Car Radio for the Model T Ford.

The receiver is used with a remote control unit and external speaker. The four pin socket to connect to the remote control can be seen here. Below are the speaker connections.

This is not the first radio to be installed in my Model T.  The original design was installed in the car just prior to the Parkes trip, Easter 2004, and worked faultlessly. It was something new and untried to have a radio in the Model T, so there was a lot to learn. My original set was a hybrid design using a TDA7000 for the receiver and valves for the audio. FM only reception was chosen because of the RFI caused by the Model T ignition system. By mid 2006, I had enough experience with this radio to make a few changes. For one thing, I was never really happy about the solid state method of reception. While the TDA7000 works very well, it was just out of place in a Model T. The new set would be all valve.
See the new set operating here

The physical aspects of the set are based on typical 1930's designs where all the circuitry is mounted in a box mounted in a remote location, usually under a seat or on the firewall. The controls are mounted on the dash or steering column, and operate the tuning condenser and volume control in the receiver box via Bowden cables.
In the case of a 1926 Model T Ford, the petrol tank occupies most of the behind dash space, so it is not practical to mount the receiver there. The best location is therefore under the front seat. With varicap diodes taking place of the tuning condenser, it is possible in the modern day to eliminate the Bowden cables, and simply connect the receiver to the remote control with ordinary wires. There is enough room under the dash for a small remote control, or alternatively, the steering column could be used. The remaining challenge is somewhere to put the speaker. For convenience, and unobtrusiveness, I mounted it on the steering column under the dash.

Limitations of a radio in the Model T
First problem and one that can't really be solved is the noise that one is subjected to in an open touring car. The only way to deal with this is the brute force method; try and make the radio louder. One thing that was immediately evident is that it is a waste of time to achieve high fidelity sound. With all the noise present, the lower level components in the audio signal aren't really heard anyway. The other thing noted with the sound was that good bass response was pointless. The low frequency components are masked by the rumbling of the engine.
The ignition system of a Model T is famed for the RF interference it causes, and initially it was thought there might be problems. This didn't turn out to be the case at all. Firstly this is because the radio was VHF FM, and again what interference finds its way into the radio is masked by the road and engine noise. So, that's a non issue, unless you want Medium Wave AM. Even then from the experiments I did try back in 2004, it seems that it might not be a lost cause. Bypass condensers around the supply for the ignition coils might just improve things.
The electrical system is a limiting factor for the design of the radio. With only 5A available without discharging the battery, the current consumption needs to be thought out carefully and minimised. Unfortunately, this means being stuck with a single ended output stage providing about 2.5W into the speaker. As much as I'd like a push pull output or my 6CM5 amplifier (3.8W) the extra current required is too much. It is true I could set the generator charge to a higher rate. This would however overcharge the battery when the radio wasn't on, and also the generator wear would be greater. The Model T uses a 3rd brush type of generator which functions as a constant current source. There is no voltage regulator. Charging current is set by the position of the 3rd brush on the commutator.
The other limitations concerning radio installation is where to put the radio. The under seat approach with a remote control under the dash works very well. That aspect of the original design would stay. The speaker position behind the steering column under the dash is acoustically poor but physically convenient and out of sight. A smaller speaker under the dash pointed at the occupants would probably be better.

The speaker box is secured above the steering column with hose clamps. This eliminates holes being drilled. The handbrake fully forward almost touches the box.

The New Design
One of the things I wanted in the new radio was a smaller remote control unit. Previously, this contained the TDA7000 receiving circuit and a slide rule dial, and did look obvious, much like a modern car radio. It was decided to move the receiver into the main unit which would simplify, and miniaturise the remote control unit. Additionally, this would tidy up the cabling, with the RG59 aerial cable only having to go so far as the main unit under the seat, instead of all the way up to the dash. Acquiring a small rectangular speaker that could handle 2.5W with tolerable quality wasn't easy, so for now the idea of speaker in the remote unit was overlooked and the existing speaker used.

Electrically, the new receiver wasn't going to be much different to the previous, as far as audio and power supply was concerned. However, the receiving circuit was now going to be my beloved 12AT7 super regenerative receiver. No matter how well it works, the TDA7000 is a solid state IC and just seemed out of place in an 80 year old car. The annoying thing I did find with the TDA7000 is how sharp the tuning is. Even with the gearing down of the dial it was still critical and an awful sound issued forth if the receiver was slightly off tune. With the super regenerative approach the sound merely becomes more distorted as tuning varies from the optimum setting. Using a super regenerative detector makes this an AM receiver (FM to AM conversion occurs by tuning the aerial coil slightly off frequency; i.e., slope detection). This might be assumed to be a problem so close to those four unsupressed ignition coils, but not so. Firstly, there's less RFI at VHF, and also a super regenerative detector has inherent noise suppression. How? It's because detection actually takes place only during the short periods when the detector builds up to oscillation. The rest of the time anything received is ignored. It actually acts as a sampling type of receiver with the short samples fed into a low pass filter to recreate the original modulation signal.
Obviously, with the new radio, tuning would have to be done by varicap diodes controlled by a pot at the remote control. Using a ten turn pot and ten turn dial eliminates the space consuming slide rule dial.

The remote control clamps to the lip under the dash avoiding having to make any holes. Power / volume is on left. Ten turn dial is locking type to prevent drift due to vibration. Next in line is the dial lamp. On the extreme right side can just be seen the regeneration control. It seldom needs adjustment.

As far as the audio section is concerned, there would be no negative feedback. The improvement in sound quality is totally wasted in this application. Additionally, it detracts from useful gain which is important when trying to minimise the amount of valves used. Furthermore, the frequency response would be concentrated with the mid and high range components. It's better to concentrate the power in the frequency range that can be heard above the engine noise.

The New Radio
Construction of the main unit chassis and enclosure was virtually a copy of the original, as it was easy to construct and served the purpose well. Layout was much the same too, but the height of the cabinet was reduced as the original was a tight fit under the seat cross member.
I also wanted the new radio to use the existing wiring harness between the main unit, remote control, and speaker. This meant six wires were available.
Let's now analyse the circuit:

The circuit of the Model T Ford radio receiver. Note the freedom from unnecessary parts...much like the car itself.

Super Regenerative Detector.
This 12AT7 circuit has been discussed elsewhere on this site so I won't repeat it all, except to mention a few slight differences. With DSE no longer stocking the 15uH RFC's I'd been using up to now, I have had to make them. They're of the same design as used in the 6C4 receivers. 76cm of 25B&S wound over a 6.3mm plastic former. While other 15uH chokes are available, they are of the axial type and do not work in this circuit. The home made chokes actually are better than the original 15uH ones and will allow the use of a 12AU7 instead of a 12AT7. Regeneration control is by means of a cathode rheostat, but in this receiver I've simplified the arrangement by not having any bleed current. An advantage of this also is the current consumption is reduced. Remember, the current consumption of anything drawing off the B+ is multiplied by about 40 times as far as the six volt supply is concerned. The other reason for doing this was to eliminate another wire in the cable up to the remote unit. To do this means a higher value of pot than used previously. A 10K pot turned out to be a good choice, shunted by a 39K to restrict the range. Furthermore, the cathode rheostat, with no bleed current, actually provides a degree of automatic regeneration control. The regeneration control, once set, does not normally need further adjustment when tuning across the band.

The receiver sub chassis undergoing testing and optimisation of the circuit.

One interesting thing that had me confused for a while was poor oscillation, especially at the 88Mc/s end of the band. Nothing made sense. I'd isolated the RF amp in case that was loading the detector. The circuit was checked. The chokes were identical to that in the 6C4 receiver which has no problem in this regard. Then something occurred to me...what if there was magnetic coupling between the cathode choke and the aerial coil and causing negative feedback. Then it wouldn't oscillate! But why in this receiver?
What a trap this was...I looked at the direction of the windings of both coils...took the choke out and reversed it. What a difference! Good solid oscillation all over the band at last.

Tuning by DC
Short of using bowden cables linked to a variable condenser, the tuning control had to use a potentiometer varying a DC voltage. The obvious choice is to use varicap diodes
for tuning and here I ran into problems. As expected the Q was less than a variable condenser and this prevented a 12AU7 being used. At first I tried a pair of BB105 diodes but couldn't get sufficient capacitance. Strange, as I'd used these in my 12V pulse counting receiver with no problems. I then tried MV2109's and at last was able to get the correct tuning range with a maximum reverse voltage of about nine. The MV2109's also had much better Q as made evident by stronger oscillation. One of the other worries about using varicaps is the tuning voltage supply drift. Rather than use a zener diode, the correct way is to use a special tuning voltage stabiliser IC, such as the TAA550 or ZTK33. While these appear electrically as a zener diode, they are constructed quite differently, and specifically designed to have good temperature stability. The regulated voltage is approximately 33V. Again, to simplify the wiring between units, the tuning pot is wired as a rheostat and shunts the tuning voltage to earth. The combination of the 82K, 4.7K and a 50K tuning pot provides the correct 88-108Mc/s coverage. The .22uF simply shunts any noise to earth and any remaining RF at the supply end of the 2.2M. Or so I thought. With the receiver up and running, I noticed the volume setting had an effect on the tuning. Even with the volume control wire disconnected at the receiver the effect was still there. Something capacitive must be happening. The only other wire going into the wiring harness was the tuning wire.
Well, despite the .22uF there must have been RF flowing where it shouldn't. The 150K decoupler fixed that and now the receiver was so much more docile and behaved just like it should.

Close up of where most of the action is. The minimum volume preset is upper left, with the power supply filtering components in front of it. Two of the RFC's are visible along with the aerial coil towards the bottom of the 12AT7 sub chassis. To the right is the 10,000uF electrolytic for filtering the incoming 6v supply. The .01uF 3KV buffer condenser is clearly visible behind the 12AT7.

The DC volume control.
This is a novel piece of circuitry I developed for the original radio. Apart from requiring only one wire to control the volume rather than three, any noise picked up on the wire would be shunted to earth by a condenser, as we only want the DC voltage.
As before, I used a 6CS6, known in Europe as EH90. It's a pentagrid valve similar to a 6BE6, and was popular for TV use in Australia as a noise gated sync separator. By feeding a signal into one of the control grids, the other control grid voltage will affect the gain.  In my circuit, an even better range of control is obtained by varying the DC on both control grids. That is simply done by earthing both grids (DC wise) and varying the cathode voltage. In case you are wondering why I didn't use a 6BA6 or other remote cutoff pentode, it's because of the range of control voltage required. The 6BA6 required about 30V to cut it off; the 6CS6 about 16V. Why does this matter? Again, a rheostat wired pot is used to control the volume. With the pot at minimum resistance, the bleed current is fairly high in order to obtain 30V at max resistance; compared to when only 16V is required. There's a preset 100K pot in the main unit to set the bleed current, and therefore the minimum volume, when the volume control is set to it's maximum 10K resistance.
The DC volume control works extremely well. From a user point of view it is indistinguishable from the conventional audio voltage divider type of circuit. Gain from this section also brings up the detected audio level to drive the 6AQ5 output.

Inside the remote control unit. From left to right is the on/off volume  control, ten turn tuning pot, dial lamp, and to the side is the regeneration control.

The power amplifier
This is a stock standard 6AQ5 output stage. I kept to the 6AQ5 as it provides the most power output for the least heater current, and I have many of them. As mentioned before, no feedback is applied. The input signal has the bass frequencies attenuated by the .0047uF grid coupler. Realistically, the output power into the speaker is about 2.5W which is typical for a cathode biassed 6AQ5 running off about 250V B+. Plate voltage is effectively lessened by the drop across the speaker transformer and cathode resistor. One thing I was determined to do is use a proper valve output transformer. Previously I'd used a 100V line transformer which did a good job, but with not an exact impedance match, and interleaved laminations, wasn't the most efficient thing to do. I decided on a chunky Rola B23 transformer. The core looks like it could easily handle 10W. The impedance ratio is 5000 to 15 ohms which is just right for a 6AQ5. It wasn't right for the 8 ohm speaker however, so I simply unwound the secondary and brought a tap out. I rewound the rest of the secondary back on so I'd still have the option of using a 15 (or 16) ohm speaker in the future.

The power supply
No prizes for guessing I'd use a vibrator power supply; after all that's part of the charm of a valve car radio. I had hoped to use a synchronous vibrator in the normal self rectifying circuit, but the only transformer I had for this purpose seemed to have a strange characteristic or some sort of fault. It was a new old stock transformer made by Astor. It was labelled as 6V which was what I needed, and indeed when I built the radio up and used it, there was the 250V at 50mA I wanted. Alas, before the radio warmed up and was drawing B+ current, the no load voltage was horrendously high. The secondary vibrator contacts were arcing over, not only between themselves, but through the mica washers in the base of the vibrator. The .01uF buffer condenser rated at 3KV eventually went up in smoke! This transformer was obviously a lost cause, so the next in turn to try was a mains transformer, with a secondary I'd rewound to give a suitable turns ratio for 6 to 250V vibrator use. It was a Telefunken transformer that looked like it was from the 1960's. I have no idea what it was out of, with a secondary voltage of about 7 and tapping of about 3.8V. Conveniently, this was rather close to what I needed, and the core size and wire gauge looked perfect for what I wanted. As it was, despite the uneven "centre tap" I powered up the transformer from a vibrator and it worked quite well. However, this winding now functioning as the primary was contributing to DC through the core with the asymmetrical tapping. It would have been tolerable if necessary, but I did the right thing and rewound it, bringing up the efficiency by a worthwhile amount.

Vibrator waveform is text book quality.

I like the idea of a self rectifying synchronous vibrator, but alas with a single 250V winding it wasn't going to happen in this radio. So, in with the 1N4007's. Not quite keeping in with the theme but they work well, and are reliable with the high unloaded voltages from a vibrator supply. Schemes involving valve rectifiers were ruled out given the extra current consumption. I retained the synchronous vibrator, and simply paralleled the primary and secondary contacts to convert it to non synchronous operation. The vibrator is an Oak V5124.
On the six volt side of things, the heaters and inverter are switched by a relay. This is done to eliminate the stress on the power switch contacts. It also eliminates another wire to the remote control unit, as the radio is turned on by earthing only one wire. Incidentally, to keep down on the wiring between units, the switch wire is common with one of the speaker wires.
The dial lamp is actually in series with the relay coil. This eliminates the extra current consumption that would come from simply wiring the dial lamp across the 6V supply.
It may be thought that a 6V relay would not function with so much voltage dropped across the lamp (about 3V). It works perfectly, making use of two well known properties about both components. Firstly, an incandescent lamp has a low filament resistance when cold; i.e., soon as the radio is switched on. Secondly, the holding current required for a relay coil is a fraction of that required to pull it in. So when turned on, the relay receives pretty much full current. The lamp warms up in less than a second by which time the relay has pulled in.

The finished radio. The 6AQ5 is between the vibrator and speaker transformer, with the 6CS6 beneath it. Note the size of the speaker transformer to right of the 6AQ5.

Wiring and components can be seen from this side. At the left is the super-regenerative detector and audio circuitry. At the right is the power supply.

Much to my delight, the vibrator interference was easy to eliminate. First thing was to add the 10,000uF condenser right at the vibrator socket. Car radio manufacturers would have loved such high values back in the 1950's! This eliminated most of it, but being a perfectionist I tried connecting a 1uF polyester against parts of the 6V supply and found another chunk of noise gone at the relay contacts. The last of it disappeared with another 1uF (the yellow thing in the pic above) from the fuseholder to earth. Elimination of vibrator interference is really an art. You can't tell the steps needed until the radio has actually been built.
After all this, the radio was putting on a really impressive performance, working as well as a mains powered unit. The case was etch primed and painted in green enamel. Installed in the car with the existing aerial there was no doubt about the sensitivity, with 2ST from the southern highlands coming in better than it does with most receivers.

The radio fits under the front seat. The black box to the right is a 6 to 12v converter used to power appliances such as a car fridge and air compressor.

The Aerial.

Aerial is a 75cm brass rod.

The aerial is a 75cm brass rod attached to a PL-259 plug. It is supported in the plug body, and insulated from it by a piece of plastic rod turned down to suit. This bites into the thread of the plug body making a very secure attachment.
The PL-259 plug is then  attached to a panel mount joiner, which is mounted on a Marvi-Plate bracket secured to the number plate mounting. From the other end of the joiner, the RG-59 coaxial cable connects via another PL-259 plug. The coax was shrouded in black shoe lace to give it the vintage appearance.

Close up of how the aerial is mounted.

How does it perform?
Sensitivity while driving turned out to be better than expected. There wasn't the fading I'd anticipated as the car moved through the standing waves. My decision to change the audio frequency response by diverting all the output power to the mid and high frequencies turned out to be worthwhile. The sound just seems to be a bit clearer and easier to listen to.
As expected, one of the main improvements was ease of tuning. Partly due to the use of a ten turn pot, but mainly due to the wide bandwidth of the super regenerative detector.
However, there is some drift during warm up.  It would seem that the inclusion of a carefully chosen thermistor could be used to offset the drift by being connected into the tuning voltage supply circuit.  When you consider this is a free running oscillator operating at 100 odd megacycles, it isn't actually too bad.
The use of a proper aerial cut to the correct length is one of the key things that makes this radio perform well. (Hint to those who insist on using bits of wire for an aerial).
Some ignition interference is evident, which I did expect, but it didn't blot out the reception, and was not annoying enough to worry about it. In any case, the road and engine noise does a good job of masking it.
The new, smaller control unit looks far more appropriate than the previous receiver, and no difficulty was found in adjusting the controls. For most reception the regeneration control is seldom adjusted.
The speaker location is not optimum, especially for freeway driving. In fact it's in the worst possible position, pointing down at the floor above the noisy exhaust system. Experiments have shown that a forward facing speaker mounted at the top of the windscreen provides the best performance. Experiments were also done with a speaker (as shown in the YouTube video) mounted at the bottom of the dash. Ultimately, I would like to rebuild the control unit with a more period correct dial, and include the speaker inside.