there is no AVC applied. This is unusual, and in this set AVC is applied only to the RF amplifier and converter stages. One can assume that the range of control provided is sufficient, with the limited signal pickup helping. Perhaps strangest of all is the so called "sensitivity" preset in the cathode circuit. Note also, that the 220R cathode bias resistor is not bypassed. Since the 12AU6 is a high gain valve, we can assume this was done to keep the stage stable by providing a degree of degeneration. The sensitivity control is a 3K rheostat which provides further degeneration (and increase in bias). Why this was included in a car radio of all things is very strange. That is the last kind of receiver, which with its short aerial, you would want to desensitise. Not only that, the control is a preset not accessible as a user control. Similar sensitivity controls were included in some domestic sets around the late 1930's - early 1940's, and were sometimes labelled as a "noise suppressor". (The Kriesler 710A console from 1937 is one such set in my possession). The idea was that the control would be backed off so that the IF amplifier was operating with less gain, and only the local stations were receivable. The so called "noise suppression" comes about simply because a low gain amplifier is less noisy that a high gain one, and any interference will also be subdued. In practice, it is a gimmick and has no effect on electrical noise, if it has similar or greater amplitude than the wanted signal.
No information was included in the alignment instructions as to how this control should be adjusted in the 66MF.

265Kc/s IF.
In this receiver, the IF is 265Kc/s. This is not uncommon for portable and car sets, because a higher gain and better selectivity can be obtained than at 455Kc/s. In the early days of superhet receivers, low IF's such as this were common, but there's two reasons they went out of fashion for domestic sets. Firstly, the selectivity can be too great and the demodulated audio is restricted in bandwidth. Also, double imaging is more likely to occur. That is, a particular station, if strong enough, might appear at two places on the dial. The possibility of this increases the closer the local oscillator is to the station frequency; i.e. as the IF is reduced. To explain this, let's revise some basic superhet theory. Forgetting any tuned circuits at the RF input, the receiver will respond to a signal that is the local oscillator frequency plus or minus the IF. For practical reasons with MW receivers, the local oscillator operates on the high side; i.e. above the receiving frequency. Selectivity of a superhet is largely determined by the IF amplifier. The RF input selectivity is broad by comparison.
Now consider a station at 1100Kc/s. For an IF of 265Kc/s, the oscillator will be at 1365Kc/s. That is the received frequency plus the IF. Now, if the oscillator is operating at the IF less than the received frequency; i.e., 835Kc/s, the same station will again be received. Normally when the local oscillator is at 835Kc/s, the radio would be receiving 835 - 265, or 570Kc/s. Thus, the wanted station appears correctly on the dial at 1100Kc/s, but also at 570Kc/s. Further thought will reveal a problem if there is actually a station on 570Kc/s. A heterodyne whistle will be the result.
In theory, the RF tuned circuit at the aerial input should prevent reception of an 1100Kc/s station with the dial set to 570Kc/s. However, anyone who has experimented with simple one tuned circuit receivers, such as a crystal set, will know that selectivity is limited in practice, and worsens as signal strength increases. In fact, if the signal is strong enough it will be received over the entire tuning range.
With car and portable radios, the aerial pickup is naturally very limited, so the problem is not likely. With domestic receivers operating with a long wire aerial, and a high signal input level, more than one tuned RF circuit is often required to eliminate the effect.

Why does a higher IF such as 455Kc/s reduce this double imaging effect? Simply put, the further the local oscillator is away from the received frequency, the more effective is the RF input selectivity. Also, the image frequency is more likely to be out of band. Going back to our example of the 1100Kc/s station, if this is received on a set with a 455Kc/s IF, the local oscillator is now 1555Kc/s. The same station would also be received with the local oscillator at 645Kc/s. This can't happen in practice since the L.O. operating at 645Kc/s would be for a receiving frequency of 190Kc/s which is well below the MW band. In fact, practical constraints prevent the local oscillator tuning down that far. Assuming the lowest frequency to be received is 530Kc/s, the lowest L.O. frequency is 985Kc/s.
From this discussion, it should be evident that the choice of IF is not just a random number, but chosen to suit the tuning range of the receiver, and the station frequencies to which it will be tuned.

Detector, AVC, & Audio.
Next, we come to the first mistake on the circuit diagram. The AVC is developed in the usual fashion with one of the 12AV6 diodes shunt fed via a capacitor from the 12AU6 plate. This is decoupled by a 1M resistor which feeds the RF and converter stages. Filtering is by a 0.05uF condenser as normal. The diode load resistor is 2.2M. The error is that the shunt feed condenser is shown to be fed from the B+ line and not the 12AU6 plate. There is no way any kind of negative voltage could be developed that way, and even if it could, it would not be dependent on signal strength. The actual set wiring, and also the Rider's circuit diagram show the correct connection. Strangely, the shunt feed capacitor is only 4.7pF. That is considerably less than the normal 50-100pF used here. It would appear the design was intended to restrict the amount of AVC voltage developed.
The set has delayed AVC by virtue of the 12AV6 cathode being set at 1.5V. This means the diode has to have more than 1.5V applied before it conducts and starts producing the negative AVC voltage. This is standard practice on anything but the cheapest receivers, so as to prevent the receiver being desensitised on weak signals. The reason this comes about is because if the AVC is not delayed, there will always be some negative voltage produced because of atmospheric noise, and also noise generated in the preceding stages of the receiver.
This permanent negative voltage fed into the AVC line therefore desensitises the receiver when tuned to a weak station - which is the opposite of what is required. Delayed AVC prevents this happening by not enabling the AVC diode to conduct, until the signal has increased to something above the noise level.
Some very cheap receivers without delayed AVC actually rely on the noise level to provide the initial bias for the IF amplifier, thus dispensing with its normal cathode bias resistor and capacitor. It's a crude system, and the IF amplifier valve can have a short life when the set is tuned to weak stations if the noise level is not sufficient.

The detector circuit is standard with the other 12AV6 diode providing this function. A standard RC filter removes most of the IF, with the resultant audio across the volume control, which also functions as the diode load. So that the detector diode is not desensitised, the earthy end of the volume control is at the same potential as the 12AV6 cathode. If it was taken directly to earth, the received signal would have to be strong to overcome the 1.5V bias otherwise on the diode.
The volume control is tapped for loudness compensation (i.e. to compensate for loss of bass response in the human ear at low volume), and the tone control is also part of this circuit.
With the tone in the "low" position, the 0.003uF condenser in the 12AB5 grid circuit is shunted to earth, and the 0.05uF in the loudness control is shunted across the loudness tap. Thus, there is a predominance of low frequency response.
When the tone is set to "high", the 0.003uF has 2M of resistance in series and has little shunting effect. The 0.05uF is shorted out, so the loudness circuit no longer boosts the low frequencies at low volume. Obviously, varying degrees of frequency response are obtained with the control in intermediate positions.

The 12AV6 triode performs as a the first audio amplifier in the usual way. It uses contact bias by virtue of the 10M grid resistor. If a valve is of such construction that the grid is close to the cathode (such as a high mu triode like the 12AV6), electron emission from the cathode is sufficient to cause the grid to go negative by itself. If there is no resistor to the cathode, the amount of bias may cut the valve off. Values of grid resistor are typically 4.7M to 22M for this kind of bias. If the resistor is too low, the electrons leak away too quickly, and the voltage developed is too low. The plate load is 1M which is rather high; normally 220K to 470K would be used here.

Audio Output.
As mentioned previously, this uses the not so common 12AB5, but operating conditions are quite standard. Negative feedback is obtained by returning the 220K grid resistor to the speaker voice coil in the inverse phase. 220K is appropriately low to reduce the effect of grid emission, particularly when the heater voltage is higher than normal.
However, somewhat unusual is that while 220K is a legitimate choice of grid resistor, the preceding plate load is 1M. This results in some loss of gain since maximum power transfer cannot occur. It is pleasing to see a ceramic condenser used as the grid coupler, and not a paper type.
The next mistake in the circuit diagram concerns the cathode circuit of the 12AB5. It operates with normal cathode bias, comprising the 220R and 39R in series. The mid point connection provides the source of AVC delay voltage. The mistake is that the 20uF bypass condenser is shown connected across only the 39R. It should be connected across the 220R as well. As shown, the circuit would work, but with less gain. There is already a negative feedback circuit, so using cathode degeneration is not required for this function. The actual set wiring is correct, as is the Rider's circuit diagram.
The speaker is connected by a small 3 pin phenolic plug and socket. The pin configuration is non-standard. Not shown on the circuit is also a ceramic condenser from the speaker voice coil output to earth. The purpose is to reduce RFI coming back into the set via the speaker wiring.

Speaker plug is visible. Supply and dial light connections feed straight in.

Power Supply.
This uses a four pin non-synchronous vibrator with a 12X4 rectifier. The vibrator socket is wired so that either shunt or separate drive types can be used without alteration. Fitted to the set is a CDE type C-4 which is shunt drive.
There is no spring loaded grounding cup to secure the vibrator and earth the can. The vibrator is lightweight with an aluminium can so isn't likely to fall out. Evidently, the non-earthed can does not cause RFI with this set design.
Timing condensers are provided across both the primary and secondary of the power transformer. Note that the secondary condenser is connected from one side of the secondary to earth; not across the whole secondary as is commonly done. This reduces the voltage stress considerably, but the value of capacitance has to be higher. An interesting feature is that the transformer centre tap is returned to the 12V supply and not earth. This provides a free extra 12V on the B+ supply. It is surprising more vibrator powered equipment does not do this, but it does mean the radio can only be used with a negative earth supply. Running this radio with a positive earth would actually reduce the B+ by 12V.
The B+ filtering is standard with two electrolytics and a series resistor.

Power supply and audio output are shown here. 12X4 is at left, and 12AB5 at right. The can electro was dried out.

The 12V supply is protected with a 5A 3AG inline fuse. As is typical with Ford, the dial lamp is fed separately from the dash light circuit. This means the radio dial lights up only when the park lights come on. Dial light life should be a lot longer therefore, as the bulb is only operating if the radio is switched on, and the park lights or headlights are also on. My experience with car radios is that the dial lights serve no function during daylight and are seldom visible.
Both the 12V radio and dial light inputs are filtered with "spark plate" capacitors. These are low value, but low inductance, capacitors which reduce ignition interference coming through the supply inputs. Construction is simply that of a metal plate separated from the chassis by a sheet of mica or paper.

Two new 22uF 350V electros replaced the dried out filter caps. A three lug tagstrip had to be installed to accommodate them, but there was plenty of room, and one of the speaker transformer screws provided a convenient mounting point. Although some people do it, I do not recommend just connecting new capacitors across dried out originals. It's particularly tempting with can types as there are usually no other tagstrips nearby to connect the replacements to. The danger is if in the future the dried out capacitor breaks down.
The timing capacitor was replaced with a pair of 0.022uF Philips polycarbonate types with a 1600V rating. For this circuit, 630V types would be OK. While the value is not exactly 0.04uF as per the original, it was on the higher side, which, given a choice, is better than a lower value. Anyway, the waveform would be checked to ensure the replacement was suitable. The 0.5uF primary buffer was also replaced. Even though this capacitor is across the transformer primary, it is exposed to much more than 12V. Surprisingly, things were not well when I powered up again!

Capacitors replaced.

Vibrator.
The vibrator sounded distressed, and the waveform was awful with lots of high voltage spikes. Clearly something was amiss, but what? It appeared the timing capacitance was insufficient, and indeed, using 0.056uF instead gave a good waveform.
I had noticed previously that the vibrator sounded distressed for the split second it takes to come up to normal frequency. With some transformers, shunt drive vibrators will not start properly if there is insufficient timing capacitance. Insufficient timing capacitance is the same as saying the contact gap is too wide since the duty cycle is now less.  As I was beginning to suspect, the vibrator contact gap had increased slightly. Sure enough, the vibrator did not start if the input voltage was gradually raised. It would only start with a sudden application of voltage. When this happens, it means the contact gap has increased whereby the contact is not making connection with the reed weight directly opposite the pole piece. Also, the voltage required to get it to start this was was around 10V. This further confirmed my suspicion.
In this condition, sudden application of supply voltage will start the vibrator because the reed weight inertia will swing the reed further past the pole piece.

Last of the CDE vibrators were very simple and compact.

Time to open the vibrator and have a look. The thin aluminium made the job of peeling back the base crimp not too difficult and the damage was minimal. The contacts were in excellent condition and it appeared the gap had only increased from the hammering, rather than any electrical wear. This vibrator is of the last generation of the technology and is a very nicely designed simple and compact unit. There are no tungsten buttons on the reed, but instead the reed material is the actual contact. Tungsten buttons are still present on the fixed contacts.
A slight tweak of the contacts brought them back to where they should be, and the waveform was now correct with the new buffer condenser. Had the vibrator been a separate drive type, the contact gap could have increased without affecting the starting. The only thing to watch out for then is that the timing capacitance is still sufficient.

Vibrator waveform across the transformer primary. The primary buffer has around 47Vp-p across it.

12BE6.
With the power supply now working, some stations could be received, although very weakly. Sensitivity was very poor. I had noticed the 2K filter resistor for the B+ had been overheated in the past, and curious I checked the B+ voltage which was a bit low. Something drawing excess current in the front end would be the likely cause. When I measured the AVC, I was surprised to see +2.4V, even on a strong station. The AVC of course should never be positive. When this was shorted to earth, performance immediately improved to something more normal. Where could the positive volts be coming from? A likely source is a leaky AVC feed capacitor from the IF amplifier plate. The 4.7pF was disconnected, but the volts were still there. Next, I had a look at the tagstrip with the 0.05uF AVC filter connected to it. The adjacent tag was a B+ connection, so if the insulating material was leaky it could be the cause of the problem. Except it wasn't when I isolated the AVC from that tag. I wondered about the tuning unit in case there was some kind of leakage there, as the RF input coil is connected to the AVC. Disconnecting that didn't get rid of the volts either.
I had noticed that the positive volts always appeared soon after the B+ had come up, not at the same time. That led to suspicion the problem was actually grid emission from one of the AVC connected valves. It didn't take long to find the 12BE6 was the cause, and another one installed brought forth negative AVC volts and normal performance. At this point I replaced the AVC filter as it was a paper type, and also the screen bypass in case it became leaky later on.

Tuning Unit.
A clean up and lubrication of the tuner got the push buttons working, but the clutch needed work. The rubber clutch surface had gone brittle, was breaking up, and had become detached from the drive gear. In units of this type I've seen before, the clutch surface has often been cork. A suitable replacement might be cork gasket material, but I didn't have this so used rubber cut from a bicycle inner tube. This was contact cemented into position and looked good. Unfortunately, there was some slippage and I took the easy way out and applied some belt anti-slip compound to the rubber. Evidently, the rubber I used was did not have enough friction. I have seen at least one radio where the clutch had been glued together to stop it slipping. This is bad practice, since the tuning mechanism is not protected if someone keeps turning the tuning knob past the end of the dial. It also prevents the push buttons working properly.

Tuning clutch was in poor condition. New clutch surfaces were made from rubber.

Getting access to the clutch was awkward, but after doing it three times, I became quite adept at it. The front part housing the dial has to come off, then the front/bottom cover plate. The tuning unit is then detached by four screws behind the dial and one underneath. Once that's done, the clutch is removed by three screws on the side of the tuner. When replacing it, it's necessary to reset the anti-backlash mechanism. This is the standard set up with two sets of gear teeth side by side, spring loaded in opposite directions so there is always pressure on the drive gear teeth.

Tuning unit exposed for service. The sensitivity preset is visible to the top left of the tuner.

Audio Output.
Given the defective B+ filter caps, and that the cathode bypass for the 12AB5 was in the same can, suspicion fell upon this also. Bridging it with a 22uF electro brought up a bit more volume, along with more bass. Again, the original was disconnected and the new one connected in its place. It was also an opportune time to replace the overheated cathode resistor - a result of a previous faulty 12AB5?
One of the unused pins on the 12X4 socket had been used as a tie point for the junction of the 220R and 39R. It's a convenient thing to do by the manufacturer, but the danger is if a replacement valve is fitted that does have a connection to this pin. I left it as it was since there was no other convenient tie point, and besides, only I would ever service this radio.

Replaced parts.

Performance.
In my workshop the set suffers badly from electrostatic interference. I had this problem with the Detrola 6R. Not all car radios suffer from it, and I can only put it down to the RF input circuit design. When I tried a tuned loop aerial, the improvement was quite marked with the interference largely gone.
Away from the interference, the 66MF performs adequately but is not a  super-DX set. Not surprisingly, the sensitivity preset is required to be set at minimum resistance. Sound quality is noticeably good however, and the feedback and loudness circuit is very worthwhile. Current consumption is very low at just over 2A.
 

• See the set operating here https://youtu.be/SKjTPDtZL7o